[ { "path": ".appveyor.yml", "content": "version: 1.0.{build}-{branch}\n\nbranches:\n except:\n - gh-pages\n\nskip_tags: true\n\nskip_commits:\n files:\n - .drone.star\n - .github/**\n - android/**\n - docs/**\n - src/**/*_posix.*\n - src/**/*_linux.*\n - src/**/*_gnulinux.*\n - src/**/*_x11.*\n - src/**/*_gtk.*\n - src/**/*_android.*\n - src/**/*_mac.*\n - LICENSE\n - README.md\n\nskip_branch_with_pr: true\n\npull_requests:\n do_not_increment_build_number: true\n\nos: Visual Studio 2019\n\ninit:\n - ps: |\n If (-Not $env:APPVEYOR_PULL_REQUEST_NUMBER) {\n $env:is_not_pr = \"true\"\n }\n If (-Not $env:APPVEYOR_REPO_COMMIT_MESSAGE_EXTENDED) {\n $env:APPVEYOR_REPO_COMMIT_MESSAGE_EXTENDED = \" \"\n }\n\ninstall:\n - |\n vcpkg integrate remove\n xb setup\n\nplatform: Windows\n\nconfiguration: [Release, Checked]\n\nbuild_script:\n - xb build --config=%CONFIGURATION% --target=src\\xenia-app --target=tests\\xenia-base-tests --target=tests\\xenia-cpu-ppc-tests --target=src\\xenia-vfs-dump\n\nafter_build:\n - |\n IF NOT \"%CONFIGURATION%\"==\"Checked\" SET \"ARCHIVE_SUFFIX=%APPVEYOR_REPO_BRANCH%\"\n IF NOT \"%CONFIGURATION%\"==\"Checked\" SET \"ARCHIVE_SWITCHES=--\"\n IF \"%CONFIGURATION%\"==\"Checked\" SET \"ARCHIVE_SUFFIX=%APPVEYOR_REPO_BRANCH%_FOR-DEVS-ONLY\"\n IF \"%CONFIGURATION%\"==\"Checked\" SET \"ARCHIVE_SWITCHES=\"-pI know what I am doing.\" --\"\n 7z a xenia_%ARCHIVE_SUFFIX%.zip %ARCHIVE_SWITCHES% LICENSE \"%APPVEYOR_BUILD_FOLDER%\\build\\bin\\%PLATFORM%\\%CONFIGURATION%\\xenia.exe\" \"%APPVEYOR_BUILD_FOLDER%\\build\\bin\\%PLATFORM%\\%CONFIGURATION%\\xenia.pdb\"\n 7z a xenia-vfs-dump_%ARCHIVE_SUFFIX%.zip %ARCHIVE_SWITCHES% LICENSE \"%APPVEYOR_BUILD_FOLDER%\\build\\bin\\%PLATFORM%\\%CONFIGURATION%\\xenia-vfs-dump.exe\" \"%APPVEYOR_BUILD_FOLDER%\\build\\bin\\%PLATFORM%\\%CONFIGURATION%\\xenia-vfs-dump.pdb\"\n\nbefore_test:\n - xb gentests\n\ntest_script:\n - xb test --config=%CONFIGURATION% --no_build\n\nartifacts:\n - path: '*.zip'\n - path: xenia-cpu-ppc-test.log\n\ndeploy:\n - provider: Environment\n name: xenia-master\n release: xenia-$(appveyor_repo_branch)-v$(appveyor_build_version)\n artifact: '*.zip'\n draft: false\n prerelease: true\n on:\n branch: master\n configuration: release\n appveyor_repo_tag: true\n is_not_pr: true\n - provider: GitHub\n name: xenia-master\n repository: xenia-project/release-builds-windows\n auth_token:\n secure: /8he47z1WnPN7LcCTe5T5KMxxX0SmqFj9QMpeWEa3aZ64kMsfupOT/jKakqTM8af\n tag: v$(appveyor_build_version)\n release: v$(appveyor_build_version)\n description: |\n Windows release build for https://github.com/xenia-project/xenia/commit/$(APPVEYOR_REPO_COMMIT).\n \n $(APPVEYOR_REPO_COMMIT_MESSAGE)\n \n $(APPVEYOR_REPO_COMMIT_MESSAGE_EXTENDED)\n draft: false\n prerelease: false\n on:\n branch: master\n configuration: release\n is_not_pr: true\n" }, { "path": ".clang-format", "content": "---\nBasedOnStyle: Google\nDerivePointerAlignment: false\nPointerAlignment: Left\nSortIncludes: true\n\n# Regroup causes unnecessary noise due to clang-format bug.\nIncludeBlocks: Preserve\n\n---\nLanguage: Java\nDisableFormat: true\nSortIncludes: false\n" }, { "path": ".drone.star", "content": "def main(ctx):\n return [\n pipeline_lint(),\n pipeline_linux_desktop('x86_64-linux-clang', image_linux_x86_64(), 'amd64', 'clang', True),\n pipeline_linux_desktop('x86_64-linux-gcc', image_linux_x86_64(), 'amd64', 'gcc', False), # GCC release linking is really slow\n pipeline_android('x86_64-android', image_linux_x86_64(), 'amd64', 'Android-x86_64'),\n pipeline_android('aarch64-android', image_linux_x86_64(), 'amd64', 'Android-ARM64'),\n ]\n\ndef image_linux_x86_64():\n return 'xeniaproject/buildenv:2022-07-15'\n\ndef volume_build(toolchain, path='/drone/src/build'):\n return {\n 'name': 'build-' + toolchain,\n 'path': path,\n }\n\ndef command_cc(cc):\n # set CC, CXX, ...\n return 'export $(cat /{}.env | sed \\'s/#.*//g\\' | xargs)'.format(cc)\n\ndef command_ndk_build(platform, configuration, target):\n return '$ANDROID_NDK_ROOT/build/ndk-build NDK_PROJECT_PATH:=./bin/{configuration} NDK_APPLICATION_MK:=./xenia.Application.mk PREMAKE_ANDROIDNDK_PLATFORMS:={platform} PREMAKE_ANDROIDNDK_CONFIGURATIONS:={configuration} -j$(nproc) {target}'.format(platform=platform, configuration=configuration, target=target)\n\ndef xenia_base_tests_filters():\n # https://github.com/xenia-project/xenia/issues/2036\n return 'exclude:\"Wait on Timer\" exclude:\"Wait on Multiple Timers\" exclude:\"HighResolutionTimer\"'\n\n# Run lint in a separate pipeline so that it will try building even if lint fails\ndef pipeline_lint():\n return {\n 'kind': 'pipeline',\n 'type': 'docker',\n 'name': 'lint',\n 'steps': [\n {\n 'name': 'lint',\n 'image': image_linux_x86_64(),\n 'commands': [\n 'clang-format --version',\n './xenia-build lint --all',\n ],\n },\n ],\n }\n\ndef pipeline_linux_desktop(name, image, arch, cc, build_release_all):\n return {\n 'kind': 'pipeline',\n 'type': 'docker',\n 'name': name,\n 'platform': {\n 'os': 'linux',\n 'arch': arch,\n },\n # These volumes will be mounted at the build directory, allowing to\n # run different premake toolchains from the same source tree\n 'volumes': [\n {\n 'name': 'build-premake',\n 'temp': {},\n },\n {\n 'name': 'build-cmake',\n 'temp': {},\n },\n ],\n\n 'steps': [\n #\n # Setup the source tree\n #\n {\n 'name': 'clone-submodules',\n 'image': image,\n 'commands': [\n 'pwd',\n # May miss recursive submodules (but faster than xb setup)\n 'git submodule update --init --depth 1 -j $(nproc)',\n ],\n },\n\n\n #\n # Setup the two build systems\n #\n\n # Native premake Makefiles for production\n {\n 'name': 'toolchain-premake',\n 'image': image,\n 'volumes': [volume_build('premake')],\n 'commands': [\n command_cc(cc),\n '$CXX --version',\n 'python3 --version',\n './xenia-build premake --cc={}'.format(cc),\n ],\n 'depends_on': ['clone-submodules'],\n },\n\n # Development toolchain\n {\n 'name': 'toolchain-cmake',\n 'image': image,\n 'volumes': [volume_build('cmake')],\n 'commands': [\n command_cc(cc),\n '''\n ./xenia-build premake --cc={} --devenv=cmake\n cd build\n for c in Debug Release\n do\n mkdir cmake-$c\n cmake -S . -B cmake-$c -G Ninja -DCMAKE_BUILD_TYPE=$c\n done\n '''.format(cc),\n ],\n # Premake itself needs to be build first:\n 'depends_on': ['toolchain-premake'],\n },\n\n #\n # Building\n #\n\n {\n 'name': 'build-premake-debug-tests',\n 'image': image,\n 'volumes': [volume_build('premake')],\n 'commands': [\n command_cc(cc),\n './xenia-build build --no_premake -j$(nproc) --config=Debug --target=xenia-base-tests',\n ],\n 'depends_on': ['toolchain-premake'],\n },\n {\n 'name': 'build-premake-debug-all',\n 'image': image,\n 'volumes': [volume_build('premake')],\n 'commands': [\n command_cc(cc),\n './xenia-build build --no_premake -j$(nproc) --config=Debug',\n ],\n 'depends_on': ['build-premake-debug-tests'],\n },\n\n {\n 'name': 'build-premake-release-tests',\n 'image': image,\n 'volumes': [volume_build('premake')],\n 'commands': [\n command_cc(cc),\n './xenia-build build --no_premake -j$(nproc) --config=Release --target=xenia-base-tests',\n ],\n 'depends_on': ['toolchain-premake'],\n },\n ] + ([\n {\n 'name': 'build-premake-release-all',\n 'image': image,\n 'volumes': [volume_build('premake')],\n 'commands': [\n command_cc(cc),\n './xenia-build build --no_premake -j$(nproc) --config=Release',\n ],\n 'depends_on': ['build-premake-release-tests'],\n },\n ] if build_release_all else []) + [\n\n {\n 'name': 'build-cmake-debug-all',\n 'image': image,\n 'volumes': [volume_build('cmake')],\n 'commands': [\n command_cc(cc),\n 'cd build/cmake-Debug',\n 'cmake --build . -j$(nproc)',\n ],\n 'depends_on': ['toolchain-cmake'],\n },\n\n {\n 'name': 'build-cmake-release-tests',\n 'image': image,\n 'volumes': [volume_build('cmake')],\n 'commands': [\n command_cc(cc),\n 'cd build/cmake-Release',\n 'cmake --build . -j$(nproc) --target xenia-base-tests',\n ],\n 'depends_on': ['toolchain-cmake'],\n },\n ] + ([\n {\n 'name': 'build-cmake-release-all',\n 'image': image,\n 'volumes': [volume_build('cmake')],\n 'commands': [\n command_cc(cc),\n 'cd build/cmake-Release',\n 'cmake --build . -j$(nproc)',\n ],\n 'depends_on': ['build-cmake-release-tests'],\n },\n ] if build_release_all else []) + [\n\n\n #\n # Tests\n #\n\n {\n 'name': 'test-premake-debug-valgrind',\n 'image': image,\n 'volumes': [volume_build('premake')],\n 'commands': [\n 'valgrind --error-exitcode=99 ./build/bin/Linux/Debug/xenia-base-tests --durations yes ' + xenia_base_tests_filters(),\n ],\n 'depends_on': ['build-premake-debug-tests'],\n },\n\n {\n 'name': 'test-premake-release',\n 'image': image,\n 'volumes': [volume_build('premake')],\n 'commands': [\n './build/bin/Linux/Release/xenia-base-tests --success --durations yes ' + xenia_base_tests_filters(),\n ],\n 'depends_on': ['build-premake-release-tests'],\n },\n\n {\n 'name': 'test-cmake-release',\n 'image': image,\n 'volumes': [volume_build('cmake')],\n 'commands': [\n './build/bin/Linux/Release/xenia-base-tests --success --durations yes ' + xenia_base_tests_filters(),\n ],\n 'depends_on': ['build-cmake-release-tests'],\n },\n\n\n #\n # Stat\n #\n\n {\n 'name': 'stat',\n 'image': image,\n 'volumes': [\n volume_build('premake', '/build-premake'),\n volume_build('cmake', '/build-cmake'),\n ],\n 'commands': [\n '''\n header() {\n SEP='============================================================'\n echo\n echo $SEP\n echo $@\n echo $SEP\n }\n\n for v in premake cmake\n do\n for c in Debug Release\n do\n header $v $c\n p=/build-$v/bin/Linux/$c\n ls -la $p\n sha256sum $p/*\n done\n done\n '''\n ],\n 'depends_on': [\n 'build-premake-debug-all',\n 'build-cmake-debug-all',\n ] + ([\n 'build-premake-release-all',\n 'build-cmake-release-all',\n ] if build_release_all else [\n 'build-premake-release-tests',\n 'build-cmake-release-tests',\n ]),\n },\n ],\n }\n\n\ndef pipeline_android(name, image, arch, platform):\n return {\n 'kind': 'pipeline',\n 'type': 'docker',\n 'name': name,\n 'platform': {\n 'os': 'linux',\n 'arch': arch,\n },\n\n 'steps': [\n #\n # Setup the source tree\n #\n {\n 'name': 'clone-submodules',\n 'image': image,\n 'commands': [\n 'pwd',\n # May miss recursive submodules (but faster than xb setup)\n 'git submodule update --init --depth 1 -j $(nproc)',\n ],\n },\n\n\n #\n # Build premake and generate NDK makefiles\n #\n\n # NDK Makefiles\n {\n 'name': 'toolchain',\n 'image': image,\n 'commands': [\n 'c++ --version',\n 'python3 --version',\n './xenia-build premake --target_os android',\n ],\n 'depends_on': ['clone-submodules'],\n },\n\n\n #\n # Building\n #\n {\n 'name': 'build-debug',\n 'image': image,\n 'commands': [\n 'cd build',\n command_ndk_build(platform, 'Debug', 'all'),\n ],\n 'depends_on': ['toolchain'],\n },\n\n {\n 'name': 'build-release',\n 'image': image,\n 'commands': [\n 'cd build',\n command_ndk_build(platform, 'Release', 'all'),\n ],\n 'depends_on': ['toolchain'],\n },\n\n\n #\n # Stat\n #\n {\n 'name': 'stat',\n 'image': image,\n 'commands': [\n '''\n header() {\n SEP='============================================================'\n echo\n echo $SEP\n echo $@\n echo $SEP\n }\n\n for c in Debug Release\n do\n header $c\n p=build/bin/$c/obj/local/*\n ls -la $p\n sha256sum $p/* || true\n done\n '''\n ],\n 'depends_on': [\n 'build-debug',\n 'build-release',\n ],\n },\n ],\n }\n" }, { "path": ".gdbinit", "content": "# Ignore HighResolutionTimer custom event\nhandle SIG34 nostop noprint\n# Ignore PosixTimer custom event\nhandle SIG35 nostop noprint\n# Ignore PosixThread exit event\nhandle SIG32 nostop noprint\n# Ignore PosixThread suspend event\nhandle SIG36 nostop noprint\n# Ignore PosixThread user callback event\nhandle SIG37 nostop noprint\n" }, { "path": ".gitattributes", "content": "* text=auto whitespace=blank-at-eol,tab-in-indent,trailing-space,tabwidth=2\n\n*.bat text eol=crlf\n*.sh text eol=lf\n\n*.sln text eol=crlf -whitespace\n*.csproj text eol=crlf -whitespace merge=union\n*.vcxproj text eol=crlf -whitespace merge=union\n*.props text eol=crlf -whitespace merge=union\n\nsrc/**/shaders/bytecode/** linguist-generated=true\n" }, { "path": ".github/CONTRIBUTING.md", "content": "# Content Guidelines\n\nThe issue tracker is exclusively for filing and discussing bugs, feature\nrequests, and tracking work items. It is not for technical support or general\ndiscussion. Avoid discussing any illegal activity, such as downloading games.\n\n**Repeated misuses will result in a permanent project ban.**\n\n## Information Sourcing\n\nAll information in xenia has been derived from reverse engineering legally-owned\ngames, hardware, and tools made public by Microsoft (such as the XNA Game Studio\ntooling), scouring documentation made public by Microsoft (such as slide decks\nand other presentations at conferences), and information from code made public\nby 3rd party companies (like the Valve SDKs).\n\nThe official Microsoft Xbox Development Kits (XDKs) are not to be used for any\ninformation added to the project. The contributors do not want the XDKs, nor do\nthey want any information derived from them. The challenge of the project is\nwhat makes it fun! Poisoning the codebase with code obtained by shady means\ncould result in the project being terminated, so just don't do it.\n\n**Posting any information directly from an XDK will result in a project ban.**\n\n# Contributing Code\n\n## Style Guide\n\nPlease read over [style_guide.md](../docs/style_guide.md) before sending pull requests\nand ensure your code is clean as the buildbot (or I) will make you to fix it :)\n[style_guide.md](../docs/style_guide.md) has information about using `xb format` and\nvarious IDE auto formatting tools so that you can avoid having to clean things\nup later, so be sure to check it out.\n\nBasically: run `xb format` before you add a commit and you won't have a problem.\n\n## Referencing Sources\n\nIn code interacting with guest interfaces or protocols, where applicable, please\nleave comments describing how the information included in it was obtained. For\ncode based on analysis of the response of the original Xbox 360 hardware or\nsoftware, please provide reproduction information, such as an outline of the\nalgorithm executed, arguments and results of function calls or processor\ninstructions involved, GPU or XMA commands and state register changes. Having\ntraceable sources helps solve multiple issues:\n\n* The legality of the submitted code can be verified easily.\n* Additional analysis based on reproduction steps from prior research can be\n done to discover more details or to research the behavior of other related\n features.\n* The accuracy and completeness of the information can be evaluated. Knowing\n whether something is ground truth about the console's behavior or an\n approximation (for example, based on similar functionality in Windows, the\n Qualcomm Adreno 200 GPU, AMD's desktop GPU architectures; the Direct3D 11.3\n functional specification, which may be used as a generic fallback when no\n information specific to the Xenos or Direct3D 9 is available) may help avoid\n redoing work that has already been done if the findings are accurate, or\n making potentially wrong conclusions about related functionality if there's no\n certainty about the correctness of the information. In addition, it's useful\n to describe how complete the implementation of a feature is — such as edge\n cases checked and covered. If you are unsure if your code accurately reflects\n the behavior of the console, or you have deliberately made deviations due to\n the lack of prerequisites for an accurate implementation or for performance\n reasons (in case of the latter, it's recommended to provide both options,\n selectable with a configuration variable), or you just want more research to\n be done in the future, please leave a TODO comment in the format provided in\n [style_guide.md](../docs/style_guide.md).\n\nIf you have verified your code by checking the correctness of the behavior of a\ngame, **do not** refer to it by its title trademark. To avoid unnecessary\ndependencies on third parties, instead, use the hexadecimal title ID number\ndisplayed in the title bar beside the name of the game. It's also recommended to\navoid using proper names of game content if they can be replaced with easily\nunderstandable pointers not including them, such as \"first mission\",\n\"protagonist\", \"enemy aircraft\".\n\nDo not leave any hard-coded references to specific games, even in title ID form,\nin any part of the user interface, including the configuration file. If you want\nto provide an example of a game where changing a configuration variable may have\na noticeable effect, use a code comment near the declaration of the variable\nrather than its description string. Any game identifiers referenced in the user\ninterface must be obtained only from information provided by the user such as\ngame data files.\n\nAlso, do not put any conditionals based on hard-coded identifiers of games — the\ntask of the project is researching the Xbox 360 console itself and documenting\nits behavior by creating open implementations of its interfaces. Game-specific\nhacks provide no help in achieving that, instead only complicating research by\nintroducing incorrect state and hiding the symptoms of actual issues. While\ntemporary workarounds, though discouraged, may be added in cases when progress\nwould be blocked otherwise in other areas, they must be expressed and reasoned\nin terms of the common interface rather than logic internal to a specific game.\n\n## Clean Git History\n\nTools such as `git bisect` are used on the repository regularly to check for and\nidentify regressions. Such tools require a clean git history to function\nproperly. Incoming pull requests must follow good git rules, the most basic of\nwhich is that individual commits add functionality in somewhat working form and\nfully compile and run on their own. Small pull requests with a single commit are\nbest and multiple commits in a pull request are allowed only if they are\nkept clean. If not clean, you will be asked to rebase your pulls (and if\nyou don't know what that means, avoid getting into that situation ;).\n\nExample of a bad commit history:\n\n* Adding audio callback, random file loading, networking, etc. (+2000 lines)\n* Whoops.\n* Fixing build break.\n* Fixing lint errors.\n* Adding audio callback, second attempt.\n* ...\n\nHistories like this make it extremely difficult to check out any individual\ncommit and know that the repository is in a good state. Rebasing,\ncherry-picking, or splitting your commits into separate branches will help keep\nthings clean and easy.\n\n# License\n\nAll xenia code is licensed under the 3-clause BSD license as detailed in\n[LICENSE](../LICENSE). Code under `third_party/` is licensed under its original\nlicense.\n\nIncoming code in pull requests are subject to the xenia [LICENSE](../LICENSE).\nOnce code comes into the codebase it is very difficult to ever fully remove so\ncopyright is ascribed to the project to prevent future disputes such as what\noccurred in [Dolphin](https://dolphin-emu.org/blog/2015/05/25/relicensing-dolphin/).\nThat said: xenia will never be sold, never be made closed source, and never\nchange to a fundamentally incompatible license.\n\nAny `third_party/` code added will be reviewed for conformance with the license.\nIn general, GPL code is forbidden unless it is used exclusively for\ndevelopment-time tooling (like compiling). LGPL code is strongly discouraged as\nit complicates building.\n" }, { "path": ".github/ISSUE_TEMPLATE/bug_report.yaml", "content": "name: Bug report\ndescription: Template for bug reports.\nlabels: bug\nbody:\n - id: disclaimer\n type: markdown\n attributes:\n value: |\n Try to create a very concise title that's straight to the point.\n\n **THIS IS NOT A SUPPORT FORUM!** For support, first read the wiki: https://github.com/xenia-project/xenia/wiki\n If your question wasn't answered there or you need help, proceed to #help in the Discord server: https://discord.gg/Q9mxZf9\n\n DO NOT CREATE ISSUES ABOUT SPECIFIC GAMES IN THIS REPOSITORY!\n A game specific issue would be e.g. \"Game X crashes after you hit a character a certain way\"\n A Xenia issue would be e.g. \"Kernel export NtDoSomething does nothing\"\n For specific games, visit https://github.com/xenia-project/game-compatibility#game-compatibility\n - id: validation\n type: checkboxes\n attributes:\n label: Validation\n options:\n - label: I've read the [FAQ](https://github.com/xenia-project/xenia/wiki/FAQ).\n required: true\n - label: The Xenia build used is from the master branch (not MLBS/AlexVS/Canary/pull requests, etc.)\n required: true\n - label: This issue isn't for tech support (help with Xenia).\n required: true\n - label: If this issue occurs in a specific game, I've done analysis to locate the faulty subsystem of the emulator and a potential reason in it.\n required: true\n - label: I've checked if this issue hasn't already been reported.\n required: true\n - label: 'My device meets the minimum requirements: https://github.com/xenia-project/xenia/wiki/Quickstart#system-requirements'\n required: true\n - label: '(If building) I have read the building doc: https://github.com/xenia-project/xenia/blob/master/docs/building.md'\n - id: problem\n type: textarea\n attributes:\n label: Describe what's going wrong\n validations:\n required: true\n - id: what-should-happen\n type: textarea\n attributes:\n label: Describe what should happen\n validations:\n required: true\n - id: callstack\n type: textarea\n attributes:\n label: If applicable, provide a callstack here, especially for crashes\n - id: logfile\n type: textarea\n attributes:\n label: If applicable, upload a logfile and link it here\n" }, { "path": ".github/ISSUE_TEMPLATE/config.yml", "content": "contact_links:\n - name: Xenia Discord server\n url: https://discord.gg/Q9mxZf9\n about: Tech support for Xenia belongs here.\n" }, { "path": ".gitignore", "content": "# ==============================================================================\n# Misc system junk\n# ==============================================================================\n\n.DS_Store\n._*\n.Spotlight-V100\n.Trashes\n.com.apple.*\nThumbs.db\nDesktop.ini\n.svn\n\n# Microprofile settings\n.microprofilepreset.*\n\n# ==============================================================================\n# Projects/IDE files\n# ==============================================================================\n\n*~\n\n# Sublime Text\n*.sublime-project\n*.sublime-workspace\n\n# VIM\n.*.sw[a-z]\n*.un~\nSession.vim\n\n# TextMate\n*.tmproj\n*.tmproject\ntmtags\n\n# Eclipse\n.project\n.metadata\n\n# WebStorm\n.idea\n\n# VS\n.vs\n*.user\n*.sdf\n*.opensdf\n*.suo\nbin/\nobj/\n\n# VSCode\n.vscode\n\n# ==============================================================================\n# Temp generated code\n# ==============================================================================\n\n*.py[co]\n.coverage\n*.o\n*.aps\n\n# ==============================================================================\n# Logs and dumps\n# ==============================================================================\n\nnpm-debug.log\nprivate/\n*.trace\nimgui.ini\n*.log\n\n# ==============================================================================\n# Build system output\n# ==============================================================================\n\n# npm/node\n.lock-wscript\nnode_modules/\nnode_modules/**/build/\nnode_modules/.bin/\n\n# coverage/etc\n/scratch/\n\n/build/\n\n# ==============================================================================\n# Local-only paths\n# ==============================================================================\n\n.vagrant\n/attic/\n/content/\n/third_party/binutils/binutils-2.24.tar.gz\n/third_party/binutils/bin/\n/third_party/binutils/powerpc-none-elf/\n/third_party/binutils/share/\n/third_party/binutils/binutils*\n/third_party/vasm/\n/tools/shader-playground/*.dll\n" }, { "path": ".gitmodules", "content": "[submodule \"third_party/xbyak\"]\n\tpath = third_party/xbyak\n\turl = https://github.com/xenia-project/xbyak.git\n[submodule \"third_party/imgui\"]\n\tpath = third_party/imgui\n\turl = https://github.com/ocornut/imgui.git\n[submodule \"third_party/binutils-ppc-cygwin\"]\n\tpath = third_party/binutils-ppc-cygwin\n\turl = https://github.com/benvanik/binutils-ppc-cygwin.git\n[submodule \"third_party/catch\"]\n\tpath = third_party/catch\n\turl = https://github.com/catchorg/Catch2.git\n[submodule \"third_party/premake-core\"]\n\tpath = third_party/premake-core\n\turl = https://github.com/xenia-project/premake-core.git\n[submodule \"third_party/snappy\"]\n\tpath = third_party/snappy\n\turl = https://github.com/xenia-project/snappy.git\n[submodule \"third_party/premake-export-compile-commands\"]\n\tpath = third_party/premake-export-compile-commands\n\turl = https://github.com/xenia-project/premake-export-compile-commands.git\n[submodule \"third_party/discord-rpc\"]\n\tpath = third_party/discord-rpc\n\turl = https://github.com/discordapp/discord-rpc.git\n[submodule \"third_party/rapidjson\"]\n\tpath = third_party/rapidjson\n\turl = https://github.com/Tencent/rapidjson.git\n[submodule \"third_party/aes_128\"]\n\tpath = third_party/aes_128\n\turl = https://github.com/openluopworld/aes_128.git\n[submodule \"third_party/capstone\"]\n\tpath = third_party/capstone\n\turl = https://github.com/xenia-project/capstone.git\n[submodule \"third_party/cpptoml\"]\n\tpath = third_party/cpptoml\n\turl = https://github.com/skystrife/cpptoml.git\n[submodule \"third_party/cxxopts\"]\n\tpath = third_party/cxxopts\n\turl = https://github.com/jarro2783/cxxopts.git\n[submodule \"third_party/SDL2\"]\n\tpath = third_party/SDL2\n\turl = https://github.com/libsdl-org/SDL.git\n[submodule \"third_party/utfcpp\"]\n\tpath = third_party/utfcpp\n\turl = https://github.com/xenia-project/utfcpp.git\n[submodule \"third_party/fmt\"]\n\tpath = third_party/fmt\n\turl = https://github.com/fmtlib/fmt.git\n[submodule \"third_party/disruptorplus\"]\n\tpath = third_party/disruptorplus\n\turl = https://github.com/xenia-project/disruptorplus.git\n[submodule \"third_party/DirectXShaderCompiler\"]\n\tpath = third_party/DirectXShaderCompiler\n\turl = https://github.com/microsoft/DirectXShaderCompiler.git\n[submodule \"third_party/premake-cmake\"]\n\tpath = third_party/premake-cmake\n\turl = https://github.com/JoelLinn/premake-cmake.git\n[submodule \"third_party/date\"]\n\tpath = third_party/date\n\turl = https://github.com/HowardHinnant/date.git\n[submodule \"third_party/xxhash\"]\n\tpath = third_party/xxhash\n\turl = https://github.com/Cyan4973/xxHash.git\n[submodule \"third_party/FFmpeg\"]\n\tpath = third_party/FFmpeg\n\turl = https://github.com/xenia-project/FFmpeg.git\n[submodule \"third_party/premake-androidndk\"]\n\tpath = third_party/premake-androidndk\n\turl = https://github.com/Triang3l/premake-androidndk.git\n[submodule \"third_party/FidelityFX-CAS\"]\n\tpath = third_party/FidelityFX-CAS\n\turl = https://github.com/GPUOpen-Effects/FidelityFX-CAS.git\n[submodule \"third_party/FidelityFX-FSR\"]\n\tpath = third_party/FidelityFX-FSR\n\turl = https://github.com/GPUOpen-Effects/FidelityFX-FSR.git\n[submodule \"third_party/Vulkan-Headers\"]\n\tpath = third_party/Vulkan-Headers\n\turl = https://github.com/KhronosGroup/Vulkan-Headers.git\n[submodule \"third_party/glslang\"]\n\tpath = third_party/glslang\n\turl = https://github.com/KhronosGroup/glslang.git\n[submodule \"third_party/SPIRV-Tools\"]\n\tpath = third_party/SPIRV-Tools\n\turl = https://github.com/KhronosGroup/SPIRV-Tools.git\n[submodule \"third_party/VulkanMemoryAllocator\"]\n\tpath = third_party/VulkanMemoryAllocator\n\turl = https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator.git\n[submodule \"third_party/DirectX-Headers\"]\n\tpath = third_party/DirectX-Headers\n\turl = https://github.com/microsoft/DirectX-Headers.git\n" }, { "path": "LICENSE", "content": "Copyright (c) 2015, Ben Vanik.\nAll rights reserved.\n\nRedistribution and use in source and binary forms, with or without\nmodification, are permitted provided that the following conditions are met:\n * Redistributions of source code must retain the above copyright\n notice, this list of conditions and the following disclaimer.\n * Redistributions in binary form must reproduce the above copyright\n notice, this list of conditions and the following disclaimer in the\n documentation and/or other materials provided with the distribution.\n * Neither the name of the project nor the\n names of its contributors may be used to endorse or promote products\n derived from this software without specific prior written permission.\n\nTHIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND\nANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED\nWARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\nDISCLAIMED. IN NO EVENT SHALL BEN VANIK BE LIABLE FOR ANY\nDIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES\n(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;\nLOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND\nON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\n(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS\nSOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n" }, { "path": "README.md", "content": "

\n \n \n \n

\n\n

Xenia - Xbox 360 Emulator

\n\nXenia is an experimental emulator for the Xbox 360. For more information, see the\n[main Xenia wiki](https://github.com/xenia-project/xenia/wiki).\n\n**Interested in supporting the core contributors?** Visit\n[Xenia Project on Patreon](https://www.patreon.com/xenia_project).\n\nCome chat with us about **emulator-related topics** on [Discord](https://discord.gg/Q9mxZf9).\nFor developer chat join `#dev` but stay on topic. Lurking is not only fine, but encouraged!\nPlease check the [FAQ](https://github.com/xenia-project/xenia/wiki/FAQ) page before asking questions.\nWe've got jobs/lives/etc, so don't expect instant answers.\n\nDiscussing illegal activities will get you banned.\n\n## Status\n\nBuildbot | Status | Releases\n-------- | ------ | --------\n[Windows](https://ci.appveyor.com/project/benvanik/xenia/branch/master) | [![Build status](https://ci.appveyor.com/api/projects/status/ftqiy86kdfawyx3a/branch/master?svg=true)](https://ci.appveyor.com/project/benvanik/xenia/branch/master) | [Latest](https://github.com/xenia-project/release-builds-windows/releases/latest) ◦ [All](https://github.com/xenia-project/release-builds-windows/releases)\n[Linux](https://cloud.drone.io/xenia-project/xenia) | [![Build status](https://cloud.drone.io/api/badges/xenia-project/xenia/status.svg)](https://cloud.drone.io/xenia-project/xenia)\n\nQuite a few real games run. Quite a few don't.\nSee the [Game compatibility list](https://github.com/xenia-project/game-compatibility/issues)\nfor currently tracked games, and feel free to contribute your own updates,\nscreenshots, and information there following the [existing conventions](https://github.com/xenia-project/game-compatibility/blob/master/README.md).\n\n## Disclaimer\n\nThe goal of this project is to experiment, research, and educate on the topic\nof emulation of modern devices and operating systems. **It is not for enabling\nillegal activity**. All information is obtained via reverse engineering of\nlegally purchased devices and games and information made public on the internet\n(you'd be surprised what's indexed on Google...).\n\n## Quickstart\n\nSee the [Quickstart](https://github.com/xenia-project/xenia/wiki/Quickstart) page.\n\n## Building\n\nSee [building.md](docs/building.md) for setup and information about the\n`xb` script. When writing code, check the [style guide](docs/style_guide.md)\nand be sure to run clang-format!\n\n## Contributors Wanted!\n\nHave some spare time, know advanced C++, and want to write an emulator?\nContribute! There's a ton of work that needs to be done, a lot of which\nis wide open greenfield fun.\n\n**For general rules and guidelines please see [CONTRIBUTING.md](.github/CONTRIBUTING.md).**\n\nFixes and optimizations are always welcome (please!), but in addition to\nthat there are some major work areas still untouched:\n\n* Help work through [missing functionality/bugs in games](https://github.com/xenia-project/xenia/labels/compat)\n* Reduce the size of Xenia's [huge log files](https://github.com/xenia-project/xenia/issues/1526)\n* Skilled with Linux? A strong contributor is needed to [help with porting](https://github.com/xenia-project/xenia/labels/platform-linux)\n\nSee more projects [good for contributors](https://github.com/xenia-project/xenia/labels/good%20first%20issue). It's a good idea to ask on Discord and check the issues page before beginning work on\nsomething.\n\n## FAQ\n\nSee the [frequently asked questions](https://github.com/xenia-project/xenia/wiki/FAQ) page.\n" }, { "path": "android/android_studio_project/.gitignore", "content": "*.iml\n.gradle\n/local.properties\n/.idea/caches\n/.idea/libraries\n/.idea/modules.xml\n/.idea/workspace.xml\n/.idea/navEditor.xml\n/.idea/assetWizardSettings.xml\n.DS_Store\n/build\n/captures\n.externalNativeBuild\n.cxx\nlocal.properties\n" }, { "path": "android/android_studio_project/app/.gitignore", "content": "/build" }, { "path": "android/android_studio_project/app/build.gradle", "content": "plugins {\n id 'com.android.application'\n}\n\nandroid {\n compileSdkVersion 33\n ndkVersion '25.0.8775105'\n\n defaultConfig {\n applicationId 'jp.xenia.emulator'\n // 24 (7.0) - Vulkan.\n minSdkVersion 24\n targetSdkVersion 33\n versionCode 1\n versionName 'Prototype'\n externalNativeBuild {\n ndkBuild {\n arguments 'NDK_APPLICATION_MK:=../../../build/xenia.Application.mk',\n 'PREMAKE_ANDROIDNDK_PLATFORMS:=Android-ARM64',\n 'PREMAKE_ANDROIDNDK_PLATFORMS+=Android-x86_64',\n // ndk.jobs doesn't work as of Gradle 7.1.0.\n \"-j${Runtime.runtime.availableProcessors()}\",\n // Work around \"Bad file descriptor\" on Windows on NDK r22+.\n '--output-sync=none'\n // For the app, don't build the executables designed for running from a terminal.\n // To build the executables, run ndk-build manually.\n targets 'xenia-app'\n }\n }\n ndk {\n abiFilters 'arm64-v8a', 'x86_64'\n stl 'c++_static'\n }\n }\n\n buildTypes {\n release {\n externalNativeBuild {\n ndkBuild {\n arguments 'PREMAKE_ANDROIDNDK_CONFIGURATIONS:=Release'\n }\n }\n minifyEnabled true\n proguardFiles getDefaultProguardFile('proguard-android-optimize.txt'), 'proguard-rules.pro'\n }\n debug {\n applicationIdSuffix '.debug'\n debuggable true\n externalNativeBuild {\n ndkBuild {\n arguments 'PREMAKE_ANDROIDNDK_CONFIGURATIONS:=Debug'\n }\n }\n }\n checked {\n applicationIdSuffix '.checked'\n debuggable true\n externalNativeBuild {\n ndkBuild {\n arguments 'PREMAKE_ANDROIDNDK_CONFIGURATIONS:=Checked'\n }\n }\n }\n }\n\n flavorDimensions 'distribution'\n productFlavors {\n github {\n dimension 'distribution'\n applicationIdSuffix '.github'\n }\n googlePlay {\n dimension 'distribution'\n // TODO(Triang3l): Provide a signing config for core contributors only.\n }\n }\n\n compileOptions {\n sourceCompatibility JavaVersion.VERSION_1_8\n targetCompatibility JavaVersion.VERSION_1_8\n }\n\n externalNativeBuild {\n ndkBuild {\n path file('../../../build/xenia.wks.Android.mk')\n }\n }\n}\n\ndependencies {\n implementation 'org.jetbrains:annotations:15.0'\n}" }, { "path": "android/android_studio_project/app/proguard-rules.pro", "content": "# Add project specific ProGuard rules here.\n# You can control the set of applied configuration files using the\n# proguardFiles setting in build.gradle.\n#\n# For more details, see\n# http://developer.android.com/guide/developing/tools/proguard.html\n\n# If your project uses WebView with JS, uncomment the following\n# and specify the fully qualified class name to the JavaScript interface\n# class:\n#-keepclassmembers class fqcn.of.javascript.interface.for.webview {\n# public *;\n#}\n\n# Uncomment this to preserve the line number information for\n# debugging stack traces.\n#-keepattributes SourceFile,LineNumberTable\n\n# If you keep the line number information, uncomment this to\n# hide the original source file name.\n#-renamesourcefileattribute SourceFile" }, { "path": "android/android_studio_project/app/src/main/AndroidManifest.xml", "content": "\n\n\n \n\n \n\n \n \n\n \n\n \n \n \n \n \n \n\n \n\n \n\n \n\n" }, { "path": "android/android_studio_project/app/src/main/java/jp/xenia/XeniaRuntimeException.java", "content": "package jp.xenia;\n\n/**\n * Base class for all unchecked exceptions thrown by the Xenia project components.\n */\npublic class XeniaRuntimeException extends RuntimeException {\n public XeniaRuntimeException() {\n }\n\n public XeniaRuntimeException(final String name) {\n super(name);\n }\n\n public XeniaRuntimeException(final String name, final Throwable cause) {\n super(name, cause);\n }\n\n public XeniaRuntimeException(final Exception cause) {\n super(cause);\n }\n}\n" }, { "path": "android/android_studio_project/app/src/main/java/jp/xenia/emulator/GpuTraceViewerActivity.java", "content": "package jp.xenia.emulator;\n\nimport android.os.Bundle;\n\npublic class GpuTraceViewerActivity extends WindowedAppActivity {\n @Override\n protected String getWindowedAppIdentifier() {\n return \"xenia_gpu_vulkan_trace_viewer\";\n }\n\n @Override\n protected void onCreate(final Bundle savedInstanceState) {\n super.onCreate(savedInstanceState);\n\n setContentView(R.layout.activity_gpu_trace_viewer);\n setWindowSurfaceView(findViewById(R.id.gpu_trace_viewer_surface_view));\n }\n}\n" }, { "path": "android/android_studio_project/app/src/main/java/jp/xenia/emulator/LauncherActivity.java", "content": "package jp.xenia.emulator;\n\nimport android.app.Activity;\nimport android.content.Intent;\nimport android.net.Uri;\nimport android.os.Bundle;\nimport android.view.View;\n\npublic class LauncherActivity extends Activity {\n private static final int REQUEST_OPEN_GPU_TRACE_VIEWER = 0;\n\n @Override\n protected void onCreate(final Bundle savedInstanceState) {\n super.onCreate(savedInstanceState);\n\n setContentView(R.layout.activity_launcher);\n }\n\n @Override\n protected void onActivityResult(\n final int requestCode, final int resultCode, final Intent data) {\n if (requestCode == REQUEST_OPEN_GPU_TRACE_VIEWER && resultCode == RESULT_OK) {\n final Uri uri = data.getData();\n if (uri != null) {\n final Intent gpuTraceViewerIntent = new Intent(this, GpuTraceViewerActivity.class);\n final Bundle gpuTraceViewerLaunchArguments = new Bundle();\n gpuTraceViewerLaunchArguments.putString(\"target_trace_file\", uri.toString());\n gpuTraceViewerIntent.putExtra(\n WindowedAppActivity.EXTRA_CVARS, gpuTraceViewerLaunchArguments);\n startActivity(gpuTraceViewerIntent);\n }\n }\n }\n\n public void onLaunchGpuTraceViewerClick(final View view) {\n final Intent intent = new Intent(Intent.ACTION_OPEN_DOCUMENT);\n intent.addCategory(Intent.CATEGORY_OPENABLE);\n intent.setType(\"application/octet-stream\");\n startActivityForResult(intent, REQUEST_OPEN_GPU_TRACE_VIEWER);\n }\n\n public void onLaunchWindowDemoClick(final View view) {\n startActivity(new Intent(this, WindowDemoActivity.class));\n }\n}\n" }, { "path": "android/android_studio_project/app/src/main/java/jp/xenia/emulator/WindowDemoActivity.java", "content": "package jp.xenia.emulator;\n\nimport android.os.Bundle;\n\npublic class WindowDemoActivity extends WindowedAppActivity {\n @Override\n protected String getWindowedAppIdentifier() {\n return \"xenia_ui_window_vulkan_demo\";\n }\n\n @Override\n protected void onCreate(final Bundle savedInstanceState) {\n super.onCreate(savedInstanceState);\n\n setContentView(R.layout.activity_window_demo);\n setWindowSurfaceView(findViewById(R.id.window_demo_surface_view));\n }\n}\n" }, { "path": "android/android_studio_project/app/src/main/java/jp/xenia/emulator/WindowSurfaceView.java", "content": "package jp.xenia.emulator;\n\nimport android.content.Context;\nimport android.graphics.Canvas;\nimport android.util.AttributeSet;\nimport android.view.SurfaceView;\n\npublic class WindowSurfaceView extends SurfaceView {\n public WindowSurfaceView(final Context context) {\n super(context);\n // Native drawing is invoked from onDraw.\n setWillNotDraw(false);\n }\n\n public WindowSurfaceView(final Context context, final AttributeSet attrs) {\n super(context, attrs);\n setWillNotDraw(false);\n }\n\n public WindowSurfaceView(\n final Context context, final AttributeSet attrs, final int defStyleAttr) {\n super(context, attrs, defStyleAttr);\n setWillNotDraw(false);\n }\n\n public WindowSurfaceView(\n final Context context, final AttributeSet attrs, final int defStyleAttr,\n final int defStyleRes) {\n super(context, attrs, defStyleAttr, defStyleRes);\n setWillNotDraw(false);\n }\n\n @Override\n protected void onDraw(final Canvas canvas) {\n final Context context = getContext();\n if (!(context instanceof WindowedAppActivity)) {\n return;\n }\n final WindowedAppActivity activity = (WindowedAppActivity) context;\n activity.onWindowSurfaceDraw(false);\n }\n}\n" }, { "path": "android/android_studio_project/app/src/main/java/jp/xenia/emulator/WindowedAppActivity.java", "content": "package jp.xenia.emulator;\n\nimport android.annotation.SuppressLint;\nimport android.app.Activity;\nimport android.content.res.AssetManager;\nimport android.os.Bundle;\nimport android.view.MotionEvent;\nimport android.view.Surface;\nimport android.view.SurfaceHolder;\nimport android.view.View;\n\nimport org.jetbrains.annotations.Nullable;\n\nimport jp.xenia.XeniaRuntimeException;\n\npublic abstract class WindowedAppActivity extends Activity {\n // The EXTRA_CVARS value literal is also used in the native code.\n\n /**\n * Name of the Bundle intent extra containing Xenia config variable launch arguments.\n */\n public static final String EXTRA_CVARS = \"jp.xenia.emulator.WindowedAppActivity.EXTRA_CVARS\";\n\n static {\n System.loadLibrary(\"xenia-app\");\n }\n\n private final WindowSurfaceListener mWindowSurfaceListener = new WindowSurfaceListener();\n\n // May be 0 while destroying (mainly while the superclass is).\n private long mAppContext = 0;\n\n @Nullable\n private WindowSurfaceView mWindowSurfaceView = null;\n\n private native long initializeWindowedAppOnCreate(\n String windowedAppIdentifier, AssetManager assetManager);\n\n private native void onDestroyNative(long appContext);\n\n private native void onWindowSurfaceLayoutChange(\n long appContext, int left, int top, int right, int bottom);\n\n private native boolean onWindowSurfaceMotionEvent(long appContext, MotionEvent event);\n\n private native void onWindowSurfaceChanged(long appContext, Surface windowSurface);\n\n private native void paintWindow(long appContext, boolean forcePaint);\n\n protected abstract String getWindowedAppIdentifier();\n\n protected void setWindowSurfaceView(@Nullable final WindowSurfaceView windowSurfaceView) {\n if (mWindowSurfaceView == windowSurfaceView) {\n return;\n }\n\n // Detach from the old surface.\n if (mWindowSurfaceView != null) {\n mWindowSurfaceView.getHolder().removeCallback(mWindowSurfaceListener);\n mWindowSurfaceView.setOnTouchListener(null);\n mWindowSurfaceView.setOnGenericMotionListener(null);\n mWindowSurfaceView.removeOnLayoutChangeListener(mWindowSurfaceListener);\n mWindowSurfaceView = null;\n if (mAppContext != 0) {\n onWindowSurfaceChanged(mAppContext, null);\n }\n }\n\n if (windowSurfaceView == null) {\n return;\n }\n\n mWindowSurfaceView = windowSurfaceView;\n // FIXME(Triang3l): This doesn't work if the layout has already been performed.\n mWindowSurfaceView.addOnLayoutChangeListener(mWindowSurfaceListener);\n mWindowSurfaceView.setOnGenericMotionListener(mWindowSurfaceListener);\n mWindowSurfaceView.setOnTouchListener(mWindowSurfaceListener);\n final SurfaceHolder windowSurfaceHolder = mWindowSurfaceView.getHolder();\n windowSurfaceHolder.addCallback(mWindowSurfaceListener);\n // If setting after the creation of the surface.\n if (mAppContext != 0) {\n final Surface windowSurface = windowSurfaceHolder.getSurface();\n if (windowSurface != null) {\n onWindowSurfaceChanged(mAppContext, windowSurface);\n }\n }\n }\n\n public void onWindowSurfaceDraw(final boolean forcePaint) {\n if (mAppContext == 0) {\n return;\n }\n paintWindow(mAppContext, forcePaint);\n }\n\n // Used from the native WindowedAppContext. May be called from non-UI threads.\n @SuppressWarnings(\"UnusedDeclaration\")\n protected void postInvalidateWindowSurface() {\n if (mWindowSurfaceView == null) {\n return;\n }\n mWindowSurfaceView.postInvalidate();\n }\n\n @Override\n protected void onCreate(final Bundle savedInstanceState) {\n super.onCreate(savedInstanceState);\n\n final String windowedAppIdentifier = getWindowedAppIdentifier();\n mAppContext = initializeWindowedAppOnCreate(windowedAppIdentifier, getAssets());\n if (mAppContext == 0) {\n finish();\n throw new XeniaRuntimeException(\n \"Error initializing the windowed app \" + windowedAppIdentifier);\n }\n }\n\n @Override\n protected void onDestroy() {\n setWindowSurfaceView(null);\n if (mAppContext != 0) {\n onDestroyNative(mAppContext);\n }\n mAppContext = 0;\n super.onDestroy();\n }\n\n private class WindowSurfaceListener implements\n View.OnGenericMotionListener,\n View.OnLayoutChangeListener,\n View.OnTouchListener,\n SurfaceHolder.Callback2 {\n @Override\n public void onLayoutChange(\n final View v, final int left, final int top, final int right, final int bottom,\n final int oldLeft, final int oldTop, final int oldRight, final int oldBottom) {\n if (mAppContext != 0) {\n onWindowSurfaceLayoutChange(mAppContext, left, top, right, bottom);\n }\n }\n\n @Override\n public boolean onGenericMotion(final View view, final MotionEvent event) {\n if (mAppContext == 0) {\n return false;\n }\n return onWindowSurfaceMotionEvent(mAppContext, event);\n }\n\n @SuppressLint(\"ClickableViewAccessibility\")\n @Override\n public boolean onTouch(final View view, final MotionEvent event) {\n if (mAppContext == 0) {\n return false;\n }\n return onWindowSurfaceMotionEvent(mAppContext, event);\n }\n\n @Override\n public void surfaceCreated(final SurfaceHolder holder) {\n if (mAppContext == 0) {\n return;\n }\n onWindowSurfaceChanged(mAppContext, holder.getSurface());\n }\n\n @Override\n public void surfaceChanged(\n final SurfaceHolder holder, final int format, final int width, final int height) {\n if (mAppContext == 0) {\n return;\n }\n onWindowSurfaceChanged(mAppContext, holder.getSurface());\n }\n\n @Override\n public void surfaceDestroyed(final SurfaceHolder holder) {\n if (mAppContext == 0) {\n return;\n }\n onWindowSurfaceChanged(mAppContext, null);\n }\n\n @Override\n public void surfaceRedrawNeeded(final SurfaceHolder holder) {\n onWindowSurfaceDraw(true);\n }\n }\n}\n" }, { "path": "android/android_studio_project/app/src/main/res/drawable/ic_launcher_background.xml", "content": "\n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n\n" }, { "path": "android/android_studio_project/app/src/main/res/drawable-v24/ic_launcher_foreground.xml", "content": "\n \n \n \n \n \n \n \n \n \n" }, { "path": "android/android_studio_project/app/src/main/res/layout/activity_gpu_trace_viewer.xml", "content": "\n\n" }, { "path": "android/android_studio_project/app/src/main/res/layout/activity_launcher.xml", "content": "\n\n\n \n\n \n\n \n\n \n\n\n" }, { "path": "android/android_studio_project/app/src/main/res/layout/activity_window_demo.xml", "content": "\n\n" }, { "path": "android/android_studio_project/app/src/main/res/mipmap-anydpi-v26/ic_launcher.xml", "content": "\n\n \n \n" }, { "path": "android/android_studio_project/app/src/main/res/mipmap-anydpi-v26/ic_launcher_round.xml", "content": "\n\n \n \n" }, { "path": "android/android_studio_project/app/src/main/res/values/strings.xml", "content": "\n Xenia\n GPU Trace Viewer\n Window Demo\n" }, { "path": "android/android_studio_project/build.gradle", "content": "// Top-level build file where you can add configuration options common to all sub-projects/modules.\nbuildscript {\n repositories {\n google()\n mavenCentral()\n }\n dependencies {\n classpath 'com.android.tools.build:gradle:7.2.1'\n\n // NOTE: Do not place your application dependencies here; they belong\n // in the individual module build.gradle files\n }\n}\n\nallprojects {\n repositories {\n google()\n mavenCentral()\n }\n}\n\ntask clean(type: Delete) {\n delete rootProject.buildDir\n}" }, { "path": "android/android_studio_project/gradle/wrapper/gradle-wrapper.properties", "content": "#Mon Nov 01 23:19:20 MSK 2021\ndistributionBase=GRADLE_USER_HOME\ndistributionUrl=https\\://services.gradle.org/distributions/gradle-7.3.3-bin.zip\ndistributionPath=wrapper/dists\nzipStorePath=wrapper/dists\nzipStoreBase=GRADLE_USER_HOME\n" }, { "path": "android/android_studio_project/gradle.properties", "content": "# Project-wide Gradle settings.\n# IDE (e.g. Android Studio) users:\n# Gradle settings configured through the IDE *will override*\n# any settings specified in this file.\n# For more details on how to configure your build environment visit\n# http://www.gradle.org/docs/current/userguide/build_environment.html\n# Specifies the JVM arguments used for the daemon process.\n# The setting is particularly useful for tweaking memory settings.\norg.gradle.jvmargs=-Xmx2048m -Dfile.encoding=UTF-8\n# When configured, Gradle will run in incubating parallel mode.\n# This option should only be used with decoupled projects. More details, visit\n# http://www.gradle.org/docs/current/userguide/multi_project_builds.html#sec:decoupled_projects\n# org.gradle.parallel=true\n# AndroidX package structure to make it clearer which packages are bundled with the\n# Android operating system, and which are packaged with your app\"s APK\n# https://developer.android.com/topic/libraries/support-library/androidx-rn\nandroid.useAndroidX=true\n# Automatically convert third-party libraries to use AndroidX\nandroid.enableJetifier=true" }, { "path": "android/android_studio_project/gradlew", "content": "#!/usr/bin/env sh\n\n##############################################################################\n##\n## Gradle start up script for UN*X\n##\n##############################################################################\n\n# Attempt to set APP_HOME\n# Resolve links: $0 may be a link\nPRG=\"$0\"\n# Need this for relative symlinks.\nwhile [ -h \"$PRG\" ] ; do\n ls=`ls -ld \"$PRG\"`\n link=`expr \"$ls\" : '.*-> \\(.*\\)$'`\n if expr \"$link\" : '/.*' > /dev/null; then\n PRG=\"$link\"\n else\n PRG=`dirname \"$PRG\"`\"/$link\"\n fi\ndone\nSAVED=\"`pwd`\"\ncd \"`dirname \\\"$PRG\\\"`/\" >/dev/null\nAPP_HOME=\"`pwd -P`\"\ncd \"$SAVED\" >/dev/null\n\nAPP_NAME=\"Gradle\"\nAPP_BASE_NAME=`basename \"$0\"`\n\n# Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.\nDEFAULT_JVM_OPTS=\"\"\n\n# Use the maximum available, or set MAX_FD != -1 to use that value.\nMAX_FD=\"maximum\"\n\nwarn () {\n echo \"$*\"\n}\n\ndie () {\n echo\n echo \"$*\"\n echo\n exit 1\n}\n\n# OS specific support (must be 'true' or 'false').\ncygwin=false\nmsys=false\ndarwin=false\nnonstop=false\ncase \"`uname`\" in\n CYGWIN* )\n cygwin=true\n ;;\n Darwin* )\n darwin=true\n ;;\n MINGW* )\n msys=true\n ;;\n NONSTOP* )\n nonstop=true\n ;;\nesac\n\nCLASSPATH=$APP_HOME/gradle/wrapper/gradle-wrapper.jar\n\n# Determine the Java command to use to start the JVM.\nif [ -n \"$JAVA_HOME\" ] ; then\n if [ -x \"$JAVA_HOME/jre/sh/java\" ] ; then\n # IBM's JDK on AIX uses strange locations for the executables\n JAVACMD=\"$JAVA_HOME/jre/sh/java\"\n else\n JAVACMD=\"$JAVA_HOME/bin/java\"\n fi\n if [ ! -x \"$JAVACMD\" ] ; then\n die \"ERROR: JAVA_HOME is set to an invalid directory: $JAVA_HOME\n\nPlease set the JAVA_HOME variable in your environment to match the\nlocation of your Java installation.\"\n fi\nelse\n JAVACMD=\"java\"\n which java >/dev/null 2>&1 || die \"ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.\n\nPlease set the JAVA_HOME variable in your environment to match the\nlocation of your Java installation.\"\nfi\n\n# Increase the maximum file descriptors if we can.\nif [ \"$cygwin\" = \"false\" -a \"$darwin\" = \"false\" -a \"$nonstop\" = \"false\" ] ; then\n MAX_FD_LIMIT=`ulimit -H -n`\n if [ $? -eq 0 ] ; then\n if [ \"$MAX_FD\" = \"maximum\" -o \"$MAX_FD\" = \"max\" ] ; then\n MAX_FD=\"$MAX_FD_LIMIT\"\n fi\n ulimit -n $MAX_FD\n if [ $? -ne 0 ] ; then\n warn \"Could not set maximum file descriptor limit: $MAX_FD\"\n fi\n else\n warn \"Could not query maximum file descriptor limit: $MAX_FD_LIMIT\"\n fi\nfi\n\n# For Darwin, add options to specify how the application appears in the dock\nif $darwin; then\n GRADLE_OPTS=\"$GRADLE_OPTS \\\"-Xdock:name=$APP_NAME\\\" \\\"-Xdock:icon=$APP_HOME/media/gradle.icns\\\"\"\nfi\n\n# For Cygwin, switch paths to Windows format before running java\nif $cygwin ; then\n APP_HOME=`cygpath --path --mixed \"$APP_HOME\"`\n CLASSPATH=`cygpath --path --mixed \"$CLASSPATH\"`\n JAVACMD=`cygpath --unix \"$JAVACMD\"`\n\n # We build the pattern for arguments to be converted via cygpath\n ROOTDIRSRAW=`find -L / -maxdepth 1 -mindepth 1 -type d 2>/dev/null`\n SEP=\"\"\n for dir in $ROOTDIRSRAW ; do\n ROOTDIRS=\"$ROOTDIRS$SEP$dir\"\n SEP=\"|\"\n done\n OURCYGPATTERN=\"(^($ROOTDIRS))\"\n # Add a user-defined pattern to the cygpath arguments\n if [ \"$GRADLE_CYGPATTERN\" != \"\" ] ; then\n OURCYGPATTERN=\"$OURCYGPATTERN|($GRADLE_CYGPATTERN)\"\n fi\n # Now convert the arguments - kludge to limit ourselves to /bin/sh\n i=0\n for arg in \"$@\" ; do\n CHECK=`echo \"$arg\"|egrep -c \"$OURCYGPATTERN\" -`\n CHECK2=`echo \"$arg\"|egrep -c \"^-\"` ### Determine if an option\n\n if [ $CHECK -ne 0 ] && [ $CHECK2 -eq 0 ] ; then ### Added a condition\n eval `echo args$i`=`cygpath --path --ignore --mixed \"$arg\"`\n else\n eval `echo args$i`=\"\\\"$arg\\\"\"\n fi\n i=$((i+1))\n done\n case $i in\n (0) set -- ;;\n (1) set -- \"$args0\" ;;\n (2) set -- \"$args0\" \"$args1\" ;;\n (3) set -- \"$args0\" \"$args1\" \"$args2\" ;;\n (4) set -- \"$args0\" \"$args1\" \"$args2\" \"$args3\" ;;\n (5) set -- \"$args0\" \"$args1\" \"$args2\" \"$args3\" \"$args4\" ;;\n (6) set -- \"$args0\" \"$args1\" \"$args2\" \"$args3\" \"$args4\" \"$args5\" ;;\n (7) set -- \"$args0\" \"$args1\" \"$args2\" \"$args3\" \"$args4\" \"$args5\" \"$args6\" ;;\n (8) set -- \"$args0\" \"$args1\" \"$args2\" \"$args3\" \"$args4\" \"$args5\" \"$args6\" \"$args7\" ;;\n (9) set -- \"$args0\" \"$args1\" \"$args2\" \"$args3\" \"$args4\" \"$args5\" \"$args6\" \"$args7\" \"$args8\" ;;\n esac\nfi\n\n# Escape application args\nsave () {\n for i do printf %s\\\\n \"$i\" | sed \"s/'/'\\\\\\\\''/g;1s/^/'/;\\$s/\\$/' \\\\\\\\/\" ; done\n echo \" \"\n}\nAPP_ARGS=$(save \"$@\")\n\n# Collect all arguments for the java command, following the shell quoting and substitution rules\neval set -- $DEFAULT_JVM_OPTS $JAVA_OPTS $GRADLE_OPTS \"\\\"-Dorg.gradle.appname=$APP_BASE_NAME\\\"\" -classpath \"\\\"$CLASSPATH\\\"\" org.gradle.wrapper.GradleWrapperMain \"$APP_ARGS\"\n\n# by default we should be in the correct project dir, but when run from Finder on Mac, the cwd is wrong\nif [ \"$(uname)\" = \"Darwin\" ] && [ \"$HOME\" = \"$PWD\" ]; then\n cd \"$(dirname \"$0\")\"\nfi\n\nexec \"$JAVACMD\" \"$@\"\n" }, { "path": "android/android_studio_project/gradlew.bat", "content": "@if \"%DEBUG%\" == \"\" @echo off\r\n@rem ##########################################################################\r\n@rem\r\n@rem Gradle startup script for Windows\r\n@rem\r\n@rem ##########################################################################\r\n\r\n@rem Set local scope for the variables with windows NT shell\r\nif \"%OS%\"==\"Windows_NT\" setlocal\r\n\r\nset DIRNAME=%~dp0\r\nif \"%DIRNAME%\" == \"\" set DIRNAME=.\r\nset APP_BASE_NAME=%~n0\r\nset APP_HOME=%DIRNAME%\r\n\r\n@rem Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.\r\nset DEFAULT_JVM_OPTS=\r\n\r\n@rem Find java.exe\r\nif defined JAVA_HOME goto findJavaFromJavaHome\r\n\r\nset JAVA_EXE=java.exe\r\n%JAVA_EXE% -version >NUL 2>&1\r\nif \"%ERRORLEVEL%\" == \"0\" goto init\r\n\r\necho.\r\necho ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.\r\necho.\r\necho Please set the JAVA_HOME variable in your environment to match the\r\necho location of your Java installation.\r\n\r\ngoto fail\r\n\r\n:findJavaFromJavaHome\r\nset JAVA_HOME=%JAVA_HOME:\"=%\r\nset JAVA_EXE=%JAVA_HOME%/bin/java.exe\r\n\r\nif exist \"%JAVA_EXE%\" goto init\r\n\r\necho.\r\necho ERROR: JAVA_HOME is set to an invalid directory: %JAVA_HOME%\r\necho.\r\necho Please set the JAVA_HOME variable in your environment to match the\r\necho location of your Java installation.\r\n\r\ngoto fail\r\n\r\n:init\r\n@rem Get command-line arguments, handling Windows variants\r\n\r\nif not \"%OS%\" == \"Windows_NT\" goto win9xME_args\r\n\r\n:win9xME_args\r\n@rem Slurp the command line arguments.\r\nset CMD_LINE_ARGS=\r\nset _SKIP=2\r\n\r\n:win9xME_args_slurp\r\nif \"x%~1\" == \"x\" goto execute\r\n\r\nset CMD_LINE_ARGS=%*\r\n\r\n:execute\r\n@rem Setup the command line\r\n\r\nset CLASSPATH=%APP_HOME%\\gradle\\wrapper\\gradle-wrapper.jar\r\n\r\n@rem Execute Gradle\r\n\"%JAVA_EXE%\" %DEFAULT_JVM_OPTS% %JAVA_OPTS% %GRADLE_OPTS% \"-Dorg.gradle.appname=%APP_BASE_NAME%\" -classpath \"%CLASSPATH%\" org.gradle.wrapper.GradleWrapperMain %CMD_LINE_ARGS%\r\n\r\n:end\r\n@rem End local scope for the variables with windows NT shell\r\nif \"%ERRORLEVEL%\"==\"0\" goto mainEnd\r\n\r\n:fail\r\nrem Set variable GRADLE_EXIT_CONSOLE if you need the _script_ return code instead of\r\nrem the _cmd.exe /c_ return code!\r\nif not \"\" == \"%GRADLE_EXIT_CONSOLE%\" exit 1\r\nexit /b 1\r\n\r\n:mainEnd\r\nif \"%OS%\"==\"Windows_NT\" endlocal\r\n\r\n:omega\r\n" }, { "path": "android/android_studio_project/settings.gradle", "content": "include ':app'\nrootProject.name = \"Xenia\"" }, { "path": "assets/icon/LICENSE", "content": "Attribution-ShareAlike 4.0 International\n\n=======================================================================\n\nCreative Commons Corporation (\"Creative Commons\") is not a law firm and\ndoes not provide legal services or legal advice. Distribution of\nCreative Commons public licenses does not create a lawyer-client or\nother relationship. Creative Commons makes its licenses and related\ninformation available on an \"as-is\" basis. Creative Commons gives no\nwarranties regarding its licenses, any material licensed under their\nterms and conditions, or any related information. Creative Commons\ndisclaims all liability for damages resulting from their use to the\nfullest extent possible.\n\nUsing Creative Commons Public Licenses\n\nCreative Commons public licenses provide a standard set of terms and\nconditions that creators and other rights holders may use to share\noriginal works of authorship and other material subject to copyright\nand certain other rights specified in the public license below. The\nfollowing considerations are for informational purposes only, are not\nexhaustive, and do not form part of our licenses.\n\n Considerations for licensors: Our public licenses are\n intended for use by those authorized to give the public\n permission to use material in ways otherwise restricted by\n copyright and certain other rights. Our licenses are\n irrevocable. Licensors should read and understand the terms\n and conditions of the license they choose before applying it.\n Licensors should also secure all rights necessary before\n applying our licenses so that the public can reuse the\n material as expected. Licensors should clearly mark any\n material not subject to the license. This includes other CC-\n licensed material, or material used under an exception or\n limitation to copyright. More considerations for licensors:\n\twiki.creativecommons.org/Considerations_for_licensors\n\n Considerations for the public: By using one of our public\n licenses, a licensor grants the public permission to use the\n licensed material under specified terms and conditions. If\n the licensor's permission is not necessary for any reason--for\n example, because of any applicable exception or limitation to\n copyright--then that use is not regulated by the license. Our\n licenses grant only permissions under copyright and certain\n other rights that a licensor has authority to grant. Use of\n the licensed material may still be restricted for other\n reasons, including because others have copyright or other\n rights in the material. A licensor may make special requests,\n such as asking that all changes be marked or described.\n Although not required by our licenses, you are encouraged to\n respect those requests where reasonable. More considerations\n for the public: \n\twiki.creativecommons.org/Considerations_for_licensees\n\n=======================================================================\n\nCreative Commons Attribution-ShareAlike 4.0 International Public\nLicense\n\nBy exercising the Licensed Rights (defined below), You accept and agree\nto be bound by the terms and conditions of this Creative Commons\nAttribution-ShareAlike 4.0 International Public License (\"Public\nLicense\"). To the extent this Public License may be interpreted as a\ncontract, You are granted the Licensed Rights in consideration of Your\nacceptance of these terms and conditions, and the Licensor grants You\nsuch rights in consideration of benefits the Licensor receives from\nmaking the Licensed Material available under these terms and\nconditions.\n\n\nSection 1 -- Definitions.\n\n a. Adapted Material means material subject to Copyright and Similar\n Rights that is derived from or based upon the Licensed Material\n and in which the Licensed Material is translated, altered,\n arranged, transformed, or otherwise modified in a manner requiring\n permission under the Copyright and Similar Rights held by the\n Licensor. For purposes of this Public License, where the Licensed\n Material is a musical work, performance, or sound recording,\n Adapted Material is always produced where the Licensed Material is\n synched in timed relation with a moving image.\n\n b. Adapter's License means the license You apply to Your Copyright\n and Similar Rights in Your contributions to Adapted Material in\n accordance with the terms and conditions of this Public License.\n\n c. BY-SA Compatible License means a license listed at\n creativecommons.org/compatiblelicenses, approved by Creative\n Commons as essentially the equivalent of this Public License.\n\n d. Copyright and Similar Rights means copyright and/or similar rights\n closely related to copyright including, without limitation,\n performance, broadcast, sound recording, and Sui Generis Database\n Rights, without regard to how the rights are labeled or\n categorized. For purposes of this Public License, the rights\n specified in Section 2(b)(1)-(2) are not Copyright and Similar\n Rights.\n\n e. Effective Technological Measures means those measures that, in the\n absence of proper authority, may not be circumvented under laws\n fulfilling obligations under Article 11 of the WIPO Copyright\n Treaty adopted on December 20, 1996, and/or similar international\n agreements.\n\n f. Exceptions and Limitations means fair use, fair dealing, and/or\n any other exception or limitation to Copyright and Similar Rights\n that applies to Your use of the Licensed Material.\n\n g. License Elements means the license attributes listed in the name\n of a Creative Commons Public License. The License Elements of this\n Public License are Attribution and ShareAlike.\n\n h. Licensed Material means the artistic or literary work, database,\n or other material to which the Licensor applied this Public\n License.\n\n i. Licensed Rights means the rights granted to You subject to the\n terms and conditions of this Public License, which are limited to\n all Copyright and Similar Rights that apply to Your use of the\n Licensed Material and that the Licensor has authority to license.\n\n j. Licensor means the individual(s) or entity(ies) granting rights\n under this Public License.\n\n k. Share means to provide material to the public by any means or\n process that requires permission under the Licensed Rights, such\n as reproduction, public display, public performance, distribution,\n dissemination, communication, or importation, and to make material\n available to the public including in ways that members of the\n public may access the material from a place and at a time\n individually chosen by them.\n\n l. Sui Generis Database Rights means rights other than copyright\n resulting from Directive 96/9/EC of the European Parliament and of\n the Council of 11 March 1996 on the legal protection of databases,\n as amended and/or succeeded, as well as other essentially\n equivalent rights anywhere in the world.\n\n m. You means the individual or entity exercising the Licensed Rights\n under this Public License. Your has a corresponding meaning.\n\n\nSection 2 -- Scope.\n\n a. License grant.\n\n 1. Subject to the terms and conditions of this Public License,\n the Licensor hereby grants You a worldwide, royalty-free,\n non-sublicensable, non-exclusive, irrevocable license to\n exercise the Licensed Rights in the Licensed Material to:\n\n a. reproduce and Share the Licensed Material, in whole or\n in part; and\n\n b. produce, reproduce, and Share Adapted Material.\n\n 2. Exceptions and Limitations. For the avoidance of doubt, where\n Exceptions and Limitations apply to Your use, this Public\n License does not apply, and You do not need to comply with\n its terms and conditions.\n\n 3. Term. The term of this Public License is specified in Section\n 6(a).\n\n 4. Media and formats; technical modifications allowed. The\n Licensor authorizes You to exercise the Licensed Rights in\n all media and formats whether now known or hereafter created,\n and to make technical modifications necessary to do so. The\n Licensor waives and/or agrees not to assert any right or\n authority to forbid You from making technical modifications\n necessary to exercise the Licensed Rights, including\n technical modifications necessary to circumvent Effective\n Technological Measures. For purposes of this Public License,\n simply making modifications authorized by this Section 2(a)\n (4) never produces Adapted Material.\n\n 5. Downstream recipients.\n\n a. Offer from the Licensor -- Licensed Material. Every\n recipient of the Licensed Material automatically\n receives an offer from the Licensor to exercise the\n Licensed Rights under the terms and conditions of this\n Public License.\n\n b. Additional offer from the Licensor -- Adapted Material.\n Every recipient of Adapted Material from You\n automatically receives an offer from the Licensor to\n exercise the Licensed Rights in the Adapted Material\n under the conditions of the Adapter's License You apply.\n\n c. No downstream restrictions. You may not offer or impose\n any additional or different terms or conditions on, or\n apply any Effective Technological Measures to, the\n Licensed Material if doing so restricts exercise of the\n Licensed Rights by any recipient of the Licensed\n Material.\n\n 6. No endorsement. Nothing in this Public License constitutes or\n may be construed as permission to assert or imply that You\n are, or that Your use of the Licensed Material is, connected\n with, or sponsored, endorsed, or granted official status by,\n the Licensor or others designated to receive attribution as\n provided in Section 3(a)(1)(A)(i).\n\n b. Other rights.\n\n 1. Moral rights, such as the right of integrity, are not\n licensed under this Public License, nor are publicity,\n privacy, and/or other similar personality rights; however, to\n the extent possible, the Licensor waives and/or agrees not to\n assert any such rights held by the Licensor to the limited\n extent necessary to allow You to exercise the Licensed\n Rights, but not otherwise.\n\n 2. Patent and trademark rights are not licensed under this\n Public License.\n\n 3. To the extent possible, the Licensor waives any right to\n collect royalties from You for the exercise of the Licensed\n Rights, whether directly or through a collecting society\n under any voluntary or waivable statutory or compulsory\n licensing scheme. In all other cases the Licensor expressly\n reserves any right to collect such royalties.\n\n\nSection 3 -- License Conditions.\n\nYour exercise of the Licensed Rights is expressly made subject to the\nfollowing conditions.\n\n a. Attribution.\n\n 1. If You Share the Licensed Material (including in modified\n form), You must:\n\n a. retain the following if it is supplied by the Licensor\n with the Licensed Material:\n\n i. identification of the creator(s) of the Licensed\n Material and any others designated to receive\n attribution, in any reasonable manner requested by\n the Licensor (including by pseudonym if\n designated);\n\n ii. a copyright notice;\n\n iii. a notice that refers to this Public License;\n\n iv. a notice that refers to the disclaimer of\n warranties;\n\n v. a URI or hyperlink to the Licensed Material to the\n extent reasonably practicable;\n\n b. indicate if You modified the Licensed Material and\n retain an indication of any previous modifications; and\n\n c. indicate the Licensed Material is licensed under this\n Public License, and include the text of, or the URI or\n hyperlink to, this Public License.\n\n 2. You may satisfy the conditions in Section 3(a)(1) in any\n reasonable manner based on the medium, means, and context in\n which You Share the Licensed Material. For example, it may be\n reasonable to satisfy the conditions by providing a URI or\n hyperlink to a resource that includes the required\n information.\n\n 3. If requested by the Licensor, You must remove any of the\n information required by Section 3(a)(1)(A) to the extent\n reasonably practicable.\n\n b. ShareAlike.\n\n In addition to the conditions in Section 3(a), if You Share\n Adapted Material You produce, the following conditions also apply.\n\n 1. The Adapter's License You apply must be a Creative Commons\n license with the same License Elements, this version or\n later, or a BY-SA Compatible License.\n\n 2. You must include the text of, or the URI or hyperlink to, the\n Adapter's License You apply. You may satisfy this condition\n in any reasonable manner based on the medium, means, and\n context in which You Share Adapted Material.\n\n 3. You may not offer or impose any additional or different terms\n or conditions on, or apply any Effective Technological\n Measures to, Adapted Material that restrict exercise of the\n rights granted under the Adapter's License You apply.\n\n\nSection 4 -- Sui Generis Database Rights.\n\nWhere the Licensed Rights include Sui Generis Database Rights that\napply to Your use of the Licensed Material:\n\n a. for the avoidance of doubt, Section 2(a)(1) grants You the right\n to extract, reuse, reproduce, and Share all or a substantial\n portion of the contents of the database;\n\n b. if You include all or a substantial portion of the database\n contents in a database in which You have Sui Generis Database\n Rights, then the database in which You have Sui Generis Database\n Rights (but not its individual contents) is Adapted Material,\n\n including for purposes of Section 3(b); and\n c. You must comply with the conditions in Section 3(a) if You Share\n all or a substantial portion of the contents of the database.\n\nFor the avoidance of doubt, this Section 4 supplements and does not\nreplace Your obligations under this Public License where the Licensed\nRights include other Copyright and Similar Rights.\n\n\nSection 5 -- Disclaimer of Warranties and Limitation of Liability.\n\n a. UNLESS OTHERWISE SEPARATELY UNDERTAKEN BY THE LICENSOR, TO THE\n EXTENT POSSIBLE, THE LICENSOR OFFERS THE LICENSED MATERIAL AS-IS\n AND AS-AVAILABLE, AND MAKES NO REPRESENTATIONS OR WARRANTIES OF\n ANY KIND CONCERNING THE LICENSED MATERIAL, WHETHER EXPRESS,\n IMPLIED, STATUTORY, OR OTHER. THIS INCLUDES, WITHOUT LIMITATION,\n WARRANTIES OF TITLE, MERCHANTABILITY, FITNESS FOR A PARTICULAR\n PURPOSE, NON-INFRINGEMENT, ABSENCE OF LATENT OR OTHER DEFECTS,\n ACCURACY, OR THE PRESENCE OR ABSENCE OF ERRORS, WHETHER OR NOT\n KNOWN OR DISCOVERABLE. WHERE DISCLAIMERS OF WARRANTIES ARE NOT\n ALLOWED IN FULL OR IN PART, THIS DISCLAIMER MAY NOT APPLY TO YOU.\n\n b. TO THE EXTENT POSSIBLE, IN NO EVENT WILL THE LICENSOR BE LIABLE\n TO YOU ON ANY LEGAL THEORY (INCLUDING, WITHOUT LIMITATION,\n NEGLIGENCE) OR OTHERWISE FOR ANY DIRECT, SPECIAL, INDIRECT,\n INCIDENTAL, CONSEQUENTIAL, PUNITIVE, EXEMPLARY, OR OTHER LOSSES,\n COSTS, EXPENSES, OR DAMAGES ARISING OUT OF THIS PUBLIC LICENSE OR\n USE OF THE LICENSED MATERIAL, EVEN IF THE LICENSOR HAS BEEN\n ADVISED OF THE POSSIBILITY OF SUCH LOSSES, COSTS, EXPENSES, OR\n DAMAGES. WHERE A LIMITATION OF LIABILITY IS NOT ALLOWED IN FULL OR\n IN PART, THIS LIMITATION MAY NOT APPLY TO YOU.\n\n c. The disclaimer of warranties and limitation of liability provided\n above shall be interpreted in a manner that, to the extent\n possible, most closely approximates an absolute disclaimer and\n waiver of all liability.\n\n\nSection 6 -- Term and Termination.\n\n a. This Public License applies for the term of the Copyright and\n Similar Rights licensed here. However, if You fail to comply with\n this Public License, then Your rights under this Public License\n terminate automatically.\n\n b. Where Your right to use the Licensed Material has terminated under\n Section 6(a), it reinstates:\n\n 1. automatically as of the date the violation is cured, provided\n it is cured within 30 days of Your discovery of the\n violation; or\n\n 2. upon express reinstatement by the Licensor.\n\n For the avoidance of doubt, this Section 6(b) does not affect any\n right the Licensor may have to seek remedies for Your violations\n of this Public License.\n\n c. For the avoidance of doubt, the Licensor may also offer the\n Licensed Material under separate terms or conditions or stop\n distributing the Licensed Material at any time; however, doing so\n will not terminate this Public License.\n\n d. Sections 1, 5, 6, 7, and 8 survive termination of this Public\n License.\n\n\nSection 7 -- Other Terms and Conditions.\n\n a. The Licensor shall not be bound by any additional or different\n terms or conditions communicated by You unless expressly agreed.\n\n b. Any arrangements, understandings, or agreements regarding the\n Licensed Material not stated herein are separate from and\n independent of the terms and conditions of this Public License.\n\n\nSection 8 -- Interpretation.\n\n a. For the avoidance of doubt, this Public License does not, and\n shall not be interpreted to, reduce, limit, restrict, or impose\n conditions on any use of the Licensed Material that could lawfully\n be made without permission under this Public License.\n\n b. To the extent possible, if any provision of this Public License is\n deemed unenforceable, it shall be automatically reformed to the\n minimum extent necessary to make it enforceable. If the provision\n cannot be reformed, it shall be severed from this Public License\n without affecting the enforceability of the remaining terms and\n conditions.\n\n c. No term or condition of this Public License will be waived and no\n failure to comply consented to unless expressly agreed to by the\n Licensor.\n\n d. Nothing in this Public License constitutes or may be interpreted\n as a limitation upon, or waiver of, any privileges and immunities\n that apply to the Licensor or You, including from the legal\n processes of any jurisdiction or authority.\n\n\n=======================================================================\n\nCreative Commons is not a party to its public\nlicenses. Notwithstanding, Creative Commons may elect to apply one of\nits public licenses to material it publishes and in those instances\nwill be considered the “Licensor.” The text of the Creative Commons\npublic licenses is dedicated to the public domain under the CC0 Public\nDomain Dedication. Except for the limited purpose of indicating that\nmaterial is shared under a Creative Commons public license or as\notherwise permitted by the Creative Commons policies published at\ncreativecommons.org/policies, Creative Commons does not authorize the\nuse of the trademark \"Creative Commons\" or any other trademark or logo\nof Creative Commons without its prior written consent including,\nwithout limitation, in connection with any unauthorized modifications\nto any of its public licenses or any other arrangements,\nunderstandings, or agreements concerning use of licensed material. For\nthe avoidance of doubt, this paragraph does not form part of the\npublic licenses.\n\nCreative Commons may be contacted at creativecommons.org.\n\n" }, { "path": "docs/building.md", "content": "# Building\n\nYou must have a 64-bit machine for building and running the project. Always\nrun your system updater before building and make sure you have the latest\ndrivers.\n\n## Setup\n\n### Windows\n\n* Windows 7 or later\n* [Visual Studio 2022, Visual Studio 2019, or Visual Studio 2017](https://www.visualstudio.com/downloads/)\n * For Visual Studio 2022, MSBuild `v142` must be used due to a compiler bug; See [#2003](https://github.com/xenia-project/xenia/issues/2003).\n* [Python 3.6+](https://www.python.org/downloads/)\n * Ensure Python is in PATH.\n* Windows 11 SDK version 10.0.22000.0 (for Visual Studio 2019, this or any newer version)\n\n```\ngit clone https://github.com/xenia-project/xenia.git\ncd xenia\nxb setup\n\n# Build on command line (add --config=release for release):\nxb build\n\n\n# Pull latest changes, rebase, update submodules, and run premake:\nxb pull\n\n# Run premake and open Visual Studio (run the 'xenia-app' project):\nxb devenv\n\n# Run premake to update the sln/vcproj's:\nxb premake\n\n# Format code to the style guide:\nxb format\n```\n\n\n#### Debugging\n\nVS behaves oddly with the debug paths. Open the 'xenia-app' project properties\nand set the 'Command' to `$(SolutionDir)$(TargetPath)` and the\n'Working Directory' to `$(SolutionDir)..\\..`. You can specify flags and\nthe file to run in the 'Command Arguments' field (or use `--flagfile=flags.txt`).\n\nBy default logs are written to a file with the name of the executable. You can\noverride this with `--log_file=log.txt`.\n\nIf running under Visual Studio and you want to look at the JIT'ed code\n(available around 0xA0000000) you should pass `--emit_source_annotations` to\nget helpful spacers/movs in the disassembly.\n\n### Linux\n\nLinux support is extremely experimental and presently incomplete.\n\nThe build script uses LLVM/Clang 9. GCC while it should work in theory, is not easily\ninterchangeable right now.\n\n* Normal building via `xb build` uses Make.\n* [CodeLite](https://codelite.org) is supported. `xb devenv` will generate a workspace and attempt to open it. Your distribution's version may be out of date so check their website.\n* Experimental CMake generation is available to facilitate use of other IDEs such as [CLion](https://www.jetbrains.com/clion/). If `clion` is available inside `$PATH`, `xb devenv` will start it. Otherwise `build/CMakeLists.txt` needs to be generated by invoking `xb premake --devenv=cmake` manually.\n\nClang-9 or newer should be available from system repositories on all up to date distributions.\nYou will also need some development libraries. To get them on an Ubuntu system:\n\n```bash\nsudo apt-get install libgtk-3-dev libpthread-stubs0-dev liblz4-dev libx11-dev libx11-xcb-dev libvulkan-dev libsdl2-dev libiberty-dev libunwind-dev libc++-dev libc++abi-dev\n```\n\nIn addition, you will need up to date Vulkan libraries and drivers for your hardware, which most distributions have in their standard repositories nowadays.\n\n## Running\n\nTo make life easier you can set the program startup arguments in your IDE to something like `--log_file=stdout /path/to/Default.xex` to log to console rather than a file and start up the emulator right away.\n" }, { "path": "docs/cpu.md", "content": "# CPU Documentation\n\n## The JIT\n\n![JIT Diagram](images/CPU-JIT.png?raw=true)\n\nThe JIT is the core of Xenia. It translates Xenon PowerPC code into native\ncode runnable on the host computer.\n\nThere are 3 phases to translation:\n1. Translation to IR (intermediate representation)\n2. IR compilation/optimization\n3. Backend emission\n\nPowerPC instructions are translated to Xenia's intermediate representation\nformat in src/xenia/cpu/ppc/ppc_emit_*.cc (e.g. processor control is done in\n[ppc_emit_control.cc](../src/xenia/cpu/ppc/ppc_emit_control.cc)). HIR opcodes\nare relatively simple opcodes such that any host can define an implementation.\n\nAfter the HIR is generated, it is ran through a compiler to prep it for generation.\nThe compiler is ran in a series of passes, the order of which is defined in\n[ppc_translator.cc](../src/xenia/cpu/ppc/ppc_translator.cc). Some passes are\nessential to the successful generation, while others are merely for optimization\npurposes. Compiler passes are defined in src/xenia/cpu/compiler/passes with\ndescriptive class names.\n\nFinally, the backend consumes the HIR and emits code that runs natively on the\nhost. Currently, the only backend that exists is the x64 backend, with all the\nemission done in\n[x64_sequences.cc](../src/xenia/cpu/backend/x64/x64_sequences.cc).\n\n## ABI\n\nXenia guest functions are not directly callable, but rather must be called\nthrough APIs provided by Xenia. Xenia will first execute a thunk to transition\nthe host context to a state dependent on the JIT backend, and that will call the\nguest code.\n\n### x64\n\nTransition thunks defined in [x64_backend.cc](../src/xenia/cpu/backend/x64/x64_backend.cc#L389).\nRegisters are stored on the stack as defined by [StackLayout::Thunk](../src/xenia/cpu/backend/x64/x64_stack_layout.h#L96)\nfor later transitioning back to the host.\n\nSome registers are reserved for usage by the JIT to store temporary variables.\nSee: [X64Emitter::gpr_reg_map_ and X64Emitter::xmm_reg_map_](../src/xenia/cpu/backend/x64/x64_emitter.cc#L57).\n\n#### Integer Registers\n\nRegister | Usage\n--- | ---\nRAX | Scratch\nRBX | JIT temp\nRCX | Scratch\nRDX | Scratch\nRSP | Stack Pointer\nRBP | Unused\nRSI | PowerPC Context\nRDI | Virtual Memory Base\nR8-R11 | Unused (parameters)\nR12-R15 | JIT temp\n\n#### Floating Point Registers\nRegister | Usage\n--- | ---\nXMM0-XMM5 | Scratch\nXMM6-XMM15 | JIT temp\n\n## Memory\n\nXenia defines virtual memory as a mapped range beginning at Memory::virtual_membase(),\nand physical memory as another mapped range from Memory::physical_membase()\n(usually 0x100000000 and 0x200000000, respectively). If the default bases are\nnot available, they are shifted left 1 bit until an available range is found.\n\nThe guest only has access to these ranges, nothing else.\n\n### Map\n```\n0x00000000 - 0x3FFFFFFF (1024mb) - virtual 4k pages\n0x40000000 - 0x7FFFFFFF (1024mb) - virtual 64k pages\n0x80000000 - 0x8BFFFFFF ( 192mb) - xex 64k pages\n0x8C000000 - 0x8FFFFFFF ( 64mb) - xex 64k pages (encrypted)\n0x90000000 - 0x9FFFFFFF ( 256mb) - xex 4k pages\n0xA0000000 - 0xBFFFFFFF ( 512mb) - physical 64k pages (overlapped)\n0xC0000000 - 0xDFFFFFFF - physical 16mb pages (overlapped)\n0xE0000000 - 0xFFFFFFFF - physical 4k pages (overlapped)\n```\n\nVirtual pages are usually allocated by NtAllocateVirtualMemory, and\nphysical pages are usually allocated by MmAllocatePhysicalMemoryEx.\n\nVirtual pages mapped to physical memory are also mapped to the physical membase,\ni.e. virtual 0xA0000000 == physical 0x00000000\n\nThe 0xE0000000-0xFFFFFFFF range is mapped to physical memory with a single 4 KB\npage offset. On Windows, memory mappings must be aligned to 64 KB, so the offset\nhas to be added when guest addresses are converted to host addresses in the\ntranslated CPU code. This can't be faked other ways because calculations\ninvolving the offset are built into games - see the following sequence:\n\n```\nsrwi r9, r10, 20 # r9 = r10 >> 20\nclrlwi r10, r10, 3 # r10 = r10 & 0x1FFFFFFF (physical address)\naddi r11, r9, 0x200\nrlwinm r11, r11, 0,19,19 # r11 = r11 & 0x1000\nadd r11, r11, r10 # add 1 page to addresses > 0xE0000000\n\n# r11 = addess passed to GPU\n```\n\n## Memory Management\n\nTODO\n\n## References\n\n### PowerPC\n\nThe processor in the 360 is a 64-bit PowerPC chip running in 32-bit mode.\nPrograms are still allowed to use 64-bit PowerPC instructions, and registers\nare 64-bit as well, but 32-bit instructions will run in 32-bit mode.\nThe CPU is largely similar to the PPC part in the PS3, so Cell documents\noften line up for the core instructions. The 360 adds some additional AltiVec\ninstructions, though, which are only documented in a few places (like the gcc source code, etc).\n\n* [Free60 Info](https://free60project.github.io/wiki/Xenon_(CPU))\n* [Power ISA docs](https://web.archive.org/web/20140603115759/https://www.power.org/wp-content/uploads/2012/07/PowerISA_V2.06B_V2_PUBLIC.pdf) (aka 'PowerISA')\n* [PowerPC Programming Environments Manual](https://web.archive.org/web/20141028181028/https://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/F7E732FF811F783187256FDD004D3797/$file/pem_64bit_v3.0.2005jul15.pdf) (aka 'pem_64')\n* [PowerPC Vector PEM](https://web.archive.org/web/20130502201029/https://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/C40E4C6133B31EE8872570B500791108/$file/vector_simd_pem_v_2.07c_26Oct2006_cell.pdf)\n* [AltiVec PEM](https://web.archive.org/web/20151110180336/https://cache.freescale.com/files/32bit/doc/ref_manual/ALTIVECPEM.pdf)\n* [VMX128 Opcodes](http://biallas.net/doc/vmx128/vmx128.txt)\n* [AltiVec Decoding](https://github.com/kakaroto/ps3ida/blob/master/plugins/PPCAltivec/src/main.cpp)\n\n### x64\n\n* [Intel Manuals](https://software.intel.com/en-us/articles/intel-sdm)\n * [Combined Intel Manuals](https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-manual-325462.pdf)\n* [Apple AltiVec/SSE Migration Guide](https://developer.apple.com/legacy/library/documentation/Performance/Conceptual/Accelerate_sse_migration/Accelerate_sse_migration.pdf)\n" }, { "path": "docs/cpu_todo.md", "content": "# CPU TODO\n\nThere are many improvements that can be done under `xe::cpu` to improve\ndebugging, performance (both to JIT and of generated code), and portability.\nSome are in various states of completion, and others are just thoughts that need\nmore exploring.\n\n## Debugging Improvements\n\n### Reproducable X64 Emission\n\nIt'd be useful to be able to run a PPC function through the entire pipeline and\nspit out x64 that is byte-for-byte identical across runs. This would allow\nautomated verification, bulk analysis, etc. Currently `X64Emitter::Emplace`\nwill relocate the x64 when placing it in memory, which will be at a different\nlocation each time. Instead it would be nice to have the xbyak `calcJmpAddress`\nthat performs the relocations use the address of our choosing.\n\n### Sampling Profiler\n\nOnce we have stack walking it'd be nice to take something like\n[micro-profiler](https://code.google.com/p/micro-profiler/) and augment it to\nsupport our system. This would let us run continuous performance analysis and\ntrack hotspots in JITed code without a large performance impact. Automatically\nshowing the top hot functions in the debugger could help track down poor\ntranslation much faster.\n\n### Intel Architecture Code Analyzer Support\n\nThe [Intel ACA](https://software.intel.com/en-us/articles/intel-architecture-code-analyzer)\nis a nifty tool that, given a kernel of x64, can detail theoretical performance\ncharacteristics on different processors down to cycle timings and potential\nbottlenecks on memory/execution units. It's designed to run on elf/obj/etc files\nhowever it simply looks for special markers in the code. Having something that\nwalks the code cache and dumps a specially formatted file with the markers\naround basic blocks could allow running the tool in bulk, or alternatively being\nable to invoke it one-off by dumping a specific x64 block to disk and processing\nit for display when looking at the code in the debugger would be useful.\n\nI've done some early experiments with this and its possible to pass just a\nbin file with the markers and the x64.\n\n### Function Tracing/Coverage Information\n\n`function_trace_data.h` contains the `FunctionTraceData` struct, which is\ncurrently partially populated by the x64 backend. This enables tracking of which\nthreads a function is called on, function call count, recent callers of the\nfunction, and even instruction-level counts.\n\nThis is all only partially implemented, though, and there's no tool to read it\nout. This would be nice to get integrated into the debugger so that it can\noverlay the information when viewing a function, but also useful in aggregate to\nfind hot functions/code paths or enhance callstacks by automatically annotating\nthread information.\n\n#### Block-level Counting\n\nCurrently the code assumes each instruction has a count, however this is\nexpensive and often unneeded as it can be done on a block level and then the\ninstruction counts can be derived from that. This can reduce the overhead (both\nin memory and accounting time) by an order of magnitude.\n\n### On-Stack Context Inspection\n\nCurrently the debugger only works with `--store_all_context_values`, as it can\nonly get the values of PPC registers when they are stored to the PPC context\nafter each instruction. As this can slow things down by ~10-20% it could be\nuseful to be able to preserve the optimized and register-allocated HIR so that\nhost registers holding context values can be derived on demand. Or, we could\njust make `--store_all_context_values` faster.\n\n## JIT Performance Improvements\n\n### Reduce HIR Size\n\nCurrently there are a lot of pointers stored within `Instr`, `Value`, and\nrelated types. These are big 8B values that eat a lot of memory and really\nhurt the cache (especially with all the block/instruction walking done).\nAligning everything to 16B values in the arena and using 16bit indices\n(or something) could shrink things a lot.\n\n### Serialize Code Cache\n\nThe x64 code cache is currently set up to use fixed memory addresses and is even\nrepresented as mapped memory. It should be fairly easy to back this with a file\nand have all code written to disk. Adding more metadata, or perhaps a side-car\nfile, would allow for the code to be written to disk. On future runs the code\ncache could load this data (by mapping the file containing the code right into\nmemory) and short cut JIT'ing entirely.\n\nIt would be possible to use a common container format (ELF/etc), however there's\nelegance in not requiring any additional steps beyond the memory mapping. Such\ncontainers could be useful for running static tools against, though.\n\n## Portability Improvements\n\n### Emulated Opcode Layer\n\nHaving a way to use emulated variants for any HIR opcode in a backend would\nhelp when writing a new backend as well as when verifying the existing backends.\nThis may look like a C library with functions for each opcode/type pairing and\nutilities to call out to them. Something like the x64 backend could then call\nout to these with CallNativeSafe (or some faster equivalent) and something like\nan interpreter backend would be fairly trivial to write.\n\n## X64 Backend Improvements\n\n### Implement Emulated Instructions\n\nThere are a ton of half-implemented HIR opcodes that call out to C++ to do their\nwork. These are extremely expensive as they incur a full guest-to-host thunk\n(~hundreds of instructions!). Basically, any of the `Emulate*`/`CallNativeSafe`\nfunctions in `x64_sequences.cc` need to be replaced with proper AVX/AVX2\nvariants.\n\n### Increase Register Availability\n\nCurrently only a few x64 registers are usable (due to reservations by the\nbackend or ABI conflicts). Though register pressure is surprisingly light in\nmost cases there are pathological cases that result in a lot of spills. By\nfreeing up some of the registers these spills could be reduced.\n\n### Constant Pooling\n\nThis may make sense as a compiler pass instead.\n\nRight now, particular sequences of instructions are nasty - such as anything\nusing `LoadConstantXmm` to load non-zero or non-1 vec128's. Instead of doing the\nsuper fat (20-30byte!) constant loads as they are done now it may be better to\nkeep a per-function constant table and instead use RIP-relative addressing (or\nsomething) to use the memory-form AVX instructions.\n\nFor example, right now this:\n```\n v82.v128 = [0,1,2,3]\n v83.v128 = or v81.v128, v82.128\n```\n\nTranslates to (something like):\n```\n mov([rsp+0x...], 0x00000000)\n mov([rsp+0x...+4], 0x00000001)\n mov([rsp+0x...+8], 0x00000002)\n mov([rsp+0x...+12], 0x00000003)\n vmovdqa(xmm2, [rsp+0x...])\n vor(xmm2, xmm2, xmm2)\n```\n\nWhere as it could be:\n```\n vor(xmm2, xmm2, [rip+0x...])\n```\n\nWhether the cost of doing the constant de-dupe is worth it remains to be seen.\nRight now it's wasting a lot of instruction cache space, increasing decode time,\nand potentially using a lot more memory bandwidth.\n\n## Optimization Improvements\n\n### Speed Up RegisterAllocationPass\n\nCurrently the slowest pass, this could be improved by requiring less use\ntracking or perhaps maintaining the use tracking in other passes. A faster\nSortUsageList (radix or something fancy?) may be helpful as well.\n\n### More Opcodes in ConstantPropagationPass\n\nThere's a few HIR opcodes with no handling, and others with minimal handling.\nIt'd be nice to know what paths need improvement and add them, as any work here\nmakes things free later on.\n\n### Cross-Block ConstantPropagationPass\n\nConstant propagation currently only occurs within a single block. This makes it\ndifficult to optimize common PPC patterns like loading the constants 0 or 1 into\na register before a loop and other loads of expensive altivec values.\n\nEither ControlFlowAnalysisPass or DataFlowAnalysisPass could be piggy-backed to\ntrack constant load_context/store_context's across block bounds and propagate\nthe values. This is simpler than dynamic values as no phi functions or anything\nfancy needs to happen.\n\n### Add TypePropagationPass\n\nThere are many extensions/truncations in generated code right now due to\nvarious load/stores of varying widths. Being able to find and short-\ncircuit the conversions early on would make following passes cleaner\nand faster as they'd have to trace through fewer value definitions and there'd\nbe less extraneous movs in the final code.\n\nExample (after ContextPromotion):\n```\n v82.i32 = truncate v81.i64\n v83.i32 = and v82.i32, 3F\n v85.i64 = zero_extend v84.i32\n```\n\nBecomes (after DCE/etc):\n```\n v85.i64 = and v81.i64, 3F\n```\n\n### Enhance MemorySequenceCombinationPass with Extend/Truncate\n\nCurrently this pass will look for byte_swap and merge that into loads/stores.\nThis allows for better final codegen at the cost of making optimization more\ndifficult, so it only happens at the end of the process.\n\nThere's currently TODOs in there for adding extend/truncate support, which\nwill extend what it does with swaps to also merge the\nsign_extend/zero_extend/truncate into the matching load/store. This allows for\nthe x64 backend to generate the proper mov's that do these operations without\nrequiring additional steps. Note that if we had a LIR and a peephole optimizer\nthis would be better done there.\n\nLoad with swap and extend:\n```\n v1.i32 = load v0\n v2.i32 = byte_swap v1.i32\n v3.i64 = zero_extend v2.i32\n```\n\nBecomes:\n```\n v1.i64 = load_convert v0, [swap|i32->i64,zero]\n```\n\nStore with truncate and swap:\n```\n v1.i64 = ...\n v2.i32 = truncate v1.i64\n v3.i32 = byte_swap v2.i32\n store v0, v3.i32\n```\n\nBecomes:\n```\n store_convert v0, v1.i64, [swap|i64->i32,trunc]\n```\n\n### Add DeadStoreEliminationPass\n\nGeneric DSE pass, removing all redundant stores. ContextPromotion may be\nable to take care of most of these, as the input assembly is generally\npretty optimized already. This pass would mainly be looking for introduced\nstores, such as those from comparisons.\n\nCurrently ControlFlowAnalysisPass will annotate blocks with incoming/outgoing\nedges as well as dominators, and that could be used to check whether stores into\nthe context are used in their destination block or instead overwritten\n(currently they almost never are).\n\nIf this pass was able to remove a good number of the stores then the comparisons\nwould also be removed with dead code elimination and dramatically reduce branch\noverhead.\n\nExample:\n```\n:\n v0 = compare_ult ... (later removed by DCE)\n v1 = compare_ugt ... (later removed by DCE)\n v2 = compare_eq ...\n store_context +300, v0 <-- removed\n store_context +301, v1 <-- removed\n store_context +302, v2 <-- removed\n branch_true v1, ...\n:\n v3 = compare_ult ...\n v4 = compare_ugt ...\n v5 = compare_eq ...\n store_context +300, v3 <-- these may be required if at end of function\n store_context +301, v4 or before a call\n store_context +302, v5\n branch_true v5, ...\n```\n\n### Add X64CanonicalizationPass\n\nFor various opcodes add copies/commute the arguments to match x64\noperand semantics. This makes code generation easier and if done\nbefore register allocation can prevent a lot of extra shuffling in\nthe emitted code.\n\nExample:\n```\n:\n v0 = ...\n v1 = ...\n v2 = add v0, v1 <-- v1 now unused\n```\n\nBecomes:\n```\n v0 = ...\n v1 = ...\n v1 = add v1, v0 <-- src1 = dest/src, so reuse for both\n by commuting and setting dest = src1\n```\n\n### Add MergeLocalSlotsPass\n\nAs the RegisterAllocationPass runs it generates load_local/store_local as it\nspills. Currently each set of locals is unique to each block, which in very\nlarge functions can result in a lot of locals that are only used briefly. It\nmay be useful to use the results of the ControlFlowAnalysisPass to track local\nliveness and merge the slots so they are reused when they cannot possibly be\nlive at the same time. This saves stack space and potentially improves cache\nbehavior.\n" }, { "path": "docs/gpu.md", "content": "# GPU Documentation\n\n## The Xenos Chip\n\nThe [Xenos](https://en.wikipedia.org/wiki/Xenos_\\(graphics_chip\\)) is a graphics\nchip designed by AMD based off of the R5xx architecture.\n\n### Command Processing\n\nThe Xenos runs commands supplied to it directly by the DirectX bare-bones driver\nvia a ringbuffer located in system memory.\n\nThe bulk of the command processing code is located at\n[src/xenia/gpu/command_processor.cc](../src/xenia/gpu/command_processor.cc)\n\n### EDRAM\n\nThe Xenos uses special high-speed memory located on the same die as the chip to \nstore framebuffers/render targets.\n\nTODO: More documentation\n\n## Options\n\n### General\n\nSee the top of [src/xenia/gpu/gpu_flags.cc](../src/xenia/gpu/gpu_flags.cc).\n\n`--vsync=false` will attempt to render the game as fast as possible instead of\nwaiting for a fixed 60hz timer.\n\n### Vulkan\n\nSee the top of [src/xenia/gpu/vulkan/vulkan_gpu_flags.cc](../src/xenia/gpu/vulkan/vulkan_gpu_flags.cc).\n\n`vulkan_dump_disasm=true` \"Dump shader disassembly. NVIDIA only supported.\"\n\n## Tools\n\n### Shaders\n\n#### Shader Dumps\n\nAdding `--dump_shaders=path/` will write all translated shaders to the given\npath with names based on input hash (so they'll be stable across runs).\nBinaries containing the original microcode will be placed side-by-side with\nthe dumped output to make it easy to pipe to `xe-gpu-shader-compiler`.\n\n#### xe-gpu-shader-compiler\n\nA standalone shader compiler exists to allow for quick shader translation\ntesting. You can pass a binary ucode shader in and get either disassembled\nucode or translated source out. This is best used through the Shader\nPlayground tool.\n\n```\n xe-gpu-shader-compiler \\\n --shader_input=input_file.bin.vs (or .fs)\n --shader_output=output_file.txt\n --shader_output_type=ucode (or spirvtext)\n```\n\n#### Shader Playground\n\nBuilt separately (for now) under [tools/shader-playground/](../tools/shader-playground/)\nis a GUI for interactive shader assembly, disassembly, validation, and\ntranslation.\n\n![Shader Playground Screenshot](images/shader_playground.png?raw=true)\n\nEntering shader microcode on the left will invoke the XNA Game Studio\nD3D compiler to translate the ucode to binary. The D3D compiler is then\nused to disassemble the binary and display the optimized form. If\n`xe-gpu-shader-compiler` has been built the ucode will be passed to that\nfor disassembly and that will then be passed through D3D compiler. If\nthe output of D3D compiler on the xenia disassembly doesn't match the\noriginal D3D compiler output the box will turn red, indicating that the\ndisassembly is broken. Finally, the right most box will show the\ntranslated shader in the desired format.\n\nFor more information and setup instructions see\n[tools/shader-playground/README.md](../tools/shader-playground/README.md).\n\n### xe-gpu-trace-viewer\n\nTo quickly iterate on graphical issues, xenia can dump frames (or sequences of\nframes) while running that can be opened and inspected in a separate app.\n\nThe basic workflow is:\n\n1. Capture the frame in game (using F4) or a stream of frames.\n2. Add the file path to the xe-gpu-trace-viewer Debugging command line in\nVisual Studio.\n3. Launch xe-gpu-trace-viewer.\n4. Poke around, find issues, etc.\n5. Modify code.\n6. Build and relaunch.\n7. Goto 4.\n\n#### Capturing Frames\n\nFirst, specify a path to capture traces to with\n`--trace_gpu_prefix=path/file_prefix_`. All files will have a randomish name\nbased on that.\n\nWhen running xenia.exe you can hit F4 at any time to capture the next frame the\ngame tries to draw (up until a VdSwap call). The file can be used immediately.\n\n#### Capturing Sequences\n\nPassing `--trace_gpu_stream` will write all frames rendered to a file, allowing\nyou to seek through them in the trace viewer. These files will get large.\n\n## References\n\n### Command Buffer/Registers\n\nRegisters documented at [src/xenia/gpu/register_table.inc](../src/xenia/gpu/register_table.inc).\n\nPM4 commands documented at [src/xenia/gpu/xenos.h](../src/xenia/gpu/xenos.h#L521).\n\n#### Performance Counters that may be read back by D3D\n\nThey are 64-bit values and have a high and low 32-bit register as well as a `SELECT` register each:\n\n- CP_PERFCOUNTER0\n\n- RBBM_PERFCOUNTER0\n- RBBM_PERFCOUNTER1\n\n- SQ_PERFCOUNTER0\n- SQ_PERFCOUNTER1\n- SQ_PERFCOUNTER2\n- SQ_PERFCOUNTER3\n\n- VGT_PERFCOUNTER0\n- VGT_PERFCOUNTER1\n- VGT_PERFCOUNTER2\n- VGT_PERFCOUNTER3\n\n- VC_PERFCOUNTER0\n- VC_PERFCOUNTER1\n- VC_PERFCOUNTER2\n- VC_PERFCOUNTER3\n\n- PA_SU_PERFCOUNTER0\n- PA_SU_PERFCOUNTER1\n- PA_SU_PERFCOUNTER2\n- PA_SU_PERFCOUNTER3\n\n- PA_SC_PERFCOUNTER0\n- PA_SC_PERFCOUNTER1\n- PA_SC_PERFCOUNTER2\n- PA_SC_PERFCOUNTER3\n\n- HZ_PERFCOUNTER0\n- HZ_PERFCOUNTER1\n\n- TCR_PERFCOUNTER0\n- TCR_PERFCOUNTER1\n\n- TCM_PERFCOUNTER0\n- TCM_PERFCOUNTER1\n\n- TCF_PERFCOUNTER0\n- TCF_PERFCOUNTER1\n- TCF_PERFCOUNTER2\n- TCF_PERFCOUNTER3\n- TCF_PERFCOUNTER4\n- TCF_PERFCOUNTER5\n- TCF_PERFCOUNTER6\n- TCF_PERFCOUNTER7\n- TCF_PERFCOUNTER8\n- TCF_PERFCOUNTER9\n- TCF_PERFCOUNTER10\n- TCF_PERFCOUNTER11\n\n- TP0_PERFCOUNTER0\n- TP0_PERFCOUNTER1\n- TP1_PERFCOUNTER0\n- TP1_PERFCOUNTER1\n- TP2_PERFCOUNTER0\n- TP2_PERFCOUNTER1\n- TP3_PERFCOUNTER0\n- TP3_PERFCOUNTER1\n\n- SX_PERFCOUNTER0\n\n- BC_PERFCOUNTER0\n- BC_PERFCOUNTER1\n- BC_PERFCOUNTER2\n- BC_PERFCOUNTER3\n\n- MC0_PERFCOUNTER0\n- MC1_PERFCOUNTER0\n\n- MH_PERFCOUNTER0\n- MH_PERFCOUNTER1\n- MH_PERFCOUNTER2\n\n- BIF_PERFCOUNTER0\n\n### Shaders\n\n* [LLVM R600 Tables](https://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/R600Instructions.td)\n** The opcode formats don't match, but the name->psuedo code is correct.\n* [xemit](https://github.com/gligli/libxemit/blob/master/xemitops.c)\n" }, { "path": "docs/instruction_tracing.md", "content": "In x64_tracers.cc:\n\nEnable tracing:\n```\n#define ITRACE 1 <--- for only ppc instructions\n#define DTRACE 1 <--- add HIR data\n```\n\nIf tracing data, run with the following flags:\n```\n--store_all_context_values\n```\n\nBy default, tracing will start at the beginning and only for the specified\nthread.\n\nChange traced thread by thread creation ID:\n```\n#define TARGET_THREAD 4\n```\n\nTo only trace at a certain point, change default trace flag to false:\n```\nbool trace_enabled = true;\n```\nAdd a breakpoint:\n```\n--break_on_instruction=0x821009A4\n```\nOn break, add the following to the Watch window and set it to true:\n```\nxe::cpu::backend::x64::trace_enabled\n```\nContinue, and watch stuff appear in the log.\n" }, { "path": "docs/kernel.md", "content": "# Kernel Documentation\n\n## Kernel shims\nXenia implements all kernel APIs as native functions under the host.\n\nWhen a module is loaded, the loader will find all kernel imports, put a syscall in\ntheir place, then lookup a kernel export and link it to each import. The JIT will\ngenerate a sequence of instructions to call into Xenia's export if it encounters a syscall.\n\nCurrently, there are two ways an export can be defined -\n[for example](../src/xenia/kernel/xboxkrnl/xboxkrnl_audio.cc):\n* `SHIM_CALL XAudioGetSpeakerConfig_shim(PPCContext* ppc_context, KernelState* kernel_state)`\n* `dword_result_t XAudioGetSpeakerConfig(lpdword_t config_ptr)`\n\nThe `SHIM_CALL` convention is deprecated, but allows a closer look at the internals of how\ncalls are done. `ppc_context` is the guest PowerPC context, which holds all guest\nregisters. Function parameters are fetched from r3...r10 (`SHIM_GET_ARG_32`), and\nadditional parameters are loaded from the stack. The return value (if there is one)\nis stored in r3 (`SHIM_SET_RETURN_32`).\n\nDetails on how calls transition from guest -> host can be found in the [cpu documentation](cpu.md).\n\nThe newer convention does the same, but uses templates to automate the process\nof loading arguments and setting a return value.\n\n## Kernel Modules\nXenia has an implementation of two xbox kernel modules, xboxkrnl.exe and xam.xex\n\n### xboxkrnl.exe - Xbox kernel\n\nDefined under src/xenia/kernel/xboxkrnl.\n\nThis is a slightly modified version of the NT kernel. Most of the APIs \nare equivalent to ones you'd find on MSDN or other online sources.\n\nSource files are organized into groups of APIs.\n\n### xam.xex - Xbox Auxiliary Methods\n\nDefined under src/xenia/kernel/xam.\n\nThis module implements functionality specific to the Xbox." }, { "path": "docs/ppc/vmx128.txt", "content": " 2006/09/01 Revision 1.2\n-----------------------------------------------------------------\n\n\n This is a description of the VMX128-type opcodes found on\n the xbox360 processor. I figured this out by looking at\n various disassmblies, so there might some errors and\n missing instructions. Some instructions have unknown\n semantics for me.\n\n See comments or corrections to sb#biallas.net\n\n\n=================================================================\nConventions:\n VD128, VS128: 5 lower bits of a VMX128 vector register \n number\n VDh:\t upper 2 bits of VD128\n (so register number is (VDh << 5 | VD128))\n\t\t \n VA128: same as VD128\n A: bit 6 of VA128\n a: bit 5 of VA128\n (so register number is (A<<6 | a<<5 | VA128))\n \n VB128: same as VD128\n VBh: same as VDh\n \n VC128: 3 bits of a VMX128 vector register number\n (you can only use vr0-vr7 here)\n \n RA, RB: general purpose register number\n\n UIMM: unsigned immediate value\n SIMM: signed immediate value\n \n PERMh: upper 3 bits of a permutation\n PERMl: lower 5 bits of a permutation\n \n x, y, z: unknown immediate values\n\n=================================================================\n lvewx128 Load Vector128 Element Word Indexed\n|0 0 0 1 0 0| VD128 | RA | RB |0 0 0 1 0 0 0|VDh|1 1|\n\n lvewx128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n lvlx128 Load Vector128 Left Indexed\n|0 0 0 1 0 0| VD128 | RA | RB |1 0 0 0 0 0 0|VDh|1 1|\n\n lvlx128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n lvrx128 Load Vector128 Right Indexed\n|0 0 0 1 0 0| VD128 | RA | RB |1 0 0 0 1 0 0|VDh|1 1|\n\n lvrx128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n lvlxl128 Load Vector128 Left Indexed LRU\n|0 0 0 1 0 0| VD128 | RA | RB |1 1 0 0 0 0 0|VDh|1 1|\n\n lvlxl128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n lvrxl128 Load Vector128 Right Indexed LRU\n|0 0 0 1 0 0| VD128 | RA | RB |1 1 0 0 1 0 0|VDh|1 1|\n\n lvrxl128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n lvsl128 Load Vector128 for Shift Left\n|0 0 0 1 0 0| VD128 | RA | RB |0 0 0 0 0 0 0|VDh|1 1|\n\n lvsl128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n lvsr128 Load Vector128 for Shift Right\n|0 0 0 1 0 0| VD128 | RA | RB |0 0 0 0 1 0 0|VDh|1 1|\n\n lvsr128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n lvx128 Load Vector128 Indexed\n|0 0 0 1 0 0| VD128 | RA | RB |0 0 0 1 1 0 0|VDh|1 1|\n\n lvx128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n lvxl128 Load Vector128 Indexed LRU\n|0 0 0 1 0 0| VD128 | RA | RB |0 1 0 1 1 0 0|VDh|1 1|\n\n lvxl128 vr(VD128), r(RA), r(RB)\n\n\n=================================================================\n stewx128 Store Vector128 Element Word Indexed\n|0 0 0 1 0 0| VS128 | RA | RB |0 1 1 0 0 0 0|VDh|1 1|\n\n stvewx128 vr(VS128), r(RA), r(RB)\n\n\n=================================================================\n stvlx128 Store Vector128 Left Indexed\n|0 0 0 1 0 0| VS128 | RA | RB |1 0 1 0 0 0 0|VDh|1 1|\n\n stvlx128 vr(VS128), r(RA), r(RB)\n\n\n=================================================================\n stvlxl128 Store Vector128 Left Indexed LRU\n|0 0 0 1 0 0| VS128 | RA | RB |1 1 1 0 0 0 0|VDh|1 1|\n\n lvlxl128 vr(VS128), r(RA), r(RB)\n\n\n=================================================================\n stvrx128 Store Vector128 Right Indexed\n|0 0 0 1 0 0| VS128 | RA | RB |1 0 1 0 1 0 0|VDh|1 1|\n\n stvrx128 vr(VS128), r(RA), r(RB)\n\n\n=================================================================\n stvrxl128 Store Vector128 Right Indexed LRU\n|0 0 0 1 0 0| VS128 | RA | RB |1 1 1 0 1 0 0|VDh|1 1|\n\n stvrxl128 vr(VS128), r(RA), r(RB)\n\n\n=================================================================\n stvx128 Store Vector128 Indexed\n|0 0 0 1 0 0| VS128 | RA | RB |0 0 1 1 1 0 0|VDh|1 1|\n\n stvx128 vr(VS128), r(RA), r(RB)\n\n\n=================================================================\n stvxl128 Store Vector128 Indexed LRU\n|0 0 0 1 0 0| VS128 | RA | RB |0 1 1 1 1 0 0|VDh|1 1|\n\n stvxl128 vr(VS128), r(RA), r(RB)\n\n\n=================================================================\n vaddfp128 Vector128 Add Floating Point\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|0 0 0 0|a|1|VDh|VBh|\n\n vaddfp128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vand128 Vector128 Logical AND\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 0 0 0|a|1|VDh|VBh|\n\n vand128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vandc128 Vector128 Logical AND \n with Complement\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 0 1 0|a|1|VDh|VBh|\n\n vandc128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vcfpsxws128 Vector128 Convert From Floating-Point to \n Signed Fixed-Point Word Saturate\n|0 0 0 1 1 0| VD128 | SIMM | VB128 |0 1 0 0 0 1 1|VDh|VBh|\n\n vcfpsxws128 vr(VD128), vr(VB128), SIMM\n\n\n=================================================================\n vcfpuxws128 Vector128 Convert From Floating-Point to \n Unsigned Fixed-Point Word Saturate\n|0 0 0 1 1 0| VD128 | UIMM | VB128 |0 1 0 0 1 1 1|VDh|VBh|\n\n vcfpuxws128 vr(VD128), vr(VB128), UIMM\n\n\n=================================================================\n vcmpbfp128 Vector128 Compare Bounds \n Floating Point\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|0 1 1|R|a|0|VDh|VBh|\n\n vcmpbfp128 vr(VD128), vr(VA128), vr(VB128) (R == 0)\n vcmpbfp128. vr(VD128), vr(VA128), vr(VB128) (R == 1)\n\n\n=================================================================\n vcmpeqfp128 Vector128 Compare Equal-to\n Floating Point\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|0 0 0|R|a|0|VDh|VBh|\n\n vcmpeqfp128 vr(VD128), vr(VA128), vr(VB128) (R == 0)\n vcmpeqfp128. vr(VD128), vr(VA128), vr(VB128) (R == 1)\n\n\n=================================================================\n vcmpequw128 Vector128 Compare Equal-to \n Unsigned Word\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|1 0 0|R|a|0|VDh|VBh|\n\n vcmpequw128 vr(VD128), vr(VA128), vr(VB128) (R == 0)\n vcmpequw128. vr(VD128), vr(VA128), vr(VB128) (R == 1)\n\n\n=================================================================\n vcmpgefp128 Vector128 Compare Greater-Than-\n or-Equal-to Floating Point\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|0 0 1|R|a|0|VDh|VBh|\n\n vcmpgefp128 vr(VD128), vr(VA128), vr(VB128) (R == 0)\n vcmpgefp128. vr(VD128), vr(VA128), vr(VB128) (R == 1)\n\n\n=================================================================\n vcmpgtfp128 Vector128 Compare Greater-Than \n Floating-Point\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|0 1 0|R|a|0|VDh|VBh|\n\n vcmpgtfp128 vr(VD128), vr(VA128), vr(VB128) (R == 0)\n vcmpgtfp128. vr(VD128), vr(VA128), vr(VB128) (R == 1)\n\n\n=================================================================\n vcsxwfp128 Vector128 Convert From Signed Fixed-Point \n Word to Floating-Point\n|0 0 0 1 1 0| VD128 | UIMM | VB128 |0 1 0 1 0 1 1|VDh|VBh|\n\n vcsxwfp128 vr(VD128), vr(VB128), SIMM\n\n\n=================================================================\n vcuxwfp128 Vector128 Convert From Unsigned Fixed-Point\n Word to Floating-Point\n|0 0 0 1 1 0| VD128 | UIMM | VB128 |0 1 0 1 1 1 1|VDh|VBh|\n\n vcuxwfp128 vr(VD128), vr(VB128), UIMM\n\n\n=================================================================\n vexptefp128 Vector128 2 Raised to the Exponent \n Estimate Floating Point\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |1 1 0 1 0 1 1|VDh|VBh|\n\n vexptefp128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vlogefp128 Vector128 Log2 Estimate \n Floating Point\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |1 1 0 1 1 1 1|VDh|VBh|\n\n vlogefp128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vmaddcfp128 Vector128 Multiply Add \n Floating Point\n|0 0 0 1 0 1| VDS128 | VA128 | VB128 |A|0 1 0 0|a|1|VDh|VBh|\n\n vmaddcfp128 vr(VDS128), vr(VA128), vr(VSD128), vr(VB128)\n\n\n=================================================================\n vmaddfp128 Vector128 Multiply Add \n Floating Point\n|0 0 0 1 0 1| VDS128 | VA128 | VB128 |A|0 0 1 1|a|1|VDh|VBh|\n\n vmaddfp128 vr(VDS128), vr(VA128), vr(VB128), vr(VDS128)\n\n\n=================================================================\n vmaxfp128 Vector128 Maximum \n Floating Point\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|1 0 1 0|a|0|VDh|VBh|\n\n vmaxfp128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vminfp128 Vector128 Minimum\n Floating Point\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|1 0 1 1|a|0|VDh|VBh|\n\n vminfp128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vmrghw128 Vector128 Merge High Word\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|1 1 0 0|a|0|VDh|VBh|\n\n vmrghw128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vmrglw128 Vector128 Merge Low Word\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|1 1 0 1|a|0|VDh|VBh|\n\n vmrglw128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vmsum3fp128 Vector128 Multiply Sum 3-way \n Floating Point\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|0 1 1 0|a|1|VDh|VBh|\n\n vmsub3fp128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vmsum4fp128 Vector128 Multiply Sum 4-way \n Floating-Point\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|0 1 1 1|a|1|VDh|VBh|\n\n vmsub4fp128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vmulfp128 Vector128 Multiply\n Floating-Point\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|0 0 1 0|a|1|VDh|VBh|\n\n vmulfp128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vnmsubfp128 Vector128 Negative Multiply-Subtract \n Floating Point\n|0 0 0 1 0 1| VDS128 | VA128 | VB128 |A|0 1 0 1|a|1|VDh|VBh|\n\n vnmsubfp128 vr(VDS128), vr(VA128), vr(VB128), vr(VDS128)\n\n\n=================================================================\n vnor128 Vector128 Logical NOR\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 0 1 0|a|1|VDh|VBh|\n\n vnor128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vor128 Vector128 Logical OR\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 0 1 1|a|1|VDh|VBh|\n\n vor128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vperm128 Vector128 Permutation\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|0| VC |a|0|VDh|VBh|\n\n vperm128 vr(VD128), vr(VA128), vr(VB128), vr(VC)\n\n\n=================================================================\n vpermwi128 Vector128 Permutate Word Immediate\n|0 0 0 1 1 0| VD128 | PERMl | VB128 |0|1|PERMh|0|1|VDh|VBh|\n\n vpermwi128 vr(VD128), vr(VB128), (PERMh << 5 | PERMl)\n\n\n=================================================================\n vpkd3d128 Vector128 Pack D3Dtype, Rotate Left \n Immediate and Mask Insert\n|0 0 0 1 1 0| VD128 | x | y | VB128 |1 1 0| z |0 1|VDh|VBh|\n\n vpkd3d128 vr(VD128), vr(VB128), x, y, z\n\n\n=================================================================\n vpkshss128 Vector128 Pack Signed Half Word \n Signed Saturate\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 0 0 0|a|0|VDh|VBh|\n\n vpkshss128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vpkshus128 Vector128 Pack Signed Half Word \n Unsigned Saturate\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 0 0 1|a|0|VDh|VBh|\n\n vpkshus128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vpkswss128 Vector128 Pack Signed Word\n Signed Saturate\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 0 1 0|a|0|VDh|VBh|\n\n vpkswss128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vpkswus128 Vector128 Pack Signed Word \n Unsigned Saturate\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 0 1 1|a|0|VDh|VBh|\n\n vpkswus128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vpkuhum128 Vector128 Pack Unsigned Half Word \n Unsigned Modulo\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 1 0 0|a|0|VDh|VBh|\n\n vpkuhum128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vpkuhus128 Vector128 Pack Unsigned Half Word \n Unsigned Saturate\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 1 0 1|a|0|VDh|VBh|\n\n vpkuhus128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vpkuwum128 Vector128 Pack Unsigned Word\n Unsigned Modulo\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 1 1 0|a|0|VDh|VBh|\n\n vpkuwum128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vpkuwus128 Vector128 Pack Unsigned Word\n Unsigned Saturate\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 1 1 1|a|0|VDh|VBh|\n\n vpkuwus128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vrefp128 Vector128 Reciprocal Estimate \n Floating Point\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |1 1 0 0 0 1 1|VDh|VBh|\n\n vrefp128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vrfim128 Vector128 Round to Floating-Point \n Integer toward -oo\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |0 1 1 0 0 1 1|VDh|VBh|\n\n vrfim128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vrfin128 Vector128 Round to Floating-Point \n Integer toward Nearest\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |0 1 1 0 1 1 1|VDh|VBh|\n\n vrfin128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vrfip128 Vector128 Round to Floating-Point \n Integer toward +oo\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |0 1 1 1 0 1 1|VDh|VBh|\n\n vrfip128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vrfiz128 Vector128 Round to Floating-Point \n Integer toward Zero\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |0 1 1 1 1 1 1|VDh|VBh|\n\n vrfiz128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vrlimi128 Vector128 Rotate Left Immediate \n and Mask Insert\n|0 0 0 1 1 0| VD128 | UIMM | VB128 |1 1 1| z |0 1|VDh|VBh|\n\n vrlimi128 vr(VD128), vr(VB128), UIMM, z\n\n\n=================================================================\n vrlw128 Vector128 Rotate Left Word\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|0 0 0 1|a|1|VDh|VBh|\n\n vrlw128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vrsqrtefp128 Vector128 Reciprocal Square Root \n Estimate Floating Point\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |1 1 0 0 1 1 1|VDh|VBh|\n\n vrsqrtefp128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vsel128 Vector128 Select\n|0 0 0 1 0 1| VDS128 | VA128 | VB128 |A|1 1 0 1|a|1|VDh|VBh|\n\n vsel128 vr(VDS128), vr(VA128), vr(VB128), vr(VDS128)\n\n\n=================================================================\n vsldoi128 Vector128 Shift Left Double \n by Octet Immediate\n|0 0 0 1 0 0| VD128 | VA128 | VB128 |A| SHB |a|1|VDh|VBh|\n\n vsldoi128 vr(VD128), vr(VA128), vr(VB128), SHB\n\n\n=================================================================\n vslo128 Vector128 Shift Left Octet\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 1 1 0|a|1|VDh|VBh|\n\n vslo128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vslw128 Vector128 Shift Left Word\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|0 0 1 1|a|1|VDh|VBh|\n\n vslw128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vspltisw128 Vector128 Splat Immediate \n Signed Word\n|0 0 0 1 1 0| VD128 | SIMM | VB128 |1 1 1 0 1 1 1|VDh|VBh|\n\n vspltisw128 vr(VD128), vr(VB128), SIMM\n\n\n=================================================================\n vspltw128 Vector128 Splat Word\n|0 0 0 1 1 0| VD128 | UIMM | VB128 |1 1 1 0 0 1 1|VDh|VBh|\n\n vspltw128 vr(VD128), vr(VB128), UIMM\n\n\n=================================================================\n vsraw128 Vector128 Shift Right \n Arithmetic Word\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|0 1 0 1|a|1|VDh|VBh|\n\n vsraw128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vsro128 Vector128 Shift Right Octet\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|1 1 1 1|a|1|VDh|VBh|\n\n vsro128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vsrw128 Vector128 Shift Right Word\n|0 0 0 1 1 0| VD128 | VA128 | VB128 |A|0 1 1 1|a|1|VDh|VBh|\n\n vsrw128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vsubfp128 Vector128 Subtract Floating Point\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|0 0 0 1|a|1|VDh|VBh|\n\n vsubfp128 vr(VD128), vr(VA128), vr(VB128)\n\n\n=================================================================\n vupkd3d128 Vector128 Unpack D3Dtype\n|0 0 0 1 1 0| VD128 | UIMM | VB128 |1 1 1 1 1 1 1|VDh|VBh|\n\n vupkd3d128 vr(VD128), vr(VB128), UIMM\n\n\n=================================================================\n vupkhsb128 Vector128 Unpack \n High Signed Byte\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |0 1 1 1 0 0 0|VDh|VBh|\n\n vupkhsb128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vupklsb128 Vector128 Unpack \n Low Signed Byte\n|0 0 0 1 1 0| VD128 |0 0 0 0 0| VB128 |0 1 1 1 1 0 0|VDh|VBh|\n\n vupkhsb128 vr(VD128), vr(VB128)\n\n\n=================================================================\n vxor128 Vector128 Logical XOR\n|0 0 0 1 0 1| VD128 | VA128 | VB128 |A|1 1 0 0|a|1|VDh|VBh|\n\n vxor128 vr(VD128), vr(VA128), vr(VB128)\n" }, { "path": "docs/style_guide.md", "content": "# C++ Style Guide\n\nThe style guide can be summed up as 'clang-format with the Google style set'.\nIn addition, the [Google Style Guide](https://google.github.io/styleguide/cppguide.html)\nis followed and cpplint is the source of truth. When in doubt, defer to what\ncode in the project already does.\n\nBase rules:\n\n* 80 column line length max\n* LF (Unix-style) line endings\n* 2-space soft tabs, no TABs!\n* [Google Style Guide](https://google.github.io/styleguide/cppguide.html) for naming/casing/etc\n* Sort includes according to the [style guide rules](https://google.github.io/styleguide/cppguide.html#Names_and_Order_of_Includes)\n* Comments are properly punctuated (that means capitalization and periods, etc)\n* TODO's must be attributed like `// TODO(yourgithubname): foo.`\n\nCode that really breaks from the formatting rules will not be accepted, as then\nno one else can use clang-format on the code without also touching all your\nlines.\n\n### Why?\n\nTo quote the [Google Style Guide](https://google.github.io/styleguide/cppguide.html):\n\n```\nOne way in which we keep the code base manageable is by enforcing consistency.\nIt is very important that any programmer be able to look at another's code and\nquickly understand it. Maintaining a uniform style and following conventions\nmeans that we can more easily use \"pattern-matching\" to infer what various\nsymbols are and what invariants are true about them. Creating common, required\nidioms and patterns makes code much easier to understand. In some cases there\nmight be good arguments for changing certain style rules, but we nonetheless\nkeep things as they are in order to preserve consistency.\n```\n\n## Buildbot Verification\n\nThe buildbot runs `xb lint --all` on the master branch, and will run\n`xb lint --origin` on pull requests. Run `xb format` before you commit each\nlocal change so that you are consistently clean, otherwise you may have to\nrebase. If you forget, run `xb format --origin` and rebase your changes (so you\ndon't end up with 5 changes and then a 6th 'whoops' one — that's nasty).\n\nThe buildbot is running LLVM 3.8.0. If you are noticing style differences\nbetween your local lint/format and the buildbot, ensure you are running that\nversion.\n\n## Tools\n\n### clang-format\n\nclang-format with the Google style is used to format all files. I recommend\ninstalling/wiring it up to your editor of choice so that you don't even have to\nthink about tabs and wrapping and such.\n\n#### Command Line\n\nTo use the `xb format` auto-formatter, you need to have a `clang-format` on your\nPATH. If you're on Windows you can do this by installing an LLVM binary package\nfrom [the LLVM downloads page](https://llvm.org/releases/download.html). If you\ninstall it to the default location the `xb format` command will find it\nautomatically even if you don't choose to put all of LLVM onto your PATH.\n\n#### Visual Studio\n\nGrab the official [experimental Visual Studio plugin](https://llvm.org/builds/).\nTo switch to the Google style go Tools -> Options -> LLVM/Clang -> ClangFormat\nand set Style to Google. Then use ctrl-r/ctrl-f to trigger the formatting.\nUnfortunately it only does the cursor by default, so you'll have to select the\nwhole doc and invoke it to get it all done.\n\nIf you have a better option, let me know!\n\n#### Xcode\n\nInstall [Alcatraz](https://github.com/alcatraz/Alcatraz) to get the [ClangFormat](https://github.com/travisjeffery/ClangFormat-Xcode)\npackage. Set it to use the Google style and format on save. Never think about\ntabs or linefeeds or whatever again.\n\n### cpplint\n\nTODO(benvanik): write a cool script to do this/editor plugins.\nIn the future, the linter will run as a git commit hook and on travis.\n\n# Android Style Guide\n\nAndroid Java and Groovy code and XML files currently don't have automatic format\nverification during builds, however, stick to the [AOSP Java Code Style Rules](https://source.android.com/setup/contribute/code-style),\nwhich contain guidelines not only for code formatting, but for the usage of\nlanguage features as well.\n\nThe formatting rules used in Xenia match the default Android Studio settings.\nThey diverge from the C++ code style rules of Xenia in many areas, such as\nindentation width and the maximum line length, however, the goal for Android\nformatting in Xenia is to ensure quick development environment setup.\n\nIn Java code, limit the length of each line to 100 characters. If an assignment\ndoesn't fit in the limit, move the right-hand side of it to a separate line with\n8-space indentation. Similarly, if the argument list of a method declaration or\na call is too long, start the entire argument list on a new line, also indented\nwith 8 spaces — this is one of the differences from the C++ code style in Xenia,\nwhere arguments may be aligned with the opening bracket. In general, follow the\n[rectangle rule](https://github.com/google/google-java-format/wiki/The-Rectangle-Rule)\nso expressions in the code constitute a hierarchy of their bounding rectangles,\nensuring that with 4-space indentation for block contents and 8-space\nindentation for subexpressions.\n\nIn XML files, if the width of the line with an element exceeds 100 characters,\nor in most cases when there are multiple attributes, each attribute should be\nplaced on a separate line with 4-space indentation, with the exception of the\nfirst `xmlns`, which should stay on the same line as the element name.\n\nIn Groovy, use 4-space indentation for blocks and 8-space indentation for\nsplitting arguments into multiple lines. String literals should be written in\nsingle quotes unless string interpolation is used.\n\nYou can use the Code -> Reformat Code and Code -> Reformat File options in\nAndroid Studio to apply coarse formatting rules for different kinds of files\nsupported by Android Studio, such as Java, XML and Groovy. While clang-format is\nvery strict and generates code with the single allowed way of formatting,\nAndroid Studio, however, preserves many style choices in the original code, so\nit's recommended to approximate the final style manually instead of relying\nentirely on automatic formatting. Also use Code -> Rearrange Code to maintain a\nconsistent structure of Java class declarations.\n" }, { "path": "premake5.lua", "content": "include(\"tools/build\")\nrequire(\"third_party/premake-export-compile-commands/export-compile-commands\")\nrequire(\"third_party/premake-androidndk/androidndk\")\nrequire(\"third_party/premake-cmake/cmake\")\n\nlocation(build_root)\ntargetdir(build_bin)\nobjdir(build_obj)\n\n-- Define an ARCH variable\n-- Only use this to enable architecture-specific functionality.\nif os.istarget(\"linux\") then\n ARCH = os.outputof(\"uname -p\")\nelse\n ARCH = \"unknown\"\nend\n\nincludedirs({\n \".\",\n \"src\",\n \"third_party\",\n})\n\ndefines({\n \"_UNICODE\",\n \"UNICODE\",\n})\n\ncppdialect(\"C++17\")\nexceptionhandling(\"On\")\nrtti(\"On\")\nsymbols(\"On\")\n\n-- TODO(DrChat): Find a way to disable this on other architectures.\nif ARCH ~= \"ppc64\" then\n filter(\"architecture:x86_64\")\n vectorextensions(\"AVX\")\n filter({})\nend\n\ncharacterset(\"Unicode\")\nflags({\n \"FatalWarnings\", -- Treat warnings as errors.\n})\n\nfilter(\"kind:StaticLib\")\n defines({\n \"_LIB\",\n })\n\nfilter(\"configurations:Checked\")\n runtime(\"Debug\")\n optimize(\"Off\")\n defines({\n \"DEBUG\",\n })\nfilter({\"configurations:Checked\", \"platforms:Windows\"})\n buildoptions({\n \"/RTCsu\", -- Full Run-Time Checks.\n })\nfilter({\"configurations:Checked\", \"platforms:Linux\"})\n defines({\n \"_GLIBCXX_DEBUG\", -- libstdc++ debug mode\n })\n\nfilter(\"configurations:Debug\")\n runtime(\"Release\")\n optimize(\"Off\")\n defines({\n \"DEBUG\",\n \"_NO_DEBUG_HEAP=1\",\n })\nfilter({\"configurations:Debug\", \"platforms:Linux\"})\n defines({\n \"_GLIBCXX_DEBUG\", -- make dbg symbols work on some distros\n })\n\nfilter(\"configurations:Release\")\n runtime(\"Release\")\n defines({\n \"NDEBUG\",\n \"_NO_DEBUG_HEAP=1\",\n })\n optimize(\"Speed\")\n inlining(\"Auto\")\n flags({\n \"LinkTimeOptimization\",\n })\n -- Not using floatingpoint(\"Fast\") - NaN checks are used in some places\n -- (though rarely), overall preferable to avoid any functional differences\n -- between debug and release builds, and to have calculations involved in GPU\n -- (especially anything that may affect vertex position invariance) and CPU\n -- (such as constant propagation) emulation as predictable as possible,\n -- including handling of specials since games make assumptions about them.\nfilter(\"platforms:Linux\")\n system(\"linux\")\n toolset(\"clang\")\n buildoptions({\n -- \"-mlzcnt\", -- (don't) Assume lzcnt is supported.\n })\n pkg_config.all(\"gtk+-x11-3.0\")\n links({\n \"stdc++fs\",\n \"dl\",\n \"lz4\",\n \"pthread\",\n \"rt\",\n })\n\nfilter({\"platforms:Linux\", \"kind:*App\"})\n linkgroups(\"On\")\n\nfilter({\"platforms:Linux\", \"language:C++\", \"toolset:gcc\"})\n disablewarnings({\n \"unused-result\"\n })\n\nfilter({\"platforms:Linux\", \"toolset:gcc\"})\n if ARCH == \"ppc64\" then\n buildoptions({\n \"-m32\",\n \"-mpowerpc64\"\n })\n linkoptions({\n \"-m32\",\n \"-mpowerpc64\"\n })\n end\n\nfilter({\"platforms:Linux\", \"language:C++\", \"toolset:clang\"})\n disablewarnings({\n \"deprecated-register\"\n })\nfilter({\"platforms:Linux\", \"language:C++\", \"toolset:clang\", \"files:*.cc or *.cpp\"})\n buildoptions({\n \"-stdlib=libstdc++\",\n })\n\nfilter(\"platforms:Android-*\")\n system(\"android\")\n systemversion(\"24\")\n cppstl(\"c++\")\n staticruntime(\"On\")\n -- Hidden visibility is needed to prevent dynamic relocations in FFmpeg\n -- AArch64 Neon libavcodec assembly with PIC (accesses extern lookup tables\n -- using `adrp` and `add`, without the Global Object Table, expecting that all\n -- FFmpeg symbols that aren't a part of the FFmpeg API are hidden by FFmpeg's\n -- original build system) by resolving those relocations at link time instead.\n visibility(\"Hidden\")\n links({\n \"android\",\n \"dl\",\n \"log\",\n })\n\nfilter(\"platforms:Windows\")\n system(\"windows\")\n toolset(\"msc\")\n buildoptions({\n \"/utf-8\", -- 'build correctly on systems with non-Latin codepages'.\n -- Mark warnings as severe\n \"/w14839\", -- non-standard use of class 'type' as an argument to a variadic function\n \"/w14840\", -- non-portable use of class 'type' as an argument to a variadic function\n -- Disable warnings\n \"/wd4100\", -- Unreferenced parameters are ok.\n \"/wd4201\", -- Nameless struct/unions are ok.\n \"/wd4512\", -- 'assignment operator was implicitly defined as deleted'.\n \"/wd4127\", -- 'conditional expression is constant'.\n \"/wd4324\", -- 'structure was padded due to alignment specifier'.\n \"/wd4189\", -- 'local variable is initialized but not referenced'.\n })\n flags({\n \"MultiProcessorCompile\", -- Multiprocessor compilation.\n \"NoMinimalRebuild\", -- Required for /MP above.\n })\n\n defines({\n \"_CRT_NONSTDC_NO_DEPRECATE\",\n \"_CRT_SECURE_NO_WARNINGS\",\n \"WIN32\",\n \"_WIN64=1\",\n \"_AMD64=1\",\n })\n linkoptions({\n \"/ignore:4006\", -- Ignores complaints about empty obj files.\n \"/ignore:4221\",\n })\n links({\n \"ntdll\",\n \"wsock32\",\n \"ws2_32\",\n \"xinput\",\n \"comctl32\",\n \"shcore\",\n \"shlwapi\",\n \"dxguid\",\n \"bcrypt\",\n })\n\n-- Embed the manifest for things like dependencies and DPI awareness.\nfilter({\"platforms:Windows\", \"kind:ConsoleApp or WindowedApp\"})\n files({\n \"src/xenia/base/app_win32.manifest\"\n })\n\n-- Create scratch/ path\nif not os.isdir(\"scratch\") then\n os.mkdir(\"scratch\")\nend\n\nworkspace(\"xenia\")\n uuid(\"931ef4b0-6170-4f7a-aaf2-0fece7632747\")\n startproject(\"xenia-app\")\n if os.istarget(\"android\") then\n platforms({\"Android-ARM64\", \"Android-x86_64\"})\n filter(\"platforms:Android-ARM64\")\n architecture(\"ARM64\")\n filter(\"platforms:Android-x86_64\")\n architecture(\"x86_64\")\n filter({})\n else\n architecture(\"x86_64\")\n if os.istarget(\"linux\") then\n platforms({\"Linux\"})\n elseif os.istarget(\"macosx\") then\n platforms({\"Mac\"})\n xcodebuildsettings({ \n [\"ARCHS\"] = \"x86_64\"\n })\n elseif os.istarget(\"windows\") then\n platforms({\"Windows\"})\n -- 10.0.15063.0: ID3D12GraphicsCommandList1::SetSamplePositions.\n -- 10.0.19041.0: D3D12_HEAP_FLAG_CREATE_NOT_ZEROED.\n -- 10.0.22000.0: DWMWA_WINDOW_CORNER_PREFERENCE.\n filter(\"action:vs2017\")\n systemversion(\"10.0.22000.0\")\n filter(\"action:vs2019\")\n systemversion(\"10.0\")\n filter({})\n end\n end\n configurations({\"Checked\", \"Debug\", \"Release\"})\n\n include(\"third_party/aes_128.lua\")\n include(\"third_party/capstone.lua\")\n include(\"third_party/dxbc.lua\")\n include(\"third_party/discord-rpc.lua\")\n include(\"third_party/cxxopts.lua\")\n include(\"third_party/cpptoml.lua\")\n include(\"third_party/FFmpeg/premake5.lua\")\n include(\"third_party/fmt.lua\")\n include(\"third_party/glslang-spirv.lua\")\n include(\"third_party/imgui.lua\")\n include(\"third_party/mspack.lua\")\n include(\"third_party/snappy.lua\")\n include(\"third_party/xxhash.lua\")\n\n if not os.istarget(\"android\") then\n -- SDL2 requires sdl2-config, and as of November 2020 isn't high-quality on\n -- Android yet, most importantly in game controllers - the keycode and axis\n -- enums are being ruined during conversion to SDL2 enums resulting in only\n -- one controller (Nvidia Shield) being supported, digital triggers are also\n -- not supported; lifecycle management (especially surface loss) is also\n -- complicated.\n include(\"third_party/SDL2.lua\")\n end\n\n -- Disable treating warnings as fatal errors for all third party projects, as\n -- well as other things relevant only to Xenia itself.\n for _, prj in ipairs(premake.api.scope.current.solution.projects) do\n project(prj.name)\n removefiles({\n \"src/xenia/base/app_win32.manifest\"\n })\n removeflags({\n \"FatalWarnings\",\n })\n end\n\n include(\"src/xenia\")\n include(\"src/xenia/app\")\n include(\"src/xenia/app/discord\")\n include(\"src/xenia/apu\")\n include(\"src/xenia/apu/nop\")\n include(\"src/xenia/base\")\n include(\"src/xenia/cpu\")\n include(\"src/xenia/cpu/backend/x64\")\n include(\"src/xenia/debug/ui\")\n include(\"src/xenia/gpu\")\n include(\"src/xenia/gpu/null\")\n include(\"src/xenia/gpu/vulkan\")\n include(\"src/xenia/hid\")\n include(\"src/xenia/hid/nop\")\n include(\"src/xenia/kernel\")\n include(\"src/xenia/ui\")\n include(\"src/xenia/ui/vulkan\")\n include(\"src/xenia/vfs\")\n\n if not os.istarget(\"android\") then\n include(\"src/xenia/apu/sdl\")\n include(\"src/xenia/helper/sdl\")\n include(\"src/xenia/hid/sdl\")\n end\n\n if os.istarget(\"windows\") then\n include(\"src/xenia/apu/xaudio2\")\n include(\"src/xenia/gpu/d3d12\")\n include(\"src/xenia/hid/winkey\")\n include(\"src/xenia/hid/xinput\")\n include(\"src/xenia/ui/d3d12\")\n end\n" }, { "path": "src/xenia/app/discord/discord_presence.cc", "content": "/**\n******************************************************************************\n* Xenia : Xbox 360 Emulator Research Project *\n******************************************************************************\n* Copyright 2020 Ben Vanik. All rights reserved. *\n* Released under the BSD license - see LICENSE in the root for more details. *\n******************************************************************************\n*/\n\n#include \"discord_presence.h\"\n#include \"third_party/discord-rpc/include/discord_rpc.h\"\n#include \"xenia/base/string.h\"\n\nnamespace xe {\nnamespace discord {\n\nvoid HandleDiscordReady(const DiscordUser* request) {}\nvoid HandleDiscordError(int errorCode, const char* message) {}\nvoid HandleDiscordJoinGame(const char* joinSecret) {}\nvoid HandleDiscordJoinRequest(const DiscordUser* request) {}\nvoid HandleDiscordSpectateGame(const char* spectateSecret) {}\n\nvoid DiscordPresence::Initialize() {\n DiscordEventHandlers handlers = {};\n handlers.ready = &HandleDiscordReady;\n handlers.errored = &HandleDiscordError;\n handlers.joinGame = &HandleDiscordJoinGame;\n handlers.joinRequest = &HandleDiscordJoinRequest;\n handlers.spectateGame = &HandleDiscordSpectateGame;\n Discord_Initialize(\"606840046649081857\", &handlers, 0, \"\");\n}\n\nvoid DiscordPresence::NotPlaying() {\n DiscordRichPresence discordPresence = {};\n discordPresence.state = \"Idle\";\n discordPresence.details = \"Standby\";\n discordPresence.largeImageKey = \"app\";\n discordPresence.instance = 1;\n Discord_UpdatePresence(&discordPresence);\n}\n\nvoid DiscordPresence::PlayingTitle(const std::string_view game_title) {\n auto details = std::string(game_title);\n DiscordRichPresence discordPresence = {};\n discordPresence.state = \"In Game\";\n discordPresence.details = details.c_str();\n // TODO(gibbed): we don't have state icons yet.\n // discordPresence.smallImageKey = \"app\";\n // discordPresence.largeImageKey = \"state_ingame\";\n discordPresence.largeImageKey = \"app\";\n discordPresence.instance = 1;\n Discord_UpdatePresence(&discordPresence);\n}\n\nvoid DiscordPresence::Shutdown() { Discord_Shutdown(); }\n\n} // namespace discord\n} // namespace xe\n" }, { "path": "src/xenia/app/discord/discord_presence.h", "content": "/**\n******************************************************************************\n* Xenia : Xbox 360 Emulator Research Project *\n******************************************************************************\n* Copyright 2020 Ben Vanik. All rights reserved. *\n* Released under the BSD license - see LICENSE in the root for more details. *\n******************************************************************************\n*/\n\n#ifndef XENIA_DISCORD_DISCORD_PRESENCE_H_\n#define XENIA_DISCORD_DISCORD_PRESENCE_H_\n\n#include \n\nnamespace xe {\nnamespace discord {\n\nclass DiscordPresence {\n public:\n static void Initialize();\n static void NotPlaying();\n static void PlayingTitle(const std::string_view game_title);\n static void Shutdown();\n};\n\n} // namespace discord\n} // namespace xe\n\n#endif // XENIA_DISCORD_DISCORD_PRESENCE_H_\n" }, { "path": "src/xenia/app/discord/premake5.lua", "content": "project_root = \"../../../..\"\ninclude(project_root..\"/tools/build\")\n\ngroup(\"src\")\nproject(\"xenia-app-discord\")\n uuid(\"d14c0885-22d2-40de-ab28-7b234ef2b949\")\n kind(\"StaticLib\")\n language(\"C++\")\n links({\n \"discord-rpc\"\n })\n defines({\n })\n includedirs({\n project_root..\"/third_party/discord-rpc/src\"\n })\n files({\n \"discord_presence.cc\",\n \"discord_presence.h\"\n })\n" }, { "path": "src/xenia/app/emulator_window.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/app/emulator_window.h\"\n\n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"third_party/imgui/imgui.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/clock.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/debugging.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/base/system.h\"\n#include \"xenia/base/threading.h\"\n#include \"xenia/cpu/processor.h\"\n#include \"xenia/emulator.h\"\n#include \"xenia/gpu/command_processor.h\"\n#include \"xenia/gpu/graphics_system.h\"\n#include \"xenia/ui/file_picker.h\"\n#include \"xenia/ui/graphics_provider.h\"\n#include \"xenia/ui/imgui_dialog.h\"\n#include \"xenia/ui/imgui_drawer.h\"\n#include \"xenia/ui/immediate_drawer.h\"\n#include \"xenia/ui/presenter.h\"\n#include \"xenia/ui/ui_event.h\"\n#include \"xenia/ui/virtual_key.h\"\n\n// Autogenerated by `xb premake`.\n#include \"build/version.h\"\n\nDECLARE_bool(debug);\n\nDEFINE_bool(fullscreen, false, \"Whether to launch the emulator in fullscreen.\",\n \"Display\");\n\nDEFINE_string(\n postprocess_antialiasing, \"\",\n \"Post-processing anti-aliasing effect to apply to the image output of the \"\n \"game.\\n\"\n \"Using post-process anti-aliasing is heavily recommended when AMD \"\n \"FidelityFX Contrast Adaptive Sharpening or Super Resolution 1.0 is \"\n \"active.\\n\"\n \"Use: [none, fxaa, fxaa_extreme]\\n\"\n \" none (or any value not listed here):\\n\"\n \" Don't alter the original image.\\n\"\n \" fxaa:\\n\"\n \" NVIDIA Fast Approximate Anti-Aliasing 3.11, normal quality preset (12).\"\n \"\\n\"\n \" fxaa_extreme:\\n\"\n \" NVIDIA Fast Approximate Anti-Aliasing 3.11, extreme quality preset \"\n \"(39).\",\n \"Display\");\nDEFINE_string(\n postprocess_scaling_and_sharpening, \"\",\n \"Post-processing effect to use for resampling and/or sharpening of the \"\n \"final display output.\\n\"\n \"Use: [bilinear, cas, fsr]\\n\"\n \" bilinear (or any value not listed here):\\n\"\n \" Original image at 1:1, simple bilinear stretching for resampling.\\n\"\n \" cas:\\n\"\n \" Use AMD FidelityFX Contrast Adaptive Sharpening (CAS) for sharpening \"\n \"at scaling factors of up to 2x2, with additional bilinear stretching for \"\n \"larger factors.\\n\"\n \" fsr:\\n\"\n \" Use AMD FidelityFX Super Resolution 1.0 (FSR) for highest-quality \"\n \"upscaling, or AMD FidelityFX Contrast Adaptive Sharpening for sharpening \"\n \"while not scaling or downsampling.\\n\"\n \" For scaling by factors of more than 2x2, multiple FSR passes are done.\",\n \"Display\");\nDEFINE_double(\n postprocess_ffx_cas_additional_sharpness,\n xe::ui::Presenter::GuestOutputPaintConfig::kCasAdditionalSharpnessDefault,\n \"Additional sharpness for AMD FidelityFX Contrast Adaptive Sharpening \"\n \"(CAS), from 0 to 1.\\n\"\n \"Higher is sharper.\",\n \"Display\");\nDEFINE_uint32(\n postprocess_ffx_fsr_max_upsampling_passes,\n xe::ui::Presenter::GuestOutputPaintConfig::kFsrMaxUpscalingPassesMax,\n \"Maximum number of upsampling passes performed in AMD FidelityFX Super \"\n \"Resolution 1.0 (FSR) before falling back to bilinear stretching after the \"\n \"final pass.\\n\"\n \"Each pass upscales only to up to 2x2 the previous size. If the game \"\n \"outputs a 1280x720 image, 1 pass will upscale it to up to 2560x1440 \"\n \"(below 4K), after 2 passes it will be upscaled to a maximum of 5120x2880 \"\n \"(including 3840x2160 for 4K), and so on.\\n\"\n \"This variable has no effect if the display resolution isn't very high, \"\n \"but may be reduced on resolutions like 4K or 8K in case the performance \"\n \"impact of multiple FSR upsampling passes is too high, or if softer edges \"\n \"are desired.\\n\"\n \"The default value is the maximum internally supported by Xenia.\",\n \"Display\");\nDEFINE_double(\n postprocess_ffx_fsr_sharpness_reduction,\n xe::ui::Presenter::GuestOutputPaintConfig::kFsrSharpnessReductionDefault,\n \"Sharpness reduction for AMD FidelityFX Super Resolution 1.0 (FSR), in \"\n \"stops.\\n\"\n \"Lower is sharper.\",\n \"Display\");\n// Dithering to 8bpc is enabled by default since the effect is minor, only\n// effects what can't be shown normally by host displays, and nothing is changed\n// by it for 8bpc source without resampling.\nDEFINE_bool(\n postprocess_dither, true,\n \"Dither the final image output from the internal precision to 8 bits per \"\n \"channel so gradients are smoother.\\n\"\n \"On a 10bpc display, the lower 2 bits will still be kept, but noise will \"\n \"be added to them - disabling may be recommended for 10bpc, but it \"\n \"depends on the 10bpc displaying capabilities of the actual display used.\",\n \"Display\");\n\nnamespace xe {\nnamespace app {\n\nusing xe::ui::FileDropEvent;\nusing xe::ui::KeyEvent;\nusing xe::ui::MenuItem;\nusing xe::ui::UIEvent;\n\nconst std::string kBaseTitle = \"Xenia\";\n\nEmulatorWindow::EmulatorWindow(Emulator* emulator,\n ui::WindowedAppContext& app_context)\n : emulator_(emulator),\n app_context_(app_context),\n window_listener_(*this),\n window_(ui::Window::Create(app_context, kBaseTitle, 1280, 720)),\n imgui_drawer_(\n std::make_unique(window_.get(), kZOrderImGui)),\n display_config_game_config_load_callback_(\n new DisplayConfigGameConfigLoadCallback(*emulator, *this)) {\n base_title_ = kBaseTitle +\n#ifdef DEBUG\n#if _NO_DEBUG_HEAP == 1\n \" DEBUG\"\n#else\n \" CHECKED\"\n#endif\n#endif\n \" (\"\n#ifdef XE_BUILD_IS_PR\n \"PR#\" XE_BUILD_PR_NUMBER \" \" XE_BUILD_PR_REPO\n \" \" XE_BUILD_PR_BRANCH \"@\" XE_BUILD_PR_COMMIT_SHORT \" against \"\n#endif\n XE_BUILD_BRANCH \"@\" XE_BUILD_COMMIT_SHORT \" on \" XE_BUILD_DATE\n \")\";\n}\n\nstd::unique_ptr EmulatorWindow::Create(\n Emulator* emulator, ui::WindowedAppContext& app_context) {\n assert_true(app_context.IsInUIThread());\n std::unique_ptr emulator_window(\n new EmulatorWindow(emulator, app_context));\n if (!emulator_window->Initialize()) {\n return nullptr;\n }\n return emulator_window;\n}\n\nEmulatorWindow::~EmulatorWindow() {\n // Notify the ImGui drawer that the immediate drawer is being destroyed.\n ShutdownGraphicsSystemPresenterPainting();\n}\n\nui::Presenter* EmulatorWindow::GetGraphicsSystemPresenter() const {\n gpu::GraphicsSystem* graphics_system = emulator_->graphics_system();\n return graphics_system ? graphics_system->presenter() : nullptr;\n}\n\nvoid EmulatorWindow::SetupGraphicsSystemPresenterPainting() {\n ShutdownGraphicsSystemPresenterPainting();\n\n if (!window_) {\n return;\n }\n\n ui::Presenter* presenter = GetGraphicsSystemPresenter();\n if (!presenter) {\n return;\n }\n\n ApplyDisplayConfigForCvars();\n\n window_->SetPresenter(presenter);\n\n immediate_drawer_ =\n emulator_->graphics_system()->provider()->CreateImmediateDrawer();\n if (immediate_drawer_) {\n immediate_drawer_->SetPresenter(presenter);\n imgui_drawer_->SetPresenterAndImmediateDrawer(presenter,\n immediate_drawer_.get());\n Profiler::SetUserIO(kZOrderProfiler, window_.get(), presenter,\n immediate_drawer_.get());\n }\n}\n\nvoid EmulatorWindow::ShutdownGraphicsSystemPresenterPainting() {\n Profiler::SetUserIO(kZOrderProfiler, window_.get(), nullptr, nullptr);\n imgui_drawer_->SetPresenterAndImmediateDrawer(nullptr, nullptr);\n immediate_drawer_.reset();\n if (window_) {\n window_->SetPresenter(nullptr);\n }\n}\n\nvoid EmulatorWindow::OnEmulatorInitialized() {\n emulator_initialized_ = true;\n window_->SetMainMenuEnabled(true);\n // When the user can see that the emulator isn't initializing anymore (the\n // menu isn't disabled), enter fullscreen if requested.\n if (cvars::fullscreen) {\n SetFullscreen(true);\n }\n}\n\nvoid EmulatorWindow::EmulatorWindowListener::OnClosing(ui::UIEvent& e) {\n emulator_window_.app_context_.QuitFromUIThread();\n}\n\nvoid EmulatorWindow::EmulatorWindowListener::OnFileDrop(ui::FileDropEvent& e) {\n emulator_window_.FileDrop(e.filename());\n}\n\nvoid EmulatorWindow::EmulatorWindowListener::OnKeyDown(ui::KeyEvent& e) {\n emulator_window_.OnKeyDown(e);\n}\n\nvoid EmulatorWindow::DisplayConfigGameConfigLoadCallback::PostGameConfigLoad() {\n emulator_window_.ApplyDisplayConfigForCvars();\n}\n\nvoid EmulatorWindow::DisplayConfigDialog::OnDraw(ImGuiIO& io) {\n gpu::GraphicsSystem* graphics_system =\n emulator_window_.emulator_->graphics_system();\n if (!graphics_system) {\n return;\n }\n\n // In the top-left corner so it's close to the menu bar from where it was\n // opened.\n // Origin Y coordinate 20 was taken from the Dear ImGui demo.\n ImGui::SetNextWindowPos(ImVec2(20, 20), ImGuiCond_FirstUseEver);\n ImGui::SetNextWindowSize(ImVec2(20, 20), ImGuiCond_FirstUseEver);\n // Alpha from Dear ImGui tooltips (0.35 from the overlay provides too low\n // visibility). Translucent so some effect of the changes can still be seen\n // through it.\n ImGui::SetNextWindowBgAlpha(0.6f);\n bool dialog_open = true;\n if (!ImGui::Begin(\"Post-processing\", &dialog_open,\n ImGuiWindowFlags_NoCollapse |\n ImGuiWindowFlags_AlwaysAutoResize |\n ImGuiWindowFlags_HorizontalScrollbar)) {\n ImGui::End();\n return;\n }\n // Even if the close button has been pressed, still paint everything not to\n // have one frame with an empty window.\n\n // Prevent user confusion which has been reported multiple times.\n ImGui::TextUnformatted(\"All effects can be used on GPUs of any brand.\");\n ImGui::Spacing();\n\n gpu::CommandProcessor* command_processor =\n graphics_system->command_processor();\n if (command_processor) {\n if (ImGui::TreeNodeEx(\n \"Anti-aliasing\",\n ImGuiTreeNodeFlags_Framed | ImGuiTreeNodeFlags_DefaultOpen)) {\n gpu::CommandProcessor::SwapPostEffect current_swap_post_effect =\n command_processor->GetDesiredSwapPostEffect();\n int new_swap_post_effect_index = int(current_swap_post_effect);\n ImGui::RadioButton(\"None\", &new_swap_post_effect_index,\n int(gpu::CommandProcessor::SwapPostEffect::kNone));\n ImGui::RadioButton(\n \"NVIDIA Fast Approximate Anti-Aliasing 3.11 (FXAA), normal quality\",\n &new_swap_post_effect_index,\n int(gpu::CommandProcessor::SwapPostEffect::kFxaa));\n ImGui::RadioButton(\n \"NVIDIA Fast Approximate Anti-Aliasing 3.11 (FXAA), extreme quality\",\n &new_swap_post_effect_index,\n int(gpu::CommandProcessor::SwapPostEffect::kFxaaExtreme));\n gpu::CommandProcessor::SwapPostEffect new_swap_post_effect =\n gpu::CommandProcessor::SwapPostEffect(new_swap_post_effect_index);\n if (current_swap_post_effect != new_swap_post_effect) {\n command_processor->SetDesiredSwapPostEffect(new_swap_post_effect);\n }\n\n // Override the values in the cvars to save them to the config at exit if\n // the user has set them to anything new.\n if (GetSwapPostEffectForCvarValue(cvars::postprocess_antialiasing) !=\n new_swap_post_effect) {\n OVERRIDE_string(postprocess_antialiasing,\n GetCvarValueForSwapPostEffect(new_swap_post_effect));\n }\n\n ImGui::TreePop();\n }\n }\n\n ui::Presenter* presenter = graphics_system->presenter();\n if (presenter) {\n const ui::Presenter::GuestOutputPaintConfig& current_presenter_config =\n presenter->GetGuestOutputPaintConfigFromUIThread();\n ui::Presenter::GuestOutputPaintConfig new_presenter_config =\n current_presenter_config;\n\n if (ImGui::TreeNodeEx(\n \"Resampling and sharpening\",\n ImGuiTreeNodeFlags_Framed | ImGuiTreeNodeFlags_DefaultOpen)) {\n // Filtering effect.\n int new_effect_index = int(new_presenter_config.GetEffect());\n ImGui::RadioButton(\n \"None / bilinear\", &new_effect_index,\n int(ui::Presenter::GuestOutputPaintConfig::Effect::kBilinear));\n ImGui::RadioButton(\n \"AMD FidelityFX Contrast Adaptive Sharpening (CAS)\",\n &new_effect_index,\n int(ui::Presenter::GuestOutputPaintConfig::Effect::kCas));\n ImGui::RadioButton(\n \"AMD FidelityFX Super Resolution 1.0 (FSR)\", &new_effect_index,\n int(ui::Presenter::GuestOutputPaintConfig::Effect::kFsr));\n new_presenter_config.SetEffect(\n ui::Presenter::GuestOutputPaintConfig::Effect(new_effect_index));\n\n // effect_description must be one complete, but short enough, sentence per\n // line, as TextWrapped doesn't work correctly in auto-resizing windows\n // (in the initial frames, the window becomes extremely tall, and widgets\n // added after the wrapped text have no effect on the width of the text).\n const char* effect_description = nullptr;\n switch (new_presenter_config.GetEffect()) {\n case ui::Presenter::GuestOutputPaintConfig::Effect::kBilinear:\n effect_description =\n \"Simple bilinear filtering is done if resampling is needed.\\n\"\n \"Otherwise, only anti-aliasing is done if enabled, or displaying \"\n \"as is.\";\n break;\n case ui::Presenter::GuestOutputPaintConfig::Effect::kCas:\n effect_description =\n \"Sharpening and resampling to up to 2x2 to improve the fidelity \"\n \"of details.\\n\"\n \"For scaling by more than 2x2, bilinear stretching is done \"\n \"afterwards.\";\n break;\n case ui::Presenter::GuestOutputPaintConfig::Effect::kFsr:\n effect_description =\n \"High-quality edge-preserving upscaling to arbitrary target \"\n \"resolutions.\\n\"\n \"For scaling by more than 2x2, multiple upsampling passes are \"\n \"done.\\n\"\n \"If not upscaling, Contrast Adaptive Sharpening (CAS) is used \"\n \"instead.\";\n break;\n }\n if (effect_description) {\n ImGui::TextUnformatted(effect_description);\n }\n\n if (new_presenter_config.GetEffect() ==\n ui::Presenter::GuestOutputPaintConfig::Effect::kCas ||\n new_presenter_config.GetEffect() ==\n ui::Presenter::GuestOutputPaintConfig::Effect::kFsr) {\n if (effect_description) {\n ImGui::Spacing();\n }\n\n ImGui::TextUnformatted(\n \"FXAA is highly recommended when using CAS or FSR.\");\n\n ImGui::Spacing();\n\n // 2 decimal places is more or less enough precision for the sharpness\n // given the minor visual effect of small changes, the width of the\n // slider, and readability convenience (2 decimal places is like an\n // integer percentage). However, because Dear ImGui parses the string\n // representation of the number and snaps the value to it internally,\n // 2 decimal places actually offer less precision than the slider itself\n // does. This is especially prominent in the low range of the non-linear\n // FSR sharpness reduction slider. 3 decimal places are optimal in this\n // case.\n\n if (new_presenter_config.GetEffect() ==\n ui::Presenter::GuestOutputPaintConfig::Effect::kFsr) {\n float fsr_sharpness_reduction =\n new_presenter_config.GetFsrSharpnessReduction();\n ImGui::TextUnformatted(\n \"FSR sharpness reduction when upscaling (lower is sharper):\");\n const auto label =\n fmt::format(\"{:.3f} stops\", fsr_sharpness_reduction);\n // Power 2.0 scaling as the reduction is in stops, used in exp2.\n fsr_sharpness_reduction = sqrt(2.f * fsr_sharpness_reduction);\n ImGui::SliderFloat(\n \"##FSRSharpnessReduction\", &fsr_sharpness_reduction,\n ui::Presenter::GuestOutputPaintConfig::kFsrSharpnessReductionMin,\n ui::Presenter::GuestOutputPaintConfig::kFsrSharpnessReductionMax,\n label.c_str(), ImGuiSliderFlags_NoInput);\n fsr_sharpness_reduction =\n .5f * fsr_sharpness_reduction * fsr_sharpness_reduction;\n ImGui::SameLine();\n if (ImGui::Button(\"Reset##ResetFSRSharpnessReduction\")) {\n fsr_sharpness_reduction = ui::Presenter::GuestOutputPaintConfig ::\n kFsrSharpnessReductionDefault;\n }\n new_presenter_config.SetFsrSharpnessReduction(\n fsr_sharpness_reduction);\n }\n\n float cas_additional_sharpness =\n new_presenter_config.GetCasAdditionalSharpness();\n ImGui::TextUnformatted(\n new_presenter_config.GetEffect() ==\n ui::Presenter::GuestOutputPaintConfig::Effect::kFsr\n ? \"CAS additional sharpness when not upscaling (higher is \"\n \"sharper):\"\n : \"CAS additional sharpness (higher is sharper):\");\n ImGui::SliderFloat(\n \"##CASAdditionalSharpness\", &cas_additional_sharpness,\n ui::Presenter::GuestOutputPaintConfig::kCasAdditionalSharpnessMin,\n ui::Presenter::GuestOutputPaintConfig::kCasAdditionalSharpnessMax,\n \"%.3f\");\n ImGui::SameLine();\n if (ImGui::Button(\"Reset##ResetCASAdditionalSharpness\")) {\n cas_additional_sharpness = ui::Presenter::GuestOutputPaintConfig ::\n kCasAdditionalSharpnessDefault;\n }\n new_presenter_config.SetCasAdditionalSharpness(\n cas_additional_sharpness);\n\n // There's no need to expose the setting for the maximum number of FSR\n // EASU passes as it's largely meaningless if the user doesn't have a\n // very high-resolution monitor compared to the original image size as\n // most of the values of the slider will have no effect, and that's just\n // very fine-grained performance control for a fixed-overhead pass only\n // for huge screen resolutions.\n }\n\n ImGui::TreePop();\n }\n\n if (ImGui::TreeNodeEx(\"Dithering\", ImGuiTreeNodeFlags_Framed |\n ImGuiTreeNodeFlags_DefaultOpen)) {\n bool dither = current_presenter_config.GetDither();\n ImGui::Checkbox(\n \"Dither the final output to 8bpc to make gradients smoother\",\n &dither);\n new_presenter_config.SetDither(dither);\n\n ImGui::TreePop();\n }\n\n presenter->SetGuestOutputPaintConfigFromUIThread(new_presenter_config);\n\n // Override the values in the cvars to save them to the config at exit if\n // the user has set them to anything new.\n ui::Presenter::GuestOutputPaintConfig cvars_presenter_config =\n GetGuestOutputPaintConfigForCvars();\n if (cvars_presenter_config.GetEffect() !=\n new_presenter_config.GetEffect()) {\n OVERRIDE_string(postprocess_scaling_and_sharpening,\n GetCvarValueForGuestOutputPaintEffect(\n new_presenter_config.GetEffect()));\n }\n if (cvars_presenter_config.GetCasAdditionalSharpness() !=\n new_presenter_config.GetCasAdditionalSharpness()) {\n OVERRIDE_double(postprocess_ffx_cas_additional_sharpness,\n new_presenter_config.GetCasAdditionalSharpness());\n }\n if (cvars_presenter_config.GetFsrSharpnessReduction() !=\n new_presenter_config.GetFsrSharpnessReduction()) {\n OVERRIDE_double(postprocess_ffx_fsr_sharpness_reduction,\n new_presenter_config.GetFsrSharpnessReduction());\n }\n if (cvars_presenter_config.GetDither() !=\n new_presenter_config.GetDither()) {\n OVERRIDE_bool(postprocess_dither, new_presenter_config.GetDither());\n }\n }\n\n ImGui::End();\n\n if (!dialog_open) {\n emulator_window_.ToggleDisplayConfigDialog();\n // `this` might have been destroyed by ToggleDisplayConfigDialog.\n return;\n }\n}\n\nbool EmulatorWindow::Initialize() {\n window_->AddListener(&window_listener_);\n window_->AddInputListener(&window_listener_, kZOrderEmulatorWindowInput);\n\n // Main menu.\n // FIXME: This code is really messy.\n auto main_menu = MenuItem::Create(MenuItem::Type::kNormal);\n auto file_menu = MenuItem::Create(MenuItem::Type::kPopup, \"&File\");\n {\n file_menu->AddChild(\n MenuItem::Create(MenuItem::Type::kString, \"&Open...\", \"Ctrl+O\",\n std::bind(&EmulatorWindow::FileOpen, this)));\n#ifdef DEBUG\n file_menu->AddChild(\n MenuItem::Create(MenuItem::Type::kString, \"Close\",\n std::bind(&EmulatorWindow::FileClose, this)));\n#endif // #ifdef DEBUG\n file_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n file_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"Show content directory...\",\n std::bind(&EmulatorWindow::ShowContentDirectory, this)));\n file_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n file_menu->AddChild(\n MenuItem::Create(MenuItem::Type::kString, \"E&xit\", \"Alt+F4\",\n [this]() { window_->RequestClose(); }));\n }\n main_menu->AddChild(std::move(file_menu));\n\n // CPU menu.\n auto cpu_menu = MenuItem::Create(MenuItem::Type::kPopup, \"&CPU\");\n {\n cpu_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"&Reset Time Scalar\", \"Numpad *\",\n std::bind(&EmulatorWindow::CpuTimeScalarReset, this)));\n cpu_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"Time Scalar /= 2\", \"Numpad -\",\n std::bind(&EmulatorWindow::CpuTimeScalarSetHalf, this)));\n cpu_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"Time Scalar *= 2\", \"Numpad +\",\n std::bind(&EmulatorWindow::CpuTimeScalarSetDouble, this)));\n }\n cpu_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n {\n cpu_menu->AddChild(MenuItem::Create(MenuItem::Type::kString,\n \"Toggle Profiler &Display\", \"F3\",\n []() { Profiler::ToggleDisplay(); }));\n cpu_menu->AddChild(MenuItem::Create(MenuItem::Type::kString,\n \"&Pause/Resume Profiler\", \"`\",\n []() { Profiler::TogglePause(); }));\n }\n cpu_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n {\n cpu_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"&Break and Show Guest Debugger\",\n \"Pause/Break\", std::bind(&EmulatorWindow::CpuBreakIntoDebugger, this)));\n cpu_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"&Break into Host Debugger\",\n \"Ctrl+Pause/Break\",\n std::bind(&EmulatorWindow::CpuBreakIntoHostDebugger, this)));\n }\n main_menu->AddChild(std::move(cpu_menu));\n\n // GPU menu.\n auto gpu_menu = MenuItem::Create(MenuItem::Type::kPopup, \"&GPU\");\n {\n gpu_menu->AddChild(\n MenuItem::Create(MenuItem::Type::kString, \"&Trace Frame\", \"F4\",\n std::bind(&EmulatorWindow::GpuTraceFrame, this)));\n }\n gpu_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n {\n gpu_menu->AddChild(\n MenuItem::Create(MenuItem::Type::kString, \"&Clear Runtime Caches\", \"F5\",\n std::bind(&EmulatorWindow::GpuClearCaches, this)));\n }\n main_menu->AddChild(std::move(gpu_menu));\n\n // Display menu.\n auto display_menu = MenuItem::Create(MenuItem::Type::kPopup, \"&Display\");\n {\n display_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"&Post-processing settings\", \"F6\",\n std::bind(&EmulatorWindow::ToggleDisplayConfigDialog, this)));\n }\n display_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n {\n display_menu->AddChild(\n MenuItem::Create(MenuItem::Type::kString, \"&Fullscreen\", \"F11\",\n std::bind(&EmulatorWindow::ToggleFullscreen, this)));\n }\n main_menu->AddChild(std::move(display_menu));\n\n // Help menu.\n auto help_menu = MenuItem::Create(MenuItem::Type::kPopup, \"&Help\");\n {\n help_menu->AddChild(\n MenuItem::Create(MenuItem::Type::kString, \"FA&Q...\", \"F1\",\n std::bind(&EmulatorWindow::ShowFAQ, this)));\n help_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n help_menu->AddChild(\n MenuItem::Create(MenuItem::Type::kString, \"Game &compatibility...\",\n std::bind(&EmulatorWindow::ShowCompatibility, this)));\n help_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n help_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"Build commit on GitHub...\", \"F2\",\n std::bind(&EmulatorWindow::ShowBuildCommit, this)));\n help_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"Recent changes on GitHub...\", [this]() {\n LaunchWebBrowser(\n \"https://github.com/xenia-project/xenia/compare/\" XE_BUILD_COMMIT\n \"...\" XE_BUILD_BRANCH);\n }));\n help_menu->AddChild(MenuItem::Create(MenuItem::Type::kSeparator));\n help_menu->AddChild(MenuItem::Create(\n MenuItem::Type::kString, \"&About...\",\n [this]() { LaunchWebBrowser(\"https://xenia.jp/about/\"); }));\n }\n main_menu->AddChild(std::move(help_menu));\n\n window_->SetMainMenu(std::move(main_menu));\n\n window_->SetMainMenuEnabled(false);\n\n UpdateTitle();\n\n if (!window_->Open()) {\n XELOGE(\"Failed to open the platform window\");\n return false;\n }\n\n Profiler::SetUserIO(kZOrderProfiler, window_.get(), nullptr, nullptr);\n\n return true;\n}\n\nconst char* EmulatorWindow::GetCvarValueForSwapPostEffect(\n gpu::CommandProcessor::SwapPostEffect effect) {\n switch (effect) {\n case gpu::CommandProcessor::SwapPostEffect::kFxaa:\n return \"fxaa\";\n case gpu::CommandProcessor::SwapPostEffect::kFxaaExtreme:\n return \"fxaa_extreme\";\n default:\n return \"\";\n }\n}\n\ngpu::CommandProcessor::SwapPostEffect\nEmulatorWindow::GetSwapPostEffectForCvarValue(const std::string& cvar_value) {\n if (cvar_value == GetCvarValueForSwapPostEffect(\n gpu::CommandProcessor::SwapPostEffect::kFxaa)) {\n return gpu::CommandProcessor::SwapPostEffect::kFxaa;\n }\n if (cvar_value == GetCvarValueForSwapPostEffect(\n gpu::CommandProcessor::SwapPostEffect::kFxaaExtreme)) {\n return gpu::CommandProcessor::SwapPostEffect::kFxaaExtreme;\n }\n return gpu::CommandProcessor::SwapPostEffect::kNone;\n}\n\nconst char* EmulatorWindow::GetCvarValueForGuestOutputPaintEffect(\n ui::Presenter::GuestOutputPaintConfig::Effect effect) {\n switch (effect) {\n case ui::Presenter::GuestOutputPaintConfig::Effect::kCas:\n return \"cas\";\n case ui::Presenter::GuestOutputPaintConfig::Effect::kFsr:\n return \"fsr\";\n default:\n return \"\";\n }\n}\n\nui::Presenter::GuestOutputPaintConfig::Effect\nEmulatorWindow::GetGuestOutputPaintEffectForCvarValue(\n const std::string& cvar_value) {\n if (cvar_value == GetCvarValueForGuestOutputPaintEffect(\n ui::Presenter::GuestOutputPaintConfig::Effect::kCas)) {\n return ui::Presenter::GuestOutputPaintConfig::Effect::kCas;\n }\n if (cvar_value == GetCvarValueForGuestOutputPaintEffect(\n ui::Presenter::GuestOutputPaintConfig::Effect::kFsr)) {\n return ui::Presenter::GuestOutputPaintConfig::Effect::kFsr;\n }\n return ui::Presenter::GuestOutputPaintConfig::Effect::kBilinear;\n}\n\nui::Presenter::GuestOutputPaintConfig\nEmulatorWindow::GetGuestOutputPaintConfigForCvars() {\n ui::Presenter::GuestOutputPaintConfig paint_config;\n paint_config.SetAllowOverscanCutoff(true);\n paint_config.SetEffect(GetGuestOutputPaintEffectForCvarValue(\n cvars::postprocess_scaling_and_sharpening));\n paint_config.SetCasAdditionalSharpness(\n float(cvars::postprocess_ffx_cas_additional_sharpness));\n paint_config.SetFsrMaxUpsamplingPasses(\n cvars::postprocess_ffx_fsr_max_upsampling_passes);\n paint_config.SetFsrSharpnessReduction(\n float(cvars::postprocess_ffx_fsr_sharpness_reduction));\n paint_config.SetDither(cvars::postprocess_dither);\n return paint_config;\n}\n\nvoid EmulatorWindow::ApplyDisplayConfigForCvars() {\n gpu::GraphicsSystem* graphics_system = emulator_->graphics_system();\n if (!graphics_system) {\n return;\n }\n\n gpu::CommandProcessor* command_processor =\n graphics_system->command_processor();\n if (command_processor) {\n command_processor->SetDesiredSwapPostEffect(\n GetSwapPostEffectForCvarValue(cvars::postprocess_antialiasing));\n }\n\n ui::Presenter* presenter = graphics_system->presenter();\n if (presenter) {\n presenter->SetGuestOutputPaintConfigFromUIThread(\n GetGuestOutputPaintConfigForCvars());\n }\n}\n\nvoid EmulatorWindow::OnKeyDown(ui::KeyEvent& e) {\n if (!emulator_initialized_) {\n return;\n }\n\n switch (e.virtual_key()) {\n case ui::VirtualKey::kO: {\n if (!e.is_ctrl_pressed()) {\n return;\n }\n FileOpen();\n } break;\n case ui::VirtualKey::kMultiply: {\n CpuTimeScalarReset();\n } break;\n case ui::VirtualKey::kSubtract: {\n CpuTimeScalarSetHalf();\n } break;\n case ui::VirtualKey::kAdd: {\n CpuTimeScalarSetDouble();\n } break;\n\n case ui::VirtualKey::kF3: {\n Profiler::ToggleDisplay();\n } break;\n\n case ui::VirtualKey::kF4: {\n GpuTraceFrame();\n } break;\n case ui::VirtualKey::kF5: {\n GpuClearCaches();\n } break;\n\n case ui::VirtualKey::kF6: {\n ToggleDisplayConfigDialog();\n } break;\n case ui::VirtualKey::kF11: {\n ToggleFullscreen();\n } break;\n case ui::VirtualKey::kEscape: {\n // Allow users to escape fullscreen (but not enter it).\n if (!window_->IsFullscreen()) {\n return;\n }\n SetFullscreen(false);\n } break;\n\n#ifdef DEBUG\n case ui::VirtualKey::kF7: {\n // Save to file\n // TODO: Choose path based on user input, or from options\n // TODO: Spawn a new thread to do this.\n emulator()->SaveToFile(\"test.sav\");\n } break;\n case ui::VirtualKey::kF8: {\n // Restore from file\n // TODO: Choose path from user\n // TODO: Spawn a new thread to do this.\n emulator()->RestoreFromFile(\"test.sav\");\n } break;\n#endif // #ifdef DEBUG\n\n case ui::VirtualKey::kPause: {\n CpuBreakIntoDebugger();\n } break;\n case ui::VirtualKey::kCancel: {\n CpuBreakIntoHostDebugger();\n } break;\n\n case ui::VirtualKey::kF1: {\n ShowFAQ();\n } break;\n\n case ui::VirtualKey::kF2: {\n ShowBuildCommit();\n } break;\n\n default:\n return;\n }\n\n e.set_handled(true);\n}\n\nvoid EmulatorWindow::FileDrop(const std::filesystem::path& filename) {\n if (!emulator_initialized_) {\n return;\n }\n auto result = emulator_->LaunchPath(filename);\n if (XFAILED(result)) {\n // TODO: Display a message box.\n XELOGE(\"Failed to launch target: {:08X}\", result);\n }\n}\n\nvoid EmulatorWindow::FileOpen() {\n std::filesystem::path path;\n\n auto file_picker = xe::ui::FilePicker::Create();\n file_picker->set_mode(ui::FilePicker::Mode::kOpen);\n file_picker->set_type(ui::FilePicker::Type::kFile);\n file_picker->set_multi_selection(false);\n file_picker->set_title(\"Select Content Package\");\n file_picker->set_extensions({\n {\"Supported Files\", \"*.iso;*.xex;*.*\"},\n {\"Disc Image (*.iso)\", \"*.iso\"},\n {\"Xbox Executable (*.xex)\", \"*.xex\"},\n //{\"Content Package (*.xcp)\", \"*.xcp\" },\n {\"All Files (*.*)\", \"*.*\"},\n });\n if (file_picker->Show(window_.get())) {\n auto selected_files = file_picker->selected_files();\n if (!selected_files.empty()) {\n path = selected_files[0];\n }\n }\n\n if (!path.empty()) {\n // Normalize the path and make absolute.\n auto abs_path = std::filesystem::absolute(path);\n auto result = emulator_->LaunchPath(abs_path);\n if (XFAILED(result)) {\n // TODO: Display a message box.\n XELOGE(\"Failed to launch target: {:08X}\", result);\n }\n }\n}\n\nvoid EmulatorWindow::FileClose() {\n if (emulator_->is_title_open()) {\n emulator_->TerminateTitle();\n }\n}\n\nvoid EmulatorWindow::ShowContentDirectory() {\n std::filesystem::path target_path;\n\n auto content_root = emulator_->content_root();\n if (!emulator_->is_title_open() || !emulator_->kernel_state()) {\n target_path = content_root;\n } else {\n // TODO(gibbed): expose this via ContentManager?\n auto title_id =\n fmt::format(\"{:08X}\", emulator_->kernel_state()->title_id());\n auto package_root = content_root / title_id;\n target_path = package_root;\n }\n\n if (!std::filesystem::exists(target_path)) {\n std::filesystem::create_directories(target_path);\n }\n\n LaunchFileExplorer(target_path);\n}\n\nvoid EmulatorWindow::CpuTimeScalarReset() {\n Clock::set_guest_time_scalar(1.0);\n UpdateTitle();\n}\n\nvoid EmulatorWindow::CpuTimeScalarSetHalf() {\n Clock::set_guest_time_scalar(Clock::guest_time_scalar() / 2.0);\n UpdateTitle();\n}\n\nvoid EmulatorWindow::CpuTimeScalarSetDouble() {\n Clock::set_guest_time_scalar(Clock::guest_time_scalar() * 2.0);\n UpdateTitle();\n}\n\nvoid EmulatorWindow::CpuBreakIntoDebugger() {\n if (!cvars::debug) {\n xe::ui::ImGuiDialog::ShowMessageBox(imgui_drawer_.get(), \"Xenia Debugger\",\n \"Xenia must be launched with the \"\n \"--debug flag in order to enable \"\n \"debugging.\");\n return;\n }\n auto processor = emulator()->processor();\n if (processor->execution_state() == cpu::ExecutionState::kRunning) {\n // Currently running, so interrupt (and show the debugger).\n processor->Pause();\n } else {\n // Not running, so just bring the debugger into focus.\n processor->ShowDebugger();\n }\n}\n\nvoid EmulatorWindow::CpuBreakIntoHostDebugger() { xe::debugging::Break(); }\n\nvoid EmulatorWindow::GpuTraceFrame() {\n emulator()->graphics_system()->RequestFrameTrace();\n}\n\nvoid EmulatorWindow::GpuClearCaches() {\n emulator()->graphics_system()->ClearCaches();\n}\n\nvoid EmulatorWindow::SetFullscreen(bool fullscreen) {\n if (window_->IsFullscreen() == fullscreen) {\n return;\n }\n window_->SetFullscreen(fullscreen);\n window_->SetCursorVisibility(fullscreen\n ? ui::Window::CursorVisibility::kAutoHidden\n : ui::Window::CursorVisibility::kVisible);\n}\n\nvoid EmulatorWindow::ToggleFullscreen() {\n SetFullscreen(!window_->IsFullscreen());\n}\n\nvoid EmulatorWindow::ToggleDisplayConfigDialog() {\n if (!display_config_dialog_) {\n display_config_dialog_ = std::unique_ptr(\n new DisplayConfigDialog(imgui_drawer_.get(), *this));\n } else {\n display_config_dialog_.reset();\n }\n}\n\nvoid EmulatorWindow::ShowCompatibility() {\n const std::string_view base_url =\n \"https://github.com/xenia-project/game-compatibility/issues\";\n std::string url;\n // Avoid searching for a title ID of \"00000000\".\n uint32_t title_id = emulator_->title_id();\n if (!title_id) {\n url = base_url;\n } else {\n url = fmt::format(\"{}?q=is%3Aissue+is%3Aopen+{:08X}\", base_url, title_id);\n }\n LaunchWebBrowser(url);\n}\n\nvoid EmulatorWindow::ShowFAQ() {\n LaunchWebBrowser(\"https://github.com/xenia-project/xenia/wiki/FAQ\");\n}\n\nvoid EmulatorWindow::ShowBuildCommit() {\n#ifdef XE_BUILD_IS_PR\n LaunchWebBrowser(\n \"https://github.com/xenia-project/xenia/pull/\" XE_BUILD_PR_NUMBER);\n#else\n LaunchWebBrowser(\n \"https://github.com/xenia-project/xenia/commit/\" XE_BUILD_COMMIT);\n#endif\n}\n\nvoid EmulatorWindow::UpdateTitle() {\n xe::StringBuffer sb;\n sb.Append(base_title_);\n\n // Title information, if available\n if (emulator()->is_title_open()) {\n sb.AppendFormat(u8\" | [{:08X}\", emulator()->title_id());\n auto title_version = emulator()->title_version();\n if (!title_version.empty()) {\n sb.Append(u8\" v\");\n sb.Append(title_version);\n }\n sb.Append(u8\"]\");\n\n auto title_name = emulator()->title_name();\n if (!title_name.empty()) {\n sb.Append(u8\" \");\n sb.Append(title_name);\n }\n }\n\n // Graphics system name, if available\n auto graphics_system = emulator()->graphics_system();\n if (graphics_system) {\n auto graphics_name = graphics_system->name();\n if (!graphics_name.empty()) {\n sb.Append(u8\" <\");\n sb.Append(graphics_name);\n sb.Append(u8\">\");\n }\n }\n\n if (Clock::guest_time_scalar() != 1.0) {\n sb.AppendFormat(u8\" (@{:.2f}x)\", Clock::guest_time_scalar());\n }\n\n if (initializing_shader_storage_) {\n sb.Append(u8\" (Preloading shaders\\u2026)\");\n }\n\n window_->SetTitle(sb.to_string_view());\n}\n\nvoid EmulatorWindow::SetInitializingShaderStorage(bool initializing) {\n if (initializing_shader_storage_ == initializing) {\n return;\n }\n initializing_shader_storage_ = initializing;\n UpdateTitle();\n}\n\n} // namespace app\n} // namespace xe\n" }, { "path": "src/xenia/app/emulator_window.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APP_EMULATOR_WINDOW_H_\n#define XENIA_APP_EMULATOR_WINDOW_H_\n\n#include \n#include \n\n#include \"xenia/emulator.h\"\n#include \"xenia/gpu/command_processor.h\"\n#include \"xenia/ui/imgui_dialog.h\"\n#include \"xenia/ui/imgui_drawer.h\"\n#include \"xenia/ui/immediate_drawer.h\"\n#include \"xenia/ui/menu_item.h\"\n#include \"xenia/ui/presenter.h\"\n#include \"xenia/ui/window.h\"\n#include \"xenia/ui/window_listener.h\"\n#include \"xenia/ui/windowed_app_context.h\"\n#include \"xenia/xbox.h\"\n\nnamespace xe {\nnamespace app {\n\nclass EmulatorWindow {\n public:\n enum : size_t {\n // The UI is on top of the game and is open in special cases, so\n // lowest-priority.\n kZOrderHidInput,\n kZOrderImGui,\n kZOrderProfiler,\n // Emulator window controls are expected to be always accessible by the\n // user, so highest-priority.\n kZOrderEmulatorWindowInput,\n };\n\n virtual ~EmulatorWindow();\n\n static std::unique_ptr Create(\n Emulator* emulator, ui::WindowedAppContext& app_context);\n\n Emulator* emulator() const { return emulator_; }\n ui::WindowedAppContext& app_context() const { return app_context_; }\n ui::Window* window() const { return window_.get(); }\n ui::ImGuiDrawer* imgui_drawer() const { return imgui_drawer_.get(); }\n\n ui::Presenter* GetGraphicsSystemPresenter() const;\n void SetupGraphicsSystemPresenterPainting();\n void ShutdownGraphicsSystemPresenterPainting();\n\n void OnEmulatorInitialized();\n\n void UpdateTitle();\n void SetFullscreen(bool fullscreen);\n void ToggleFullscreen();\n void SetInitializingShaderStorage(bool initializing);\n\n private:\n class EmulatorWindowListener final : public ui::WindowListener,\n public ui::WindowInputListener {\n public:\n explicit EmulatorWindowListener(EmulatorWindow& emulator_window)\n : emulator_window_(emulator_window) {}\n\n void OnClosing(ui::UIEvent& e) override;\n void OnFileDrop(ui::FileDropEvent& e) override;\n\n void OnKeyDown(ui::KeyEvent& e) override;\n\n private:\n EmulatorWindow& emulator_window_;\n };\n\n class DisplayConfigGameConfigLoadCallback\n : public Emulator::GameConfigLoadCallback {\n public:\n DisplayConfigGameConfigLoadCallback(Emulator& emulator,\n EmulatorWindow& emulator_window)\n : Emulator::GameConfigLoadCallback(emulator),\n emulator_window_(emulator_window) {}\n\n void PostGameConfigLoad() override;\n\n private:\n EmulatorWindow& emulator_window_;\n };\n\n class DisplayConfigDialog final : public ui::ImGuiDialog {\n public:\n DisplayConfigDialog(ui::ImGuiDrawer* imgui_drawer,\n EmulatorWindow& emulator_window)\n : ui::ImGuiDialog(imgui_drawer), emulator_window_(emulator_window) {}\n\n protected:\n void OnDraw(ImGuiIO& io) override;\n\n private:\n EmulatorWindow& emulator_window_;\n };\n\n explicit EmulatorWindow(Emulator* emulator,\n ui::WindowedAppContext& app_context);\n\n bool Initialize();\n\n // For comparisons, use GetSwapPostEffectForCvarValue instead as the default\n // fallback may be used for multiple values.\n static const char* GetCvarValueForSwapPostEffect(\n gpu::CommandProcessor::SwapPostEffect effect);\n static gpu::CommandProcessor::SwapPostEffect GetSwapPostEffectForCvarValue(\n const std::string& cvar_value);\n // For comparisons, use GetGuestOutputPaintEffectForCvarValue instead as the\n // default fallback may be used for multiple values.\n static const char* GetCvarValueForGuestOutputPaintEffect(\n ui::Presenter::GuestOutputPaintConfig::Effect effect);\n static ui::Presenter::GuestOutputPaintConfig::Effect\n GetGuestOutputPaintEffectForCvarValue(const std::string& cvar_value);\n static ui::Presenter::GuestOutputPaintConfig\n GetGuestOutputPaintConfigForCvars();\n void ApplyDisplayConfigForCvars();\n\n void OnKeyDown(ui::KeyEvent& e);\n void FileDrop(const std::filesystem::path& filename);\n void FileOpen();\n void FileClose();\n void ShowContentDirectory();\n void CpuTimeScalarReset();\n void CpuTimeScalarSetHalf();\n void CpuTimeScalarSetDouble();\n void CpuBreakIntoDebugger();\n void CpuBreakIntoHostDebugger();\n void GpuTraceFrame();\n void GpuClearCaches();\n void ToggleDisplayConfigDialog();\n void ShowCompatibility();\n void ShowFAQ();\n void ShowBuildCommit();\n\n Emulator* emulator_;\n ui::WindowedAppContext& app_context_;\n EmulatorWindowListener window_listener_;\n std::unique_ptr window_;\n std::unique_ptr imgui_drawer_;\n std::unique_ptr\n display_config_game_config_load_callback_;\n // Creation may fail, in this case immediate drawer UI must not be drawn.\n std::unique_ptr immediate_drawer_;\n\n bool emulator_initialized_ = false;\n\n std::string base_title_;\n bool initializing_shader_storage_ = false;\n\n std::unique_ptr display_config_dialog_;\n};\n\n} // namespace app\n} // namespace xe\n\n#endif // XENIA_APP_EMULATOR_WINDOW_H_\n" }, { "path": "src/xenia/app/main_resources.rc", "content": "//{{NO_DEPENDENCIES}}\n\nMAINICON ICON \"..\\\\..\\\\..\\\\assets\\\\icon\\\\icon.ico\"\n" }, { "path": "src/xenia/app/premake5.lua", "content": "project_root = \"../../..\"\ninclude(project_root..\"/tools/build\")\n\ngroup(\"src\")\nproject(\"xenia-app\")\n uuid(\"d7e98620-d007-4ad8-9dbd-b47c8853a17f\")\n language(\"C++\")\n links({\n \"xenia-apu\",\n \"xenia-apu-nop\",\n \"xenia-base\",\n \"xenia-core\",\n \"xenia-cpu\",\n \"xenia-gpu\",\n \"xenia-gpu-null\",\n \"xenia-gpu-vulkan\",\n \"xenia-hid\",\n \"xenia-hid-nop\",\n \"xenia-kernel\",\n \"xenia-ui\",\n \"xenia-ui-vulkan\",\n \"xenia-vfs\",\n })\n links({\n \"aes_128\",\n \"capstone\",\n \"fmt\",\n \"dxbc\",\n \"discord-rpc\",\n \"glslang-spirv\",\n \"imgui\",\n \"libavcodec\",\n \"libavutil\",\n \"mspack\",\n \"snappy\",\n \"xxhash\",\n })\n defines({\n \"XBYAK_NO_OP_NAMES\",\n \"XBYAK_ENABLE_OMITTED_OPERAND\",\n })\n local_platform_files()\n files({\n \"../base/main_init_\"..platform_suffix..\".cc\",\n \"../ui/windowed_app_main_\"..platform_suffix..\".cc\",\n })\n\n resincludedirs({\n project_root,\n })\n\n filter(SINGLE_LIBRARY_FILTER)\n -- Unified library containing all apps as StaticLibs, not just the main\n -- emulator windowed app.\n kind(\"SharedLib\")\n links({\n \"xenia-gpu-vulkan-trace-viewer\",\n \"xenia-hid-demo\",\n \"xenia-ui-window-vulkan-demo\",\n })\n filter(NOT_SINGLE_LIBRARY_FILTER)\n kind(\"WindowedApp\")\n\n -- `targetname` is broken if building from Gradle, works only for toggling the\n -- `lib` prefix, as Gradle uses LOCAL_MODULE_FILENAME, not a derivative of\n -- LOCAL_MODULE, to specify the targets to build when executing ndk-build.\n filter(\"platforms:not Android-*\")\n targetname(\"xenia\")\n\n filter(\"architecture:x86_64\")\n links({\n \"xenia-cpu-backend-x64\",\n })\n\n -- TODO(Triang3l): The emulator itself on Android.\n filter(\"platforms:not Android-*\")\n files({\n \"xenia_main.cc\",\n })\n\n filter(\"platforms:Windows\")\n files({\n \"main_resources.rc\",\n })\n\n filter({\"architecture:x86_64\", \"files:../base/main_init_\"..platform_suffix..\".cc\"})\n vectorextensions(\"IA32\") -- Disable AVX for main_init_win.cc so our AVX check doesn't use AVX instructions.\n\n filter(\"platforms:not Android-*\")\n links({\n \"xenia-app-discord\",\n \"xenia-apu-sdl\",\n -- TODO(Triang3l): CPU debugger on Android.\n \"xenia-debug-ui\",\n \"xenia-helper-sdl\",\n \"xenia-hid-sdl\",\n })\n\n filter(\"platforms:Linux\")\n links({\n \"X11\",\n \"xcb\",\n \"X11-xcb\",\n \"SDL2\",\n })\n\n filter(\"platforms:Windows\")\n links({\n \"xenia-apu-xaudio2\",\n \"xenia-gpu-d3d12\",\n \"xenia-hid-winkey\",\n \"xenia-hid-xinput\",\n \"xenia-ui-d3d12\",\n })\n\n filter({\"platforms:Windows\", SINGLE_LIBRARY_FILTER})\n links({\n \"xenia-gpu-d3d12-trace-viewer\",\n \"xenia-ui-window-d3d12-demo\",\n })\n\n filter(\"platforms:Windows\")\n -- Only create the .user file if it doesn't already exist.\n local user_file = project_root..\"/build/xenia-app.vcxproj.user\"\n if not os.isfile(user_file) then\n debugdir(project_root)\n debugargs({\n })\n end\n" }, { "path": "src/xenia/app/xenia_main.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/app/discord/discord_presence.h\"\n#include \"xenia/app/emulator_window.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/debugging.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/base/threading.h\"\n#include \"xenia/config.h\"\n#include \"xenia/debug/ui/debug_window.h\"\n#include \"xenia/emulator.h\"\n#include \"xenia/ui/file_picker.h\"\n#include \"xenia/ui/window.h\"\n#include \"xenia/ui/window_listener.h\"\n#include \"xenia/ui/windowed_app.h\"\n#include \"xenia/ui/windowed_app_context.h\"\n#include \"xenia/vfs/devices/host_path_device.h\"\n\n// Available audio systems:\n#include \"xenia/apu/nop/nop_audio_system.h\"\n#if !XE_PLATFORM_ANDROID\n#include \"xenia/apu/sdl/sdl_audio_system.h\"\n#endif // !XE_PLATFORM_ANDROID\n#if XE_PLATFORM_WIN32\n#include \"xenia/apu/xaudio2/xaudio2_audio_system.h\"\n#endif // XE_PLATFORM_WIN32\n\n// Available graphics systems:\n#include \"xenia/gpu/null/null_graphics_system.h\"\n#include \"xenia/gpu/vulkan/vulkan_graphics_system.h\"\n#if XE_PLATFORM_WIN32\n#include \"xenia/gpu/d3d12/d3d12_graphics_system.h\"\n#endif // XE_PLATFORM_WIN32\n\n// Available input drivers:\n#include \"xenia/hid/nop/nop_hid.h\"\n#if !XE_PLATFORM_ANDROID\n#include \"xenia/hid/sdl/sdl_hid.h\"\n#endif // !XE_PLATFORM_ANDROID\n#if XE_PLATFORM_WIN32\n#include \"xenia/hid/winkey/winkey_hid.h\"\n#include \"xenia/hid/xinput/xinput_hid.h\"\n#endif // XE_PLATFORM_WIN32\n\n#include \"third_party/fmt/include/fmt/format.h\"\n\nDEFINE_string(apu, \"any\", \"Audio system. Use: [any, nop, sdl, xaudio2]\", \"APU\");\nDEFINE_string(gpu, \"any\", \"Graphics system. Use: [any, d3d12, vulkan, null]\",\n \"GPU\");\nDEFINE_string(hid, \"any\", \"Input system. Use: [any, nop, sdl, winkey, xinput]\",\n \"HID\");\n\nDEFINE_path(\n storage_root, \"\",\n \"Root path for persistent internal data storage (config, etc.), or empty \"\n \"to use the path preferred for the OS, such as the documents folder, or \"\n \"the emulator executable directory if portable.txt is present in it.\",\n \"Storage\");\nDEFINE_path(\n content_root, \"\",\n \"Root path for guest content storage (saves, etc.), or empty to use the \"\n \"content folder under the storage root.\",\n \"Storage\");\nDEFINE_path(\n cache_root, \"\",\n \"Root path for files used to speed up certain parts of the emulator or the \"\n \"game. These files may be persistent, but they can be deleted without \"\n \"major side effects such as progress loss. If empty, the cache folder \"\n \"under the storage root, or, if available, the cache directory preferred \"\n \"for the OS, will be used.\",\n \"Storage\");\n\nDEFINE_bool(mount_scratch, false, \"Enable scratch mount\", \"Storage\");\nDEFINE_bool(mount_cache, false, \"Enable cache mount\", \"Storage\");\n\nDEFINE_transient_path(target, \"\",\n \"Specifies the target .xex or .iso to execute.\",\n \"General\");\nDEFINE_transient_bool(portable, false,\n \"Specifies if Xenia should run in portable mode.\",\n \"General\");\n\nDECLARE_bool(debug);\n\nDEFINE_bool(discord, true, \"Enable Discord rich presence\", \"General\");\n\nnamespace xe {\nnamespace app {\n\nclass EmulatorApp final : public xe::ui::WindowedApp {\n public:\n static std::unique_ptr Create(\n xe::ui::WindowedAppContext& app_context) {\n return std::unique_ptr(new EmulatorApp(app_context));\n }\n\n ~EmulatorApp();\n\n bool OnInitialize() override;\n\n protected:\n void OnDestroy() override;\n\n private:\n template \n class Factory {\n private:\n struct Creator {\n std::string name;\n std::function is_available;\n std::function(Args...)> instantiate;\n };\n\n std::vector creators_;\n\n public:\n void Add(const std::string_view name, std::function is_available,\n std::function(Args...)> instantiate) {\n creators_.push_back({std::string(name), is_available, instantiate});\n }\n\n void Add(const std::string_view name,\n std::function(Args...)> instantiate) {\n auto always_available = []() { return true; };\n Add(name, always_available, instantiate);\n }\n\n template \n void Add(const std::string_view name) {\n Add(name, DT::IsAvailable, [](Args... args) {\n return std::make_unique
(std::forward(args)...);\n });\n }\n\n std::unique_ptr Create(const std::string_view name, Args... args) {\n if (!name.empty() && name != \"any\") {\n auto it = std::find_if(\n creators_.cbegin(), creators_.cend(),\n [&name](const auto& f) { return name.compare(f.name) == 0; });\n if (it != creators_.cend() && (*it).is_available()) {\n return (*it).instantiate(std::forward(args)...);\n }\n return nullptr;\n } else {\n for (const auto& creator : creators_) {\n if (!creator.is_available()) continue;\n auto instance = creator.instantiate(std::forward(args)...);\n if (!instance) continue;\n return instance;\n }\n return nullptr;\n }\n }\n\n std::vector> CreateAll(const std::string_view name,\n Args... args) {\n std::vector> instances;\n if (!name.empty() && name != \"any\") {\n auto it = std::find_if(\n creators_.cbegin(), creators_.cend(),\n [&name](const auto& f) { return name.compare(f.name) == 0; });\n if (it != creators_.cend() && (*it).is_available()) {\n auto instance = (*it).instantiate(std::forward(args)...);\n if (instance) {\n instances.emplace_back(std::move(instance));\n }\n }\n } else {\n for (const auto& creator : creators_) {\n if (!creator.is_available()) continue;\n auto instance = creator.instantiate(std::forward(args)...);\n if (instance) {\n instances.emplace_back(std::move(instance));\n }\n }\n }\n return instances;\n }\n };\n\n class DebugWindowClosedListener final : public xe::ui::WindowListener {\n public:\n explicit DebugWindowClosedListener(EmulatorApp& emulator_app)\n : emulator_app_(emulator_app) {}\n\n void OnClosing(xe::ui::UIEvent& e) override;\n\n private:\n EmulatorApp& emulator_app_;\n };\n\n explicit EmulatorApp(xe::ui::WindowedAppContext& app_context);\n\n static std::unique_ptr CreateAudioSystem(\n cpu::Processor* processor);\n static std::unique_ptr CreateGraphicsSystem();\n static std::vector> CreateInputDrivers(\n ui::Window* window);\n\n void EmulatorThread();\n void ShutdownEmulatorThreadFromUIThread();\n\n DebugWindowClosedListener debug_window_closed_listener_;\n\n std::unique_ptr emulator_;\n std::unique_ptr emulator_window_;\n\n // Created on demand, used by the emulator.\n std::unique_ptr debug_window_;\n\n // Refreshing the emulator - placed after its dependencies.\n std::atomic emulator_thread_quit_requested_;\n std::unique_ptr emulator_thread_event_;\n std::thread emulator_thread_;\n};\n\nvoid EmulatorApp::DebugWindowClosedListener::OnClosing(xe::ui::UIEvent& e) {\n EmulatorApp* emulator_app = &emulator_app_;\n emulator_app->emulator_->processor()->set_debug_listener(nullptr);\n emulator_app->debug_window_.reset();\n}\n\nEmulatorApp::EmulatorApp(xe::ui::WindowedAppContext& app_context)\n : xe::ui::WindowedApp(app_context, \"xenia\", \"[Path to .iso/.xex]\"),\n debug_window_closed_listener_(*this) {\n AddPositionalOption(\"target\");\n}\n\nEmulatorApp::~EmulatorApp() {\n // Should be shut down from OnDestroy if OnInitialize has ever been done, but\n // for the most safety as a running thread may be destroyed only after\n // joining.\n ShutdownEmulatorThreadFromUIThread();\n}\n\nstd::unique_ptr EmulatorApp::CreateAudioSystem(\n cpu::Processor* processor) {\n Factory factory;\n#if XE_PLATFORM_WIN32\n factory.Add(\"xaudio2\");\n#endif // XE_PLATFORM_WIN32\n#if !XE_PLATFORM_ANDROID\n factory.Add(\"sdl\");\n#endif // !XE_PLATFORM_ANDROID\n factory.Add(\"nop\");\n return factory.Create(cvars::apu, processor);\n}\n\nstd::unique_ptr EmulatorApp::CreateGraphicsSystem() {\n // While Vulkan is supported by a large variety of operating systems (Windows,\n // GNU/Linux, Android, also via the MoltenVK translation layer on top of Metal\n // on macOS and iOS), please don't remove platform-specific GPU backends from\n // Xenia.\n //\n // Regardless of the operating system, having multiple options provides more\n // stability to users. In case of driver issues, users may try switching\n // between the available backends. For example, in June 2022, on Nvidia Ampere\n // (RTX 30xx), Xenia had synchronization issues that resulted in flickering,\n // most prominently in 4D5307E6, on Direct3D 12 - but the same issue was not\n // reproducible in the Vulkan backend, however, it used ImageSampleExplicitLod\n // with explicit gradients for cubemaps, which triggered a different driver\n // bug on Nvidia (every 1 out of 2x2 pixels receiving junk).\n //\n // Specifically on Microsoft platforms, there are a few reasons why supporting\n // Direct3D 12 is desirable rather than limiting Xenia to Vulkan only:\n // - Wider hardware support for Direct3D 12 on x86 Windows desktops.\n // Direct3D 12 requires the minimum of Nvidia Fermi, or, with a pre-2021\n // driver version, Intel HD Graphics 4200. Vulkan, however, is supported\n // only starting with Nvidia Kepler and a much more recent Intel UHD\n // Graphics generation.\n // - Wider hardware support on other kinds of Microsoft devices. The Xbox One\n // and the Xbox Series X|S only support Direct3D as the GPU API in their UWP\n // runtime, and only version 12 can be granted expanded resource access.\n // Qualcomm, as of June 2022, also doesn't provide a Vulkan implementation\n // for their Arm-based Windows devices, while Direct3D 12 is available.\n // - Both older Intel GPUs and the Xbox One apparently, as well as earlier\n // Windows 10 versions, also require Shader Model 5.1 DXBC shaders rather\n // than Shader Model 6 DXIL ones, so a DXBC shader translator should be\n // available in Xenia too, a DXIL one doesn't fully replace it.\n // - As of June 2022, AMD also refuses to implement the\n // VK_EXT_fragment_shader_interlock Vulkan extension in their drivers, as\n // well as its OpenGL counterpart, which is heavily utilized for accurate\n // support of Xenos render target formats that don't have PC equivalents\n // (8_8_8_8_GAMMA, 2_10_10_10_FLOAT, 16_16 and 16_16_16_16 with -32 to 32\n // range, D24FS8) with correct blending. Direct3D 12, however, requires\n // support for similar functionality (rasterizer-ordered views) on the\n // feature level 12_1, and the AMD driver implements it on Direct3D, as well\n // as raster order groups in their Metal driver.\n //\n // Additionally, different host GPU APIs receive feature support at different\n // paces. VK_EXT_fragment_shader_interlock first appeared in 2019, for\n // instance, while Xenia had been taking advantage of rasterizer-ordered views\n // on Direct3D 12 for over half a year at that point (they have existed in\n // Direct3D 12 since the first version).\n //\n // MoltenVK on top Metal also has its flaws and limitations. Metal, for\n // instance, as of June 2022, doesn't provide a switch for primitive restart,\n // while Vulkan does - so MoltenVK is not completely transparent to Xenia,\n // many of its issues that may be not very obvious (unlike when the Metal API\n // is used directly) should be taken into account in Xenia. Also, as of June\n // 2022, MoltenVK translates SPIR-V shaders into the C++-based Metal Shading\n // Language rather than AIR directly, which likely massively increases\n // pipeline object creation time - and Xenia translates shaders and creates\n // pipelines when they're first actually used for a draw command by the game,\n // thus it can't precompile anything that hasn't ever been encountered before\n // there's already no time to waste.\n //\n // Very old hardware (Direct3D 10 level) is also not supported by most Vulkan\n // drivers. However, in the future, Xenia may be ported to it using the\n // Direct3D 11 API with the feature level 10_1 or 10_0. OpenGL, however, had\n // been lagging behind Direct3D prior to versions 4.x, and didn't receive\n // compute shaders until a 4.2 extension (while 4.2 already corresponds\n // roughly to Direct3D 11 features) - and replacing Xenia compute shaders with\n // transform feedback / stream output is not always trivial (in particular,\n // will need to rely on GL_ARB_transform_feedback3 for skipping over memory\n // locations that shouldn't be overwritten).\n //\n // For maintainability, as much implementation code as possible should be\n // placed in `xe::gpu` and shared between the backends rather than duplicated\n // between them.\n const std::string gpu_implementation_name = cvars::gpu;\n if (gpu_implementation_name == \"null\") {\n return std::make_unique();\n }\n Factory factory;\n#if XE_PLATFORM_WIN32\n factory.Add(\"d3d12\");\n#endif // XE_PLATFORM_WIN32\n factory.Add(\"vulkan\");\n std::unique_ptr gpu_implementation =\n factory.Create(gpu_implementation_name);\n if (!gpu_implementation) {\n xe::FatalError(\n \"Unable to initialize the graphics subsystem.\\n\"\n \"\\n\"\n#if XE_PLATFORM_ANDROID\n \"The GPU must support at least Vulkan 1.0 with the 'independentBlend' \"\n \"feature.\\n\"\n \"\\n\"\n#else\n#if XE_PLATFORM_WIN32\n \"For Direct3D 12, at least Windows 10 is required, and the GPU must be \"\n \"compatible with Direct3D 12 feature level 11_0.\\n\"\n \"\\n\"\n#endif // XE_PLATFORM_WIN32\n \"For Vulkan, the Vulkan runtime must be installed, and the GPU must \"\n \"support at least Vulkan 1.0. The Vulkan runtime can be downloaded at \"\n \"https://vulkan.lunarg.com/sdk/home.\\n\"\n \"\\n\"\n \"Also, ensure that you have the latest driver installed for your GPU.\\n\"\n \"\\n\"\n#endif // XE_PLATFORM_ANDROID\n \"See https://xenia.jp/faq/ for more information and the system \"\n \"requirements.\");\n }\n return gpu_implementation;\n}\n\nstd::vector> EmulatorApp::CreateInputDrivers(\n ui::Window* window) {\n std::vector> drivers;\n if (cvars::hid.compare(\"nop\") == 0) {\n drivers.emplace_back(\n xe::hid::nop::Create(window, EmulatorWindow::kZOrderHidInput));\n } else {\n Factory factory;\n#if XE_PLATFORM_WIN32\n factory.Add(\"xinput\", xe::hid::xinput::Create);\n#endif // XE_PLATFORM_WIN32\n#if !XE_PLATFORM_ANDROID\n factory.Add(\"sdl\", xe::hid::sdl::Create);\n#endif // !XE_PLATFORM_ANDROID\n#if XE_PLATFORM_WIN32\n // WinKey input driver should always be the last input driver added!\n factory.Add(\"winkey\", xe::hid::winkey::Create);\n#endif // XE_PLATFORM_WIN32\n for (auto& driver : factory.CreateAll(cvars::hid, window,\n EmulatorWindow::kZOrderHidInput)) {\n if (XSUCCEEDED(driver->Setup())) {\n drivers.emplace_back(std::move(driver));\n }\n }\n if (drivers.empty()) {\n // Fallback to nop if none created.\n drivers.emplace_back(\n xe::hid::nop::Create(window, EmulatorWindow::kZOrderHidInput));\n }\n }\n return drivers;\n}\n\nbool EmulatorApp::OnInitialize() {\n Profiler::Initialize();\n Profiler::ThreadEnter(\"Main\");\n\n // Figure out where internal files and content should go.\n std::filesystem::path storage_root = cvars::storage_root;\n if (storage_root.empty()) {\n storage_root = xe::filesystem::GetExecutableFolder();\n if (!cvars::portable &&\n !std::filesystem::exists(storage_root / \"portable.txt\")) {\n storage_root = xe::filesystem::GetUserFolder();\n#if defined(XE_PLATFORM_WIN32) || defined(XE_PLATFORM_GNU_LINUX)\n storage_root = storage_root / \"Xenia\";\n#else\n // TODO(Triang3l): Point to the app's external storage \"files\" directory\n // on Android.\n#warning Unhandled platform for the data root.\n storage_root = storage_root / \"Xenia\";\n#endif\n }\n }\n storage_root = std::filesystem::absolute(storage_root);\n XELOGI(\"Storage root: {}\", xe::path_to_utf8(storage_root));\n\n config::SetupConfig(storage_root);\n\n std::filesystem::path content_root = cvars::content_root;\n if (content_root.empty()) {\n content_root = storage_root / \"content\";\n } else {\n // If content root isn't an absolute path, then it should be relative to the\n // storage root.\n if (!content_root.is_absolute()) {\n content_root = storage_root / content_root;\n }\n }\n content_root = std::filesystem::absolute(content_root);\n XELOGI(\"Content root: {}\", xe::path_to_utf8(content_root));\n\n std::filesystem::path cache_root = cvars::cache_root;\n if (cache_root.empty()) {\n cache_root = storage_root / \"cache\";\n // TODO(Triang3l): Point to the app's external storage \"cache\" directory on\n // Android.\n } else {\n // If content root isn't an absolute path, then it should be relative to the\n // storage root.\n if (!cache_root.is_absolute()) {\n cache_root = storage_root / cache_root;\n }\n }\n cache_root = std::filesystem::absolute(cache_root);\n XELOGI(\"Cache root: {}\", xe::path_to_utf8(cache_root));\n\n if (cvars::discord) {\n discord::DiscordPresence::Initialize();\n discord::DiscordPresence::NotPlaying();\n }\n\n // Create the emulator but don't initialize so we can setup the window.\n emulator_ =\n std::make_unique(\"\", storage_root, content_root, cache_root);\n\n // Main emulator display window.\n emulator_window_ = EmulatorWindow::Create(emulator_.get(), app_context());\n if (!emulator_window_) {\n XELOGE(\"Failed to create the main emulator window\");\n return false;\n }\n\n // Setup the emulator and run its loop in a separate thread.\n emulator_thread_quit_requested_.store(false, std::memory_order_relaxed);\n emulator_thread_event_ = xe::threading::Event::CreateAutoResetEvent(false);\n assert_not_null(emulator_thread_event_);\n emulator_thread_ = std::thread(&EmulatorApp::EmulatorThread, this);\n\n return true;\n}\n\nvoid EmulatorApp::OnDestroy() {\n ShutdownEmulatorThreadFromUIThread();\n\n if (cvars::discord) {\n discord::DiscordPresence::Shutdown();\n }\n\n Profiler::Dump();\n // The profiler needs to shut down before the graphics context.\n Profiler::Shutdown();\n\n // Write all cvar overrides to the config.\n config::SaveConfig();\n\n // TODO(DrChat): Remove this code and do a proper exit.\n XELOGI(\"Cheap-skate exit!\");\n std::quick_exit(EXIT_SUCCESS);\n}\n\nvoid EmulatorApp::EmulatorThread() {\n assert_not_null(emulator_thread_event_);\n\n xe::threading::set_name(\"Emulator\");\n Profiler::ThreadEnter(\"Emulator\");\n\n // Setup and initialize all subsystems. If we can't do something\n // (unsupported system, memory issues, etc) this will fail early.\n X_STATUS result = emulator_->Setup(\n emulator_window_->window(), emulator_window_->imgui_drawer(), true,\n CreateAudioSystem, CreateGraphicsSystem, CreateInputDrivers);\n if (XFAILED(result)) {\n XELOGE(\"Failed to setup emulator: {:08X}\", result);\n app_context().RequestDeferredQuit();\n return;\n }\n\n app_context().CallInUIThread(\n [this]() { emulator_window_->SetupGraphicsSystemPresenterPainting(); });\n\n if (cvars::mount_scratch) {\n auto scratch_device = std::make_unique(\n \"\\\\SCRATCH\", \"scratch\", false);\n if (!scratch_device->Initialize()) {\n XELOGE(\"Unable to scan scratch path\");\n } else {\n if (!emulator_->file_system()->RegisterDevice(\n std::move(scratch_device))) {\n XELOGE(\"Unable to register scratch path\");\n } else {\n emulator_->file_system()->RegisterSymbolicLink(\"scratch:\", \"\\\\SCRATCH\");\n }\n }\n }\n\n if (cvars::mount_cache) {\n auto cache0_device =\n std::make_unique(\"\\\\CACHE0\", \"cache0\", false);\n if (!cache0_device->Initialize()) {\n XELOGE(\"Unable to scan cache0 path\");\n } else {\n if (!emulator_->file_system()->RegisterDevice(std::move(cache0_device))) {\n XELOGE(\"Unable to register cache0 path\");\n } else {\n emulator_->file_system()->RegisterSymbolicLink(\"cache0:\", \"\\\\CACHE0\");\n }\n }\n\n auto cache1_device =\n std::make_unique(\"\\\\CACHE1\", \"cache1\", false);\n if (!cache1_device->Initialize()) {\n XELOGE(\"Unable to scan cache1 path\");\n } else {\n if (!emulator_->file_system()->RegisterDevice(std::move(cache1_device))) {\n XELOGE(\"Unable to register cache1 path\");\n } else {\n emulator_->file_system()->RegisterSymbolicLink(\"cache1:\", \"\\\\CACHE1\");\n }\n }\n\n // Some (older?) games try accessing cache:\\ too\n // NOTE: this must be registered _after_ the cache0/cache1 devices, due to\n // substring/start_with logic inside VirtualFileSystem::ResolvePath, else\n // accesses to those devices will go here instead\n auto cache_device =\n std::make_unique(\"\\\\CACHE\", \"cache\", false);\n if (!cache_device->Initialize()) {\n XELOGE(\"Unable to scan cache path\");\n } else {\n if (!emulator_->file_system()->RegisterDevice(std::move(cache_device))) {\n XELOGE(\"Unable to register cache path\");\n } else {\n emulator_->file_system()->RegisterSymbolicLink(\"cache:\", \"\\\\CACHE\");\n }\n }\n }\n\n // Set a debug handler.\n // This will respond to debugging requests so we can open the debug UI.\n if (cvars::debug) {\n emulator_->processor()->set_debug_listener_request_handler(\n [this](xe::cpu::Processor* processor) {\n if (debug_window_) {\n return debug_window_.get();\n }\n app_context().CallInUIThreadSynchronous([this]() {\n debug_window_ = xe::debug::ui::DebugWindow::Create(emulator_.get(),\n app_context());\n debug_window_->window()->AddListener(\n &debug_window_closed_listener_);\n });\n // If failed to enqueue the UI thread call, this will just be null.\n return debug_window_.get();\n });\n }\n\n emulator_->on_launch.AddListener([&](auto title_id, const auto& game_title) {\n if (cvars::discord) {\n discord::DiscordPresence::PlayingTitle(\n game_title.empty() ? \"Unknown Title\" : std::string(game_title));\n }\n app_context().CallInUIThread([this]() { emulator_window_->UpdateTitle(); });\n emulator_thread_event_->Set();\n });\n\n emulator_->on_shader_storage_initialization.AddListener(\n [this](bool initializing) {\n app_context().CallInUIThread([this, initializing]() {\n emulator_window_->SetInitializingShaderStorage(initializing);\n });\n });\n\n emulator_->on_terminate.AddListener([]() {\n if (cvars::discord) {\n discord::DiscordPresence::NotPlaying();\n }\n });\n\n // Enable emulator input now that the emulator is properly loaded.\n app_context().CallInUIThread(\n [this]() { emulator_window_->OnEmulatorInitialized(); });\n\n // Grab path from the flag or unnamed argument.\n std::filesystem::path path;\n if (!cvars::target.empty()) {\n path = cvars::target;\n }\n\n if (!path.empty()) {\n // Normalize the path and make absolute.\n auto abs_path = std::filesystem::absolute(path);\n result = emulator_->LaunchPath(abs_path);\n if (XFAILED(result)) {\n xe::FatalError(fmt::format(\"Failed to launch target: {:08X}\", result));\n app_context().RequestDeferredQuit();\n return;\n }\n }\n\n // Now, we're going to use this thread to drive events related to emulation.\n while (!emulator_thread_quit_requested_.load(std::memory_order_relaxed)) {\n xe::threading::Wait(emulator_thread_event_.get(), false);\n while (true) {\n emulator_->WaitUntilExit();\n if (emulator_->TitleRequested()) {\n emulator_->LaunchNextTitle();\n } else {\n break;\n }\n }\n }\n}\n\nvoid EmulatorApp::ShutdownEmulatorThreadFromUIThread() {\n // TODO(Triang3l): Proper shutdown of the emulator (relying on std::quick_exit\n // for now) - currently WaitUntilExit loops forever otherwise (plus possibly\n // lots of other things not shutting down correctly now). Some parts of the\n // code call the regular std::exit, which seems to be calling destructors (at\n // least on Linux), so the entire join is currently commented out.\n#if 0\n // Same thread as the one created it, to make sure there's zero possibility of\n // a race with the creation of the emulator thread.\n assert_true(app_context().IsInUIThread());\n emulator_thread_quit_requested_.store(true, std::memory_order_relaxed);\n if (!emulator_thread_.joinable()) {\n return;\n }\n emulator_thread_event_->Set();\n emulator_thread_.join();\n#endif\n}\n\n} // namespace app\n} // namespace xe\n\nXE_DEFINE_WINDOWED_APP(xenia, xe::app::EmulatorApp::Create);\n" }, { "path": "src/xenia/apu/apu_flags.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/apu_flags.h\"\n\nDEFINE_bool(mute, false, \"Mutes all audio output.\", \"APU\")\n" }, { "path": "src/xenia/apu/apu_flags.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_APU_FLAGS_H_\n#define XENIA_APU_APU_FLAGS_H_\n\n#include \"xenia/base/cvar.h\"\nDECLARE_bool(mute)\n\n#endif // XENIA_APU_APU_FLAGS_H_\n" }, { "path": "src/xenia/apu/audio_driver.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/audio_driver.h\"\n\nnamespace xe {\nnamespace apu {\n\nAudioDriver::AudioDriver(Memory* memory) : memory_(memory) {}\n\nAudioDriver::~AudioDriver() = default;\n\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/audio_driver.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_AUDIO_DRIVER_H_\n#define XENIA_APU_AUDIO_DRIVER_H_\n\n#include \"xenia/memory.h\"\n#include \"xenia/xbox.h\"\n\nnamespace xe {\nnamespace apu {\n\nclass AudioDriver {\n public:\n explicit AudioDriver(Memory* memory);\n virtual ~AudioDriver();\n\n virtual void SubmitFrame(uint32_t samples_ptr) = 0;\n\n protected:\n inline uint8_t* TranslatePhysical(uint32_t guest_address) const {\n return memory_->TranslatePhysical(guest_address);\n }\n\n Memory* memory_ = nullptr;\n};\n\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_AUDIO_DRIVER_H_\n" }, { "path": "src/xenia/apu/audio_system.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/audio_system.h\"\n\n#include \"xenia/apu/apu_flags.h\"\n#include \"xenia/apu/audio_driver.h\"\n#include \"xenia/apu/xma_decoder.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/byte_stream.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/base/ring_buffer.h\"\n#include \"xenia/base/string_buffer.h\"\n#include \"xenia/base/threading.h\"\n#include \"xenia/cpu/thread_state.h\"\n\n// As with normal Microsoft, there are like twelve different ways to access\n// the audio APIs. Early games use XMA*() methods almost exclusively to touch\n// decoders. Later games use XAudio*() and direct memory writes to the XMA\n// structures (as opposed to the XMA* calls), meaning that we have to support\n// both.\n//\n// For ease of implementation, most audio related processing is handled in\n// AudioSystem, and the functions here call off to it.\n// The XMA*() functions just manipulate the audio system in the guest context\n// and let the normal AudioSystem handling take it, to prevent duplicate\n// implementations. They can be found in xboxkrnl_audio_xma.cc\n\nnamespace xe {\nnamespace apu {\n\nAudioSystem::AudioSystem(cpu::Processor* processor)\n : memory_(processor->memory()),\n processor_(processor),\n worker_running_(false) {\n std::memset(clients_, 0, sizeof(clients_));\n\n for (size_t i = 0; i < kMaximumClientCount; ++i) {\n client_semaphores_[i] =\n xe::threading::Semaphore::Create(0, kMaximumQueuedFrames);\n assert_not_null(client_semaphores_[i]);\n wait_handles_[i] = client_semaphores_[i].get();\n }\n shutdown_event_ = xe::threading::Event::CreateAutoResetEvent(false);\n assert_not_null(shutdown_event_);\n wait_handles_[kMaximumClientCount] = shutdown_event_.get();\n\n xma_decoder_ = std::make_unique(processor_);\n\n resume_event_ = xe::threading::Event::CreateAutoResetEvent(false);\n assert_not_null(resume_event_);\n}\n\nAudioSystem::~AudioSystem() {\n if (xma_decoder_) {\n xma_decoder_->Shutdown();\n }\n}\n\nX_STATUS AudioSystem::Setup(kernel::KernelState* kernel_state) {\n X_STATUS result = xma_decoder_->Setup(kernel_state);\n if (result) {\n return result;\n }\n\n worker_running_ = true;\n worker_thread_ = kernel::object_ref(\n new kernel::XHostThread(kernel_state, 128 * 1024, 0, [this]() {\n WorkerThreadMain();\n return 0;\n }));\n // As we run audio callbacks the debugger must be able to suspend us.\n worker_thread_->set_can_debugger_suspend(true);\n worker_thread_->set_name(\"Audio Worker\");\n worker_thread_->Create();\n\n return X_STATUS_SUCCESS;\n}\n\nvoid AudioSystem::WorkerThreadMain() {\n // Initialize driver and ringbuffer.\n Initialize();\n\n // Main run loop.\n while (worker_running_) {\n // These handles signify the number of submitted samples. Once we reach\n // 64 samples, we wait until our audio backend releases a semaphore\n // (signaling a sample has finished playing)\n auto result =\n xe::threading::WaitAny(wait_handles_, xe::countof(wait_handles_), true);\n if (result.first == xe::threading::WaitResult::kFailed) {\n // TODO: Assert?\n continue;\n }\n\n if (result.first == threading::WaitResult::kSuccess &&\n result.second == kMaximumClientCount) {\n // Shutdown event signaled.\n if (paused_) {\n pause_fence_.Signal();\n threading::Wait(resume_event_.get(), false);\n }\n\n continue;\n }\n\n // Number of clients pumped\n bool pumped = false;\n if (result.first == xe::threading::WaitResult::kSuccess) {\n auto index = result.second;\n\n auto global_lock = global_critical_region_.Acquire();\n uint32_t client_callback = clients_[index].callback;\n uint32_t client_callback_arg = clients_[index].wrapped_callback_arg;\n global_lock.unlock();\n\n if (client_callback) {\n SCOPE_profile_cpu_i(\"apu\", \"xe::apu::AudioSystem->client_callback\");\n uint64_t args[] = {client_callback_arg};\n processor_->Execute(worker_thread_->thread_state(), client_callback,\n args, xe::countof(args));\n }\n\n pumped = true;\n }\n\n if (!worker_running_) {\n break;\n }\n\n if (!pumped) {\n SCOPE_profile_cpu_i(\"apu\", \"Sleep\");\n xe::threading::Sleep(std::chrono::milliseconds(500));\n }\n }\n worker_running_ = false;\n\n // TODO(benvanik): call module API to kill?\n}\n\nint AudioSystem::FindFreeClient() {\n for (int i = 0; i < kMaximumClientCount; i++) {\n auto& client = clients_[i];\n if (!client.in_use) {\n return i;\n }\n }\n\n return -1;\n}\n\nvoid AudioSystem::Initialize() {}\n\nvoid AudioSystem::Shutdown() {\n worker_running_ = false;\n shutdown_event_->Set();\n if (worker_thread_) {\n worker_thread_->Wait(0, 0, 0, nullptr);\n worker_thread_.reset();\n }\n}\n\nX_STATUS AudioSystem::RegisterClient(uint32_t callback, uint32_t callback_arg,\n size_t* out_index) {\n auto global_lock = global_critical_region_.Acquire();\n\n auto index = FindFreeClient();\n assert_true(index >= 0);\n\n auto client_semaphore = client_semaphores_[index].get();\n auto ret = client_semaphore->Release(kMaximumQueuedFrames, nullptr);\n assert_true(ret);\n\n AudioDriver* driver;\n auto result = CreateDriver(index, client_semaphore, &driver);\n if (XFAILED(result)) {\n return result;\n }\n assert_not_null(driver);\n\n uint32_t ptr = memory()->SystemHeapAlloc(0x4);\n xe::store_and_swap(memory()->TranslateVirtual(ptr), callback_arg);\n\n clients_[index] = {driver, callback, callback_arg, ptr, true};\n\n if (out_index) {\n *out_index = index;\n }\n\n return X_STATUS_SUCCESS;\n}\n\nvoid AudioSystem::SubmitFrame(size_t index, uint32_t samples_ptr) {\n SCOPE_profile_cpu_f(\"apu\");\n\n auto global_lock = global_critical_region_.Acquire();\n assert_true(index < kMaximumClientCount);\n assert_true(clients_[index].driver != NULL);\n (clients_[index].driver)->SubmitFrame(samples_ptr);\n}\n\nvoid AudioSystem::UnregisterClient(size_t index) {\n SCOPE_profile_cpu_f(\"apu\");\n\n auto global_lock = global_critical_region_.Acquire();\n assert_true(index < kMaximumClientCount);\n DestroyDriver(clients_[index].driver);\n memory()->SystemHeapFree(clients_[index].wrapped_callback_arg);\n clients_[index] = {0};\n\n // Drain the semaphore of its count.\n auto client_semaphore = client_semaphores_[index].get();\n xe::threading::WaitResult wait_result;\n do {\n wait_result = xe::threading::Wait(client_semaphore, false,\n std::chrono::milliseconds(0));\n } while (wait_result == xe::threading::WaitResult::kSuccess);\n assert_true(wait_result == xe::threading::WaitResult::kTimeout);\n}\n\nbool AudioSystem::Save(ByteStream* stream) {\n stream->Write(kAudioSaveSignature);\n\n // Count the number of used clients first.\n // Any gaps should be handled gracefully.\n uint32_t used_clients = 0;\n for (int i = 0; i < kMaximumClientCount; i++) {\n if (clients_[i].in_use) {\n used_clients++;\n }\n }\n\n stream->Write(used_clients);\n for (uint32_t i = 0; i < kMaximumClientCount; i++) {\n auto& client = clients_[i];\n if (!client.in_use) {\n continue;\n }\n\n stream->Write(i);\n stream->Write(client.callback);\n stream->Write(client.callback_arg);\n stream->Write(client.wrapped_callback_arg);\n }\n\n return true;\n}\n\nbool AudioSystem::Restore(ByteStream* stream) {\n if (stream->Read() != kAudioSaveSignature) {\n XELOGE(\"AudioSystem::Restore - Invalid magic value!\");\n return false;\n }\n\n uint32_t num_clients = stream->Read();\n for (uint32_t i = 0; i < num_clients; i++) {\n auto id = stream->Read();\n assert_true(id < kMaximumClientCount);\n\n auto& client = clients_[id];\n\n // Reset the semaphore and recreate the driver ourselves.\n if (client.driver) {\n UnregisterClient(id);\n }\n\n client.callback = stream->Read();\n client.callback_arg = stream->Read();\n client.wrapped_callback_arg = stream->Read();\n\n client.in_use = true;\n\n auto client_semaphore = client_semaphores_[id].get();\n auto ret = client_semaphore->Release(kMaximumQueuedFrames, nullptr);\n assert_true(ret);\n\n AudioDriver* driver = nullptr;\n auto status = CreateDriver(id, client_semaphore, &driver);\n if (XFAILED(status)) {\n XELOGE(\n \"AudioSystem::Restore - Call to CreateDriver failed with status \"\n \"{:08X}\",\n status);\n return false;\n }\n\n assert_not_null(driver);\n client.driver = driver;\n }\n\n return true;\n}\n\nvoid AudioSystem::Pause() {\n if (paused_) {\n return;\n }\n paused_ = true;\n\n // Kind of a hack, but it works.\n shutdown_event_->Set();\n pause_fence_.Wait();\n\n xma_decoder_->Pause();\n}\n\nvoid AudioSystem::Resume() {\n if (!paused_) {\n return;\n }\n paused_ = false;\n\n resume_event_->Set();\n\n xma_decoder_->Resume();\n}\n\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/audio_system.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_AUDIO_SYSTEM_H_\n#define XENIA_APU_AUDIO_SYSTEM_H_\n\n#include \n#include \n\n#include \"xenia/base/mutex.h\"\n#include \"xenia/base/threading.h\"\n#include \"xenia/cpu/processor.h\"\n#include \"xenia/kernel/xthread.h\"\n#include \"xenia/memory.h\"\n#include \"xenia/xbox.h\"\n\nnamespace xe {\nnamespace apu {\n\nconstexpr fourcc_t kAudioSaveSignature = make_fourcc(\"XAUD\");\n\nclass AudioDriver;\nclass XmaDecoder;\n\nclass AudioSystem {\n public:\n virtual ~AudioSystem();\n\n Memory* memory() const { return memory_; }\n cpu::Processor* processor() const { return processor_; }\n XmaDecoder* xma_decoder() const { return xma_decoder_.get(); }\n\n virtual X_STATUS Setup(kernel::KernelState* kernel_state);\n virtual void Shutdown();\n\n X_STATUS RegisterClient(uint32_t callback, uint32_t callback_arg,\n size_t* out_index);\n void UnregisterClient(size_t index);\n void SubmitFrame(size_t index, uint32_t samples_ptr);\n\n bool Save(ByteStream* stream);\n bool Restore(ByteStream* stream);\n\n bool is_paused() const { return paused_; }\n void Pause();\n void Resume();\n\n protected:\n explicit AudioSystem(cpu::Processor* processor);\n\n virtual void Initialize();\n\n void WorkerThreadMain();\n\n virtual X_STATUS CreateDriver(size_t index,\n xe::threading::Semaphore* semaphore,\n AudioDriver** out_driver) = 0;\n virtual void DestroyDriver(AudioDriver* driver) = 0;\n\n // TODO(gibbed): respect XAUDIO2_MAX_QUEUED_BUFFERS somehow (ie min(64,\n // XAUDIO2_MAX_QUEUED_BUFFERS))\n static const size_t kMaximumQueuedFrames = 64;\n\n Memory* memory_ = nullptr;\n cpu::Processor* processor_ = nullptr;\n std::unique_ptr xma_decoder_;\n\n std::atomic worker_running_ = {false};\n kernel::object_ref worker_thread_;\n\n xe::global_critical_region global_critical_region_;\n static const size_t kMaximumClientCount = 8;\n struct {\n AudioDriver* driver;\n uint32_t callback;\n uint32_t callback_arg;\n uint32_t wrapped_callback_arg;\n bool in_use;\n } clients_[kMaximumClientCount];\n\n int FindFreeClient();\n\n std::unique_ptr\n client_semaphores_[kMaximumClientCount];\n // Event is always there in case we have no clients.\n std::unique_ptr shutdown_event_;\n xe::threading::WaitHandle* wait_handles_[kMaximumClientCount + 1];\n\n bool paused_ = false;\n threading::Fence pause_fence_;\n std::unique_ptr resume_event_;\n};\n\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_AUDIO_SYSTEM_H_\n" }, { "path": "src/xenia/apu/conversion.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_CONVERSION_H_\n#define XENIA_APU_CONVERSION_H_\n\n#include \n\n#include \"xenia/base/byte_order.h\"\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace conversion {\n\n#if XE_ARCH_AMD64\ninline void sequential_6_BE_to_interleaved_6_LE(float* output,\n const float* input,\n size_t ch_sample_count) {\n const uint32_t* in = reinterpret_cast(input);\n uint32_t* out = reinterpret_cast(output);\n const __m128i byte_swap_shuffle =\n _mm_set_epi8(12, 13, 14, 15, 8, 9, 10, 11, 4, 5, 6, 7, 0, 1, 2, 3);\n for (size_t sample = 0; sample < ch_sample_count; sample++) {\n __m128i sample0 = _mm_set_epi32(\n in[3 * ch_sample_count + sample], in[2 * ch_sample_count + sample],\n in[1 * ch_sample_count + sample], in[0 * ch_sample_count + sample]);\n uint32_t sample1 = in[4 * ch_sample_count + sample];\n uint32_t sample2 = in[5 * ch_sample_count + sample];\n sample0 = _mm_shuffle_epi8(sample0, byte_swap_shuffle);\n _mm_storeu_si128(reinterpret_cast<__m128i*>(&out[sample * 6]), sample0);\n sample1 = xe::byte_swap(sample1);\n out[sample * 6 + 4] = sample1;\n sample2 = xe::byte_swap(sample2);\n out[sample * 6 + 5] = sample2;\n }\n}\n\ninline void sequential_6_BE_to_interleaved_2_LE(float* output,\n const float* input,\n size_t ch_sample_count) {\n assert_true(ch_sample_count % 4 == 0);\n const uint32_t* in = reinterpret_cast(input);\n uint32_t* out = reinterpret_cast(output);\n const __m128i byte_swap_shuffle =\n _mm_set_epi8(12, 13, 14, 15, 8, 9, 10, 11, 4, 5, 6, 7, 0, 1, 2, 3);\n const __m128 half = _mm_set1_ps(0.5f);\n const __m128 two_fifths = _mm_set1_ps(1.0f / 2.5f);\n\n // put center on left and right, discard low frequency\n for (size_t sample = 0; sample < ch_sample_count; sample += 4) {\n // load 4 samples from 6 channels each\n __m128 fl = _mm_loadu_ps(&input[0 * ch_sample_count + sample]);\n __m128 fr = _mm_loadu_ps(&input[1 * ch_sample_count + sample]);\n __m128 fc = _mm_loadu_ps(&input[2 * ch_sample_count + sample]);\n __m128 bl = _mm_loadu_ps(&input[4 * ch_sample_count + sample]);\n __m128 br = _mm_loadu_ps(&input[5 * ch_sample_count + sample]);\n // byte swap\n fl = _mm_castsi128_ps(\n _mm_shuffle_epi8(_mm_castps_si128(fl), byte_swap_shuffle));\n fr = _mm_castsi128_ps(\n _mm_shuffle_epi8(_mm_castps_si128(fr), byte_swap_shuffle));\n fc = _mm_castsi128_ps(\n _mm_shuffle_epi8(_mm_castps_si128(fc), byte_swap_shuffle));\n bl = _mm_castsi128_ps(\n _mm_shuffle_epi8(_mm_castps_si128(bl), byte_swap_shuffle));\n br = _mm_castsi128_ps(\n _mm_shuffle_epi8(_mm_castps_si128(br), byte_swap_shuffle));\n\n __m128 center_halved = _mm_mul_ps(fc, half);\n __m128 left = _mm_add_ps(_mm_add_ps(fl, bl), center_halved);\n __m128 right = _mm_add_ps(_mm_add_ps(fr, br), center_halved);\n left = _mm_mul_ps(left, two_fifths);\n right = _mm_mul_ps(right, two_fifths);\n _mm_storeu_ps(&output[sample * 2], _mm_unpacklo_ps(left, right));\n _mm_storeu_ps(&output[(sample + 2) * 2], _mm_unpackhi_ps(left, right));\n }\n}\n#else\ninline void sequential_6_BE_to_interleaved_6_LE(float* output,\n const float* input,\n size_t ch_sample_count) {\n for (size_t sample = 0; sample < ch_sample_count; sample++) {\n for (size_t channel = 0; channel < 6; channel++) {\n output[sample * 6 + channel] =\n xe::byte_swap(input[channel * ch_sample_count + sample]);\n }\n }\n}\ninline void sequential_6_BE_to_interleaved_2_LE(float* output,\n const float* input,\n size_t ch_sample_count) {\n // Default 5.1 channel mapping is fl, fr, fc, lf, bl, br\n // https://docs.microsoft.com/en-us/windows/win32/xaudio2/xaudio2-default-channel-mapping\n for (size_t sample = 0; sample < ch_sample_count; sample++) {\n // put center on left and right, discard low frequency\n float fl = xe::byte_swap(input[0 * ch_sample_count + sample]);\n float fr = xe::byte_swap(input[1 * ch_sample_count + sample]);\n float fc = xe::byte_swap(input[2 * ch_sample_count + sample]);\n float br = xe::byte_swap(input[4 * ch_sample_count + sample]);\n float bl = xe::byte_swap(input[5 * ch_sample_count + sample]);\n float center_halved = fc * 0.5f;\n output[sample * 2] = (fl + bl + center_halved) * (1.0f / 2.5f);\n output[sample * 2 + 1] = (fr + br + center_halved) * (1.0f / 2.5f);\n }\n}\n#endif\n\n} // namespace conversion\n} // namespace apu\n} // namespace xe\n\n#endif\n" }, { "path": "src/xenia/apu/debug_visualizers.natvis", "content": "\n\n\n \n \n id={id_}, allocated={is_allocated_}, enabled={is_enabled_}\n \n\n\n" }, { "path": "src/xenia/apu/nop/nop_apu_flags.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/nop/nop_apu_flags.h\"\n" }, { "path": "src/xenia/apu/nop/nop_apu_flags.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_NOP_NOP_APU_FLAGS_H_\n#define XENIA_APU_NOP_NOP_APU_FLAGS_H_\n\n#endif // XENIA_APU_NOP_NOP_APU_FLAGS_H_\n" }, { "path": "src/xenia/apu/nop/nop_audio_system.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/nop/nop_audio_system.h\"\n\n#include \"xenia/apu/apu_flags.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace nop {\n\nstd::unique_ptr NopAudioSystem::Create(cpu::Processor* processor) {\n return std::make_unique(processor);\n}\n\nNopAudioSystem::NopAudioSystem(cpu::Processor* processor)\n : AudioSystem(processor) {}\n\nNopAudioSystem::~NopAudioSystem() = default;\n\nX_STATUS NopAudioSystem::CreateDriver(size_t index,\n xe::threading::Semaphore* semaphore,\n AudioDriver** out_driver) {\n return X_STATUS_NOT_IMPLEMENTED;\n}\n\nvoid NopAudioSystem::DestroyDriver(AudioDriver* driver) { assert_always(); }\n\n} // namespace nop\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/nop/nop_audio_system.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_NOP_NOP_AUDIO_SYSTEM_H_\n#define XENIA_APU_NOP_NOP_AUDIO_SYSTEM_H_\n\n#include \"xenia/apu/audio_system.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace nop {\n\nclass NopAudioSystem : public AudioSystem {\n public:\n explicit NopAudioSystem(cpu::Processor* processor);\n ~NopAudioSystem() override;\n\n static bool IsAvailable() { return true; }\n\n static std::unique_ptr Create(cpu::Processor* processor);\n\n X_STATUS CreateDriver(size_t index, xe::threading::Semaphore* semaphore,\n AudioDriver** out_driver) override;\n void DestroyDriver(AudioDriver* driver) override;\n};\n\n} // namespace nop\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_NOP_NOP_AUDIO_SYSTEM_H_\n" }, { "path": "src/xenia/apu/nop/premake5.lua", "content": "project_root = \"../../../..\"\ninclude(project_root..\"/tools/build\")\n\ngroup(\"src\")\nproject(\"xenia-apu-nop\")\n uuid(\"f37dbf3a-d200-4cc0-83f0-f801b1bdd862\")\n kind(\"StaticLib\")\n language(\"C++\")\n links({\n \"xenia-base\",\n \"xenia-apu\",\n })\n defines({\n })\n local_platform_files()\n" }, { "path": "src/xenia/apu/premake5.lua", "content": "project_root = \"../../..\"\ninclude(project_root..\"/tools/build\")\n\ngroup(\"src\")\nproject(\"xenia-apu\")\n uuid(\"f4df01f0-50e4-4c67-8f54-61660696cc79\")\n kind(\"StaticLib\")\n language(\"C++\")\n links({\n \"libavcodec\",\n \"libavutil\",\n \"xenia-base\",\n })\n defines({\n })\n includedirs({\n project_root..\"/third_party/FFmpeg/\",\n })\n local_platform_files()\n" }, { "path": "src/xenia/apu/sdl/premake5.lua", "content": "project_root = \"../../../..\"\ninclude(project_root..\"/tools/build\")\n\ngroup(\"src\")\nproject(\"xenia-apu-sdl\")\n uuid(\"153b4e8b-813a-40e6-9366-4b51abc73c45\")\n kind(\"StaticLib\")\n language(\"C++\")\n links({\n \"xenia-apu\",\n \"xenia-base\",\n \"xenia-helper-sdl\",\n \"SDL2\",\n })\n defines({\n })\n local_platform_files()\n sdl2_include()\n" }, { "path": "src/xenia/apu/sdl/sdl_audio_driver.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/sdl/sdl_audio_driver.h\"\n\n#include \n#include \n\n#include \"xenia/apu/apu_flags.h\"\n#include \"xenia/apu/conversion.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/helper/sdl/sdl_helper.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace sdl {\n\nSDLAudioDriver::SDLAudioDriver(Memory* memory,\n xe::threading::Semaphore* semaphore)\n : AudioDriver(memory), semaphore_(semaphore) {}\n\nSDLAudioDriver::~SDLAudioDriver() {\n assert_true(frames_queued_.empty());\n assert_true(frames_unused_.empty());\n};\n\nbool SDLAudioDriver::Initialize() {\n SDL_version ver = {};\n SDL_GetVersion(&ver);\n if ((ver.major < 2) || (ver.major == 2 && ver.minor == 0 && ver.patch < 8)) {\n XELOGW(\n \"SDL library version {}.{}.{} is outdated. \"\n \"You may experience choppy audio.\",\n ver.major, ver.minor, ver.patch);\n }\n\n if (!xe::helper::sdl::SDLHelper::Prepare()) {\n return false;\n }\n if (SDL_InitSubSystem(SDL_INIT_AUDIO) < 0) {\n return false;\n }\n sdl_initialized_ = true;\n\n SDL_AudioSpec desired_spec = {};\n SDL_AudioSpec obtained_spec;\n desired_spec.freq = frame_frequency_;\n desired_spec.format = AUDIO_F32;\n desired_spec.channels = frame_channels_;\n desired_spec.samples = channel_samples_;\n desired_spec.callback = SDLCallback;\n desired_spec.userdata = this;\n // Allow the hardware to decide between 5.1 and stereo\n int allowed_change = SDL_AUDIO_ALLOW_CHANNELS_CHANGE;\n for (int i = 0; i < 2; i++) {\n sdl_device_id_ = SDL_OpenAudioDevice(nullptr, 0, &desired_spec,\n &obtained_spec, allowed_change);\n if (sdl_device_id_ <= 0) {\n XELOGE(\"SDL_OpenAudioDevice() failed.\");\n return false;\n }\n if (obtained_spec.channels == 2 || obtained_spec.channels == 6) {\n break;\n }\n // If the system is 4 or 7.1, let SDL convert\n allowed_change = 0;\n SDL_CloseAudioDevice(sdl_device_id_);\n sdl_device_id_ = -1;\n }\n if (sdl_device_id_ <= 0) {\n XELOGE(\"Failed to get a compatible SDL Audio Device.\");\n return false;\n }\n sdl_device_channels_ = obtained_spec.channels;\n\n SDL_PauseAudioDevice(sdl_device_id_, 0);\n\n return true;\n}\n\nvoid SDLAudioDriver::SubmitFrame(uint32_t frame_ptr) {\n const auto input_frame = memory_->TranslateVirtual(frame_ptr);\n float* output_frame;\n {\n std::unique_lock guard(frames_mutex_);\n if (frames_unused_.empty()) {\n output_frame = new float[frame_samples_];\n } else {\n output_frame = frames_unused_.top();\n frames_unused_.pop();\n }\n }\n\n std::memcpy(output_frame, input_frame, frame_samples_ * sizeof(float));\n\n {\n std::unique_lock guard(frames_mutex_);\n frames_queued_.push(output_frame);\n }\n}\n\nvoid SDLAudioDriver::Shutdown() {\n if (sdl_device_id_ > 0) {\n SDL_CloseAudioDevice(sdl_device_id_);\n sdl_device_id_ = -1;\n }\n if (sdl_initialized_) {\n SDL_QuitSubSystem(SDL_INIT_AUDIO);\n sdl_initialized_ = false;\n }\n std::unique_lock guard(frames_mutex_);\n while (!frames_unused_.empty()) {\n delete[] frames_unused_.top();\n frames_unused_.pop();\n };\n while (!frames_queued_.empty()) {\n delete[] frames_queued_.front();\n frames_queued_.pop();\n };\n}\n\nvoid SDLAudioDriver::SDLCallback(void* userdata, Uint8* stream, int len) {\n SCOPE_profile_cpu_f(\"apu\");\n if (!userdata || !stream) {\n XELOGE(\"SDLAudioDriver::sdl_callback called with nullptr.\");\n return;\n }\n const auto driver = static_cast(userdata);\n assert_true(len ==\n sizeof(float) * channel_samples_ * driver->sdl_device_channels_);\n\n std::unique_lock guard(driver->frames_mutex_);\n if (driver->frames_queued_.empty()) {\n std::memset(stream, 0, len);\n } else {\n auto buffer = driver->frames_queued_.front();\n driver->frames_queued_.pop();\n if (cvars::mute) {\n std::memset(stream, 0, len);\n } else {\n switch (driver->sdl_device_channels_) {\n case 2:\n conversion::sequential_6_BE_to_interleaved_2_LE(\n reinterpret_cast(stream), buffer, channel_samples_);\n break;\n case 6:\n conversion::sequential_6_BE_to_interleaved_6_LE(\n reinterpret_cast(stream), buffer, channel_samples_);\n break;\n default:\n assert_unhandled_case(driver->sdl_device_channels_);\n break;\n }\n }\n driver->frames_unused_.push(buffer);\n\n auto ret = driver->semaphore_->Release(1, nullptr);\n assert_true(ret);\n }\n};\n} // namespace sdl\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/sdl/sdl_audio_driver.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_SDL_SDL_AUDIO_DRIVER_H_\n#define XENIA_APU_SDL_SDL_AUDIO_DRIVER_H_\n\n#include \n#include \n#include \n\n#include \"SDL.h\"\n#include \"xenia/apu/audio_driver.h\"\n#include \"xenia/base/threading.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace sdl {\n\nclass SDLAudioDriver : public AudioDriver {\n public:\n SDLAudioDriver(Memory* memory, xe::threading::Semaphore* semaphore);\n ~SDLAudioDriver() override;\n\n bool Initialize();\n void SubmitFrame(uint32_t frame_ptr) override;\n void Shutdown();\n\n protected:\n static void SDLCallback(void* userdata, Uint8* stream, int len);\n\n xe::threading::Semaphore* semaphore_ = nullptr;\n\n SDL_AudioDeviceID sdl_device_id_ = -1;\n bool sdl_initialized_ = false;\n uint8_t sdl_device_channels_ = 0;\n\n static const uint32_t frame_frequency_ = 48000;\n static const uint32_t frame_channels_ = 6;\n static const uint32_t channel_samples_ = 256;\n static const uint32_t frame_samples_ = frame_channels_ * channel_samples_;\n static const uint32_t frame_size_ = sizeof(float) * frame_samples_;\n std::queue frames_queued_ = {};\n std::stack frames_unused_ = {};\n std::mutex frames_mutex_ = {};\n};\n\n} // namespace sdl\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_SDL_SDL_AUDIO_DRIVER_H_\n" }, { "path": "src/xenia/apu/sdl/sdl_audio_system.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/sdl/sdl_audio_system.h\"\n\n#include \"xenia/apu/apu_flags.h\"\n#include \"xenia/apu/sdl/sdl_audio_driver.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace sdl {\n\nstd::unique_ptr SDLAudioSystem::Create(cpu::Processor* processor) {\n return std::make_unique(processor);\n}\n\nSDLAudioSystem::SDLAudioSystem(cpu::Processor* processor)\n : AudioSystem(processor) {}\n\nSDLAudioSystem::~SDLAudioSystem() {}\n\nvoid SDLAudioSystem::Initialize() { AudioSystem::Initialize(); }\n\nX_STATUS SDLAudioSystem::CreateDriver(size_t index,\n xe::threading::Semaphore* semaphore,\n AudioDriver** out_driver) {\n assert_not_null(out_driver);\n auto driver = new SDLAudioDriver(memory_, semaphore);\n if (!driver->Initialize()) {\n driver->Shutdown();\n return X_STATUS_UNSUCCESSFUL;\n }\n\n *out_driver = driver;\n return X_STATUS_SUCCESS;\n}\n\nvoid SDLAudioSystem::DestroyDriver(AudioDriver* driver) {\n assert_not_null(driver);\n auto sdldriver = dynamic_cast(driver);\n assert_not_null(sdldriver);\n sdldriver->Shutdown();\n delete sdldriver;\n}\n\n} // namespace sdl\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/sdl/sdl_audio_system.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_SDL_SDL_AUDIO_SYSTEM_H_\n#define XENIA_APU_SDL_SDL_AUDIO_SYSTEM_H_\n\n#include \"xenia/apu/audio_system.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace sdl {\n\nclass SDLAudioSystem : public AudioSystem {\n public:\n explicit SDLAudioSystem(cpu::Processor* processor);\n ~SDLAudioSystem() override;\n\n static bool IsAvailable() { return true; }\n\n static std::unique_ptr Create(cpu::Processor* processor);\n\n X_RESULT CreateDriver(size_t index, xe::threading::Semaphore* semaphore,\n AudioDriver** out_driver) override;\n void DestroyDriver(AudioDriver* driver) override;\n\n protected:\n void Initialize() override;\n};\n\n} // namespace sdl\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_SDL_SDL_AUDIO_SYSTEM_H_\n" }, { "path": "src/xenia/apu/xaudio2/premake5.lua", "content": "project_root = \"../../../..\"\ninclude(project_root..\"/tools/build\")\n\ngroup(\"src\")\nproject(\"xenia-apu-xaudio2\")\n uuid(\"7a54a497-24d9-4c0e-a013-8507a04231f9\")\n kind(\"StaticLib\")\n language(\"C++\")\n links({\n \"xenia-base\",\n \"xenia-apu\",\n })\n defines({\n })\n local_platform_files()\n" }, { "path": "src/xenia/apu/xaudio2/xaudio2_api.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2019 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_XAUDIO2_XAUDIO2_API_H_\n#define XENIA_APU_XAUDIO2_XAUDIO2_API_H_\n\n#include \n\n#include \"xenia/base/platform_win.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace xaudio2 {\nnamespace api {\n\n// Parts of XAudio2.h needed for Xenia. The header in the Windows SDK (for 2.8+)\n// or the DirectX SDK (for 2.7 and below) is for a single version that is chosen\n// for the target OS version at compile time, and cannot be useful for different\n// DLL versions.\n//\n// XAudio 2.8 is also not available on Windows 7, but including the 2.7 header\n// is complicated because it has the same name and include guard as the 2.8 one,\n// and also 2.7 is outdated - support already dropped in Microsoft Store apps,\n// for instance.\n//\n// Copyright (c) Microsoft Corporation. All rights reserved.\n\nclass __declspec(uuid(\"5a508685-a254-4fba-9b82-9a24b00306af\")) XAudio2_7;\ninterface __declspec(uuid(\"8bcf1f58-9fe7-4583-8ac6-e2adc465c8bb\")) IXAudio2_7;\ninterface __declspec(uuid(\"60d8dac8-5aa1-4e8e-b597-2f5e2883d484\")) IXAudio2_8;\n\n#pragma pack(push, 1)\n\nstatic constexpr UINT32 XE_XAUDIO2_MAX_QUEUED_BUFFERS = 64;\nstatic constexpr float XE_XAUDIO2_MAX_FREQ_RATIO = 1024.0f;\nstatic constexpr float XE_XAUDIO2_DEFAULT_FREQ_RATIO = 2.0f;\n\nstatic constexpr UINT32 XE_XAUDIO2_COMMIT_NOW = 0;\nstatic constexpr UINT32 XE_XAUDIO2_NO_LOOP_REGION = 0;\nstatic constexpr UINT32 XE_XAUDIO2_DEFAULT_CHANNELS = 0;\nstatic constexpr UINT32 XE_XAUDIO2_DEFAULT_SAMPLERATE = 0;\n\n// 2.8+ only.\nstatic constexpr UINT32 XE_XAUDIO2_VOICE_NOSAMPLESPLAYED = 0x0100;\n\ninterface IXAudio2_7;\ninterface IXAudio2_8;\ninterface IXAudio2Voice;\ninterface IXAudio2_7SourceVoice;\ninterface IXAudio2_8SourceVoice;\ninterface IXAudio2SubmixVoice;\ninterface IXAudio2_7MasteringVoice;\ninterface IXAudio2_8MasteringVoice;\ninterface IXAudio2EngineCallback;\ninterface IXAudio2VoiceCallback;\n\ntypedef UINT32 XAUDIO2_PROCESSOR;\nstatic constexpr XAUDIO2_PROCESSOR XE_XAUDIO2_ANY_PROCESSOR = 0xFFFFFFFF;\nstatic constexpr XAUDIO2_PROCESSOR XE_XAUDIO2_7_DEFAULT_PROCESSOR =\n XE_XAUDIO2_ANY_PROCESSOR;\nstatic constexpr XAUDIO2_PROCESSOR XE_XAUDIO2_8_DEFAULT_PROCESSOR = 0x00000001;\n\n// 2.7 only.\nstruct XAUDIO2_DEVICE_DETAILS;\n\nstruct XAUDIO2_VOICE_DETAILS;\nstruct XAUDIO2_VOICE_SENDS;\nstruct XAUDIO2_EFFECT_CHAIN;\nstruct XAUDIO2_FILTER_PARAMETERS;\n\nstruct XAUDIO2_BUFFER {\n UINT32 Flags;\n UINT32 AudioBytes;\n const BYTE* pAudioData;\n UINT32 PlayBegin;\n UINT32 PlayLength;\n UINT32 LoopBegin;\n UINT32 LoopLength;\n UINT32 LoopCount;\n void* pContext;\n};\n\nstruct XAUDIO2_BUFFER_WMA;\n\nstruct XAUDIO2_VOICE_STATE {\n void* pCurrentBufferContext;\n UINT32 BuffersQueued;\n UINT64 SamplesPlayed;\n};\n\nstruct XAUDIO2_PERFORMANCE_DATA;\n\nstruct XAUDIO2_DEBUG_CONFIGURATION {\n UINT32 TraceMask;\n UINT32 BreakMask;\n BOOL LogThreadID;\n BOOL LogFileline;\n BOOL LogFunctionName;\n BOOL LogTiming;\n};\n\nstatic constexpr UINT32 XE_XAUDIO2_LOG_ERRORS = 0x0001;\nstatic constexpr UINT32 XE_XAUDIO2_LOG_WARNINGS = 0x0002;\n\n// clang-format off\n\n// IXAudio2 2.7.\nDECLARE_INTERFACE_(IXAudio2_7, IUnknown) {\n STDMETHOD(QueryInterface)(REFIID riid, void** ppvInterface) = 0;\n STDMETHOD_(ULONG, AddRef)() = 0;\n STDMETHOD_(ULONG, Release)() = 0;\n // 2.7 only.\n STDMETHOD(GetDeviceCount)(UINT32* pCount) = 0;\n // 2.7 only.\n STDMETHOD(GetDeviceDetails)(UINT32 Index,\n XAUDIO2_DEVICE_DETAILS* pDeviceDetails) = 0;\n // 2.7 only.\n STDMETHOD(Initialize)(\n UINT32 Flags = 0,\n XAUDIO2_PROCESSOR XAudio2Processor = XE_XAUDIO2_7_DEFAULT_PROCESSOR) = 0;\n STDMETHOD(RegisterForCallbacks)(IXAudio2EngineCallback* pCallback) = 0;\n STDMETHOD_(void, UnregisterForCallbacks)(\n IXAudio2EngineCallback* pCallback) = 0;\n STDMETHOD(CreateSourceVoice)(\n IXAudio2_7SourceVoice** ppSourceVoice, const WAVEFORMATEX* pSourceFormat,\n UINT32 Flags = 0, float MaxFrequencyRatio = XE_XAUDIO2_DEFAULT_FREQ_RATIO,\n IXAudio2VoiceCallback* pCallback = nullptr,\n const XAUDIO2_VOICE_SENDS* pSendList = nullptr,\n const XAUDIO2_EFFECT_CHAIN* pEffectChain = nullptr) = 0;\n STDMETHOD(CreateSubmixVoice)(\n IXAudio2SubmixVoice** ppSubmixVoice, UINT32 InputChannels,\n UINT32 InputSampleRate, UINT32 Flags = 0, UINT32 ProcessingStage = 0,\n const XAUDIO2_VOICE_SENDS* pSendList = nullptr,\n const XAUDIO2_EFFECT_CHAIN* pEffectChain = nullptr) = 0;\n // 2.7: Device index instead of device ID, no stream category.\n STDMETHOD(CreateMasteringVoice)(\n IXAudio2_7MasteringVoice** ppMasteringVoice,\n UINT32 InputChannels = XE_XAUDIO2_DEFAULT_CHANNELS,\n UINT32 InputSampleRate = XE_XAUDIO2_DEFAULT_SAMPLERATE, UINT32 Flags = 0,\n UINT32 DeviceIndex = 0,\n const XAUDIO2_EFFECT_CHAIN* pEffectChain = nullptr) = 0;\n STDMETHOD(StartEngine)() = 0;\n STDMETHOD_(void, StopEngine)() = 0;\n STDMETHOD(CommitChanges)(UINT32 OperationSet) = 0;\n STDMETHOD_(void, GetPerformanceData)(XAUDIO2_PERFORMANCE_DATA* pPerfData) = 0;\n STDMETHOD_(void, SetDebugConfiguration)(\n const XAUDIO2_DEBUG_CONFIGURATION* pDebugConfiguration,\n void* pReserved = nullptr) = 0;\n};\n\n// IXAudio2 2.8.\nDECLARE_INTERFACE_(IXAudio2_8, IUnknown) {\n STDMETHOD(QueryInterface)(REFIID riid, void** ppvInterface) = 0;\n STDMETHOD_(ULONG, AddRef)() = 0;\n STDMETHOD_(ULONG, Release)() = 0;\n // 2.8: No GetDeviceCount, GetDeviceDetails and Initialize.\n STDMETHOD(RegisterForCallbacks)(IXAudio2EngineCallback* pCallback) = 0;\n STDMETHOD_(void, UnregisterForCallbacks)(\n IXAudio2EngineCallback* pCallback) = 0;\n STDMETHOD(CreateSourceVoice)(\n IXAudio2_8SourceVoice** ppSourceVoice, const WAVEFORMATEX* pSourceFormat,\n UINT32 Flags = 0, float MaxFrequencyRatio = XE_XAUDIO2_DEFAULT_FREQ_RATIO,\n IXAudio2VoiceCallback* pCallback = nullptr,\n const XAUDIO2_VOICE_SENDS* pSendList = nullptr,\n const XAUDIO2_EFFECT_CHAIN* pEffectChain = nullptr) = 0;\n STDMETHOD(CreateSubmixVoice)(\n IXAudio2SubmixVoice** ppSubmixVoice, UINT32 InputChannels,\n UINT32 InputSampleRate, UINT32 Flags = 0, UINT32 ProcessingStage = 0,\n const XAUDIO2_VOICE_SENDS* pSendList = nullptr,\n const XAUDIO2_EFFECT_CHAIN* pEffectChain = nullptr) = 0;\n // 2.8: Device ID instead of device index, added stream category.\n STDMETHOD(CreateMasteringVoice)(\n IXAudio2_8MasteringVoice** ppMasteringVoice,\n UINT32 InputChannels = XE_XAUDIO2_DEFAULT_CHANNELS,\n UINT32 InputSampleRate = XE_XAUDIO2_DEFAULT_SAMPLERATE, UINT32 Flags = 0,\n LPCWSTR szDeviceId = nullptr,\n const XAUDIO2_EFFECT_CHAIN* pEffectChain = nullptr,\n AUDIO_STREAM_CATEGORY StreamCategory = AudioCategory_GameEffects) = 0;\n STDMETHOD(StartEngine)() = 0;\n STDMETHOD_(void, StopEngine)() = 0;\n STDMETHOD(CommitChanges)(UINT32 OperationSet) = 0;\n STDMETHOD_(void, GetPerformanceData)(XAUDIO2_PERFORMANCE_DATA* pPerfData) = 0;\n STDMETHOD_(void, SetDebugConfiguration)(\n const XAUDIO2_DEBUG_CONFIGURATION* pDebugConfiguration,\n void* pReserved = nullptr) = 0;\n};\n\nDECLARE_INTERFACE(IXAudio2Voice) {\n#define XE_APU_XAUDIO2_API_DECLARE_IXAUDIO2VOICE_METHODS \\\n STDMETHOD_(void, GetVoiceDetails)(XAUDIO2_VOICE_DETAILS* pVoiceDetails) = 0; \\\n STDMETHOD(SetOutputVoices)(const XAUDIO2_VOICE_SENDS* pSendList) = 0; \\\n STDMETHOD(SetEffectChain)(const XAUDIO2_EFFECT_CHAIN* pEffectChain) = 0; \\\n STDMETHOD(EnableEffect)(UINT32 EffectIndex, \\\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0; \\\n STDMETHOD(DisableEffect)(UINT32 EffectIndex, \\\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0; \\\n STDMETHOD_(void, GetEffectState)(UINT32 EffectIndex, BOOL* pEnabled) = 0; \\\n STDMETHOD(SetEffectParameters)( \\\n UINT32 EffectIndex, const void* pParameters, UINT32 ParametersByteSize, \\\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0; \\\n STDMETHOD(GetEffectParameters)(UINT32 EffectIndex, void* pParameters, \\\n UINT32 ParametersByteSize) = 0; \\\n STDMETHOD(SetFilterParameters)( \\\n const XAUDIO2_FILTER_PARAMETERS* pParameters, \\\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0; \\\n STDMETHOD_(void, GetFilterParameters)( \\\n XAUDIO2_FILTER_PARAMETERS* pParameters) = 0; \\\n STDMETHOD(SetOutputFilterParameters)( \\\n IXAudio2Voice* pDestinationVoice, \\\n const XAUDIO2_FILTER_PARAMETERS* pParameters, \\\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0; \\\n STDMETHOD_(void, GetOutputFilterParameters)( \\\n IXAudio2Voice* pDestinationVoice, \\\n XAUDIO2_FILTER_PARAMETERS* pParameters) = 0; \\\n STDMETHOD(SetVolume)(float Volume, \\\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0; \\\n STDMETHOD_(void, GetVolume)(float* pVolume) = 0; \\\n STDMETHOD(SetChannelVolumes)( \\\n UINT32 Channels, const float* pVolumes, \\\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0; \\\n STDMETHOD_(void, GetChannelVolumes)(UINT32 Channels, float* pVolumes) = 0; \\\n STDMETHOD(SetOutputMatrix)(IXAudio2Voice* pDestinationVoice, \\\n UINT32 SourceChannels, \\\n UINT32 DestinationChannels, \\\n const float* pLevelMatrix, \\\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0; \\\n STDMETHOD_(void, GetOutputMatrix)(IXAudio2Voice* pDestinationVoice, \\\n UINT32 SourceChannels, \\\n UINT32 DestinationChannels, \\\n float* pLevelMatrix) = 0; \\\n STDMETHOD_(void, DestroyVoice)() = 0;\n XE_APU_XAUDIO2_API_DECLARE_IXAUDIO2VOICE_METHODS\n};\n\n// IXAudio2SourceVoice 2.7.\nDECLARE_INTERFACE_(IXAudio2_7SourceVoice, IXAudio2Voice) {\n XE_APU_XAUDIO2_API_DECLARE_IXAUDIO2VOICE_METHODS\n STDMETHOD(Start)(UINT32 Flags = 0,\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0;\n STDMETHOD(Stop)(UINT32 Flags = 0,\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0;\n STDMETHOD(SubmitSourceBuffer)(\n const XAUDIO2_BUFFER* pBuffer,\n const XAUDIO2_BUFFER_WMA* pBufferWMA = nullptr) = 0;\n STDMETHOD(FlushSourceBuffers)() = 0;\n STDMETHOD(Discontinuity)() = 0;\n STDMETHOD(ExitLoop)(UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0;\n // 2.7: No Flags.\n STDMETHOD_(void, GetState)(XAUDIO2_VOICE_STATE* pVoiceState) = 0;\n STDMETHOD(SetFrequencyRatio)(float Ratio,\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0;\n STDMETHOD_(void, GetFrequencyRatio)(float* pRatio) = 0;\n STDMETHOD(SetSourceSampleRate)(UINT32 NewSourceSampleRate) = 0;\n};\n\n// IXAudio2SourceVoice 2.8.\nDECLARE_INTERFACE_(IXAudio2_8SourceVoice, IXAudio2Voice) {\n XE_APU_XAUDIO2_API_DECLARE_IXAUDIO2VOICE_METHODS\n STDMETHOD(Start)(UINT32 Flags = 0,\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0;\n STDMETHOD(Stop)(UINT32 Flags = 0,\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0;\n STDMETHOD(SubmitSourceBuffer)(\n const XAUDIO2_BUFFER* pBuffer,\n const XAUDIO2_BUFFER_WMA* pBufferWMA = nullptr) = 0;\n STDMETHOD(FlushSourceBuffers)() = 0;\n STDMETHOD(Discontinuity)() = 0;\n STDMETHOD(ExitLoop)(UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0;\n // 2.8: Flags added.\n STDMETHOD_(void, GetState)(XAUDIO2_VOICE_STATE* pVoiceState,\n UINT32 Flags = 0) = 0;\n STDMETHOD(SetFrequencyRatio)(float Ratio,\n UINT32 OperationSet = XE_XAUDIO2_COMMIT_NOW) = 0;\n STDMETHOD_(void, GetFrequencyRatio)(float* pRatio) = 0;\n STDMETHOD(SetSourceSampleRate)(UINT32 NewSourceSampleRate) = 0;\n};\n\n// IXAudio2MasteringVoice 2.7.\nDECLARE_INTERFACE_(IXAudio2_7MasteringVoice, IXAudio2Voice) {\n XE_APU_XAUDIO2_API_DECLARE_IXAUDIO2VOICE_METHODS\n // 2.7: No GetChannelMask.\n};\n\n// IXAudio2MasteringVoice 2.8.\nDECLARE_INTERFACE_(IXAudio2_8MasteringVoice, IXAudio2Voice) {\n XE_APU_XAUDIO2_API_DECLARE_IXAUDIO2VOICE_METHODS\n // 2.8: GetChannelMask added.\n STDMETHOD(GetChannelMask)(DWORD* pChannelmask) = 0;\n};\n\nDECLARE_INTERFACE(IXAudio2VoiceCallback) {\n // Called just before this voice's processing pass begins.\n STDMETHOD_(void, OnVoiceProcessingPassStart)(UINT32 BytesRequired) = 0;\n STDMETHOD_(void, OnVoiceProcessingPassEnd)() = 0;\n STDMETHOD_(void, OnStreamEnd)() = 0;\n STDMETHOD_(void, OnBufferStart)(void* pBufferContext) = 0;\n STDMETHOD_(void, OnBufferEnd)(void* pBufferContext) = 0;\n STDMETHOD_(void, OnLoopEnd)(void* pBufferContext) = 0;\n STDMETHOD_(void, OnVoiceError)(void* pBufferContext, HRESULT Error) = 0;\n};\n\n#pragma pack(pop)\n\n} // namespace api\n} // namespace xaudio2\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_XAUDIO2_XAUDIO2_API_H_\n" }, { "path": "src/xenia/apu/xaudio2/xaudio2_apu_flags.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/xaudio2/xaudio2_apu_flags.h\"\n" }, { "path": "src/xenia/apu/xaudio2/xaudio2_apu_flags.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_XAUDIO2_XAUDIO2_APU_FLAGS_H_\n#define XENIA_APU_XAUDIO2_XAUDIO2_APU_FLAGS_H_\n\n#endif // XENIA_APU_XAUDIO2_XAUDIO2_APU_FLAGS_H_\n" }, { "path": "src/xenia/apu/xaudio2/xaudio2_audio_driver.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/xaudio2/xaudio2_audio_driver.h\"\n\n#include \"xenia/apu/apu_flags.h\"\n#include \"xenia/apu/conversion.h\"\n#include \"xenia/apu/xaudio2/xaudio2_api.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/clock.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/platform_win.h\"\n#include \"xenia/base/threading.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace xaudio2 {\n\nclass XAudio2AudioDriver::VoiceCallback : public api::IXAudio2VoiceCallback {\n public:\n explicit VoiceCallback(xe::threading::Semaphore* semaphore)\n : semaphore_(semaphore) {}\n ~VoiceCallback() {}\n\n void OnStreamEnd() {}\n void OnVoiceProcessingPassEnd() {}\n void OnVoiceProcessingPassStart(uint32_t samples_required) {}\n void OnBufferEnd(void* context) {\n auto ret = semaphore_->Release(1, nullptr);\n assert_true(ret);\n }\n void OnBufferStart(void* context) {}\n void OnLoopEnd(void* context) {}\n void OnVoiceError(void* context, HRESULT result) {}\n\n private:\n xe::threading::Semaphore* semaphore_ = nullptr;\n};\n\nXAudio2AudioDriver::XAudio2AudioDriver(Memory* memory,\n xe::threading::Semaphore* semaphore)\n : AudioDriver(memory), semaphore_(semaphore) {}\n\nXAudio2AudioDriver::~XAudio2AudioDriver() = default;\n\nbool XAudio2AudioDriver::Initialize() {\n voice_callback_ = new VoiceCallback(semaphore_);\n\n // Load the XAudio2 DLL dynamically. Needed both for 2.7 and for\n // differentiating between 2.8 and later versions. Windows 8.1 SDK references\n // XAudio2_8.dll in xaudio2.lib, which is available on Windows 8.1, however,\n // Windows 10 SDK references XAudio2_9.dll in it, which is only available in\n // Windows 10, and XAudio2_8.dll is linked through a different .lib -\n // xaudio2_8.lib, so easier not to link the .lib at all.\n xaudio2_module_ = static_cast(LoadLibraryW(L\"XAudio2_8.dll\"));\n if (xaudio2_module_) {\n xaudio2_create_ = reinterpret_cast(\n GetProcAddress(static_cast(xaudio2_module_), \"XAudio2Create\"));\n if (xaudio2_create_) {\n api_minor_version_ = 8;\n } else {\n XELOGE(\"XAudio2Create not found in XAudio2_8.dll\");\n FreeLibrary(static_cast(xaudio2_module_));\n xaudio2_module_ = nullptr;\n }\n }\n if (!xaudio2_module_) {\n xaudio2_module_ = static_cast(LoadLibraryW(L\"XAudio2_7.dll\"));\n if (xaudio2_module_) {\n api_minor_version_ = 7;\n } else {\n XELOGE(\"Failed to load XAudio 2.8 or 2.7 library DLL\");\n return false;\n }\n }\n\n // We need to be able to accept frames from any non-STA thread - primarily\n // from any guest thread, so MTA needs to be used. The AudioDriver, however,\n // may be initialized from the UI thread, which has the STA concurrency model,\n // or from another thread regardless of its concurrency model. So, all\n // management of the objects needs to be performed in MTA. Launch the\n // lifecycle management thread, which will handle initialization and shutdown,\n // and also provide a scope for implicit MTA in threads that have never\n // initialized COM explicitly, which lasts until all threads that have\n // initialized MTA explicitly have uninitialized it - the thread that holds\n // the MTA scope needs to be running while other threads are able to submit\n // frames as they might have not initialized MTA explicitly.\n // https://devblogs.microsoft.com/oldnewthing/?p=4613\n // This is needed for both XAudio 2.7 (created via explicit CoCreateInstance)\n // and 2.8 (using XAudio2Create, but still requiring CoInitializeEx).\n // https://docs.microsoft.com/en-us/windows/win32/xaudio2/how-to--initialize-xaudio2\n assert_false(mta_thread_.joinable());\n mta_thread_initialization_attempt_completed_ = false;\n mta_thread_shutdown_requested_ = false;\n mta_thread_ = std::thread(&XAudio2AudioDriver::MTAThread, this);\n {\n std::unique_lock mta_thread_initialization_completion_lock(\n mta_thread_initialization_completion_mutex_);\n while (true) {\n if (mta_thread_initialization_attempt_completed_) {\n break;\n }\n mta_thread_initialization_completion_cond_.wait(\n mta_thread_initialization_completion_lock);\n }\n }\n if (!mta_thread_initialization_completion_result_) {\n mta_thread_.join();\n return false;\n }\n return true;\n}\n\ntemplate \nbool XAudio2AudioDriver::InitializeObjects(Objects& objects) {\n HRESULT hr;\n\n api::XAUDIO2_DEBUG_CONFIGURATION config;\n config.TraceMask = api::XE_XAUDIO2_LOG_ERRORS | api::XE_XAUDIO2_LOG_WARNINGS;\n config.BreakMask = 0;\n config.LogThreadID = FALSE;\n config.LogTiming = TRUE;\n config.LogFunctionName = TRUE;\n config.LogFileline = TRUE;\n objects.audio->SetDebugConfiguration(&config);\n\n hr = objects.audio->CreateMasteringVoice(&objects.mastering_voice);\n if (FAILED(hr)) {\n XELOGE(\"IXAudio2::CreateMasteringVoice failed with 0x{:08X}\", hr);\n ShutdownObjects(objects);\n return false;\n }\n\n WAVEFORMATIEEEFLOATEX waveformat;\n\n waveformat.Format.wFormatTag = WAVE_FORMAT_EXTENSIBLE;\n waveformat.Format.nChannels = frame_channels_;\n waveformat.Format.nSamplesPerSec = 48000;\n waveformat.Format.wBitsPerSample = 32;\n waveformat.Format.nBlockAlign =\n (waveformat.Format.nChannels * waveformat.Format.wBitsPerSample) / 8;\n waveformat.Format.nAvgBytesPerSec =\n waveformat.Format.nSamplesPerSec * waveformat.Format.nBlockAlign;\n waveformat.Format.cbSize =\n sizeof(WAVEFORMATIEEEFLOATEX) - sizeof(WAVEFORMATEX);\n\n waveformat.SubFormat = KSDATAFORMAT_SUBTYPE_IEEE_FLOAT;\n waveformat.Samples.wValidBitsPerSample = waveformat.Format.wBitsPerSample;\n static const DWORD kChannelMasks[] = {\n 0,\n 0,\n SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT,\n 0,\n 0,\n 0,\n SPEAKER_FRONT_LEFT | SPEAKER_FRONT_CENTER | SPEAKER_FRONT_RIGHT |\n SPEAKER_LOW_FREQUENCY | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT,\n 0,\n };\n waveformat.dwChannelMask = kChannelMasks[waveformat.Format.nChannels];\n\n hr = objects.audio->CreateSourceVoice(\n &objects.pcm_voice, &waveformat.Format,\n 0, // api::XE_XAUDIO2_VOICE_NOSRC | api::XE_XAUDIO2_VOICE_NOPITCH,\n api::XE_XAUDIO2_MAX_FREQ_RATIO, voice_callback_);\n if (FAILED(hr)) {\n XELOGE(\"IXAudio2::CreateSourceVoice failed with 0x{:08X}\", hr);\n ShutdownObjects(objects);\n return false;\n }\n\n hr = objects.pcm_voice->Start();\n if (FAILED(hr)) {\n XELOGE(\"IXAudio2SourceVoice::Start failed with 0x{:08X}\", hr);\n ShutdownObjects(objects);\n return false;\n }\n\n if (cvars::mute) {\n objects.pcm_voice->SetVolume(0.0f);\n }\n\n return true;\n}\n\nvoid XAudio2AudioDriver::SubmitFrame(uint32_t frame_ptr) {\n // Process samples! They are big-endian floats.\n HRESULT hr;\n\n api::XAUDIO2_VOICE_STATE state;\n if (api_minor_version_ >= 8) {\n objects_.api_2_8.pcm_voice->GetState(&state,\n api::XE_XAUDIO2_VOICE_NOSAMPLESPLAYED);\n } else {\n objects_.api_2_7.pcm_voice->GetState(&state);\n }\n assert_true(state.BuffersQueued < frame_count_);\n\n auto input_frame = memory_->TranslateVirtual(frame_ptr);\n auto output_frame = reinterpret_cast(frames_[current_frame_]);\n auto interleave_channels = frame_channels_;\n\n // interleave the data\n conversion::sequential_6_BE_to_interleaved_6_LE(output_frame, input_frame,\n channel_samples_);\n\n api::XAUDIO2_BUFFER buffer;\n buffer.Flags = 0;\n buffer.pAudioData = reinterpret_cast(output_frame);\n buffer.AudioBytes = frame_size_;\n buffer.PlayBegin = 0;\n buffer.PlayLength = channel_samples_;\n buffer.LoopBegin = api::XE_XAUDIO2_NO_LOOP_REGION;\n buffer.LoopLength = 0;\n buffer.LoopCount = 0;\n buffer.pContext = 0;\n if (api_minor_version_ >= 8) {\n hr = objects_.api_2_8.pcm_voice->SubmitSourceBuffer(&buffer);\n } else {\n hr = objects_.api_2_7.pcm_voice->SubmitSourceBuffer(&buffer);\n }\n if (FAILED(hr)) {\n XELOGE(\"SubmitSourceBuffer failed with 0x{:08X}\", hr);\n return;\n }\n\n current_frame_ = (current_frame_ + 1) % frame_count_;\n\n // Update playback ratio to our time scalar.\n // This will keep audio in sync with the game clock.\n float frequency_ratio = static_cast(xe::Clock::guest_time_scalar());\n if (api_minor_version_ >= 8) {\n objects_.api_2_8.pcm_voice->SetFrequencyRatio(frequency_ratio);\n } else {\n objects_.api_2_7.pcm_voice->SetFrequencyRatio(frequency_ratio);\n }\n}\n\nvoid XAudio2AudioDriver::Shutdown() {\n // XAudio2 lifecycle is managed by the MTA thread.\n if (mta_thread_.joinable()) {\n {\n std::unique_lock mta_thread_shutdown_request_lock(\n mta_thread_shutdown_request_mutex_);\n mta_thread_shutdown_requested_ = true;\n }\n mta_thread_shutdown_request_cond_.notify_all();\n mta_thread_.join();\n }\n\n xaudio2_create_ = nullptr;\n if (xaudio2_module_) {\n FreeLibrary(static_cast(xaudio2_module_));\n xaudio2_module_ = nullptr;\n }\n\n if (voice_callback_) {\n delete voice_callback_;\n voice_callback_ = nullptr;\n }\n}\n\ntemplate \nvoid XAudio2AudioDriver::ShutdownObjects(Objects& objects) {\n if (objects.pcm_voice) {\n objects.pcm_voice->Stop();\n objects.pcm_voice->DestroyVoice();\n objects.pcm_voice = nullptr;\n }\n\n if (objects.mastering_voice) {\n objects.mastering_voice->DestroyVoice();\n objects.mastering_voice = nullptr;\n }\n\n if (objects.audio) {\n objects.audio->StopEngine();\n objects.audio->Release();\n objects.audio = nullptr;\n }\n}\n\nvoid XAudio2AudioDriver::MTAThread() {\n xe::threading::set_name(\"XAudio2 MTA\");\n\n assert_false(mta_thread_initialization_attempt_completed_);\n\n bool initialized = false;\n\n // Initializing MTA COM in this thread, as well making other (guest) threads\n // that don't explicitly call CoInitializeEx implicitly MTA for the period of\n // time when they can interact with XAudio through the XAudio2AudioDriver,\n // until the CoUninitialize (to be more precise, the CoUninitialize for the\n // last remaining MTA thread, but we need implicit MTA for the audio here).\n // https://devblogs.microsoft.com/oldnewthing/?p=4613\n HRESULT hr = CoInitializeEx(nullptr, COINIT_MULTITHREADED);\n bool com_initialized = SUCCEEDED(hr);\n if (!com_initialized) {\n XELOGE(\"XAudio2 MTA thread CoInitializeEx failed with 0x{:08X}\", hr);\n } else {\n if (api_minor_version_ >= 8) {\n hr = xaudio2_create_(&objects_.api_2_8.audio, 0, kProcessor);\n if (FAILED(hr)) {\n XELOGE(\"XAudio2Create failed with 0x{:08X}\", hr);\n } else {\n initialized = InitializeObjects(objects_.api_2_8);\n }\n } else {\n hr = CoCreateInstance(__uuidof(api::XAudio2_7), nullptr,\n CLSCTX_INPROC_SERVER,\n IID_PPV_ARGS(&objects_.api_2_7.audio));\n if (FAILED(hr)) {\n XELOGE(\"CoCreateInstance for XAudio2 failed with 0x{:08X}\", hr);\n } else {\n hr = objects_.api_2_7.audio->Initialize(0, kProcessor);\n if (FAILED(hr)) {\n XELOGE(\"IXAudio2::Initialize failed with 0x{:08X}\", hr);\n } else {\n initialized = InitializeObjects(objects_.api_2_7);\n }\n }\n }\n }\n\n // Notify the threads waiting for the initialization of the result.\n mta_thread_initialization_completion_result_ = initialized;\n {\n std::unique_lock mta_thread_initialization_completion_lock(\n mta_thread_initialization_completion_mutex_);\n mta_thread_initialization_attempt_completed_ = true;\n }\n mta_thread_initialization_completion_cond_.notify_all();\n\n if (initialized) {\n // Initialized successfully, await a shutdown request while keeping an\n // implicit COM MTA scope.\n {\n std::unique_lock mta_thread_shutdown_request_lock(\n mta_thread_shutdown_request_mutex_);\n while (true) {\n if (mta_thread_shutdown_requested_) {\n break;\n }\n mta_thread_shutdown_request_cond_.wait(\n mta_thread_shutdown_request_lock);\n }\n }\n\n if (api_minor_version_ >= 8) {\n ShutdownObjects(objects_.api_2_8);\n } else {\n ShutdownObjects(objects_.api_2_7);\n }\n }\n\n if (com_initialized) {\n CoUninitialize();\n }\n}\n\n} // namespace xaudio2\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/xaudio2/xaudio2_audio_driver.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_XAUDIO2_XAUDIO2_AUDIO_DRIVER_H_\n#define XENIA_APU_XAUDIO2_XAUDIO2_AUDIO_DRIVER_H_\n\n#include \n#include \n#include \n\n#include \"xenia/apu/audio_driver.h\"\n#include \"xenia/apu/xaudio2/xaudio2_api.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/threading.h\"\n\nstruct IXAudio2;\nstruct IXAudio2MasteringVoice;\nstruct IXAudio2SourceVoice;\n\nnamespace xe {\nnamespace apu {\nnamespace xaudio2 {\n\nclass XAudio2AudioDriver : public AudioDriver {\n public:\n XAudio2AudioDriver(Memory* memory, xe::threading::Semaphore* semaphore);\n ~XAudio2AudioDriver() override;\n\n bool Initialize();\n // Must not be called from COM STA threads as MTA XAudio2 will be used. It's\n // fine to call this from threads that have never initialized COM as\n // initializing MTA for any thread implicitly initializes it for all threads\n // not explicitly requesting STA (until COM is uninitialized all threads that\n // have initialized MTA).\n // https://devblogs.microsoft.com/oldnewthing/?p=4613\n void SubmitFrame(uint32_t frame_ptr) override;\n void Shutdown();\n\n private:\n // First CPU (2.8 default). XAUDIO2_ANY_PROCESSOR (2.7 default) steals too\n // much time from other things. Ideally should process audio on what roughly\n // represents thread 4 (5th) on the Xbox 360 (2.7 default on the console), or\n // even beyond the 6 guest cores.\n api::XAUDIO2_PROCESSOR kProcessor = 0x00000001;\n\n // InitializeObjects and ShutdownObjects must be called only in the lifecycle\n // management thread with COM MTA initialized.\n template \n bool InitializeObjects(Objects& objects);\n template \n void ShutdownObjects(Objects& objects);\n\n void MTAThread();\n\n void* xaudio2_module_ = nullptr;\n // clang-format off\n HRESULT (__stdcall* xaudio2_create_)(\n api::IXAudio2_8** xaudio2_out, UINT32 flags,\n api::XAUDIO2_PROCESSOR xaudio2_processor) = nullptr;\n // clang-format on\n uint32_t api_minor_version_ = 7;\n\n bool mta_thread_initialization_completion_result_;\n std::mutex mta_thread_initialization_completion_mutex_;\n std::condition_variable mta_thread_initialization_completion_cond_;\n bool mta_thread_initialization_attempt_completed_;\n\n std::mutex mta_thread_shutdown_request_mutex_;\n std::condition_variable mta_thread_shutdown_request_cond_;\n bool mta_thread_shutdown_requested_;\n\n std::thread mta_thread_;\n\n union {\n struct {\n api::IXAudio2_7* audio;\n api::IXAudio2_7MasteringVoice* mastering_voice;\n api::IXAudio2_7SourceVoice* pcm_voice;\n } api_2_7;\n struct {\n api::IXAudio2_8* audio;\n api::IXAudio2_8MasteringVoice* mastering_voice;\n api::IXAudio2_8SourceVoice* pcm_voice;\n } api_2_8;\n } objects_ = {};\n xe::threading::Semaphore* semaphore_ = nullptr;\n\n class VoiceCallback;\n VoiceCallback* voice_callback_ = nullptr;\n\n static const uint32_t frame_count_ = api::XE_XAUDIO2_MAX_QUEUED_BUFFERS;\n static const uint32_t frame_channels_ = 6;\n static const uint32_t channel_samples_ = 256;\n static const uint32_t frame_samples_ = frame_channels_ * channel_samples_;\n static const uint32_t frame_size_ = sizeof(float) * frame_samples_;\n float frames_[frame_count_][frame_samples_];\n uint32_t current_frame_ = 0;\n};\n\n} // namespace xaudio2\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_XAUDIO2_XAUDIO2_AUDIO_DRIVER_H_\n" }, { "path": "src/xenia/apu/xaudio2/xaudio2_audio_system.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/xaudio2/xaudio2_audio_system.h\"\n\n#include \"xenia/apu/apu_flags.h\"\n#include \"xenia/apu/xaudio2/xaudio2_audio_driver.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace xaudio2 {\n\nstd::unique_ptr XAudio2AudioSystem::Create(\n cpu::Processor* processor) {\n return std::make_unique(processor);\n}\n\nXAudio2AudioSystem::XAudio2AudioSystem(cpu::Processor* processor)\n : AudioSystem(processor) {}\n\nXAudio2AudioSystem::~XAudio2AudioSystem() {}\n\nvoid XAudio2AudioSystem::Initialize() { AudioSystem::Initialize(); }\n\nX_STATUS XAudio2AudioSystem::CreateDriver(size_t index,\n xe::threading::Semaphore* semaphore,\n AudioDriver** out_driver) {\n assert_not_null(out_driver);\n auto driver = new XAudio2AudioDriver(memory_, semaphore);\n if (!driver->Initialize()) {\n driver->Shutdown();\n return X_STATUS_UNSUCCESSFUL;\n }\n\n *out_driver = driver;\n return X_STATUS_SUCCESS;\n}\n\nvoid XAudio2AudioSystem::DestroyDriver(AudioDriver* driver) {\n assert_not_null(driver);\n auto xdriver = static_cast(driver);\n xdriver->Shutdown();\n assert_not_null(xdriver);\n delete xdriver;\n}\n\n} // namespace xaudio2\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/xaudio2/xaudio2_audio_system.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_XAUDIO2_XAUDIO2_AUDIO_SYSTEM_H_\n#define XENIA_APU_XAUDIO2_XAUDIO2_AUDIO_SYSTEM_H_\n\n#include \"xenia/apu/audio_system.h\"\n\nnamespace xe {\nnamespace apu {\nnamespace xaudio2 {\n\nclass XAudio2AudioSystem : public AudioSystem {\n public:\n explicit XAudio2AudioSystem(cpu::Processor* processor);\n ~XAudio2AudioSystem() override;\n\n static bool IsAvailable() { return true; }\n\n static std::unique_ptr Create(cpu::Processor* processor);\n\n X_RESULT CreateDriver(size_t index, xe::threading::Semaphore* semaphore,\n AudioDriver** out_driver) override;\n void DestroyDriver(AudioDriver* driver) override;\n\n protected:\n void Initialize() override;\n};\n\n} // namespace xaudio2\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_XAUDIO2_XAUDIO2_AUDIO_SYSTEM_H_\n" }, { "path": "src/xenia/apu/xma_context.cc", "content": "/**\n******************************************************************************\n* Xenia : Xbox 360 Emulator Research Project *\n******************************************************************************\n* Copyright 2021 Ben Vanik. All rights reserved. *\n* Released under the BSD license - see LICENSE in the root for more details. *\n******************************************************************************\n*/\n\n#include \"xenia/apu/xma_context.h\"\n\n#include \n#include \n\n#include \"xenia/apu/xma_decoder.h\"\n#include \"xenia/apu/xma_helpers.h\"\n#include \"xenia/base/bit_stream.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/base/ring_buffer.h\"\n\nextern \"C\" {\n#if XE_COMPILER_MSVC\n#pragma warning(push)\n#pragma warning(disable : 4101 4244 5033)\n#endif\n#include \"third_party/FFmpeg/libavcodec/avcodec.h\"\n#if XE_COMPILER_MSVC\n#pragma warning(pop)\n#endif\n} // extern \"C\"\n\n// Credits for most of this code goes to:\n// https://github.com/koolkdev/libertyv/blob/master/libav_wrapper/xma2dec.c\n\nnamespace xe {\nnamespace apu {\n\nXmaContext::XmaContext() = default;\n\nXmaContext::~XmaContext() {\n if (av_context_) {\n if (avcodec_is_open(av_context_)) {\n avcodec_close(av_context_);\n }\n av_free(av_context_);\n }\n if (av_frame_) {\n av_frame_free(&av_frame_);\n }\n // if (current_frame_) {\n // delete[] current_frame_;\n // }\n}\n\nint XmaContext::Setup(uint32_t id, Memory* memory, uint32_t guest_ptr) {\n id_ = id;\n memory_ = memory;\n guest_ptr_ = guest_ptr;\n\n // Allocate ffmpeg stuff:\n av_packet_ = av_packet_alloc();\n assert_not_null(av_packet_);\n\n // find the XMA2 audio decoder\n av_codec_ = avcodec_find_decoder(AV_CODEC_ID_XMAFRAMES);\n if (!av_codec_) {\n XELOGE(\"XmaContext {}: Codec not found\", id);\n return 1;\n }\n\n av_context_ = avcodec_alloc_context3(av_codec_);\n if (!av_context_) {\n XELOGE(\"XmaContext {}: Couldn't allocate context\", id);\n return 1;\n }\n\n // Initialize these to 0. They'll actually be set later.\n av_context_->channels = 0;\n av_context_->sample_rate = 0;\n\n av_frame_ = av_frame_alloc();\n if (!av_frame_) {\n XELOGE(\"XmaContext {}: Couldn't allocate frame\", id);\n return 1;\n }\n\n // FYI: We're purposely not opening the codec here. That is done later.\n return 0;\n}\n\nbool XmaContext::Work() {\n std::lock_guard lock(lock_);\n if (!is_allocated() || !is_enabled()) {\n return false;\n }\n\n set_is_enabled(false);\n\n auto context_ptr = memory()->TranslateVirtual(guest_ptr());\n XMA_CONTEXT_DATA data(context_ptr);\n Decode(&data);\n data.Store(context_ptr);\n return true;\n}\n\nvoid XmaContext::Enable() {\n std::lock_guard lock(lock_);\n\n auto context_ptr = memory()->TranslateVirtual(guest_ptr());\n XMA_CONTEXT_DATA data(context_ptr);\n\n XELOGAPU(\"XmaContext: kicking context {} (buffer {} {}/{} bits)\", id(),\n data.current_buffer, data.input_buffer_read_offset,\n (data.current_buffer == 0 ? data.input_buffer_0_packet_count\n : data.input_buffer_1_packet_count) *\n kBitsPerPacket);\n\n data.Store(context_ptr);\n\n set_is_enabled(true);\n}\n\nbool XmaContext::Block(bool poll) {\n if (!lock_.try_lock()) {\n if (poll) {\n return false;\n }\n lock_.lock();\n }\n lock_.unlock();\n return true;\n}\n\nvoid XmaContext::Clear() {\n std::lock_guard lock(lock_);\n XELOGAPU(\"XmaContext: reset context {}\", id());\n\n auto context_ptr = memory()->TranslateVirtual(guest_ptr());\n XMA_CONTEXT_DATA data(context_ptr);\n\n data.input_buffer_0_valid = 0;\n data.input_buffer_1_valid = 0;\n data.output_buffer_valid = 0;\n\n data.output_buffer_read_offset = 0;\n data.output_buffer_write_offset = 0;\n\n data.Store(context_ptr);\n}\n\nvoid XmaContext::Disable() {\n std::lock_guard lock(lock_);\n XELOGAPU(\"XmaContext: disabling context {}\", id());\n set_is_enabled(false);\n}\n\nvoid XmaContext::Release() {\n // Lock it in case the decoder thread is working on it now.\n std::lock_guard lock(lock_);\n assert_true(is_allocated_ == true);\n\n set_is_allocated(false);\n auto context_ptr = memory()->TranslateVirtual(guest_ptr());\n std::memset(context_ptr, 0, sizeof(XMA_CONTEXT_DATA)); // Zero it.\n}\n\nvoid XmaContext::SwapInputBuffer(XMA_CONTEXT_DATA* data) {\n // No more frames.\n if (data->current_buffer == 0) {\n data->input_buffer_0_valid = 0;\n } else {\n data->input_buffer_1_valid = 0;\n }\n data->current_buffer ^= 1;\n data->input_buffer_read_offset = 0;\n}\n\nbool XmaContext::TrySetupNextLoop(XMA_CONTEXT_DATA* data,\n bool ignore_input_buffer_offset) {\n // Setup the input buffer offset if next loop exists.\n // TODO(Pseudo-Kernel): Need to handle loop in the following cases.\n // 1. loop_start == loop_end == 0\n // 2. loop_start > loop_end && loop_count > 0\n if (data->loop_count > 0 && data->loop_start < data->loop_end &&\n (ignore_input_buffer_offset ||\n data->input_buffer_read_offset >= data->loop_end)) {\n // Loop back to the beginning.\n data->input_buffer_read_offset = data->loop_start;\n if (data->loop_count < 255) {\n data->loop_count--;\n }\n return true;\n }\n return false;\n}\n\n/*\nvoid XmaContext::NextPacket(\n uint8_t* input_buffer,\n uint32_t input_size,\n uint32_t input_buffer_read_offset) {\n*/\nvoid XmaContext::NextPacket(XMA_CONTEXT_DATA* data) {\n // auto packet_idx = GetFramePacketNumber(input_buffer, input_size,\n // input_buffer_read_offset);\n\n // packet_idx++;\n // if (packet_idx++ >= input_size)\n}\n\nint XmaContext::GetSampleRate(int id) {\n switch (id) {\n case 0:\n return 24000;\n case 1:\n return 32000;\n case 2:\n return 44100;\n case 3:\n return 48000;\n }\n assert_always();\n return 0;\n}\n\nbool XmaContext::ValidFrameOffset(uint8_t* block, size_t size_bytes,\n size_t frame_offset_bits) {\n uint32_t packet_num =\n GetFramePacketNumber(block, size_bytes, frame_offset_bits);\n if (packet_num == -1) {\n // Invalid packet number\n return false;\n }\n\n uint8_t* packet = block + (packet_num * kBytesPerPacket);\n size_t relative_offset_bits = frame_offset_bits % kBitsPerPacket;\n\n uint32_t first_frame_offset = xma::GetPacketFrameOffset(packet);\n if (first_frame_offset == -1 || first_frame_offset > kBitsPerPacket) {\n // Packet only contains a partial frame, so no frames can start here.\n return false;\n }\n\n BitStream stream(packet, kBitsPerPacket);\n stream.SetOffset(first_frame_offset);\n while (true) {\n if (stream.offset_bits() == relative_offset_bits) {\n return true;\n }\n\n if (stream.BitsRemaining() < 15) {\n // Not enough room for another frame header.\n return false;\n }\n\n uint64_t size = stream.Read(15);\n if ((size - 15) > stream.BitsRemaining()) {\n // Last frame.\n return false;\n } else if (size == 0x7FFF) {\n // Invalid frame (and last of this packet)\n return false;\n }\n\n stream.Advance(size - 16);\n\n // Read the trailing bit to see if frames follow\n if (stream.Read(1) == 0) {\n break;\n }\n }\n\n return false;\n}\n\nstatic void dump_raw(AVFrame* frame, int id) {\n FILE* outfile = fopen(fmt::format(\"out{}.raw\", id).c_str(), \"ab\");\n if (!outfile) {\n return;\n }\n size_t data_size = sizeof(float);\n for (int i = 0; i < frame->nb_samples; i++) {\n for (int ch = 0; ch < frame->channels; ch++) {\n fwrite(frame->data[ch] + data_size * i, 1, data_size, outfile);\n }\n }\n fclose(outfile);\n}\n\nvoid XmaContext::Decode(XMA_CONTEXT_DATA* data) {\n SCOPE_profile_cpu_f(\"apu\");\n\n // What I see:\n // XMA outputs 2 bytes per sample\n // 512 samples per frame (128 per subframe)\n // Max output size is data.output_buffer_block_count * 256\n\n // This decoder is fed packets (max 4095 per buffer)\n // Packets contain \"some\" frames\n // 32bit header (big endian)\n\n // Frames are the smallest thing the SPUs can decode.\n // They can and usually will span packets.\n\n // Sample rates (data.sample_rate):\n // 0 - 24 kHz\n // 1 - 32 kHz\n // 2 - 44.1 kHz\n // 3 - 48 kHz\n\n // SPUs also support stereo decoding. (data.is_stereo)\n\n // Check the output buffer - we cannot decode anything else if it's\n // unavailable.\n if (!data->output_buffer_valid) {\n return;\n }\n\n // No available data.\n if (!data->input_buffer_0_valid && !data->input_buffer_1_valid) {\n data->output_buffer_valid = 0;\n return;\n }\n\n // XAudio Loops\n // loop_count:\n // - XAUDIO2_MAX_LOOP_COUNT = 254\n // - XAUDIO2_LOOP_INFINITE = 255\n // loop_start/loop_end are bit offsets to a specific frame\n\n // Translate pointers for future use.\n // Sometimes the game will use rolling input buffers. If they do, we cannot\n // assume they form a complete block! In addition, the buffers DO NOT have\n // to be contiguous!\n uint8_t* in0 = data->input_buffer_0_valid\n ? memory()->TranslatePhysical(data->input_buffer_0_ptr)\n : nullptr;\n uint8_t* in1 = data->input_buffer_1_valid\n ? memory()->TranslatePhysical(data->input_buffer_1_ptr)\n : nullptr;\n uint8_t* current_input_buffer = data->current_buffer ? in1 : in0;\n\n XELOGAPU(\"Processing context {} (offset {}, buffer {}, ptr {:p})\", id(),\n data->input_buffer_read_offset, data->current_buffer,\n current_input_buffer);\n\n size_t input_buffer_0_size =\n data->input_buffer_0_packet_count * kBytesPerPacket;\n size_t input_buffer_1_size =\n data->input_buffer_1_packet_count * kBytesPerPacket;\n size_t input_total_size = input_buffer_0_size + input_buffer_1_size;\n\n size_t current_input_size =\n data->current_buffer ? input_buffer_1_size : input_buffer_0_size;\n size_t current_input_packet_count = current_input_size / kBytesPerPacket;\n\n // Output buffers are in raw PCM samples, 256 bytes per block.\n // Output buffer is a ring buffer. We need to write from the write offset\n // to the read offset.\n uint8_t* output_buffer = memory()->TranslatePhysical(data->output_buffer_ptr);\n uint32_t output_capacity =\n data->output_buffer_block_count * kBytesPerSubframeChannel;\n uint32_t output_read_offset =\n data->output_buffer_read_offset * kBytesPerSubframeChannel;\n uint32_t output_write_offset =\n data->output_buffer_write_offset * kBytesPerSubframeChannel;\n\n RingBuffer output_rb(output_buffer, output_capacity);\n output_rb.set_read_offset(output_read_offset);\n output_rb.set_write_offset(output_write_offset);\n\n // We can only decode an entire frame and write it out at a time, so\n // don't save any samples.\n // TODO(JoelLinn): subframes when looping\n size_t output_remaining_bytes = output_rb.write_count();\n output_remaining_bytes -=\n output_remaining_bytes % (kBytesPerFrameChannel << data->is_stereo);\n\n // is_dirty_ = true; // TODO\n // is_dirty_ = false; // TODO\n assert_false(data->stop_when_done);\n assert_false(data->interrupt_when_done);\n static int total_samples = 0;\n bool reuse_input_buffer = false;\n // Decode until we can't write any more data.\n while (output_remaining_bytes > 0) {\n if (!data->input_buffer_0_valid && !data->input_buffer_1_valid) {\n // Out of data.\n break;\n }\n\n // Setup the input buffer if we are at loop_end.\n // The input buffer must not be swapped out until all loops are processed.\n reuse_input_buffer = TrySetupNextLoop(data, false);\n\n // assert_true(packets_skip_ == 0);\n // assert_true(split_frame_len_ == 0);\n // assert_true(split_frame_len_partial_ == 0);\n\n // Where are we in the buffer (in XMA jargon)\n int packet_idx, frame_idx, frame_count;\n uint8_t* packet;\n bool frame_last_split;\n\n BitStream stream(current_input_buffer, current_input_size * 8);\n stream.SetOffset(data->input_buffer_read_offset);\n\n // if we had a buffer swap try to skip packets first\n if (packets_skip_ > 0) {\n packet_idx =\n GetFramePacketNumber(current_input_buffer, current_input_size,\n data->input_buffer_read_offset);\n while (packets_skip_ > 0) {\n packets_skip_--;\n packet_idx++;\n if (packet_idx >= current_input_packet_count) {\n if (!reuse_input_buffer) {\n // Last packet. Try setup once more.\n reuse_input_buffer = TrySetupNextLoop(data, true);\n }\n if (!reuse_input_buffer) {\n SwapInputBuffer(data);\n }\n return;\n }\n }\n // invalid frame pointer but needed for us\n data->input_buffer_read_offset = packet_idx * kBitsPerPacket;\n // continue;\n }\n\n if (split_frame_len_) {\n // handle a frame that was split over two packages\n packet_idx =\n GetFramePacketNumber(current_input_buffer, current_input_size,\n data->input_buffer_read_offset);\n packet = current_input_buffer + packet_idx * kBytesPerPacket;\n std::tie(frame_count, frame_last_split) = GetPacketFrameCount(packet);\n frame_idx = -1;\n\n stream =\n BitStream(current_input_buffer, (packet_idx + 1) * kBitsPerPacket);\n stream.SetOffset(packet_idx * kBitsPerPacket + 32);\n\n if (split_frame_len_ > xma::kMaxFrameLength) {\n // TODO write CopyPeekMethod\n auto offset = stream.offset_bits();\n stream.Copy(\n xma_frame_.data() + 1 +\n ((split_frame_len_partial_ + split_frame_padding_start_) / 8),\n 15 - split_frame_len_partial_);\n stream.SetOffset(offset);\n BitStream slen(xma_frame_.data() + 1, 15 + split_frame_padding_start_);\n slen.Advance(split_frame_padding_start_);\n split_frame_len_ = static_cast(slen.Read(15));\n }\n\n if (frame_count > 0) {\n assert_true(xma::GetPacketFrameOffset(packet) - 32 ==\n split_frame_len_ - split_frame_len_partial_);\n }\n\n auto offset = stream.Copy(\n xma_frame_.data() + 1 +\n ((split_frame_len_partial_ + split_frame_padding_start_) / 8),\n split_frame_len_ - split_frame_len_partial_);\n assert_true(offset ==\n (split_frame_padding_start_ + split_frame_len_partial_) % 8);\n } else {\n if (data->input_buffer_read_offset % kBitsPerPacket == 0) {\n // Invalid offset. Go ahead and set it.\n int packet_number =\n GetFramePacketNumber(current_input_buffer, current_input_size,\n data->input_buffer_read_offset);\n\n if (packet_number == -1) {\n return;\n }\n\n auto offset =\n xma::GetPacketFrameOffset(current_input_buffer +\n kBytesPerPacket * packet_number) +\n data->input_buffer_read_offset;\n if (offset == -1) {\n // No more frames.\n SwapInputBuffer(data);\n // TODO partial frames? end?\n XELOGE(\"XmaContext {}: TODO partial frames? end?\", id());\n assert_always(\"TODO\");\n return;\n } else {\n data->input_buffer_read_offset = offset;\n }\n }\n\n if (!ValidFrameOffset(current_input_buffer, current_input_size,\n data->input_buffer_read_offset)) {\n XELOGAPU(\"XmaContext {}: Invalid read offset {}!\", id(),\n data->input_buffer_read_offset);\n SwapInputBuffer(data);\n return;\n }\n\n // Where are we in the buffer (in XMA jargon)\n std::tie(packet_idx, frame_idx) =\n GetFrameNumber(current_input_buffer, current_input_size,\n data->input_buffer_read_offset);\n // TODO handle\n assert_true(packet_idx >= 0);\n assert_true(frame_idx >= 0);\n packet = current_input_buffer + packet_idx * kBytesPerPacket;\n // frames that belong to this packet\n std::tie(frame_count, frame_last_split) = GetPacketFrameCount(packet);\n assert_true(frame_count >= 0); // TODO end\n\n PrepareDecoder(packet, data->sample_rate, bool(data->is_stereo));\n\n // Current frame is split to next packet:\n bool frame_is_split = frame_last_split && (frame_idx >= frame_count - 1);\n\n stream =\n BitStream(current_input_buffer, (packet_idx + 1) * kBitsPerPacket);\n stream.SetOffset(data->input_buffer_read_offset);\n // int frame_len;\n // int frame_len_partial\n split_frame_len_partial_ = static_cast(stream.BitsRemaining());\n if (split_frame_len_partial_ >= 15) {\n split_frame_len_ = static_cast(stream.Peek(15));\n } else {\n // assert_always();\n split_frame_len_ = xma::kMaxFrameLength + 1;\n }\n assert_true(frame_is_split ==\n (split_frame_len_ > split_frame_len_partial_));\n\n // TODO fix bitstream copy\n std::memset(xma_frame_.data(), 0, xma_frame_.size());\n\n {\n auto offset =\n stream.Copy(xma_frame_.data() + 1,\n std::min(split_frame_len_, split_frame_len_partial_));\n assert_true(offset < 8);\n split_frame_padding_start_ = static_cast(offset);\n }\n\n if (frame_is_split) {\n // go to next xma packet of this stream\n packets_skip_ = xma::GetPacketSkipCount(packet) + 1;\n while (packets_skip_ > 0) {\n packets_skip_--;\n packet += kBytesPerPacket;\n packet_idx++;\n if (packet_idx >= current_input_packet_count) {\n if (!reuse_input_buffer) {\n // Last packet. Try setup once more.\n reuse_input_buffer = TrySetupNextLoop(data, true);\n }\n if (!reuse_input_buffer) {\n SwapInputBuffer(data);\n }\n return;\n }\n }\n // TODO guest might read this:\n data->input_buffer_read_offset = packet_idx * kBitsPerPacket;\n continue;\n }\n }\n\n av_packet_->data = xma_frame_.data();\n av_packet_->size = static_cast(\n 1 + ((split_frame_padding_start_ + split_frame_len_) / 8) +\n (((split_frame_padding_start_ + split_frame_len_) % 8) ? 1 : 0));\n\n auto padding_end = av_packet_->size * 8 -\n (8 + split_frame_padding_start_ + split_frame_len_);\n assert_true(padding_end < 8);\n xma_frame_[0] =\n ((split_frame_padding_start_ & 7) << 5) | ((padding_end & 7) << 2);\n\n split_frame_len_ = 0;\n split_frame_len_partial_ = 0;\n split_frame_padding_start_ = 0;\n\n auto ret = avcodec_send_packet(av_context_, av_packet_);\n if (ret < 0) {\n XELOGE(\"XmaContext {}: Error sending packet for decoding\", id());\n // TODO bail out\n assert_always();\n }\n ret = avcodec_receive_frame(av_context_, av_frame_);\n /*\n if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF)\n // TODO AVERROR_EOF???\n break;\n else\n */\n if (ret < 0) {\n XELOGE(\"XmaContext {}: Error during decoding\", id());\n assert_always();\n return; // TODO bail out\n }\n assert_true(ret == 0);\n\n {\n // copy over 1 frame\n // update input buffer read offset\n\n // assert(decoded_consumed_samples_ + kSamplesPerFrame <=\n // current_frame_.size());\n assert_true(av_context_->sample_fmt == AV_SAMPLE_FMT_FLTP);\n // assert_true(frame_is_split == (frame_idx == -1));\n\n //\t\t\tdump_raw(av_frame_, id());\n ConvertFrame((const uint8_t**)av_frame_->data, bool(data->is_stereo),\n raw_frame_.data());\n // decoded_consumed_samples_ += kSamplesPerFrame;\n\n auto byte_count = kBytesPerFrameChannel << data->is_stereo;\n assert_true(output_remaining_bytes >= byte_count);\n output_rb.Write(raw_frame_.data(), byte_count);\n output_remaining_bytes -= byte_count;\n data->output_buffer_write_offset = output_rb.write_offset() / 256;\n\n total_samples += id_ == 0 ? kSamplesPerFrame : 0;\n\n uint32_t offset = data->input_buffer_read_offset;\n // if (offset % (kBytesPerSample * 8) == 0) {\n // offset = xma::GetPacketFrameOffset(packet);\n //}\n offset = static_cast(\n GetNextFrame(current_input_buffer, current_input_size, offset));\n // assert_true((offset == 0) ==\n // (frame_is_split || (frame_idx + 1 >= frame_count)));\n if (frame_idx + 1 >= frame_count) {\n // Skip to next packet (no split frame)\n packets_skip_ = xma::GetPacketSkipCount(packet) + 1;\n while (packets_skip_ > 0) {\n packets_skip_--;\n packet_idx++;\n if (packet_idx >= current_input_packet_count) {\n if (!reuse_input_buffer) {\n // Last packet. Try setup once more.\n reuse_input_buffer = TrySetupNextLoop(data, true);\n }\n if (!reuse_input_buffer) {\n SwapInputBuffer(data);\n }\n return;\n }\n }\n packet = current_input_buffer + packet_idx * kBytesPerPacket;\n offset =\n xma::GetPacketFrameOffset(packet) + packet_idx * kBitsPerPacket;\n }\n if (offset == 0 || frame_idx == -1) {\n // Next packet but we already skipped to it\n if (packet_idx >= current_input_packet_count) {\n // Buffer is fully used\n if (!reuse_input_buffer) {\n // Last packet. Try setup once more.\n reuse_input_buffer = TrySetupNextLoop(data, true);\n }\n if (!reuse_input_buffer) {\n SwapInputBuffer(data);\n }\n break;\n }\n offset =\n xma::GetPacketFrameOffset(packet) + packet_idx * kBitsPerPacket;\n }\n // TODO buffer bounds check\n assert_true(data->input_buffer_read_offset < offset);\n data->input_buffer_read_offset = offset;\n }\n }\n\n // assert_true((split_frame_len_ != 0) == (data->input_buffer_read_offset ==\n // 0));\n\n // The game will kick us again with a new output buffer later.\n // It's important that we only invalidate this if we actually wrote to it!!\n if (output_rb.write_offset() == output_rb.read_offset()) {\n data->output_buffer_valid = 0;\n }\n}\n\nsize_t XmaContext::GetNextFrame(uint8_t* block, size_t size,\n size_t bit_offset) {\n // offset = xma::GetPacketFrameOffset(packet);\n // TODO meh\n // auto next_packet = bit_offset - bit_offset % kBitsPerPacket +\n // kBitsPerPacket;\n auto packet_idx = GetFramePacketNumber(block, size, bit_offset);\n\n BitStream stream(block, size * 8);\n stream.SetOffset(bit_offset);\n\n if (stream.BitsRemaining() < 15) {\n return 0;\n }\n\n uint64_t len = stream.Read(15);\n if ((len - 15) > stream.BitsRemaining()) {\n assert_always(\"TODO\");\n // *bit_offset = next_packet;\n // return false;\n // return next_packet;\n return 0;\n } else if (len >= xma::kMaxFrameLength) {\n assert_always(\"TODO\");\n // *bit_offset = next_packet;\n // return false;\n return 0;\n // return next_packet;\n }\n\n stream.Advance(len - (15 + 1));\n // Read the trailing bit to see if frames follow\n if (stream.Read(1) == 0) {\n return 0;\n }\n\n bit_offset += len;\n if (packet_idx < GetFramePacketNumber(block, size, bit_offset)) {\n return 0;\n }\n return bit_offset;\n}\n\nint XmaContext::GetFramePacketNumber(uint8_t* block, size_t size,\n size_t bit_offset) {\n size *= 8;\n if (bit_offset >= size) {\n // Not good :(\n assert_always();\n return -1;\n }\n\n size_t byte_offset = bit_offset >> 3;\n size_t packet_number = byte_offset / kBytesPerPacket;\n\n return (uint32_t)packet_number;\n}\n\nstd::tuple XmaContext::GetFrameNumber(uint8_t* block, size_t size,\n size_t bit_offset) {\n auto packet_idx = GetFramePacketNumber(block, size, bit_offset);\n\n if (packet_idx < 0 || (packet_idx + 1) * kBytesPerPacket > size) {\n assert_always();\n return {packet_idx, -2};\n }\n\n if (bit_offset == 0) {\n return {packet_idx, -1};\n }\n\n uint8_t* packet = block + (packet_idx * kBytesPerPacket);\n auto first_frame_offset = xma::GetPacketFrameOffset(packet);\n BitStream stream(block, size * 8);\n stream.SetOffset(packet_idx * kBitsPerPacket + first_frame_offset);\n\n int frame_idx = 0;\n while (true) {\n if (stream.BitsRemaining() < 15) {\n break;\n }\n\n if (stream.offset_bits() == bit_offset) {\n break;\n }\n\n uint64_t size = stream.Read(15);\n if ((size - 15) > stream.BitsRemaining()) {\n // Last frame.\n break;\n } else if (size == 0x7FFF) {\n // Invalid frame (and last of this packet)\n break;\n }\n\n stream.Advance(size - (15 + 1));\n\n // Read the trailing bit to see if frames follow\n if (stream.Read(1) == 0) {\n break;\n }\n frame_idx++;\n }\n return {packet_idx, frame_idx};\n}\n\nstd::tuple XmaContext::GetPacketFrameCount(uint8_t* packet) {\n auto first_frame_offset = xma::GetPacketFrameOffset(packet);\n if (first_frame_offset > kBitsPerPacket - kBitsPerHeader) {\n // frame offset is beyond packet end\n return {0, false};\n }\n\n BitStream stream(packet, kBitsPerPacket);\n stream.SetOffset(first_frame_offset);\n int frame_count = 0;\n\n while (true) {\n frame_count++;\n if (stream.BitsRemaining() < 15) {\n return {frame_count, true};\n }\n\n uint64_t size = stream.Read(15);\n if ((size - 15) > stream.BitsRemaining()) {\n return {frame_count, true};\n } else if (size == 0x7FFF) {\n assert_always();\n return {frame_count, true};\n }\n\n stream.Advance(size - (15 + 1));\n\n if (stream.Read(1) == 0) {\n return {frame_count, false};\n }\n }\n}\n\nint XmaContext::PrepareDecoder(uint8_t* packet, int sample_rate,\n bool is_two_channel) {\n // Sanity check: Packet metadata is always 1 for XMA2/0 for XMA\n assert_true((packet[2] & 0x7) == 1 || (packet[2] & 0x7) == 0);\n\n sample_rate = GetSampleRate(sample_rate);\n\n // Re-initialize the context with new sample rate and channels.\n uint32_t channels = is_two_channel ? 2 : 1;\n if (av_context_->sample_rate != sample_rate ||\n av_context_->channels != channels) {\n // We have to reopen the codec so it'll realloc whatever data it needs.\n // TODO(DrChat): Find a better way.\n avcodec_close(av_context_);\n\n av_context_->sample_rate = sample_rate;\n av_context_->channels = channels;\n\n if (avcodec_open2(av_context_, av_codec_, NULL) < 0) {\n XELOGE(\"XmaContext: Failed to reopen FFmpeg context\");\n return -1;\n }\n return 1;\n }\n return 0;\n}\n\nvoid XmaContext::ConvertFrame(const uint8_t** samples, bool is_two_channel,\n uint8_t* output_buffer) {\n // Loop through every sample, convert and drop it into the output array.\n // If more than one channel, we need to interleave the samples from each\n // channel next to each other. Always saturate because FFmpeg output is\n // not limited to [-1, 1] (for example 1.095 as seen in 5454082B).\n constexpr float scale = (1 << 15) - 1;\n auto out = reinterpret_cast(output_buffer);\n\n // For testing of vectorized versions, stereo audio is common in 4D5307E6,\n // since the first menu frame; the intro cutscene also has more than 2\n // channels.\n#if XE_ARCH_AMD64\n static_assert(kSamplesPerFrame % 8 == 0);\n const auto in_channel_0 = reinterpret_cast(samples[0]);\n const __m128 scale_mm = _mm_set1_ps(scale);\n if (is_two_channel) {\n const auto in_channel_1 = reinterpret_cast(samples[1]);\n const __m128i shufmask =\n _mm_set_epi8(14, 15, 6, 7, 12, 13, 4, 5, 10, 11, 2, 3, 8, 9, 0, 1);\n for (uint32_t i = 0; i < kSamplesPerFrame; i += 4) {\n // Load 8 samples, 4 for each channel.\n __m128 in_mm0 = _mm_loadu_ps(&in_channel_0[i]);\n __m128 in_mm1 = _mm_loadu_ps(&in_channel_1[i]);\n // Rescale.\n in_mm0 = _mm_mul_ps(in_mm0, scale_mm);\n in_mm1 = _mm_mul_ps(in_mm1, scale_mm);\n // Cast to int32.\n __m128i out_mm0 = _mm_cvtps_epi32(in_mm0);\n __m128i out_mm1 = _mm_cvtps_epi32(in_mm1);\n // Saturated cast and pack to int16.\n __m128i out_mm = _mm_packs_epi32(out_mm0, out_mm1);\n // Interleave channels and byte swap.\n out_mm = _mm_shuffle_epi8(out_mm, shufmask);\n // Store, as [out + i * 4] movdqu.\n _mm_storeu_si128(reinterpret_cast<__m128i*>(&out[i * 2]), out_mm);\n }\n } else {\n const __m128i shufmask =\n _mm_set_epi8(14, 15, 12, 13, 10, 11, 8, 9, 6, 7, 4, 5, 2, 3, 0, 1);\n for (uint32_t i = 0; i < kSamplesPerFrame; i += 8) {\n // Load 8 samples, as [in_channel_0 + i * 4] and\n // [in_channel_0 + i * 4 + 16] movups.\n __m128 in_mm0 = _mm_loadu_ps(&in_channel_0[i]);\n __m128 in_mm1 = _mm_loadu_ps(&in_channel_0[i + 4]);\n // Rescale.\n in_mm0 = _mm_mul_ps(in_mm0, scale_mm);\n in_mm1 = _mm_mul_ps(in_mm1, scale_mm);\n // Cast to int32.\n __m128i out_mm0 = _mm_cvtps_epi32(in_mm0);\n __m128i out_mm1 = _mm_cvtps_epi32(in_mm1);\n // Saturated cast and pack to int16.\n __m128i out_mm = _mm_packs_epi32(out_mm0, out_mm1);\n // Byte swap.\n out_mm = _mm_shuffle_epi8(out_mm, shufmask);\n // Store, as [out + i * 2] movdqu.\n _mm_storeu_si128(reinterpret_cast<__m128i*>(&out[i]), out_mm);\n }\n }\n#else\n uint32_t o = 0;\n for (uint32_t i = 0; i < kSamplesPerFrame; i++) {\n for (uint32_t j = 0; j <= uint32_t(is_two_channel); j++) {\n // Select the appropriate array based on the current channel.\n auto in = reinterpret_cast(samples[j]);\n\n // Raw samples sometimes aren't within [-1, 1]\n float scaled_sample = xe::clamp_float(in[i], -1.0f, 1.0f) * scale;\n\n // Convert the sample and output it in big endian.\n auto sample = static_cast(scaled_sample);\n out[o++] = xe::byte_swap(sample);\n }\n }\n#endif\n}\n\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/xma_context.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_XMA_CONTEXT_H_\n#define XENIA_APU_XMA_CONTEXT_H_\n\n#include \n#include \n#include \n#include \n//#include \n\n#include \"xenia/memory.h\"\n#include \"xenia/xbox.h\"\n\n// XMA audio format:\n// From research, XMA appears to be based on WMA Pro with\n// a few (very slight) modifications.\n// XMA2 is fully backwards-compatible with XMA1.\n\n// Helpful resources:\n// https://github.com/koolkdev/libertyv/blob/master/libav_wrapper/xma2dec.c\n// https://hcs64.com/mboard/forum.php?showthread=14818\n// https://github.com/hrydgard/minidx9/blob/master/Include/xma2defs.h\n\n// Forward declarations\nstruct AVCodec;\nstruct AVCodecParserContext;\nstruct AVCodecContext;\nstruct AVFrame;\nstruct AVPacket;\n\nnamespace xe {\nnamespace apu {\n\n// This is stored in guest space in big-endian order.\n// We load and swap the whole thing to splat here so that we can\n// use bitfields.\n// This could be important:\n// https://www.fmod.org/questions/question/forum-15859\n// Appears to be dumped in order (for the most part)\n\nstruct XMA_CONTEXT_DATA {\n // DWORD 0\n uint32_t input_buffer_0_packet_count : 12; // XMASetInputBuffer0, number of\n // 2KB packets. Max 4095 packets.\n // These packets form a block.\n uint32_t loop_count : 8; // +12bit, XMASetLoopData NumLoops\n uint32_t input_buffer_0_valid : 1; // +20bit, XMAIsInputBuffer0Valid\n uint32_t input_buffer_1_valid : 1; // +21bit, XMAIsInputBuffer1Valid\n uint32_t output_buffer_block_count : 5; // +22bit SizeWrite 256byte blocks\n uint32_t output_buffer_write_offset : 5; // +27bit\n // XMAGetOutputBufferWriteOffset\n // AKA OffsetWrite\n\n // DWORD 1\n uint32_t input_buffer_1_packet_count : 12; // XMASetInputBuffer1, number of\n // 2KB packets. Max 4095 packets.\n // These packets form a block.\n uint32_t loop_subframe_start : 2; // +12bit, XMASetLoopData\n uint32_t loop_subframe_end : 3; // +14bit, XMASetLoopData\n uint32_t loop_subframe_skip : 3; // +17bit, XMASetLoopData might be\n // subframe_decode_count\n uint32_t subframe_decode_count : 4; // +20bit\n uint32_t subframe_skip_count : 3; // +24bit\n uint32_t sample_rate : 2; // +27bit enum of sample rates\n uint32_t is_stereo : 1; // +29bit\n uint32_t unk_dword_1_c : 1; // +30bit\n uint32_t output_buffer_valid : 1; // +31bit, XMAIsOutputBufferValid\n\n // DWORD 2\n uint32_t input_buffer_read_offset : 26; // XMAGetInputBufferReadOffset\n uint32_t unk_dword_2 : 6; // ErrorStatus/ErrorSet (?)\n\n // DWORD 3\n uint32_t loop_start : 26; // XMASetLoopData LoopStartOffset\n // frame offset in bits\n uint32_t unk_dword_3 : 6; // ? ParserErrorStatus/ParserErrorSet(?)\n\n // DWORD 4\n uint32_t loop_end : 26; // XMASetLoopData LoopEndOffset\n // frame offset in bits\n uint32_t packet_metadata : 5; // XMAGetPacketMetadata\n uint32_t current_buffer : 1; // ?\n\n // DWORD 5\n uint32_t input_buffer_0_ptr; // physical address\n // DWORD 6\n uint32_t input_buffer_1_ptr; // physical address\n // DWORD 7\n uint32_t output_buffer_ptr; // physical address\n // DWORD 8\n uint32_t work_buffer_ptr; // PtrOverlapAdd(?)\n\n // DWORD 9\n // +0bit, XMAGetOutputBufferReadOffset AKA WriteBufferOffsetRead\n uint32_t output_buffer_read_offset : 5;\n uint32_t : 25;\n uint32_t stop_when_done : 1; // +30bit\n uint32_t interrupt_when_done : 1; // +31bit\n\n // DWORD 10-15\n uint32_t unk_dwords_10_15[6]; // reserved?\n\n explicit XMA_CONTEXT_DATA(const void* ptr) {\n xe::copy_and_swap(reinterpret_cast(this),\n reinterpret_cast(ptr),\n sizeof(XMA_CONTEXT_DATA) / 4);\n }\n\n void Store(void* ptr) {\n xe::copy_and_swap(reinterpret_cast(ptr),\n reinterpret_cast(this),\n sizeof(XMA_CONTEXT_DATA) / 4);\n }\n};\nstatic_assert_size(XMA_CONTEXT_DATA, 64);\n\n#pragma pack(push, 1)\n// XMA2WAVEFORMATEX\nstruct Xma2ExtraData {\n uint8_t raw[34];\n};\nstatic_assert_size(Xma2ExtraData, 34);\n#pragma pack(pop)\n\nclass XmaContext {\n public:\n static const uint32_t kBytesPerPacket = 2048;\n static const uint32_t kBitsPerPacket = kBytesPerPacket * 8;\n static const uint32_t kBitsPerHeader = 33;\n\n static const uint32_t kBytesPerSample = 2;\n static const uint32_t kSamplesPerFrame = 512;\n static const uint32_t kSamplesPerSubframe = 128;\n static const uint32_t kBytesPerFrameChannel =\n kSamplesPerFrame * kBytesPerSample;\n static const uint32_t kBytesPerSubframeChannel =\n kSamplesPerSubframe * kBytesPerSample;\n\n // static const uint32_t kOutputBytesPerBlock = 256;\n // static const uint32_t kOutputMaxSizeBytes = 31 * kOutputBytesPerBlock;\n\n explicit XmaContext();\n ~XmaContext();\n\n int Setup(uint32_t id, Memory* memory, uint32_t guest_ptr);\n bool Work();\n\n void Enable();\n bool Block(bool poll);\n void Clear();\n void Disable();\n void Release();\n\n Memory* memory() const { return memory_; }\n\n uint32_t id() { return id_; }\n uint32_t guest_ptr() { return guest_ptr_; }\n bool is_allocated() { return is_allocated_; }\n bool is_enabled() { return is_enabled_; }\n\n void set_is_allocated(bool is_allocated) { is_allocated_ = is_allocated; }\n void set_is_enabled(bool is_enabled) { is_enabled_ = is_enabled; }\n\n private:\n static void SwapInputBuffer(XMA_CONTEXT_DATA* data);\n static bool TrySetupNextLoop(XMA_CONTEXT_DATA* data,\n bool ignore_input_buffer_offset);\n static void NextPacket(XMA_CONTEXT_DATA* data);\n static int GetSampleRate(int id);\n // Get the offset of the next frame. Does not traverse packets.\n static size_t GetNextFrame(uint8_t* block, size_t size, size_t bit_offset);\n // Get the containing packet number of the frame pointed to by the offset.\n static int GetFramePacketNumber(uint8_t* block, size_t size,\n size_t bit_offset);\n // Get the packet number and the index of the frame inside that packet\n static std::tuple GetFrameNumber(uint8_t* block, size_t size,\n size_t bit_offset);\n // Get the number of frames contained in the packet (including truncated) and\n // if the last frame is split.\n static std::tuple GetPacketFrameCount(uint8_t* packet);\n\n // Convert sample format and swap bytes\n static void ConvertFrame(const uint8_t** samples, bool is_two_channel,\n uint8_t* output_buffer);\n\n bool ValidFrameOffset(uint8_t* block, size_t size_bytes,\n size_t frame_offset_bits);\n void Decode(XMA_CONTEXT_DATA* data);\n int PrepareDecoder(uint8_t* packet, int sample_rate, bool is_two_channel);\n\n Memory* memory_ = nullptr;\n\n uint32_t id_ = 0;\n uint32_t guest_ptr_ = 0;\n std::mutex lock_;\n bool is_allocated_ = false;\n bool is_enabled_ = false;\n // bool is_dirty_ = true;\n\n // ffmpeg structures\n AVPacket* av_packet_ = nullptr;\n AVCodec* av_codec_ = nullptr;\n AVCodecContext* av_context_ = nullptr;\n AVFrame* av_frame_ = nullptr;\n // uint32_t decoded_consumed_samples_ = 0; // TODO do this dynamically\n // int decoded_idx_ = -1;\n\n // bool partial_frame_saved_ = false;\n // bool partial_frame_size_known_ = false;\n // size_t partial_frame_total_size_bits_ = 0;\n // size_t partial_frame_start_offset_bits_ = 0;\n // size_t partial_frame_offset_bits_ = 0; // blah internal don't use this\n // std::vector partial_frame_buffer_;\n uint32_t packets_skip_ = 0;\n\n // bool split_frame_pending_ = false;\n uint32_t split_frame_len_ = 0;\n uint32_t split_frame_len_partial_ = 0;\n uint8_t split_frame_padding_start_ = 0;\n // first byte contains bit offset information\n std::array xma_frame_;\n\n // uint8_t* current_frame_ = nullptr;\n // conversion buffer for 2 channel frame\n std::array raw_frame_;\n // std::vector current_frame_ = std::vector(0);\n};\n\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_XMA_CONTEXT_H_\n" }, { "path": "src/xenia/apu/xma_decoder.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/xma_decoder.h\"\n\n#include \"xenia/apu/xma_context.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/base/ring_buffer.h\"\n#include \"xenia/base/string_buffer.h\"\n#include \"xenia/cpu/processor.h\"\n#include \"xenia/cpu/thread_state.h\"\n#include \"xenia/kernel/xthread.h\"\n\nextern \"C\" {\n#include \"third_party/FFmpeg/libavutil/log.h\"\n} // extern \"C\"\n\n// As with normal Microsoft, there are like twelve different ways to access\n// the audio APIs. Early games use XMA*() methods almost exclusively to touch\n// decoders. Later games use XAudio*() and direct memory writes to the XMA\n// structures (as opposed to the XMA* calls), meaning that we have to support\n// both.\n//\n// The XMA*() functions just manipulate the audio system in the guest context\n// and let the normal XmaDecoder handling take it, to prevent duplicate\n// implementations. They can be found in xboxkrnl_audio_xma.cc\n//\n// XMA details:\n// https://devel.nuclex.org/external/svn/directx/trunk/include/xma2defs.h\n// https://github.com/gdawg/fsbext/blob/master/src/xma_header.h\n//\n// XAudio2 uses XMA under the covers, and seems to map with the same\n// restrictions of frame/subframe/etc:\n// https://msdn.microsoft.com/en-us/library/windows/desktop/microsoft.directx_sdk.xaudio2.xaudio2_buffer(v=vs.85).aspx\n//\n// XMA contexts are 64b in size and tight bitfields. They are in physical\n// memory not usually available to games. Games will use MmMapIoSpace to get\n// the 64b pointer in user memory so they can party on it. If the game doesn't\n// do this, it's likely they are either passing the context to XAudio or\n// using the XMA* functions.\n\nDEFINE_bool(ffmpeg_verbose, false, \"Verbose FFmpeg output (debug and above)\",\n \"APU\");\n\nnamespace xe {\nnamespace apu {\n\nXmaDecoder::XmaDecoder(cpu::Processor* processor)\n : memory_(processor->memory()), processor_(processor) {}\n\nXmaDecoder::~XmaDecoder() = default;\n\nvoid av_log_callback(void* avcl, int level, const char* fmt, va_list va) {\n if (!cvars::ffmpeg_verbose && level > AV_LOG_WARNING) {\n return;\n }\n\n char level_char = '?';\n LogLevel log_level;\n switch (level) {\n case AV_LOG_ERROR: {\n level_char = '!';\n log_level = xe::LogLevel::Error;\n break;\n }\n case AV_LOG_WARNING: {\n level_char = 'w';\n log_level = xe::LogLevel::Warning;\n break;\n }\n case AV_LOG_INFO: {\n level_char = 'i';\n log_level = xe::LogLevel::Info;\n break;\n }\n case AV_LOG_VERBOSE: {\n level_char = 'v';\n log_level = xe::LogLevel::Debug;\n break;\n }\n case AV_LOG_DEBUG: {\n level_char = 'd';\n log_level = xe::LogLevel::Debug;\n break;\n }\n default: {\n level_char = '?';\n log_level = xe::LogLevel::Debug;\n break;\n }\n }\n\n StringBuffer buff;\n buff.AppendVarargs(fmt, va);\n xe::logging::AppendLogLineFormat(log_level, level_char, \"ffmpeg: {}\",\n buff.to_string_view());\n}\n\nX_STATUS XmaDecoder::Setup(kernel::KernelState* kernel_state) {\n // Setup ffmpeg logging callback\n av_log_set_callback(av_log_callback);\n\n // Let the processor know we want register access callbacks.\n memory_->AddVirtualMappedRange(\n 0x7FEA0000, 0xFFFF0000, 0x0000FFFF, this,\n reinterpret_cast(MMIOReadRegisterThunk),\n reinterpret_cast(MMIOWriteRegisterThunk));\n\n // Setup XMA context data.\n // The Xbox 360 kernel allocates the contexts with X_PAGE_NOCACHE |\n // X_PAGE_READWRITE and writes MmGetPhysicalAddress for the address to the\n // register.\n context_data_first_ptr_ = memory()->SystemHeapAlloc(\n sizeof(XMA_CONTEXT_DATA) * kContextCount, 256, kSystemHeapPhysical);\n context_data_last_ptr_ =\n context_data_first_ptr_ + (sizeof(XMA_CONTEXT_DATA) * kContextCount - 1);\n register_file_[XmaRegister::ContextArrayAddress] =\n memory()->GetPhysicalAddress(context_data_first_ptr_);\n\n // Setup XMA contexts.\n for (int i = 0; i < kContextCount; ++i) {\n uint32_t guest_ptr = context_data_first_ptr_ + i * sizeof(XMA_CONTEXT_DATA);\n XmaContext& context = contexts_[i];\n if (context.Setup(i, memory(), guest_ptr)) {\n assert_always();\n }\n }\n register_file_[XmaRegister::NextContextIndex] = 1;\n context_bitmap_.Resize(kContextCount);\n\n worker_running_ = true;\n work_event_ = xe::threading::Event::CreateAutoResetEvent(false);\n assert_not_null(work_event_);\n worker_thread_ = kernel::object_ref(\n new kernel::XHostThread(kernel_state, 128 * 1024, 0, [this]() {\n WorkerThreadMain();\n return 0;\n }));\n worker_thread_->set_name(\"XMA Decoder\");\n worker_thread_->set_can_debugger_suspend(true);\n worker_thread_->Create();\n\n return X_STATUS_SUCCESS;\n}\n\nvoid XmaDecoder::WorkerThreadMain() {\n uint32_t idle_loop_count = 0;\n while (worker_running_) {\n // Okay, let's loop through XMA contexts to find ones we need to decode!\n bool did_work = false;\n for (uint32_t n = 0; n < kContextCount; n++) {\n XmaContext& context = contexts_[n];\n did_work = context.Work() || did_work;\n\n // TODO: Need thread safety to do this.\n // Probably not too important though.\n // registers_.current_context = n;\n // registers_.next_context = (n + 1) % kContextCount;\n }\n\n if (paused_) {\n pause_fence_.Signal();\n resume_fence_.Wait();\n }\n\n if (!did_work) {\n idle_loop_count++;\n } else {\n idle_loop_count = 0;\n }\n\n if (idle_loop_count > 500) {\n // Idle for an extended period. Introduce a 20ms wait.\n xe::threading::Wait(work_event_.get(), false,\n std::chrono::milliseconds(20));\n }\n\n xe::threading::MaybeYield();\n }\n}\n\nvoid XmaDecoder::Shutdown() {\n worker_running_ = false;\n\n if (work_event_) {\n work_event_->Set();\n }\n\n if (paused_) {\n Resume();\n }\n\n if (worker_thread_) {\n // Wait for work thread.\n xe::threading::Wait(worker_thread_->thread(), false);\n worker_thread_.reset();\n }\n\n if (context_data_first_ptr_) {\n memory()->SystemHeapFree(context_data_first_ptr_);\n }\n\n context_data_first_ptr_ = 0;\n context_data_last_ptr_ = 0;\n}\n\nint XmaDecoder::GetContextId(uint32_t guest_ptr) {\n static_assert_size(XMA_CONTEXT_DATA, 64);\n if (guest_ptr < context_data_first_ptr_ ||\n guest_ptr > context_data_last_ptr_) {\n return -1;\n }\n assert_zero(guest_ptr & 0x3F);\n return (guest_ptr - context_data_first_ptr_) >> 6;\n}\n\nuint32_t XmaDecoder::AllocateContext() {\n size_t index = context_bitmap_.Acquire();\n if (index == -1) {\n // Out of contexts.\n return 0;\n }\n\n XmaContext& context = contexts_[index];\n assert_false(context.is_allocated());\n context.set_is_allocated(true);\n return context.guest_ptr();\n}\n\nvoid XmaDecoder::ReleaseContext(uint32_t guest_ptr) {\n auto context_id = GetContextId(guest_ptr);\n assert_true(context_id >= 0);\n\n XmaContext& context = contexts_[context_id];\n assert_true(context.is_allocated());\n context.Release();\n context_bitmap_.Release(context_id);\n}\n\nbool XmaDecoder::BlockOnContext(uint32_t guest_ptr, bool poll) {\n auto context_id = GetContextId(guest_ptr);\n assert_true(context_id >= 0);\n\n XmaContext& context = contexts_[context_id];\n return context.Block(poll);\n}\n\nuint32_t XmaDecoder::ReadRegister(uint32_t addr) {\n auto r = (addr & 0xFFFF) / 4;\n\n assert_true(r < XmaRegisterFile::kRegisterCount);\n\n switch (r) {\n case XmaRegister::ContextArrayAddress:\n break;\n case XmaRegister::CurrentContextIndex: {\n // 0606h (1818h) is rotating context processing # set to hardware ID of\n // context being processed.\n // If bit 200h is set, the locking code will possibly collide on hardware\n // IDs and error out, so we should never set it (I think?).\n uint32_t& current_context_index =\n register_file_[XmaRegister::CurrentContextIndex];\n uint32_t& next_context_index =\n register_file_[XmaRegister::NextContextIndex];\n // To prevent games from seeing a stuck XMA context, return a rotating\n // number.\n current_context_index = next_context_index;\n next_context_index = (next_context_index + 1) % kContextCount;\n break;\n }\n default:\n const auto register_info = register_file_.GetRegisterInfo(r);\n if (register_info) {\n XELOGW(\"XMA: Read from unhandled register ({:04X}, {})\", r,\n register_info->name);\n } else {\n XELOGW(\"XMA: Read from unknown register ({:04X})\", r);\n }\n break;\n }\n\n return xe::byte_swap(register_file_[r]);\n}\n\nvoid XmaDecoder::WriteRegister(uint32_t addr, uint32_t value) {\n SCOPE_profile_cpu_f(\"apu\");\n\n uint32_t r = (addr & 0xFFFF) / 4;\n value = xe::byte_swap(value);\n\n assert_true(r < XmaRegisterFile::kRegisterCount);\n register_file_[r] = value;\n\n if (r >= XmaRegister::Context0Kick && r <= XmaRegister::Context9Kick) {\n // Context kick command.\n // This will kick off the given hardware contexts.\n // Basically, this kicks the SPU and says \"hey, decode that audio!\"\n // XMAEnableContext\n\n // The context ID is a bit in the range of the entire context array.\n uint32_t base_context_id = (r - XmaRegister::Context0Kick) * 32;\n for (int i = 0; value && i < 32; ++i, value >>= 1) {\n if (value & 1) {\n uint32_t context_id = base_context_id + i;\n auto& context = contexts_[context_id];\n context.Enable();\n }\n }\n // Signal the decoder thread to start processing.\n work_event_->Set();\n } else if (r >= XmaRegister::Context0Lock && r <= XmaRegister::Context9Lock) {\n // Context lock command.\n // This requests a lock by flagging the context.\n // XMADisableContext\n uint32_t base_context_id = (r - XmaRegister::Context0Lock) * 32;\n for (int i = 0; value && i < 32; ++i, value >>= 1) {\n if (value & 1) {\n uint32_t context_id = base_context_id + i;\n auto& context = contexts_[context_id];\n context.Disable();\n }\n }\n // Signal the decoder thread to start processing.\n // work_event_->Set();\n } else if (r >= XmaRegister::Context0Clear &&\n r <= XmaRegister::Context9Clear) {\n // Context clear command.\n // This will reset the given hardware contexts.\n uint32_t base_context_id = (r - XmaRegister::Context0Clear) * 32;\n for (int i = 0; value && i < 32; ++i, value >>= 1) {\n if (value & 1) {\n uint32_t context_id = base_context_id + i;\n XmaContext& context = contexts_[context_id];\n context.Clear();\n }\n }\n } else {\n // 0601h (1804h) is written to with 0x02000000 and 0x03000000 around a lock\n // operation\n switch (r) {\n default: {\n const auto register_info = register_file_.GetRegisterInfo(r);\n if (register_info) {\n XELOGW(\"XMA: Write to unhandled register ({:04X}, {}): {:08X}\", r,\n register_info->name, value);\n } else {\n XELOGW(\"XMA: Write to unknown register ({:04X}): {:08X}\", r, value);\n }\n break;\n }\n#pragma warning(suppress : 4065)\n }\n }\n}\n\nvoid XmaDecoder::Pause() {\n if (paused_) {\n return;\n }\n paused_ = true;\n\n pause_fence_.Wait();\n}\n\nvoid XmaDecoder::Resume() {\n if (!paused_) {\n return;\n }\n paused_ = false;\n\n resume_fence_.Signal();\n}\n\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/xma_decoder.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_XMA_DECODER_H_\n#define XENIA_APU_XMA_DECODER_H_\n\n#include \n#include \n#include \n\n#include \"xenia/apu/xma_context.h\"\n#include \"xenia/apu/xma_register_file.h\"\n#include \"xenia/base/bit_map.h\"\n#include \"xenia/kernel/xthread.h\"\n#include \"xenia/xbox.h\"\n\nnamespace xe {\nnamespace apu {\n\nstruct XMA_CONTEXT_DATA;\n\nclass XmaDecoder {\n public:\n explicit XmaDecoder(cpu::Processor* processor);\n ~XmaDecoder();\n\n Memory* memory() const { return memory_; }\n cpu::Processor* processor() const { return processor_; }\n\n X_STATUS Setup(kernel::KernelState* kernel_state);\n void Shutdown();\n\n uint32_t context_array_ptr() const {\n return register_file_[XmaRegister::ContextArrayAddress];\n }\n\n uint32_t AllocateContext();\n void ReleaseContext(uint32_t guest_ptr);\n bool BlockOnContext(uint32_t guest_ptr, bool poll);\n\n uint32_t ReadRegister(uint32_t addr);\n void WriteRegister(uint32_t addr, uint32_t value);\n\n bool is_paused() const { return paused_; }\n void Pause();\n void Resume();\n\n protected:\n int GetContextId(uint32_t guest_ptr);\n\n private:\n void WorkerThreadMain();\n\n static uint32_t MMIOReadRegisterThunk(void* ppc_context, XmaDecoder* as,\n uint32_t addr) {\n return as->ReadRegister(addr);\n }\n static void MMIOWriteRegisterThunk(void* ppc_context, XmaDecoder* as,\n uint32_t addr, uint32_t value) {\n as->WriteRegister(addr, value);\n }\n\n protected:\n Memory* memory_ = nullptr;\n cpu::Processor* processor_ = nullptr;\n\n std::atomic worker_running_ = {false};\n kernel::object_ref worker_thread_;\n std::unique_ptr work_event_ = nullptr;\n\n bool paused_ = false;\n xe::threading::Fence pause_fence_; // Signaled when worker paused.\n xe::threading::Fence resume_fence_; // Signaled when resume requested.\n\n XmaRegisterFile register_file_;\n\n static const uint32_t kContextCount = 320;\n XmaContext contexts_[kContextCount];\n BitMap context_bitmap_;\n\n uint32_t context_data_first_ptr_ = 0;\n uint32_t context_data_last_ptr_ = 0;\n};\n\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_XMA_DECODER_H_\n" }, { "path": "src/xenia/apu/xma_helpers.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n// This file contains some functions used to help parse XMA data.\n\n#ifndef XENIA_APU_XMA_HELPERS_H_\n#define XENIA_APU_XMA_HELPERS_H_\n\n#include \n\nnamespace xe {\nnamespace apu {\nnamespace xma {\n\nstatic const uint32_t kMaxFrameLength = 0x7FFF;\n\n// Get number of frames that /begin/ in this packet.\nuint32_t GetPacketFrameCount(uint8_t* packet) {\n return (uint8_t)(packet[0] >> 2);\n}\n\n// Get the first frame offset in bits\nuint32_t GetPacketFrameOffset(uint8_t* packet) {\n uint32_t val = (uint16_t)(((packet[0] & 0x3) << 13) | (packet[1] << 5) |\n (packet[2] >> 3));\n // if (val > kBitsPerPacket - kBitsPerHeader) {\n // // There is no data in this packet\n // return -1;\n // } else {\n return val + 32;\n // }\n}\n\nuint32_t GetPacketMetadata(uint8_t* packet) {\n return (uint8_t)(packet[2] & 0x7);\n}\n\nuint32_t GetPacketSkipCount(uint8_t* packet) { return (uint8_t)(packet[3]); }\n\n} // namespace xma\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_XMA_HELPERS_H_\n" }, { "path": "src/xenia/apu/xma_register_file.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/apu/xma_register_file.h\"\n\n#include \n\n#include \"xenia/base/math.h\"\n\nnamespace xe {\nnamespace apu {\n\nXmaRegisterFile::XmaRegisterFile() { std::memset(values, 0, sizeof(values)); }\n\nconst XmaRegisterInfo* XmaRegisterFile::GetRegisterInfo(uint32_t index) {\n switch (index) {\n#define XE_XMA_REGISTER(index, name) \\\n case index: { \\\n static const XmaRegisterInfo reg_info = { \\\n #name, \\\n }; \\\n return ®_info; \\\n }\n#include \"xenia/apu/xma_register_table.inc\"\n#undef XE_XMA_REGISTER\n default:\n return nullptr;\n }\n}\n\n} // namespace apu\n} // namespace xe\n" }, { "path": "src/xenia/apu/xma_register_file.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_APU_XMA_REGISTER_FILE_H_\n#define XENIA_APU_XMA_REGISTER_FILE_H_\n\n#include \n#include \n\nnamespace xe {\nnamespace apu {\n\nstruct XmaRegister {\n#define XE_XMA_REGISTER(index, name) static const uint32_t name = index;\n#include \"xenia/apu/xma_register_table.inc\"\n#undef XE_XMA_REGISTER\n};\n\nstruct XmaRegisterInfo {\n const char* name;\n};\n\nclass XmaRegisterFile {\n public:\n XmaRegisterFile();\n\n static const XmaRegisterInfo* GetRegisterInfo(uint32_t index);\n\n static const size_t kRegisterCount = (0xFFFF + 1) / 4;\n uint32_t values[kRegisterCount];\n\n uint32_t operator[](uint32_t reg) const { return values[reg]; }\n uint32_t& operator[](uint32_t reg) { return values[reg]; }\n};\n\n} // namespace apu\n} // namespace xe\n\n#endif // XENIA_APU_XMA_REGISTER_FILE_H_\n" }, { "path": "src/xenia/apu/xma_register_table.inc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n// This is a partial file designed to be included by other files when\n// constructing various tables.\n\n#ifndef XE_XMA_REGISTER\n#define XE_XMA_REGISTER(index, name)\n#define __XE_XMA_REGISTER_UNSET\n#endif\n\n#ifndef XE_XMA_REGISTER_CONTEXT_GROUP\n#define XE_XMA_REGISTER_CONTEXT_GROUP(index, suffix) \\\n XE_XMA_REGISTER(index + 0, Context0##suffix) \\\n XE_XMA_REGISTER(index + 1, Context1##suffix) \\\n XE_XMA_REGISTER(index + 2, Context2##suffix) \\\n XE_XMA_REGISTER(index + 3, Context3##suffix) \\\n XE_XMA_REGISTER(index + 4, Context4##suffix) \\\n XE_XMA_REGISTER(index + 5, Context5##suffix) \\\n XE_XMA_REGISTER(index + 6, Context6##suffix) \\\n XE_XMA_REGISTER(index + 7, Context7##suffix) \\\n XE_XMA_REGISTER(index + 8, Context8##suffix) \\\n XE_XMA_REGISTER(index + 9, Context9##suffix)\n#endif\n\n// 0x0000..0x001F : ???\n// 0x0020..0x03FF : all 0xFFs?\n// 0x0400..0x043F : ???\n// 0x0440..0x047F : all 0xFFs?\n// 0x0480..0x048B : ???\n// 0x048C..0x04C0 : all 0xFFs?\n// 0x04C1..0x04CB : ???\n// 0x04CC..0x04FF : all 0xFFs?\n// 0x0500..0x051F : ???\n// 0x0520..0x057F : all 0xFFs?\n// 0x0580..0x058F : ???\n// 0x0590..0x05FF : all 0xFFs?\n\n// XMA stuff is probably only 0x0600..0x06FF\n//---------------------------------------------------------------------------//\n\nXE_XMA_REGISTER(0x0600, ContextArrayAddress)\n// 0x0601..0x0605 : ???\nXE_XMA_REGISTER(0x0606, CurrentContextIndex)\nXE_XMA_REGISTER(0x0607, NextContextIndex)\n// 0x0608 : ???\n// 0x0609..0x060F : zero?\nXE_XMA_REGISTER_CONTEXT_GROUP(0x0610, Unknown610)\n// 0x061A..0x061F : zero?\nXE_XMA_REGISTER_CONTEXT_GROUP(0x0620, Unknown620)\n// 0x062A..0x0641 : zero?\n// 0x0642..0x0644 : ???\n// 0x0645..0x064F : zero?\nXE_XMA_REGISTER_CONTEXT_GROUP(0x0650, Kick)\n// 0x065A..0x065F : zero?\nXE_XMA_REGISTER_CONTEXT_GROUP(0x0660, Unknown660)\n// 0x066A..0x0681 : zero?\n// 0x0682..0x0684 : ???\n// 0x0685..0x068F : zero?\nXE_XMA_REGISTER_CONTEXT_GROUP(0x0690, Lock)\n// 0x069A..0x069F : zero?\nXE_XMA_REGISTER_CONTEXT_GROUP(0x06A0, Clear)\n\n//---------------------------------------------------------------------------//\n\n// 0x0700..0x07FF : all 0xFFs?\n// 0x0800..0x17FF : ???\n// 0x1800..0x2FFF : all 0xFFs?\n// 0x3000..0x30FF : ???\n// 0x3100..0x3FFF : all 0xFFs?\n\n#ifdef __XE_XMA_REGISTER_UNSET\n#undef __XE_XMA_REGISTER_UNSET\n#undef XE_XMA_REGISTER\n#endif\n" }, { "path": "src/xenia/base/README.md", "content": "A lightweight cross-platform/compiler compatibility library.\n\nThis library presupposes C++11/14 support. As more compilers get C++14 it will\nassume that.\n\nOther parts of the project use this to avoid creating spaghetti linkage. Code\nspecific to the emulator should be kept out, as not all of the projects that\ndepend on this need it.\n\nWhere possible, C++11/14 STL should be used instead of adding any code in here,\nand the code should be kept as small as possible (by reusing STL/etc). Third\nparty dependencies should be kept to a minimum.\n\nTarget compilers:\n* MSVC++ 2013+\n* Clang 3.4+\n* GCC 4.8+.\n\nTarget platforms:\n* Windows 8+ (`_win.cc` suffix)\n* Mac OSX 10.9+ (`_mac.cc` suffix, falling back to `_posix.cc`)\n* Linux ? (`_posix.cc` suffix)\n\nAvoid the use of platform-specific #ifdefs and instead try to put all\nplatform-specific code in the appropriately suffixed cc files.\n" }, { "path": "src/xenia/base/app_win32.manifest", "content": "\n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n True/PM\n \n \n \n PerMonitorV2,PerMonitor\n \n \n\n" }, { "path": "src/xenia/base/arena.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/arena.h\"\n\n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/math.h\"\n\nnamespace xe {\n\nArena::Arena(size_t chunk_size)\n : chunk_size_(chunk_size), head_chunk_(nullptr), active_chunk_(nullptr) {}\n\nArena::~Arena() {\n Reset();\n Chunk* chunk = head_chunk_;\n while (chunk) {\n Chunk* next = chunk->next;\n delete chunk;\n chunk = next;\n }\n head_chunk_ = nullptr;\n}\n\nvoid Arena::Reset() {\n active_chunk_ = head_chunk_;\n if (active_chunk_) {\n active_chunk_->offset = 0;\n }\n}\n\nvoid Arena::DebugFill() {\n auto chunk = head_chunk_;\n while (chunk) {\n std::memset(chunk->buffer, 0xCD, chunk->capacity);\n chunk = chunk->next;\n }\n}\n\nvoid* Arena::Alloc(size_t size, size_t align) {\n assert_true(\n align > 0 && xe::is_pow2(align) && align <= 16,\n \"align needs to be a power of 2 and not greater than Chunk alignment\");\n\n // for alignment\n const auto get_padding = [this, align]() -> size_t {\n const size_t mask = align - 1;\n size_t deviation = active_chunk_->offset & mask;\n return (align - deviation) & mask;\n };\n\n if (active_chunk_) {\n if (active_chunk_->capacity - active_chunk_->offset <\n size + get_padding() + 4_KiB) {\n Chunk* next = active_chunk_->next;\n if (!next) {\n assert_true(size + get_padding() < chunk_size_,\n \"need to support larger chunks\");\n next = new Chunk(chunk_size_);\n active_chunk_->next = next;\n }\n next->offset = 0;\n active_chunk_ = next;\n }\n } else {\n head_chunk_ = active_chunk_ = new Chunk(chunk_size_);\n }\n\n active_chunk_->offset += get_padding();\n uint8_t* p = active_chunk_->buffer + active_chunk_->offset;\n active_chunk_->offset += size;\n assert_true((reinterpret_cast(p) & (align - 1)) == 0,\n \"alignment failed\");\n return p;\n}\n\nvoid Arena::Rewind(size_t size) { active_chunk_->offset -= size; }\n\nsize_t Arena::CalculateSize() {\n size_t total_length = 0;\n Chunk* chunk = head_chunk_;\n while (chunk) {\n total_length += chunk->offset;\n if (chunk == active_chunk_) {\n break;\n }\n chunk = chunk->next;\n }\n return total_length;\n}\n\nvoid* Arena::CloneContents() {\n size_t total_length = CalculateSize();\n auto result = reinterpret_cast(malloc(total_length));\n auto p = result;\n Chunk* chunk = head_chunk_;\n while (chunk) {\n std::memcpy(p, chunk->buffer, chunk->offset);\n p += chunk->offset;\n if (chunk == active_chunk_) {\n break;\n }\n chunk = chunk->next;\n }\n return result;\n}\n\nvoid Arena::CloneContents(void* buffer, size_t buffer_length) {\n uint8_t* p = reinterpret_cast(buffer);\n Chunk* chunk = head_chunk_;\n while (chunk) {\n std::memcpy(p, chunk->buffer, chunk->offset);\n p += chunk->offset;\n if (chunk == active_chunk_) {\n break;\n }\n chunk = chunk->next;\n }\n}\n\nArena::Chunk::Chunk(size_t chunk_size)\n : next(nullptr), capacity(chunk_size), buffer(0), offset(0) {\n buffer = reinterpret_cast(malloc(capacity));\n assert_true((reinterpret_cast(buffer) & size_t(15)) == 0,\n \"16 byte alignment required\");\n}\n\nArena::Chunk::~Chunk() {\n if (buffer) {\n free(buffer);\n }\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/arena.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_ARENA_H_\n#define XENIA_BASE_ARENA_H_\n\n#include \n#include \n#include \n\n#include \"xenia/base/literals.h\"\n\nnamespace xe {\n\nusing namespace xe::literals;\n\nclass Arena {\n public:\n explicit Arena(size_t chunk_size = 4_MiB);\n ~Arena();\n\n void Reset();\n void DebugFill();\n\n void* Alloc(size_t size, size_t align);\n template \n T* Alloc() {\n return reinterpret_cast(Alloc(sizeof(T), alignof(T)));\n }\n // When rewinding aligned allocations, any padding that was applied during\n // allocation will be leaked\n void Rewind(size_t size);\n\n void* CloneContents();\n template \n void CloneContents(std::vector* buffer) {\n buffer->resize(CalculateSize() / sizeof(T));\n CloneContents(buffer->data(), buffer->size() * sizeof(T));\n }\n\n private:\n class Chunk {\n public:\n explicit Chunk(size_t chunk_size);\n ~Chunk();\n\n Chunk* next;\n\n size_t capacity;\n uint8_t* buffer;\n size_t offset;\n };\n\n size_t CalculateSize();\n void CloneContents(void* buffer, size_t buffer_length);\n\n size_t chunk_size_;\n Chunk* head_chunk_;\n Chunk* active_chunk_;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_ARENA_H_\n" }, { "path": "src/xenia/base/assert.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_ASSERT_H_\n#define XENIA_BASE_ASSERT_H_\n\n#include \n\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\n\n#define static_assert_size(type, size) \\\n static_assert(sizeof(type) == size, \\\n \"bad definition for \" #type \": must be \" #size \" bytes\")\n\n// We rely on assert being compiled out in NDEBUG.\n#define xenia_assert assert\n\n#define __XENIA_EXPAND(x) x\n#define __XENIA_ARGC(...) \\\n __XENIA_EXPAND(__XENIA_ARGC_IMPL(__VA_ARGS__, 15, 14, 13, 12, 11, 10, 9, 8, \\\n 7, 6, 5, 4, 3, 2, 1, 0))\n#define __XENIA_ARGC_IMPL(x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, \\\n x13, x14, x15, N, ...) \\\n N\n#define __XENIA_MACRO_DISPATCH(func, ...) \\\n __XENIA_MACRO_DISPATCH_(func, __XENIA_ARGC(__VA_ARGS__))\n#define __XENIA_MACRO_DISPATCH_(func, nargs) \\\n __XENIA_MACRO_DISPATCH__(func, nargs)\n#define __XENIA_MACRO_DISPATCH__(func, nargs) func##nargs\n\n#define assert_always(...) xenia_assert(false)\n\n#define assert_true(...) \\\n __XENIA_MACRO_DISPATCH(assert_true, __VA_ARGS__)(__VA_ARGS__)\n#define assert_true1(expr) xenia_assert(expr)\n#define assert_true2(expr, message) xenia_assert((expr) || !message)\n\n#define assert_false(...) \\\n __XENIA_MACRO_DISPATCH(assert_false, __VA_ARGS__)(__VA_ARGS__)\n#define assert_false1(expr) xenia_assert(!(expr))\n#define assert_false2(expr, message) xenia_assert(!(expr) || !message)\n\n#define assert_zero(...) \\\n __XENIA_MACRO_DISPATCH(assert_zero, __VA_ARGS__)(__VA_ARGS__)\n#define assert_zero1(expr) xenia_assert((expr) == 0)\n#define assert_zero2(expr, message) xenia_assert((expr) == 0 || !message)\n\n#define assert_not_zero(...) \\\n __XENIA_MACRO_DISPATCH(assert_not_zero, __VA_ARGS__)(__VA_ARGS__)\n#define assert_not_zero1(expr) xenia_assert((expr) != 0)\n#define assert_not_zero2(expr, message) xenia_assert((expr) != 0 || !message)\n\n#define assert_null(...) \\\n __XENIA_MACRO_DISPATCH(assert_null, __VA_ARGS__)(__VA_ARGS__)\n#define assert_null1(expr) xenia_assert((expr) == nullptr)\n#define assert_null2(expr, message) xenia_assert((expr) == nullptr || !message)\n\n#define assert_not_null(...) \\\n __XENIA_MACRO_DISPATCH(assert_not_null, __VA_ARGS__)(__VA_ARGS__)\n#define assert_not_null1(expr) xenia_assert((expr) != nullptr)\n#define assert_not_null2(expr, message) \\\n xenia_assert((expr) != nullptr || !message)\n\n#define assert_unhandled_case(variable) \\\n assert_always(\"unhandled switch(\" #variable \") case\")\n\n} // namespace xe\n\n#endif // XENIA_BASE_ASSERT_H_\n" }, { "path": "src/xenia/base/atomic.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_ATOMIC_H_\n#define XENIA_BASE_ATOMIC_H_\n\n#include \n\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\n\n#if XE_PLATFORM_WIN32\n\ninline int32_t atomic_inc(volatile int32_t* value) {\n return _InterlockedIncrement(reinterpret_cast(value));\n}\ninline int32_t atomic_dec(volatile int32_t* value) {\n return _InterlockedDecrement(reinterpret_cast(value));\n}\n\ninline int32_t atomic_exchange(int32_t new_value, volatile int32_t* value) {\n return _InterlockedExchange(reinterpret_cast(value),\n new_value);\n}\ninline int64_t atomic_exchange(int64_t new_value, volatile int64_t* value) {\n return _InterlockedExchange64(reinterpret_cast(value),\n new_value);\n}\n\ninline int32_t atomic_exchange_add(int32_t amount, volatile int32_t* value) {\n return _InterlockedExchangeAdd(reinterpret_cast(value),\n amount);\n}\ninline int64_t atomic_exchange_add(int64_t amount, volatile int64_t* value) {\n return _InterlockedExchangeAdd64(reinterpret_cast(value),\n amount);\n}\n\ninline bool atomic_cas(int32_t old_value, int32_t new_value,\n volatile int32_t* value) {\n return _InterlockedCompareExchange(reinterpret_cast(value),\n new_value, old_value) == old_value;\n}\ninline bool atomic_cas(int64_t old_value, int64_t new_value,\n volatile int64_t* value) {\n return _InterlockedCompareExchange64(\n reinterpret_cast(value), new_value,\n old_value) == old_value;\n}\n\n#elif XE_PLATFORM_LINUX || XE_PLATFORM_MAC\n\ninline int32_t atomic_inc(volatile int32_t* value) {\n return __sync_add_and_fetch(value, 1);\n}\ninline int32_t atomic_dec(volatile int32_t* value) {\n return __sync_sub_and_fetch(value, 1);\n}\n\ninline int32_t atomic_exchange(int32_t new_value, volatile int32_t* value) {\n return __sync_val_compare_and_swap(value, *value, new_value);\n}\ninline int64_t atomic_exchange(int64_t new_value, volatile int64_t* value) {\n return __sync_val_compare_and_swap(value, *value, new_value);\n}\n\ninline int32_t atomic_exchange_add(int32_t amount, volatile int32_t* value) {\n return __sync_fetch_and_add(value, amount);\n}\ninline int64_t atomic_exchange_add(int64_t amount, volatile int64_t* value) {\n return __sync_fetch_and_add(value, amount);\n}\n\ninline bool atomic_cas(int32_t old_value, int32_t new_value,\n volatile int32_t* value) {\n return __sync_bool_compare_and_swap(\n reinterpret_cast(value), old_value, new_value);\n}\ninline bool atomic_cas(int64_t old_value, int64_t new_value,\n volatile int64_t* value) {\n return __sync_bool_compare_and_swap(\n reinterpret_cast(value), old_value, new_value);\n}\n\n#else\n\n#error No atomic primitives defined for this platform/cpu combination.\n\n#endif // XE_PLATFORM\n\ninline uint32_t atomic_inc(volatile uint32_t* value) {\n return static_cast(\n atomic_inc(reinterpret_cast(value)));\n}\ninline uint32_t atomic_dec(volatile uint32_t* value) {\n return static_cast(\n atomic_dec(reinterpret_cast(value)));\n}\n\ninline uint32_t atomic_exchange(uint32_t new_value, volatile uint32_t* value) {\n return static_cast(\n atomic_exchange(static_cast(new_value),\n reinterpret_cast(value)));\n}\ninline uint64_t atomic_exchange(uint64_t new_value, volatile uint64_t* value) {\n return static_cast(\n atomic_exchange(static_cast(new_value),\n reinterpret_cast(value)));\n}\n\ninline uint32_t atomic_exchange_add(uint32_t amount, volatile uint32_t* value) {\n return static_cast(\n atomic_exchange_add(static_cast(amount),\n reinterpret_cast(value)));\n}\ninline uint64_t atomic_exchange_add(uint64_t amount, volatile uint64_t* value) {\n return static_cast(\n atomic_exchange_add(static_cast(amount),\n reinterpret_cast(value)));\n}\n\ninline bool atomic_cas(uint32_t old_value, uint32_t new_value,\n volatile uint32_t* value) {\n return atomic_cas(static_cast(old_value),\n static_cast(new_value),\n reinterpret_cast(value));\n}\ninline bool atomic_cas(uint64_t old_value, uint64_t new_value,\n volatile uint64_t* value) {\n return atomic_cas(static_cast(old_value),\n static_cast(new_value),\n reinterpret_cast(value));\n}\n\n} // namespace xe\n\n#endif // XENIA_BASE_ATOMIC_H_\n" }, { "path": "src/xenia/base/bit_map.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/bit_map.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/atomic.h\"\n#include \"xenia/base/math.h\"\n\nnamespace xe {\n\nBitMap::BitMap() = default;\n\nBitMap::BitMap(size_t size_bits) { Resize(size_bits); }\n\nBitMap::BitMap(uint64_t* data, size_t size_bits) {\n assert_true(size_bits % kDataSizeBits == 0);\n\n data_.resize(size_bits / kDataSizeBits);\n std::memcpy(data_.data(), data, size_bits / kDataSizeBits);\n}\n\nsize_t BitMap::Acquire() {\n for (size_t i = 0; i < data_.size(); i++) {\n uint64_t entry = 0;\n uint64_t new_entry = 0;\n int64_t acquired_idx = -1;\n\n do {\n entry = data_[i];\n uint8_t index = lzcnt(entry);\n if (index == kDataSizeBits) {\n // None free.\n acquired_idx = -1;\n break;\n }\n\n // Entry has a free bit. Acquire it.\n uint64_t bit = 1ull << (kDataSizeBits - index - 1);\n new_entry = entry & ~bit;\n assert_not_zero(entry & bit);\n\n acquired_idx = index;\n } while (!atomic_cas(entry, new_entry, &data_[i]));\n\n if (acquired_idx != -1) {\n // Acquired.\n return (i * kDataSizeBits) + acquired_idx;\n }\n }\n\n return -1;\n}\n\nvoid BitMap::Release(size_t index) {\n auto slot = index / kDataSizeBits;\n index -= slot * kDataSizeBits;\n\n uint64_t bit = 1ull << (kDataSizeBits - index - 1);\n\n uint64_t entry = 0;\n uint64_t new_entry = 0;\n do {\n entry = data_[slot];\n assert_zero(entry & bit);\n\n new_entry = entry | bit;\n } while (!atomic_cas(entry, new_entry, &data_[slot]));\n}\n\nvoid BitMap::Resize(size_t new_size_bits) {\n auto old_size = data_.size();\n assert_true(new_size_bits % kDataSizeBits == 0);\n data_.resize(new_size_bits / kDataSizeBits);\n\n // Initialize new entries.\n if (data_.size() > old_size) {\n for (size_t i = old_size; i < data_.size(); i++) {\n data_[i] = -1;\n }\n }\n}\n\nvoid BitMap::Reset() {\n for (size_t i = 0; i < data_.size(); i++) {\n data_[i] = -1;\n }\n}\n\n} // namespace xe" }, { "path": "src/xenia/base/bit_map.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_BIT_MAP_H_\n#define XENIA_BASE_BIT_MAP_H_\n\n#include \n#include \n#include \n\nnamespace xe {\n\n// Bit Map: Efficient lookup of free/used entries.\nclass BitMap {\n public:\n BitMap();\n\n // Size is the number of entries, must be a multiple of 64.\n BitMap(size_t size_bits);\n\n // Data does not have to be aligned to a 4-byte boundary, but it is\n // preferable.\n // Size is the number of entries, must be a multiple of 64.\n BitMap(uint64_t* data, size_t size_bits);\n\n // (threadsafe) Acquires an entry and returns its index. Returns -1 if there\n // are no more free entries.\n size_t Acquire();\n\n // (threadsafe) Releases an entry by an index.\n void Release(size_t index);\n\n // Resize the bitmap. Size is the number of entries, must be a multiple of 64.\n void Resize(size_t new_size_bits);\n\n // Sets all entries to free.\n void Reset();\n\n const std::vector data() const { return data_; }\n std::vector& data() { return data_; }\n\n private:\n const static size_t kDataSize = 8;\n const static size_t kDataSizeBits = kDataSize * 8;\n std::vector data_;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_BIT_MAP_H_\n" }, { "path": "src/xenia/base/bit_range.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2019 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_BIT_RANGE_H_\n#define XENIA_BASE_BIT_RANGE_H_\n\n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/math.h\"\n\nnamespace xe {\nnamespace bit_range {\n\n// Provided length is in bits since the first. Returns of the\n// range in bits, with length == 0 if not found.\ntemplate \nstd::pair NextUnsetRange(const Block* bits, size_t first,\n size_t length) {\n if (!length) {\n return std::make_pair(size_t(first), size_t(0));\n }\n size_t last = first + length - 1;\n const size_t block_bits = sizeof(Block) * CHAR_BIT;\n size_t block_first = first / block_bits;\n size_t block_last = last / block_bits;\n size_t range_start = SIZE_MAX;\n for (size_t i = block_first; i <= block_last; ++i) {\n Block block = bits[i];\n // Ignore bits in the block outside the specified range by considering them\n // set.\n if (i == block_first) {\n block |= (Block(1) << (first & (block_bits - 1))) - 1;\n }\n if (i == block_last && (last & (block_bits - 1)) != block_bits - 1) {\n block |= ~((Block(1) << ((last & (block_bits - 1)) + 1)) - 1);\n }\n while (true) {\n uint32_t block_bit;\n if (range_start == SIZE_MAX) {\n // Check if need to open a new range.\n if (!xe::bit_scan_forward(~block, &block_bit)) {\n break;\n }\n range_start = i * block_bits + block_bit;\n } else {\n // Check if need to close the range.\n // Ignore the set bits before the beginning of the range.\n Block block_bits_set_from_start = block;\n if (i == range_start / block_bits) {\n block_bits_set_from_start &=\n ~((Block(1) << (range_start & (block_bits - 1))) - 1);\n }\n if (!xe::bit_scan_forward(block_bits_set_from_start, &block_bit)) {\n break;\n }\n return std::make_pair(range_start,\n (i * block_bits) + block_bit - range_start);\n }\n }\n }\n if (range_start != SIZE_MAX) {\n return std::make_pair(range_start, last + size_t(1) - range_start);\n }\n return std::make_pair(first + length, size_t(0));\n}\n\ntemplate \nvoid SetRange(Block* bits, size_t first, size_t length) {\n if (!length) {\n return;\n }\n size_t last = first + length - 1;\n const size_t block_bits = sizeof(Block) * CHAR_BIT;\n size_t block_first = first / block_bits;\n size_t block_last = last / block_bits;\n Block set_first = ~((Block(1) << (first & (block_bits - 1))) - 1);\n Block set_last = ~Block(0);\n if ((last & (block_bits - 1)) != (block_bits - 1)) {\n set_last &= (Block(1) << ((last & (block_bits - 1)) + 1)) - 1;\n }\n if (block_first == block_last) {\n bits[block_first] |= set_first & set_last;\n return;\n }\n bits[block_first] |= set_first;\n if (block_first + 1 < block_last) {\n std::memset(bits + block_first + 1, CHAR_MAX,\n (block_last - (block_first + 1)) * sizeof(Block));\n }\n bits[block_last] |= set_last;\n}\n\n} // namespace bit_range\n} // namespace xe\n\n#endif // XENIA_BASE_BIT_RANGE_H_\n" }, { "path": "src/xenia/base/bit_stream.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/bit_stream.h\"\n\n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/byte_order.h\"\n\nnamespace xe {\n\nBitStream::BitStream(uint8_t* buffer, size_t size_in_bits)\n : buffer_(buffer), size_bits_(size_in_bits) {}\n\nBitStream::~BitStream() = default;\n\nvoid BitStream::SetOffset(size_t offset_bits) {\n assert_false(offset_bits > size_bits_);\n offset_bits_ = std::min(offset_bits, size_bits_);\n}\n\nsize_t BitStream::BitsRemaining() { return size_bits_ - offset_bits_; }\n\nuint64_t BitStream::Peek(size_t num_bits) {\n // FYI: The reason we can't copy more than 57 bits is:\n // 57 = 7 * 8 + 1 - that can only span a maximum of 8 bytes.\n // We can't read in 9 bytes (easily), so we limit it.\n assert_false(num_bits > 57);\n assert_false(offset_bits_ + num_bits > size_bits_);\n\n size_t offset_bytes = offset_bits_ >> 3;\n size_t rel_offset_bits = offset_bits_ - (offset_bytes << 3);\n\n // offset -->\n // ..[junk]..| target bits |....[junk].............\n uint64_t bits = *(uint64_t*)(buffer_ + offset_bytes);\n\n // We need the data in little endian.\n // TODO: Have a flag specifying endianness of data?\n bits = xe::byte_swap(bits);\n\n // Shift right\n // .....[junk]........| target bits |\n bits >>= 64 - (rel_offset_bits + num_bits);\n\n // AND with mask\n // ...................| target bits |\n bits &= (1ULL << num_bits) - 1;\n\n return bits;\n}\n\nuint64_t BitStream::Read(size_t num_bits) {\n uint64_t val = Peek(num_bits);\n Advance(num_bits);\n\n return val;\n}\n\n// TODO: This is totally not tested!\nbool BitStream::Write(uint64_t val, size_t num_bits) {\n assert_false(num_bits > 57);\n assert_false(offset_bits_ + num_bits >= size_bits_);\n\n size_t offset_bytes = offset_bits_ >> 3;\n size_t rel_offset_bits = offset_bits_ - (offset_bytes << 3);\n\n // Construct a mask\n uint64_t mask = (1ULL << num_bits) - 1;\n mask <<= 64 - (rel_offset_bits + num_bits);\n mask = ~mask;\n\n // Shift the value left into position.\n val <<= 64 - (rel_offset_bits + num_bits);\n\n // offset ----->\n // ....[junk]...| target bits w/ junk |....[junk]......\n uint64_t bits = *(uint64_t*)(buffer_ + offset_bytes);\n\n // AND with mask\n // ....[junk]...| target bits (0) |........[junk]......\n bits &= mask;\n\n // OR with val\n // ....[junk]...| target bits (val) |......[junk]......\n bits |= val;\n\n // Store into the bitstream.\n *(uint64_t*)(buffer_ + offset_bytes) = bits;\n\n // Advance the bitstream forward.\n Advance(num_bits);\n\n return true;\n}\n\nsize_t BitStream::Copy(uint8_t* dest_buffer, size_t num_bits) {\n size_t offset_bytes = offset_bits_ >> 3;\n size_t rel_offset_bits = offset_bits_ - (offset_bytes << 3);\n size_t bits_left = num_bits;\n size_t out_offset_bytes = 0;\n\n // First: Copy the first few bits up to a byte boundary.\n if (rel_offset_bits) {\n uint64_t bits = Peek(8 - rel_offset_bits);\n uint8_t clear_mask = ~((uint8_t(1) << rel_offset_bits) - 1);\n dest_buffer[out_offset_bytes] &= clear_mask;\n dest_buffer[out_offset_bytes] |= (uint8_t)bits;\n\n bits_left -= 8 - rel_offset_bits;\n Advance(8 - rel_offset_bits);\n out_offset_bytes++;\n }\n\n // Second: Use memcpy for the bytes left.\n if (bits_left >= 8) {\n std::memcpy(dest_buffer + out_offset_bytes,\n buffer_ + offset_bytes + out_offset_bytes, bits_left / 8);\n out_offset_bytes += (bits_left / 8);\n Advance((bits_left / 8) * 8);\n bits_left -= (bits_left / 8) * 8;\n }\n\n // Third: Copy the last few bits.\n if (bits_left) {\n uint64_t bits = Peek(bits_left);\n bits <<= 8 - bits_left;\n\n uint8_t clear_mask = ((uint8_t(1) << bits_left) - 1);\n dest_buffer[out_offset_bytes] &= clear_mask;\n dest_buffer[out_offset_bytes] |= (uint8_t)bits;\n Advance(bits_left);\n }\n\n // Return the bit offset to the copied bits.\n return rel_offset_bits;\n}\n\nvoid BitStream::Advance(size_t num_bits) { SetOffset(offset_bits_ + num_bits); }\n\n} // namespace xe\n" }, { "path": "src/xenia/base/bit_stream.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_BIT_STREAM_H_\n#define XENIA_BASE_BIT_STREAM_H_\n\n#include \n#include \n\nnamespace xe {\n\nclass BitStream {\n public:\n BitStream(uint8_t* buffer, size_t size_in_bits);\n ~BitStream();\n\n const uint8_t* buffer() const { return buffer_; }\n uint8_t* buffer() { return buffer_; }\n size_t offset_bits() const { return offset_bits_; }\n size_t size_bits() const { return size_bits_; }\n\n void Advance(size_t num_bits);\n void SetOffset(size_t offset_bits);\n size_t BitsRemaining();\n\n // Note: num_bits MUST be in the range 0-57 (inclusive)\n uint64_t Peek(size_t num_bits);\n uint64_t Read(size_t num_bits);\n bool Write(uint64_t val, size_t num_bits); // TODO(DrChat): Not tested!\n\n size_t Copy(uint8_t* dest_buffer, size_t num_bits);\n\n private:\n uint8_t* buffer_ = nullptr;\n size_t offset_bits_ = 0;\n size_t size_bits_ = 0;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_BIT_STREAM_H_\n" }, { "path": "src/xenia/base/byte_order.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_BYTE_ORDER_H_\n#define XENIA_BASE_BYTE_ORDER_H_\n\n#include \n#if defined __has_include\n#if __has_include()\n#include \n#endif\n#endif\n#if __cpp_lib_endian\n#include \n#endif\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/platform.h\"\n\n#if !__cpp_lib_endian\n// Polyfill\n#ifdef __BYTE_ORDER__\nnamespace std {\nenum class endian {\n little = __ORDER_LITTLE_ENDIAN__,\n big = __ORDER_BIG_ENDIAN__,\n native = __BYTE_ORDER__\n};\n}\n#else\n// Hardcode to little endian for now\nnamespace std {\nenum class endian { little = 0, big = 1, native = 0 };\n}\n#endif\n#endif\n// Check for mixed endian\nstatic_assert((std::endian::native == std::endian::big) ||\n (std::endian::native == std::endian::little));\n\nnamespace xe {\n\n#if XE_COMPILER_MSVC\n#define XENIA_BASE_BYTE_SWAP_16 _byteswap_ushort\n#define XENIA_BASE_BYTE_SWAP_32 _byteswap_ulong\n#define XENIA_BASE_BYTE_SWAP_64 _byteswap_uint64\n#else\n#define XENIA_BASE_BYTE_SWAP_16 __builtin_bswap16\n#define XENIA_BASE_BYTE_SWAP_32 __builtin_bswap32\n#define XENIA_BASE_BYTE_SWAP_64 __builtin_bswap64\n#endif // XE_PLATFORM_WIN32\n\ntemplate \ninline T byte_swap(T value) {\n static_assert(\n sizeof(T) == 8 || sizeof(T) == 4 || sizeof(T) == 2 || sizeof(T) == 1,\n \"byte_swap(T value): Type T has illegal size\");\n if constexpr (sizeof(T) == 8) {\n uint64_t temp =\n XENIA_BASE_BYTE_SWAP_64(*reinterpret_cast(&value));\n return *reinterpret_cast(&temp);\n } else if constexpr (sizeof(T) == 4) {\n uint32_t temp =\n XENIA_BASE_BYTE_SWAP_32(*reinterpret_cast(&value));\n return *reinterpret_cast(&temp);\n } else if constexpr (sizeof(T) == 2) {\n uint16_t temp =\n XENIA_BASE_BYTE_SWAP_16(*reinterpret_cast(&value));\n return *reinterpret_cast(&temp);\n } else if constexpr (sizeof(T) == 1) {\n return value;\n }\n}\n\ntemplate \nstruct endian_store {\n endian_store() = default;\n endian_store(const T& src) { set(src); }\n endian_store(const endian_store& other) { set(other); }\n operator T() const { return get(); }\n\n void set(const T& src) {\n if constexpr (std::endian::native == E) {\n value = src;\n } else {\n value = xe::byte_swap(src);\n }\n }\n void set(const endian_store& other) { value = other.value; }\n T get() const {\n if constexpr (std::endian::native == E) {\n return value;\n }\n return xe::byte_swap(value);\n }\n\n endian_store& operator+=(int a) {\n *this = *this + a;\n return *this;\n }\n endian_store& operator-=(int a) {\n *this = *this - a;\n return *this;\n }\n endian_store& operator++() {\n *this += 1;\n return *this;\n } // ++a\n endian_store operator++(int) {\n *this += 1;\n return (*this - 1);\n } // a++\n endian_store& operator--() {\n *this -= 1;\n return *this;\n } // --a\n endian_store operator--(int) {\n *this -= 1;\n return (*this + 1);\n } // a--\n\n T value;\n};\n\ntemplate \nusing be = endian_store;\ntemplate \nusing le = endian_store;\n\n} // namespace xe\n\n#endif // XENIA_BASE_BYTE_ORDER_H_\n" }, { "path": "src/xenia/base/byte_stream.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/byte_stream.h\"\n\n#include \n\n#include \"xenia/base/assert.h\"\n\nnamespace xe {\n\nByteStream::ByteStream(uint8_t* data, size_t data_length, size_t offset)\n : data_(data), data_length_(data_length), offset_(offset) {}\n\nByteStream::~ByteStream() = default;\n\nvoid ByteStream::Advance(size_t num_bytes) {\n assert_true(offset_ + num_bytes <= data_length_);\n offset_ += num_bytes;\n}\n\nvoid ByteStream::Read(uint8_t* buf, size_t len) {\n assert_true(offset_ + len <= data_length_);\n std::memcpy(buf, data_ + offset_, len);\n Advance(len);\n}\n\nvoid ByteStream::Write(const uint8_t* buf, size_t len) {\n assert_true(offset_ + len <= data_length_);\n std::memcpy(data_ + offset_, buf, len);\n Advance(len);\n}\n\ntemplate <>\nstd::string ByteStream::Read() {\n std::string str;\n uint32_t len = Read();\n str.resize(len);\n Read(reinterpret_cast(&str[0]), len);\n return str;\n}\n\ntemplate <>\nstd::u16string ByteStream::Read() {\n std::u16string str;\n size_t len = Read();\n str.resize(len);\n Read(reinterpret_cast(&str[0]), len * 2);\n return str;\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/byte_stream.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_BYTE_STREAM_H_\n#define XENIA_BASE_BYTE_STREAM_H_\n\n#include \n#include \n\nnamespace xe {\n\nclass ByteStream {\n public:\n ByteStream(uint8_t* data, size_t data_length, size_t offset = 0);\n ~ByteStream();\n\n void Advance(size_t num_bytes);\n void Read(uint8_t* buf, size_t len);\n void Write(const uint8_t* buf, size_t len);\n\n void Read(void* buf, size_t len) {\n return Read(reinterpret_cast(buf), len);\n }\n void Write(const void* buf, size_t len) {\n return Write(reinterpret_cast(buf), len);\n }\n\n const uint8_t* data() const { return data_; }\n uint8_t* data() { return data_; }\n size_t data_length() const { return data_length_; }\n\n size_t offset() const { return offset_; }\n void set_offset(size_t offset) { offset_ = offset; }\n\n template \n T Read() {\n T data;\n Read(reinterpret_cast(&data), sizeof(T));\n\n return data;\n }\n\n template \n void Write(T data) {\n Write(reinterpret_cast(&data), sizeof(T));\n }\n\n void Write(const std::string_view str) {\n Write(uint32_t(str.length()));\n Write(str.data(), str.length() * sizeof(char));\n }\n\n void Write(const std::u16string_view str) {\n Write(uint32_t(str.length()));\n Write(str.data(), str.length() * sizeof(char16_t));\n }\n\n private:\n uint8_t* data_ = nullptr;\n size_t data_length_ = 0;\n size_t offset_ = 0;\n};\n\ntemplate <>\nstd::string ByteStream::Read();\n\ntemplate <>\nstd::u16string ByteStream::Read();\n\n} // namespace xe\n\n#endif // XENIA_BASE_BYTE_STREAM_H_\n" }, { "path": "src/xenia/base/chrono.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_CHRONO_H_\n#define XENIA_BASE_CHRONO_H_\n\n#include \n#include \n\n// https://github.com/HowardHinnant/date/commit/5ba1c1ad8514362dba596f228eb20eb13f63d948#r33275526\n#define HAS_UNCAUGHT_EXCEPTIONS 1\n#include \"third_party/date/include/date/tz.h\"\n\n#include \"xenia/base/clock.h\"\n\nnamespace xe {\nusing hundrednano = std::ratio<1, 10000000>;\n\nnamespace chrono {\n\nusing hundrednanoseconds = std::chrono::duration;\n\n// https://docs.microsoft.com/en-us/windows/win32/sysinfo/converting-a-time-t-value-to-a-file-time\n// Don't forget the 89 leap days.\nstatic constexpr std::chrono::seconds seconds_1601_to_1970 =\n (396 * 365 + 89) * std::chrono::seconds(60 * 60 * 24);\n\n// TODO(JoelLinn) define xstead_clock xsystem_clock etc.\n\nnamespace internal {\n// Trick to reduce code duplication and keep all the chrono template magic\n// working\nenum class Domain {\n // boring host clock:\n Host,\n // adheres to guest scaling (differrent speed, changing clock drift etc):\n Guest\n};\n\ntemplate \nstruct NtSystemClock {\n using rep = int64_t;\n using period = hundrednano;\n using duration = hundrednanoseconds;\n using time_point = std::chrono::time_point>;\n // This really depends on the context the clock is used in:\n // static constexpr bool is_steady = false;\n\n public:\n // The delta between std::chrono::system_clock (Jan 1 1970) and NT file\n // time (Jan 1 1601), in seconds. In the spec std::chrono::system_clock's\n // epoch is undefined, but C++20 cements it as Jan 1 1970.\n static constexpr std::chrono::seconds unix_epoch_delta() {\n using std::chrono::steady_clock;\n auto filetime_epoch = date::year{1601} / date::month{1} / date::day{1};\n auto system_clock_epoch = date::year{1970} / date::month{1} / date::day{1};\n auto fp = static_cast(\n static_cast(filetime_epoch));\n auto sp = static_cast(\n static_cast(system_clock_epoch));\n return fp.time_since_epoch() - sp.time_since_epoch();\n }\n\n public:\n static constexpr uint64_t to_file_time(time_point const& tp) noexcept {\n return static_cast(tp.time_since_epoch().count());\n }\n\n static constexpr time_point from_file_time(uint64_t const& tp) noexcept {\n return time_point{duration{tp}};\n }\n\n // To convert XSystemClock to sys, do clock_cast(tp) first\n // SFINAE hack https://stackoverflow.com/a/58813009\n template \n static constexpr std::enable_if_t\n to_sys(const time_point& tp) {\n using sys_duration = std::chrono::system_clock::duration;\n using sys_time = std::chrono::system_clock::time_point;\n\n auto dp = tp;\n dp += unix_epoch_delta();\n auto cdp = std::chrono::time_point_cast(dp);\n return sys_time{cdp.time_since_epoch()};\n }\n\n template \n static constexpr std::enable_if_t\n from_sys(const std::chrono::system_clock::time_point& tp) {\n auto ctp = std::chrono::time_point_cast(tp);\n auto dp = time_point{ctp.time_since_epoch()};\n dp -= unix_epoch_delta();\n return dp;\n }\n\n [[nodiscard]] static time_point now() noexcept {\n if constexpr (domain_ == Domain::Host) {\n // QueryHostSystemTime() returns windows epoch times even on POSIX\n return from_file_time(Clock::QueryHostSystemTime());\n } else if constexpr (domain_ == Domain::Guest) {\n return from_file_time(Clock::QueryGuestSystemTime());\n }\n }\n};\n} // namespace internal\n\n// Unscaled system clock which can be used for filetime <-> system_clock\n// conversion\nusing WinSystemClock = internal::NtSystemClock;\n\n// Guest system clock, scaled\nusing XSystemClock = internal::NtSystemClock;\n\n} // namespace chrono\n} // namespace xe\n\nnamespace date {\n\ntemplate <>\nstruct clock_time_conversion<::xe::chrono::WinSystemClock,\n ::xe::chrono::XSystemClock> {\n using WClock_ = ::xe::chrono::WinSystemClock;\n using XClock_ = ::xe::chrono::XSystemClock;\n\n template \n typename WClock_::time_point operator()(\n const std::chrono::time_point& t) const {\n // Consult chrono_steady_cast.h for explanation on this:\n std::atomic_thread_fence(std::memory_order_acq_rel);\n auto w_now = WClock_::now();\n auto x_now = XClock_::now();\n std::atomic_thread_fence(std::memory_order_acq_rel);\n\n auto delta = (t - x_now);\n if (!::cvars::clock_no_scaling) {\n delta = std::chrono::floor(\n delta * xe::Clock::guest_time_scalar());\n }\n return w_now + delta;\n }\n};\n\ntemplate <>\nstruct clock_time_conversion<::xe::chrono::XSystemClock,\n ::xe::chrono::WinSystemClock> {\n using WClock_ = ::xe::chrono::WinSystemClock;\n using XClock_ = ::xe::chrono::XSystemClock;\n\n template \n typename XClock_::time_point operator()(\n const std::chrono::time_point& t) const {\n // Consult chrono_steady_cast.h for explanation on this:\n std::atomic_thread_fence(std::memory_order_acq_rel);\n auto w_now = WClock_::now();\n auto x_now = XClock_::now();\n std::atomic_thread_fence(std::memory_order_acq_rel);\n\n xe::chrono::hundrednanoseconds delta = (t - w_now);\n if (!::cvars::clock_no_scaling) {\n delta = std::chrono::floor(\n delta / xe::Clock::guest_time_scalar());\n }\n return x_now + delta;\n }\n};\n\n} // namespace date\n\n#endif\n" }, { "path": "src/xenia/base/chrono_steady_cast.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_CHRONO_STEADY_CAST_H_\n#define XENIA_BASE_CHRONO_STEADY_CAST_H_\n\n#include \n\n#include \"xenia/base/chrono.h\"\n\n// This is in a separate header because casting to and from steady time points\n// usually doesn't make sense and is imprecise. However, NT uses the FileTime\n// epoch as a steady clock in waits. In such cases, include this header and use\n// clock_cast<>().\n\nnamespace date {\n\n// This conveniently works only for Host time domain because Guest needs\n// additional scaling. Convert XSystemClock to WinSystemClock first if\n// necessary.\ntemplate <>\nstruct clock_time_conversion<::xe::chrono::WinSystemClock,\n std::chrono::steady_clock> {\n // using NtSystemClock_ = ::xe::chrono::internal::NtSystemClock;\n using WinSystemClock_ = ::xe::chrono::WinSystemClock;\n using steady_clock_ = std::chrono::steady_clock;\n\n template \n typename WinSystemClock_::time_point operator()(\n const std::chrono::time_point& t) const {\n // Since there is no known epoch for steady_clock and even if, since it can\n // progress differently than other common clocks (e.g. stopping when the\n // computer is suspended), we need to use now() which introduces\n // imprecision.\n // Memory fences to keep the clock fetches close together to\n // minimize drift. This pattern was benchmarked to give the lowest\n // conversion error: error = sty_tpoint -\n // clock_cast(clock_cast(sty_tpoint));\n std::atomic_thread_fence(std::memory_order_acq_rel);\n auto steady_now = steady_clock_::now();\n auto nt_now = WinSystemClock_::now();\n std::atomic_thread_fence(std::memory_order_acq_rel);\n\n auto delta = std::chrono::floor(t - steady_now);\n return nt_now + delta;\n }\n};\n\ntemplate <>\nstruct clock_time_conversion {\n using WinSystemClock_ = ::xe::chrono::WinSystemClock;\n using steady_clock_ = std::chrono::steady_clock;\n\n template \n steady_clock_::time_point operator()(\n const std::chrono::time_point& t) const {\n std::atomic_thread_fence(std::memory_order_acq_rel);\n auto steady_now = steady_clock_::now();\n auto nt_now = WinSystemClock_::now();\n std::atomic_thread_fence(std::memory_order_acq_rel);\n\n auto delta = t - nt_now;\n return steady_now + delta;\n }\n};\n\n} // namespace date\n\n#endif\n" }, { "path": "src/xenia/base/clock.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2019 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/clock.h\"\n\n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/math.h\"\n\nDEFINE_bool(clock_no_scaling, false,\n \"Disable scaling code. Time management and locking is bypassed. \"\n \"Guest system time is directly pulled from host.\",\n \"CPU\");\nDEFINE_bool(clock_source_raw, false,\n \"Use the RDTSC instruction as the time source. \"\n \"Host CPU must support invariant TSC.\",\n \"CPU\");\n\nnamespace xe {\n\n// Time scalar applied to all time operations.\ndouble guest_time_scalar_ = 1.0;\n// Tick frequency of guest.\nuint64_t guest_tick_frequency_ = Clock::host_tick_frequency_platform();\n// Base FILETIME of the guest system from app start.\nuint64_t guest_system_time_base_ = Clock::QueryHostSystemTime();\n// Combined time and frequency ratio between host and guest.\n// Split in numerator (first) and denominator (second).\n// Computed by RecomputeGuestTickScalar.\nstd::pair guest_tick_ratio_ = std::make_pair(1, 1);\n\n// Native guest ticks.\nuint64_t last_guest_tick_count_ = 0;\n// Last sampled host tick count.\nuint64_t last_host_tick_count_ = Clock::QueryHostTickCount();\n// Mutex to ensure last_host_tick_count_ and last_guest_tick_count_ are in sync\nstd::mutex tick_mutex_;\n\nvoid RecomputeGuestTickScalar() {\n // Create a rational number with numerator (first) and denominator (second)\n auto frac =\n std::make_pair(guest_tick_frequency_, Clock::QueryHostTickFrequency());\n // Doing it this way ensures we don't mess up our frequency scaling and\n // precisely controls the precision the guest_time_scalar_ can have.\n if (guest_time_scalar_ > 1.0) {\n frac.first *= static_cast(guest_time_scalar_ * 10.0);\n frac.second *= 10;\n } else {\n frac.first *= 10;\n frac.second *= static_cast(10.0 / guest_time_scalar_);\n }\n // Keep this a rational calculation and reduce the fraction\n reduce_fraction(frac);\n\n std::lock_guard lock(tick_mutex_);\n guest_tick_ratio_ = frac;\n}\n\n// Update the guest timer for all threads.\n// Return a copy of the value so locking is reduced.\nuint64_t UpdateGuestClock() {\n uint64_t host_tick_count = Clock::QueryHostTickCount();\n\n if (cvars::clock_no_scaling) {\n // Nothing to update, calculate on the fly\n return host_tick_count * guest_tick_ratio_.first / guest_tick_ratio_.second;\n }\n\n std::unique_lock lock(tick_mutex_, std::defer_lock);\n if (lock.try_lock()) {\n // Translate host tick count to guest tick count.\n uint64_t host_tick_delta = host_tick_count > last_host_tick_count_\n ? host_tick_count - last_host_tick_count_\n : 0;\n last_host_tick_count_ = host_tick_count;\n uint64_t guest_tick_delta =\n host_tick_delta * guest_tick_ratio_.first / guest_tick_ratio_.second;\n last_guest_tick_count_ += guest_tick_delta;\n return last_guest_tick_count_;\n } else {\n // Wait until another thread has finished updating the clock.\n lock.lock();\n return last_guest_tick_count_;\n }\n}\n\n// Offset of the current guest system file time relative to the guest base time.\ninline uint64_t QueryGuestSystemTimeOffset() {\n if (cvars::clock_no_scaling) {\n return Clock::QueryHostSystemTime() - guest_system_time_base_;\n }\n\n auto guest_tick_count = UpdateGuestClock();\n\n uint64_t numerator = 10000000; // 100ns/10MHz resolution\n uint64_t denominator = guest_tick_frequency_;\n reduce_fraction(numerator, denominator);\n\n return guest_tick_count * numerator / denominator;\n}\n\nuint64_t Clock::QueryHostTickFrequency() {\n#if XE_CLOCK_RAW_AVAILABLE\n if (cvars::clock_source_raw) {\n return host_tick_frequency_raw();\n }\n#endif\n return host_tick_frequency_platform();\n}\nuint64_t Clock::QueryHostTickCount() {\n#if XE_CLOCK_RAW_AVAILABLE\n if (cvars::clock_source_raw) {\n return host_tick_count_raw();\n }\n#endif\n return host_tick_count_platform();\n}\n\ndouble Clock::guest_time_scalar() { return guest_time_scalar_; }\n\nvoid Clock::set_guest_time_scalar(double scalar) {\n if (cvars::clock_no_scaling) {\n return;\n }\n\n guest_time_scalar_ = scalar;\n RecomputeGuestTickScalar();\n}\n\nstd::pair Clock::guest_tick_ratio() {\n std::lock_guard lock(tick_mutex_);\n return guest_tick_ratio_;\n}\n\nuint64_t Clock::guest_tick_frequency() { return guest_tick_frequency_; }\n\nvoid Clock::set_guest_tick_frequency(uint64_t frequency) {\n guest_tick_frequency_ = frequency;\n RecomputeGuestTickScalar();\n}\n\nuint64_t Clock::guest_system_time_base() { return guest_system_time_base_; }\n\nvoid Clock::set_guest_system_time_base(uint64_t time_base) {\n guest_system_time_base_ = time_base;\n}\n\nuint64_t Clock::QueryGuestTickCount() {\n auto guest_tick_count = UpdateGuestClock();\n return guest_tick_count;\n}\n\nuint64_t Clock::QueryGuestSystemTime() {\n if (cvars::clock_no_scaling) {\n return Clock::QueryHostSystemTime();\n }\n\n auto guest_system_time_offset = QueryGuestSystemTimeOffset();\n return guest_system_time_base_ + guest_system_time_offset;\n}\n\nuint32_t Clock::QueryGuestUptimeMillis() {\n return static_cast(\n std::min(QueryGuestSystemTimeOffset() / 10000,\n std::numeric_limits::max()));\n}\n\nvoid Clock::SetGuestSystemTime(uint64_t system_time) {\n if (cvars::clock_no_scaling) {\n // Time is fixed to host time.\n return;\n }\n\n // Query the filetime offset to calculate a new base time.\n auto guest_system_time_offset = QueryGuestSystemTimeOffset();\n guest_system_time_base_ = system_time - guest_system_time_offset;\n}\n\nuint32_t Clock::ScaleGuestDurationMillis(uint32_t guest_ms) {\n if (cvars::clock_no_scaling) {\n return guest_ms;\n }\n\n constexpr uint64_t max = std::numeric_limits::max();\n\n if (guest_ms >= max) {\n return max;\n } else if (!guest_ms) {\n return 0;\n }\n uint64_t scaled_ms = static_cast(\n (static_cast(guest_ms) * guest_time_scalar_));\n return static_cast(std::min(scaled_ms, max));\n}\n\nint64_t Clock::ScaleGuestDurationFileTime(int64_t guest_file_time) {\n if (cvars::clock_no_scaling) {\n return static_cast(guest_file_time);\n }\n\n if (!guest_file_time) {\n return 0;\n } else if (guest_file_time > 0) {\n // Absolute time.\n uint64_t guest_time = Clock::QueryGuestSystemTime();\n int64_t relative_time = guest_file_time - static_cast(guest_time);\n int64_t scaled_time =\n static_cast(relative_time * guest_time_scalar_);\n return static_cast(guest_time) + scaled_time;\n } else {\n // Relative time.\n uint64_t scaled_file_time = static_cast(\n (static_cast(guest_file_time) * guest_time_scalar_));\n // TODO(benvanik): check for overflow?\n return scaled_file_time;\n }\n}\n\nvoid Clock::ScaleGuestDurationTimeval(int32_t* tv_sec, int32_t* tv_usec) {\n if (cvars::clock_no_scaling) {\n return;\n }\n\n uint64_t scaled_sec = static_cast(static_cast(*tv_sec) *\n guest_time_scalar_);\n uint64_t scaled_usec = static_cast(static_cast(*tv_usec) *\n guest_time_scalar_);\n if (scaled_usec > std::numeric_limits::max()) {\n uint64_t overflow_sec = scaled_usec / 1000000;\n scaled_usec -= overflow_sec * 1000000;\n scaled_sec += overflow_sec;\n }\n *tv_sec = int32_t(scaled_sec);\n *tv_usec = int32_t(scaled_usec);\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/clock.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_CLOCK_H_\n#define XENIA_BASE_CLOCK_H_\n\n#include \n#include \n\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/platform.h\"\n\n#if XE_ARCH_AMD64\n#define XE_CLOCK_RAW_AVAILABLE 1\n#endif\n\nDECLARE_bool(clock_no_scaling);\nDECLARE_bool(clock_source_raw);\n\nnamespace xe {\n\n// chrono APIs in xenia/base/chrono.h are preferred\n\nclass Clock {\n public:\n // Host ticks-per-second. Generally QueryHostTickFrequency should be used.\n // Either from platform suplied time source or from hardware directly.\n static uint64_t host_tick_frequency_platform();\n#if XE_CLOCK_RAW_AVAILABLE\n static uint64_t host_tick_frequency_raw();\n#endif\n // Host tick count. Generally QueryHostTickCount() should be used.\n static uint64_t host_tick_count_platform();\n#if XE_CLOCK_RAW_AVAILABLE\n static uint64_t host_tick_count_raw();\n#endif\n\n // Queries the host tick frequency.\n static uint64_t QueryHostTickFrequency();\n // Queries the current host tick count.\n static uint64_t QueryHostTickCount();\n // Host time, in FILETIME format.\n static uint64_t QueryHostSystemTime();\n // Queries the milliseconds since the host began.\n static uint64_t QueryHostUptimeMillis();\n\n // Guest time scalar.\n static double guest_time_scalar();\n // Sets the guest time scalar, adjusting tick and wall clock speed.\n // Ex: 1x=normal, 2x=double speed, 1/2x=half speed.\n static void set_guest_time_scalar(double scalar);\n // Get the tick ration between host and guest including time scaling if set.\n static std::pair guest_tick_ratio();\n // Guest ticks-per-second.\n static uint64_t guest_tick_frequency();\n // Sets the guest ticks-per-second.\n static void set_guest_tick_frequency(uint64_t frequency);\n // Time based used for the guest system time.\n static uint64_t guest_system_time_base();\n // Sets the guest time base, used for computing the system time.\n // By default this is the current system time.\n static void set_guest_system_time_base(uint64_t time_base);\n\n // Queries the current guest tick count, accounting for frequency adjustment\n // and scaling.\n static uint64_t QueryGuestTickCount();\n // Queries the guest time, in FILETIME format, accounting for scaling.\n static uint64_t QueryGuestSystemTime();\n // Queries the milliseconds since the guest began, accounting for scaling.\n static uint32_t QueryGuestUptimeMillis();\n\n // Sets the system time of the guest.\n static void SetGuestSystemTime(uint64_t system_time);\n\n // Scales a time duration in milliseconds, from guest time.\n static uint32_t ScaleGuestDurationMillis(uint32_t guest_ms);\n // Scales a time duration in 100ns ticks like FILETIME, from guest time.\n static int64_t ScaleGuestDurationFileTime(int64_t guest_file_time);\n // Scales a time duration represented as a timeval, from guest time.\n static void ScaleGuestDurationTimeval(int32_t* tv_sec, int32_t* tv_usec);\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_CLOCK_H_\n" }, { "path": "src/xenia/base/clock_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/clock.h\"\n\nnamespace xe {\n\nuint64_t Clock::host_tick_frequency_platform() {\n timespec res;\n int error = clock_getres(CLOCK_MONOTONIC_RAW, &res);\n assert_zero(error);\n assert_zero(res.tv_sec); // Sub second resolution is required.\n\n // Convert nano seconds to hertz. Resolution is 1ns on most systems.\n return 1000000000ull / res.tv_nsec;\n}\n\nuint64_t Clock::host_tick_count_platform() {\n timespec tp;\n int error = clock_gettime(CLOCK_MONOTONIC_RAW, &tp);\n assert_zero(error);\n\n return tp.tv_nsec + tp.tv_sec * 1000000000ull;\n}\n\nuint64_t Clock::QueryHostSystemTime() {\n // https://docs.microsoft.com/en-us/windows/win32/sysinfo/converting-a-time-t-value-to-a-file-time\n constexpr uint64_t seconds_per_day = 3600 * 24;\n // Don't forget the 89 leap days.\n constexpr uint64_t seconds_1601_to_1970 =\n ((369 * 365 + 89) * seconds_per_day);\n\n timeval now;\n int error = gettimeofday(&now, nullptr);\n assert_zero(error);\n\n // NT systems use 100ns intervals.\n return static_cast(\n (static_cast(now.tv_sec) + seconds_1601_to_1970) * 10000000ull +\n now.tv_usec * 10);\n}\n\nuint64_t Clock::QueryHostUptimeMillis() {\n return host_tick_count_platform() * 1000 / host_tick_frequency_platform();\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/clock_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2019 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/clock.h\"\n\n#include \"xenia/base/platform_win.h\"\n\nnamespace xe {\n\nuint64_t Clock::host_tick_frequency_platform() {\n LARGE_INTEGER frequency;\n QueryPerformanceFrequency(&frequency);\n return frequency.QuadPart;\n}\n\nuint64_t Clock::host_tick_count_platform() {\n LARGE_INTEGER counter;\n uint64_t time = 0;\n if (QueryPerformanceCounter(&counter)) {\n time = counter.QuadPart;\n }\n return time;\n}\n\nuint64_t Clock::QueryHostSystemTime() {\n FILETIME t;\n GetSystemTimeAsFileTime(&t);\n return (uint64_t(t.dwHighDateTime) << 32) | t.dwLowDateTime;\n}\n\nuint64_t Clock::QueryHostUptimeMillis() {\n return host_tick_count_platform() * 1000 / host_tick_frequency_platform();\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/clock_x64.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/clock.h\"\n#include \"xenia/base/platform.h\"\n\n#if XE_ARCH_AMD64 && XE_CLOCK_RAW_AVAILABLE\n\n#include \"xenia/base/logging.h\"\n\n// Wrap all these different cpu compiler intrinsics.\n// So no inline assembler here and the compiler will remove the clutter.\n#if XE_COMPILER_MSVC\n#define xe_cpu_cpuid(level, eax, ebx, ecx, edx) \\\n { \\\n int __xe_cpuid_registers_[4]; \\\n __cpuid(__xe_cpuid_registers_, (level)); \\\n (eax) = static_cast(__xe_cpuid_registers_[0]); \\\n (ebx) = static_cast(__xe_cpuid_registers_[1]); \\\n (ecx) = static_cast(__xe_cpuid_registers_[2]); \\\n (edx) = static_cast(__xe_cpuid_registers_[3]); \\\n }\n#define xe_cpu_rdtsc() __rdtsc()\n#elif XE_COMPILER_CLANG || XE_COMPILER_GNUC\n#include \n#define xe_cpu_cpuid(level, eax, ebx, ecx, edx) \\\n __cpuid((level), (eax), (ebx), (ecx), (edx));\n#define xe_cpu_rdtsc() __rdtsc()\n#else\n#error \"No cpu instruction wrappers for current compiler implemented.\"\n#endif\n\n#define CLOCK_FATAL(msg) \\\n xe::FatalError(\"The raw clock source is not supported on your CPU.\\n\" msg \\\n \"\\n\" \\\n \"Set the cvar 'clock_source_raw' to 'false'.\");\n\nnamespace xe {\n// Getting the TSC frequency can be a bit tricky. This method here only works on\n// Intel as it seems. There is no easy way to get the frequency outside of ring0\n// on AMD, so we fail gracefully if not possible.\nuint64_t Clock::host_tick_frequency_raw() {\n uint32_t eax, ebx, ecx, edx;\n\n // 00H Get max supported cpuid level.\n xe_cpu_cpuid(0x0, eax, ebx, ecx, edx);\n auto max_cpuid = eax;\n // 80000000H Get max extended cpuid level\n xe_cpu_cpuid(0x80000000, eax, ebx, ecx, edx);\n auto max_cpuid_ex = eax;\n\n // 80000007H Get extended power feature info\n if (max_cpuid_ex >= 0x80000007) {\n xe_cpu_cpuid(0x80000007, eax, ebx, ecx, edx);\n // Invariant TSC bit at position 8\n auto tsc_invariant = edx & (1 << 8);\n // If the TSC is not invariant it will change its frequency with power\n // states and across cores.\n if (!tsc_invariant) {\n CLOCK_FATAL(\"The CPU has no invariant TSC.\");\n return 0;\n }\n } else {\n CLOCK_FATAL(\"Unclear if the CPU has an invariant TSC.\")\n return 0;\n }\n\n if (max_cpuid >= 0x15) {\n // 15H Get TSC/Crystal ratio and Crystal Hz.\n xe_cpu_cpuid(0x15, eax, ebx, ecx, edx);\n uint64_t ratio_num = ebx;\n uint64_t ratio_den = eax;\n uint64_t cryst_freq = ecx;\n // For some CPUs, Crystal frequency is not reported.\n if (ratio_num && ratio_den && cryst_freq) {\n // If it is, calculate the TSC frequency\n return cryst_freq * ratio_num / ratio_den;\n }\n }\n\n if (max_cpuid >= 0x16) {\n // 16H Get CPU base frequency MHz in EAX.\n xe_cpu_cpuid(0x16, eax, ebx, ecx, edx);\n uint64_t cpu_base_freq = static_cast(eax) * 1000000;\n assert(cpu_base_freq);\n return cpu_base_freq;\n }\n\n CLOCK_FATAL(\"The clock frequency could not be determined.\");\n return 0;\n}\n\nuint64_t Clock::host_tick_count_raw() { return xe_cpu_rdtsc(); }\n\n} // namespace xe\n\n#endif\n" }, { "path": "src/xenia/base/console.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_CONSOLE_H_\n#define XENIA_BASE_CONSOLE_H_\n\nnamespace xe {\n\n// Returns true if there is a user-visible console attached to receive stdout.\nbool has_console_attached();\n\nvoid AttachConsole();\n\n} // namespace xe\n\n#endif // XENIA_BASE_CONSOLE_H_\n" }, { "path": "src/xenia/base/console_app_main.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_CONSOLE_APP_MAIN_H_\n#define XENIA_BASE_CONSOLE_APP_MAIN_H_\n\n#include \n#include \n\nnamespace xe {\n\n// Extern defined by user code. This must be present for the application to\n// launch.\nstruct ConsoleAppEntryInfo {\n std::string name;\n int (*entry_point)(const std::vector& args);\n bool transparent_options; // no argument parsing\n std::string positional_usage;\n std::vector positional_options;\n};\nConsoleAppEntryInfo GetConsoleAppEntryInfo();\n\n// TODO(Triang3l): Multiple console app entry points on Android when a console\n// activity is added. This is currently for individual executables running on\n// Termux.\n\n#define XE_DEFINE_CONSOLE_APP(name, entry_point, positional_usage, ...) \\\n xe::ConsoleAppEntryInfo xe::GetConsoleAppEntryInfo() { \\\n std::initializer_list positional_options = {__VA_ARGS__}; \\\n return xe::ConsoleAppEntryInfo{ \\\n name, entry_point, false, positional_usage, \\\n std::vector(std::move(positional_options))}; \\\n }\n\n// TODO(Joel Linn): Add some way to filter consumed arguments in\n// cvar::ParseLaunchArguments()\n#define XE_DEFINE_CONSOLE_APP_TRANSPARENT(name, entry_point) \\\n xe::ConsoleAppEntryInfo xe::GetConsoleAppEntryInfo() { \\\n return xe::ConsoleAppEntryInfo{name, entry_point, true, std::string(), \\\n std::vector()}; \\\n }\n\n} // namespace xe\n\n#endif // XENIA_BASE_CONSOLE_APP_MAIN_H_\n" }, { "path": "src/xenia/base/console_app_main_android.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n#include \"xenia/base/console_app_main.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/main_android.h\"\n\nextern \"C\" int main(int argc, char** argv) {\n xe::ConsoleAppEntryInfo entry_info = xe::GetConsoleAppEntryInfo();\n\n if (!entry_info.transparent_options) {\n cvar::ParseLaunchArguments(argc, argv, entry_info.positional_usage,\n entry_info.positional_options);\n }\n\n // Initialize Android globals, including logging. Needs parsed cvars.\n // TODO(Triang3l): Obtain the actual API level.\n xe::InitializeAndroidAppFromMainThread(__ANDROID_API__, nullptr, nullptr,\n nullptr);\n\n std::vector args;\n for (int n = 0; n < argc; n++) {\n args.emplace_back(argv[n]);\n }\n\n int result = entry_info.entry_point(args);\n\n xe::ShutdownAndroidAppFromMainThread();\n\n return result;\n}\n" }, { "path": "src/xenia/base/console_app_main_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n#include \"xenia/base/console_app_main.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/logging.h\"\n\nextern \"C\" int main(int argc, char** argv) {\n xe::ConsoleAppEntryInfo entry_info = xe::GetConsoleAppEntryInfo();\n\n if (!entry_info.transparent_options) {\n cvar::ParseLaunchArguments(argc, argv, entry_info.positional_usage,\n entry_info.positional_options);\n }\n\n // Initialize logging. Needs parsed cvars.\n xe::InitializeLogging(entry_info.name);\n\n std::vector args;\n for (int n = 0; n < argc; n++) {\n args.emplace_back(argv[n]);\n }\n\n int result = entry_info.entry_point(args);\n\n xe::ShutdownLogging();\n\n return result;\n}\n" }, { "path": "src/xenia/base/console_app_main_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n#include \"xenia/base/console_app_main.h\"\n#include \"xenia/base/main_win.h\"\n\n// A wide character entry point is required for functions like _get_wpgmptr.\nint wmain(int argc_ignored, wchar_t** argv_ignored) {\n xe::ConsoleAppEntryInfo entry_info = xe::GetConsoleAppEntryInfo();\n\n std::vector args;\n if (!xe::ParseWin32LaunchArguments(entry_info.transparent_options,\n entry_info.positional_usage,\n entry_info.positional_options, &args)) {\n return EXIT_FAILURE;\n }\n\n int result = xe::InitializeWin32App(entry_info.name);\n if (result) {\n return result;\n }\n\n result = entry_info.entry_point(args);\n\n xe::ShutdownWin32App();\n\n return result;\n}\n" }, { "path": "src/xenia/base/console_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n#include \"xenia/base/console.h\"\n\nnamespace xe {\n\nbool has_console_attached() { return isatty(fileno(stdin)) == 1; }\n\nvoid AttachConsole() {}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/console_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n#include \n#include \n\n#include \"xenia/base/console.h\"\n#include \"xenia/base/platform_win.h\"\n\nnamespace xe {\n\n// TODO(Triang3l): Set the default depending on the actual subsystem. Currently\n// it inhibits message boxes.\nstatic bool has_console_attached_ = true;\n\nbool has_console_attached() { return has_console_attached_; }\n\nstatic bool has_shell_environment_variable() {\n size_t size = 0;\n // Check if SHELL exists\n // If it doesn't, then we are in a Windows Terminal\n auto error = _wgetenv_s(&size, nullptr, 0, L\"SHELL\");\n if (error) {\n return false;\n }\n return !!size;\n}\n\nvoid AttachConsole() {\n bool has_console = ::AttachConsole(ATTACH_PARENT_PROCESS) == TRUE;\n if (!has_console || !has_shell_environment_variable()) {\n // We weren't launched from a console, so just return.\n has_console_attached_ = false;\n return;\n }\n\n AllocConsole();\n\n has_console_attached_ = true;\n\n auto std_handle = (intptr_t)GetStdHandle(STD_OUTPUT_HANDLE);\n auto con_handle = _open_osfhandle(std_handle, _O_TEXT);\n auto fp = _fdopen(con_handle, \"w\");\n freopen_s(&fp, \"CONOUT$\", \"w\", stdout);\n\n std_handle = (intptr_t)GetStdHandle(STD_ERROR_HANDLE);\n con_handle = _open_osfhandle(std_handle, _O_TEXT);\n fp = _fdopen(con_handle, \"w\");\n freopen_s(&fp, \"CONOUT$\", \"w\", stderr);\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/cvar.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/cvar.h\"\n\n#define UTF_CPP_CPLUSPLUS 201703L\n#include \"third_party/utfcpp/source/utf8.h\"\n\n#include \"xenia/base/console.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/system.h\"\n#include \"xenia/base/utf8.h\"\n\nnamespace utfcpp = utf8;\n\nusing u8_citer = utfcpp::iterator;\n\nnamespace cvar {\n\ncxxopts::Options options(\"xenia\", \"Xbox 360 Emulator\");\nstd::map* CmdVars;\nstd::map* ConfigVars;\nstd::multimap* IConfigVarUpdate::updates_;\n\nvoid PrintHelpAndExit() {\n std::cout << options.help({\"\"}) << std::endl;\n std::cout << \"For the full list of command line arguments, see xenia.cfg.\"\n << std::endl;\n exit(0);\n}\n\nvoid ParseLaunchArguments(int& argc, char**& argv,\n const std::string_view positional_help,\n const std::vector& positional_options) {\n options.add_options()(\"help\", \"Prints help and exit.\");\n\n if (!CmdVars) {\n CmdVars = new std::map();\n }\n\n if (!ConfigVars) {\n ConfigVars = new std::map();\n }\n\n for (auto& it : *CmdVars) {\n auto cmdVar = it.second;\n cmdVar->AddToLaunchOptions(&options);\n }\n\n for (const auto& it : *ConfigVars) {\n auto configVar = it.second;\n configVar->AddToLaunchOptions(&options);\n }\n\n try {\n options.positional_help(std::string(positional_help));\n options.parse_positional(positional_options);\n\n auto result = options.parse(argc, argv);\n if (result.count(\"help\")) {\n xe::AttachConsole();\n if (xe::has_console_attached()) {\n PrintHelpAndExit();\n } else {\n xe::ShowSimpleMessageBox(xe::SimpleMessageBoxType::Help,\n options.help({\"\"}));\n exit(0);\n }\n }\n\n for (auto& it : *CmdVars) {\n auto cmdVar = static_cast(it.second);\n if (result.count(cmdVar->name())) {\n cmdVar->LoadFromLaunchOptions(&result);\n }\n }\n\n for (auto& it : *ConfigVars) {\n auto configVar = static_cast(it.second);\n if (result.count(configVar->name())) {\n configVar->LoadFromLaunchOptions(&result);\n }\n }\n } catch (const cxxopts::OptionException& e) {\n xe::AttachConsole();\n if (xe::has_console_attached()) {\n std::cout << e.what() << std::endl;\n PrintHelpAndExit();\n } else {\n std::string m =\n \"Invalid launch options were given.\\n\" + options.help({\"\"});\n xe::ShowSimpleMessageBox(xe::SimpleMessageBoxType::Error, m);\n exit(0);\n }\n }\n}\n\nnamespace toml {\n\nstd::string EscapeBasicString(const std::string_view view) {\n std::string result;\n auto begin = u8_citer(view.cbegin(), view.cbegin(), view.cend());\n auto end = u8_citer(view.cend(), view.cbegin(), view.cend());\n for (auto it = begin; it != end; ++it) {\n auto c = *it;\n if (c == '\\b') {\n result += u8\"\\\\b\";\n } else if (c == '\\t') {\n result += u8\"\\\\t\";\n } else if (c == '\\n') {\n result += u8\"\\\\n\";\n } else if (c == '\\f') {\n result += u8\"\\\\f\";\n } else if (c == '\\r') {\n result += u8\"\\\\r\";\n } else if (c == '\"') {\n result += u8\"\\\\\\\"\";\n } else if (c == '\\\\') {\n result += u8\"\\\\\\\\\";\n } else if (c < 0x20 || c == 0x7F) {\n if (c <= 0xFFFF) {\n result += fmt::format(u8\"\\\\u{:04X}\", c);\n } else {\n result += fmt::format(u8\"\\\\u{:08X}\", c);\n }\n } else {\n utfcpp::append(static_cast(c), result);\n }\n }\n return result;\n}\n\nstd::string EscapeMultilineBasicString(const std::string_view view) {\n std::string result;\n int quote_run = 0;\n auto begin = u8_citer(view.cbegin(), view.cbegin(), view.cend());\n auto end = u8_citer(view.cend(), view.cbegin(), view.cend());\n for (auto it = begin; it != end; ++it) {\n auto c = *it;\n if (quote_run > 0) {\n if (c == '\"') {\n ++quote_run;\n continue;\n }\n for (int i = 0; i < quote_run; ++i) {\n if ((i % 3) == 2) {\n result += u8\"\\\\\";\n }\n result += u8\"\\\"\";\n }\n quote_run = 0;\n }\n if (c == '\\b') {\n result += u8\"\\\\b\";\n } else if (c == '\\t' || c == '\\n') {\n result += c;\n } else if (c == '\\f') {\n result += u8\"\\\\f\";\n } else if (c == '\\r') {\n // Silently drop \\r.\n // result += c;\n } else if (c == '\"') {\n quote_run = 1;\n } else if (c == '\\\\') {\n result += u8\"\\\\\\\\\";\n } else if (c < 0x20 || c == 0x7F) {\n if (c <= 0xFFFF) {\n result += fmt::format(u8\"\\\\u{:04X}\", c);\n } else {\n result += fmt::format(u8\"\\\\u{:08X}\", c);\n }\n } else {\n utfcpp::append(static_cast(c), result);\n }\n }\n for (int i = 0; i < quote_run; ++i) {\n if ((i % 3) == 2) {\n result += u8\"\\\\\";\n }\n result += u8\"\\\"\";\n }\n return result;\n}\n\nstd::string EscapeString(const std::string_view view) {\n const auto multiline_chars = std::string_view(\"\\r\\n\");\n const auto escape_chars = std::string_view(\n \"\\0\\b\\v\\f\"\n \"\\x01\\x02\\x03\\x04\\x05\\x06\\x07\\x0E\\x0F\"\n \"\\x10\\x11\\x12\\x13\\x14\\x15\\x16\\x17\\x18\\x19\\x1A\\x1B\\x1C\\x1D\\x1E\\x1F\"\n \"'\"\n \"\\x7F\");\n\n if (xe::utf8::find_any_of(view, multiline_chars) == std::string_view::npos) {\n // single line\n if (xe::utf8::find_any_of(view, escape_chars) == std::string_view::npos) {\n return \"'\" + std::string(view) + \"'\";\n } else {\n return \"\\\"\" + toml::EscapeBasicString(view) + \"\\\"\";\n }\n } else {\n // multi line\n if (xe::utf8::find_any_of(view, escape_chars) == std::string_view::npos &&\n xe::utf8::find_first_of(view, u8\"'''\") == std::string_view::npos) {\n return \"'''\\n\" + std::string(view) + \"'''\";\n } else {\n return u8\"\\\"\\\"\\\"\\n\" + toml::EscapeMultilineBasicString(view) + u8\"\\\"\\\"\\\"\";\n }\n }\n}\n\n} // namespace toml\n\n} // namespace cvar\n" }, { "path": "src/xenia/base/cvar.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CVAR_H_\n#define XENIA_CVAR_H_\n\n#include \n#include \n#include \n#include \n\n#include \"third_party/cpptoml/include/cpptoml.h\"\n#include \"third_party/cxxopts/include/cxxopts.hpp\"\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/string_util.h\"\n\n#if XE_PLATFORM_ANDROID\n#include \n#endif // XE_PLATFORM_ANDROID\n\nnamespace cvar {\n\nnamespace toml {\nstd::string EscapeString(const std::string_view str);\n}\n\nclass ICommandVar {\n public:\n virtual ~ICommandVar() = default;\n virtual const std::string& name() const = 0;\n virtual const std::string& description() const = 0;\n virtual void UpdateValue() = 0;\n virtual void AddToLaunchOptions(cxxopts::Options* options) = 0;\n virtual void LoadFromLaunchOptions(cxxopts::ParseResult* result) = 0;\n};\n\nclass IConfigVar : virtual public ICommandVar {\n public:\n virtual const std::string& category() const = 0;\n virtual bool is_transient() const = 0;\n virtual std::string config_value() const = 0;\n virtual void LoadConfigValue(std::shared_ptr result) = 0;\n virtual void LoadGameConfigValue(std::shared_ptr result) = 0;\n virtual void ResetConfigValueToDefault() = 0;\n};\n\ntemplate \nclass CommandVar : virtual public ICommandVar {\n public:\n CommandVar(const char* name, T* default_value, const char* description);\n const std::string& name() const override;\n const std::string& description() const override;\n void AddToLaunchOptions(cxxopts::Options* options) override;\n void LoadFromLaunchOptions(cxxopts::ParseResult* result) override;\n void SetCommandLineValue(T val);\n T* current_value() { return current_value_; }\n\n protected:\n std::string name_;\n T default_value_;\n T* current_value_;\n std::unique_ptr commandline_value_;\n std::string description_;\n T Convert(std::string val);\n static std::string ToString(T val);\n void SetValue(T val);\n void UpdateValue() override;\n};\n#pragma warning(push)\n#pragma warning(disable : 4250)\ntemplate \nclass ConfigVar : public CommandVar, virtual public IConfigVar {\n public:\n ConfigVar(const char* name, T* default_value, const char* description,\n const char* category, bool is_transient);\n std::string config_value() const override;\n const T& GetTypedConfigValue() const;\n const std::string& category() const override;\n bool is_transient() const override;\n void AddToLaunchOptions(cxxopts::Options* options) override;\n void LoadConfigValue(std::shared_ptr result) override;\n void LoadGameConfigValue(std::shared_ptr result) override;\n void SetConfigValue(T val);\n void SetGameConfigValue(T val);\n // Changes the actual value used to the one specified, and also makes it the\n // one that will be stored when the global config is written next time. After\n // overriding, however, the next game config loaded may still change it.\n void OverrideConfigValue(T val);\n\n private:\n std::string category_;\n bool is_transient_;\n std::unique_ptr config_value_ = nullptr;\n std::unique_ptr game_config_value_ = nullptr;\n void UpdateValue() override;\n void ResetConfigValueToDefault() override;\n};\n\n#pragma warning(pop)\ntemplate \nconst std::string& CommandVar::name() const {\n return name_;\n}\ntemplate \nconst std::string& CommandVar::description() const {\n return description_;\n}\ntemplate \nvoid CommandVar::AddToLaunchOptions(cxxopts::Options* options) {\n options->add_options()(name_, description_, cxxopts::value());\n}\ntemplate <>\ninline void CommandVar::AddToLaunchOptions(\n cxxopts::Options* options) {\n options->add_options()(name_, description_, cxxopts::value());\n}\ntemplate \nvoid ConfigVar::AddToLaunchOptions(cxxopts::Options* options) {\n options->add_options(category_)(this->name_, this->description_,\n cxxopts::value());\n}\ntemplate <>\ninline void ConfigVar::AddToLaunchOptions(\n cxxopts::Options* options) {\n options->add_options(category_)(this->name_, this->description_,\n cxxopts::value());\n}\ntemplate \nvoid CommandVar::LoadFromLaunchOptions(cxxopts::ParseResult* result) {\n T value = (*result)[name_].template as();\n SetCommandLineValue(value);\n}\ntemplate <>\ninline void CommandVar::LoadFromLaunchOptions(\n cxxopts::ParseResult* result) {\n std::string value = (*result)[name_].template as();\n SetCommandLineValue(value);\n}\ntemplate \nvoid ConfigVar::LoadConfigValue(std::shared_ptr result) {\n SetConfigValue(*cpptoml::get_impl(result));\n}\ntemplate <>\ninline void ConfigVar::LoadConfigValue(\n std::shared_ptr result) {\n SetConfigValue(\n xe::utf8::fix_path_separators(*cpptoml::get_impl(result)));\n}\ntemplate \nvoid ConfigVar::LoadGameConfigValue(std::shared_ptr result) {\n SetGameConfigValue(*cpptoml::get_impl(result));\n}\ntemplate <>\ninline void ConfigVar::LoadGameConfigValue(\n std::shared_ptr result) {\n SetGameConfigValue(\n xe::utf8::fix_path_separators(*cpptoml::get_impl(result)));\n}\ntemplate \nCommandVar::CommandVar(const char* name, T* default_value,\n const char* description)\n : name_(name),\n default_value_(*default_value),\n description_(description),\n current_value_(default_value) {}\n\ntemplate \nConfigVar::ConfigVar(const char* name, T* default_value,\n const char* description, const char* category,\n bool is_transient)\n : CommandVar(name, default_value, description),\n category_(category),\n is_transient_(is_transient) {}\n\ntemplate \nvoid CommandVar::UpdateValue() {\n if (commandline_value_) return SetValue(*commandline_value_);\n return SetValue(default_value_);\n}\ntemplate \nvoid ConfigVar::UpdateValue() {\n if (this->commandline_value_) {\n return this->SetValue(*this->commandline_value_);\n }\n if (game_config_value_) return this->SetValue(*game_config_value_);\n if (config_value_) return this->SetValue(*config_value_);\n return this->SetValue(this->default_value_);\n}\ntemplate \nT CommandVar::Convert(std::string val) {\n return xe::string_util::from_string(val);\n}\ntemplate <>\ninline std::string CommandVar::Convert(std::string val) {\n return val;\n}\ntemplate <>\ninline std::filesystem::path CommandVar::Convert(\n std::string val) {\n return xe::to_path(val);\n}\n\ntemplate <>\ninline std::string CommandVar::ToString(bool val) {\n return val ? \"true\" : \"false\";\n}\ntemplate <>\ninline std::string CommandVar::ToString(std::string val) {\n return toml::EscapeString(val);\n}\ntemplate <>\ninline std::string CommandVar::ToString(\n std::filesystem::path val) {\n return toml::EscapeString(\n xe::utf8::fix_path_separators(xe::path_to_utf8(val), '/'));\n}\n\ntemplate \nstd::string CommandVar::ToString(T val) {\n // Use fmt::format instead of std::to_string for locale-neutral formatting of\n // floats, always with a period rather than a comma, which is treated as an\n // unidentified trailing character by cpptoml.\n return fmt::format(\"{}\", val);\n}\n\ntemplate \nvoid CommandVar::SetValue(T val) {\n *current_value_ = val;\n}\ntemplate \nconst std::string& ConfigVar::category() const {\n return category_;\n}\ntemplate \nbool ConfigVar::is_transient() const {\n return is_transient_;\n}\ntemplate \nstd::string ConfigVar::config_value() const {\n if (config_value_) return this->ToString(*config_value_);\n return this->ToString(this->default_value_);\n}\ntemplate \nconst T& ConfigVar::GetTypedConfigValue() const {\n return config_value_ ? *config_value_ : this->default_value_;\n}\ntemplate \nvoid CommandVar::SetCommandLineValue(const T val) {\n commandline_value_ = std::make_unique(val);\n UpdateValue();\n}\ntemplate \nvoid ConfigVar::SetConfigValue(T val) {\n config_value_ = std::make_unique(val);\n UpdateValue();\n}\ntemplate \nvoid ConfigVar::SetGameConfigValue(T val) {\n game_config_value_ = std::make_unique(val);\n UpdateValue();\n}\ntemplate \nvoid ConfigVar::OverrideConfigValue(T val) {\n config_value_ = std::make_unique(val);\n // The user explicitly changes the value at runtime and wants it to take\n // effect immediately. Drop everything with a higher priority. The next game\n // config load, however, may still change it.\n game_config_value_.reset();\n this->commandline_value_.reset();\n UpdateValue();\n}\ntemplate \nvoid ConfigVar::ResetConfigValueToDefault() {\n SetConfigValue(this->default_value_);\n}\n\n// CVars can be initialized before these, thus initialized on-demand using new.\nextern std::map* CmdVars;\nextern std::map* ConfigVars;\n\ninline void AddConfigVar(IConfigVar* cv) {\n if (!ConfigVars) {\n ConfigVars = new std::map;\n }\n ConfigVars->emplace(cv->name(), cv);\n}\ninline void AddCommandVar(ICommandVar* cv) {\n if (!CmdVars) {\n CmdVars = new std::map;\n }\n CmdVars->emplace(cv->name(), cv);\n}\nvoid ParseLaunchArguments(int& argc, char**& argv,\n const std::string_view positional_help,\n const std::vector& positional_options);\n#if XE_PLATFORM_ANDROID\nvoid ParseLaunchArgumentsFromAndroidBundle(jobject bundle);\n#endif // XE_PLATFORM_ANDROID\n\ntemplate \nIConfigVar* define_configvar(const char* name, T* default_value,\n const char* description, const char* category,\n bool is_transient) {\n IConfigVar* cfgvar = new ConfigVar(name, default_value, description,\n category, is_transient);\n AddConfigVar(cfgvar);\n return cfgvar;\n}\n\ntemplate \nICommandVar* define_cmdvar(const char* name, T* default_value,\n const char* description) {\n ICommandVar* cmdvar = new CommandVar(name, default_value, description);\n AddCommandVar(cmdvar);\n return cmdvar;\n}\n\n#define DEFINE_bool(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, false, bool)\n\n#define DEFINE_int32(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, false, int32_t)\n\n#define DEFINE_uint32(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, false, uint32_t)\n\n#define DEFINE_uint64(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, false, uint64_t)\n\n#define DEFINE_double(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, false, double)\n\n#define DEFINE_string(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, false, std::string)\n\n#define DEFINE_path(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, false, \\\n std::filesystem::path)\n\n#define DEFINE_transient_bool(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, true, bool)\n\n#define DEFINE_transient_string(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, true, std::string)\n\n#define DEFINE_transient_path(name, default_value, description, category) \\\n DEFINE_CVar(name, default_value, description, category, true, \\\n std::filesystem::path)\n\n#define DEFINE_CVar(name, default_value, description, category, is_transient, \\\n type) \\\n namespace cvars { \\\n type name = default_value; \\\n } \\\n namespace cv { \\\n static cvar::IConfigVar* const cv_##name = cvar::define_configvar( \\\n #name, &cvars::name, description, category, is_transient); \\\n }\n\n// CmdVars can only be strings for now, we don't need any others\n#define CmdVar(name, default_value, description) \\\n namespace cvars { \\\n std::string name = default_value; \\\n } \\\n namespace cv { \\\n static cvar::ICommandVar* const cv_##name = \\\n cvar::define_cmdvar(#name, &cvars::name, description); \\\n }\n\n#define DECLARE_bool(name) DECLARE_CVar(name, bool)\n\n#define DECLARE_int32(name) DECLARE_CVar(name, int32_t)\n\n#define DECLARE_uint32(name) DECLARE_CVar(name, uint32_t)\n\n#define DECLARE_uint64(name) DECLARE_CVar(name, uint64_t)\n\n#define DECLARE_double(name) DECLARE_CVar(name, double)\n\n#define DECLARE_string(name) DECLARE_CVar(name, std::string)\n\n#define DECLARE_path(name) DECLARE_CVar(name, std::filesystem::path)\n\n#define DECLARE_CVar(name, type) \\\n namespace cvars { \\\n extern type name; \\\n }\n\n#define ACCESS_CVar(name) (*cv::cv_##name)\n\n// dynamic_cast is needed because of virtual inheritance.\n#define OVERRIDE_CVar(name, type, value) \\\n dynamic_cast*>(&ACCESS_CVar(name)) \\\n ->OverrideConfigValue(value);\n\n#define OVERRIDE_bool(name, value) OVERRIDE_CVar(name, bool, value)\n\n#define OVERRIDE_int32(name, value) OVERRIDE_CVar(name, int32_t, value)\n\n#define OVERRIDE_uint32(name, value) OVERRIDE_CVar(name, uint32_t, value)\n\n#define OVERRIDE_uint64(name, value) OVERRIDE_CVar(name, uint64_t, value)\n\n#define OVERRIDE_double(name, value) OVERRIDE_CVar(name, double, value)\n\n#define OVERRIDE_string(name, value) OVERRIDE_CVar(name, std::string, value)\n\n#define OVERRIDE_path(name, value) \\\n OVERRIDE_CVar(name, std::filesystem::path, value)\n\n// Interface for changing the default value of a variable with auto-upgrading of\n// users' configs (to distinguish between a leftover old default and an explicit\n// override), without having to rename the variable.\n//\n// Two types of updates are supported:\n// - Changing the value of the variable (UPDATE_from_type) from an explicitly\n// specified previous default value to a new one, but keeping the\n// user-specified value if it was not the default, and thus explicitly\n// overridden.\n// - Changing the meaning / domain of the variable (UPDATE_from_any), when\n// previous user-specified overrides also stop making sense. Config variable\n// type changes are also considered this type of updates (though\n// UPDATE_from_type, if the new type doesn't match the previous one, is also\n// safe to use - it behaves like UPDATE_from_any in this case).\n//\n// Rules of using UPDATE_:\n// - Do not remove previous UPDATE_ entries (both typed and from-any) if you're\n// adding a new UPDATE_from_type.\n// This ensures that if the default was changed from 1 to 2 and then to 3,\n// both users who last launched Xenia when it was 1 and when it was 2 receive\n// the update (however, those who have explicitly changed it from 2 to 1 when\n// 2 was the default will have it kept at 1).\n// It's safe to remove the history before a new UPDATE_from_any, however.\n// - The date should preferably be in UTC+0 timezone.\n// - No other pull recent pull requests should have the same date (since builds\n// are made after every commit).\n// - IConfigVarUpdate::kLastCommittedUpdateDate must be updated - see the\n// comment near its declaration.\n\nconstexpr uint32_t MakeConfigVarUpdateDate(uint32_t year, uint32_t month,\n uint32_t day, uint32_t utc_hour) {\n // Written to the config as a decimal number - pack as decimal for user\n // readability.\n // Using 31 bits in the 3rd millennium already - don't add more digits.\n return utc_hour + day * 100 + month * 10000 + year * 1000000;\n}\n\nclass IConfigVarUpdate {\n public:\n // This global highest version constant is used to ensure that version (which\n // is stored as one value for the whole config file) is monotonically\n // increased when commits - primarily pull requests - are pushed to the main\n // branch.\n //\n // This is to prevent the following situation:\n // - Pull request #1 created on day 1.\n // - Pull request #2 created on day 2.\n // - Pull request #2 from day 2 merged on day 3.\n // - User launches the latest version on day 4.\n // CVar default changes from PR #2 (day 2) applied because the user's config\n // version is day 0, which is < 2.\n // User's config has day 2 version now.\n // - Pull request #1 from day 1 merged on day 5.\n // - User launches the latest version on day 5.\n // CVar default changes from PR #1 (day 1) IGNORED because the user's config\n // version is day 2, which is >= 1.\n //\n // If this constant is not updated, static_assert will be triggered for a new\n // DEFINE_, requiring this constant to be raised. But changing this will\n // result in merge conflicts in all other pull requests also changing cvar\n // defaults - before they're merged, they will need to be updated, which will\n // ensure monotonic growth of the versions of all cvars on the main branch. In\n // the example above, PR #1 will need to be updated before it's merged.\n //\n // If you've encountered a merge conflict here in your pull request:\n // 1) Update any UPDATE_s you've added in the pull request to the current\n // date.\n // 2) Change this value to the same date.\n // If you're reviewing a pull request with a change here, check if 1) has been\n // done by the submitter before merging.\n static constexpr uint32_t kLastCommittedUpdateDate =\n MakeConfigVarUpdateDate(2020, 12, 31, 13);\n\n virtual ~IConfigVarUpdate() = default;\n\n virtual void Apply() const = 0;\n\n static void ApplyUpdates(uint32_t config_date) {\n if (!updates_) {\n return;\n }\n auto it_end = updates_->end();\n for (auto it = updates_->upper_bound(config_date); it != it_end; ++it) {\n it->second->Apply();\n }\n }\n\n // More reliable than kLastCommittedUpdateDate for actual usage\n // (kLastCommittedUpdateDate is just a pull request merge order guard), though\n // usually should be the same, but kLastCommittedUpdateDate may not include\n // removal of cvars.\n static uint32_t GetLastUpdateDate() {\n return (updates_ && !updates_->empty()) ? updates_->crbegin()->first : 0;\n }\n\n protected:\n IConfigVarUpdate(IConfigVar* const& config_var, uint32_t year, uint32_t month,\n uint32_t day, uint32_t utc_hour)\n : config_var_(config_var) {\n if (!updates_) {\n updates_ = new std::multimap;\n }\n updates_->emplace(MakeConfigVarUpdateDate(year, month, day, utc_hour),\n this);\n }\n\n IConfigVar& config_var() const {\n assert_not_null(config_var_);\n return *config_var_;\n }\n\n private:\n // Reference to pointer to loosen initialization order requirements.\n IConfigVar* const& config_var_;\n\n // Updates can be initialized before these, thus initialized on demand using\n // `new`.\n static std::multimap* updates_;\n};\n\nclass ConfigVarUpdateFromAny : public IConfigVarUpdate {\n public:\n ConfigVarUpdateFromAny(IConfigVar* const& config_var, uint32_t year,\n uint32_t month, uint32_t day, uint32_t utc_hour)\n : IConfigVarUpdate(config_var, year, month, day, utc_hour) {}\n void Apply() const override { config_var().ResetConfigValueToDefault(); }\n};\n\ntemplate \nclass ConfigVarUpdate : public IConfigVarUpdate {\n public:\n ConfigVarUpdate(IConfigVar* const& config_var, uint32_t year, uint32_t month,\n uint32_t day, uint32_t utc_hour, const T& old_default_value)\n : IConfigVarUpdate(config_var, year, month, day, utc_hour),\n old_default_value_(old_default_value) {}\n void Apply() const override {\n IConfigVar& config_var_untyped = config_var();\n ConfigVar* config_var_typed =\n dynamic_cast*>(&config_var_untyped);\n // Update only from the previous default value if the same type,\n // unconditionally reset if the type has been changed.\n if (!config_var_typed ||\n config_var_typed->GetTypedConfigValue() == old_default_value_) {\n config_var_untyped.ResetConfigValueToDefault();\n }\n }\n\n private:\n T old_default_value_;\n};\n\n#define UPDATE_from_any(name, year, month, day, utc_hour) \\\n static_assert( \\\n cvar::MakeConfigVarUpdateDate(year, month, day, utc_hour) <= \\\n cvar::IConfigVarUpdate::kLastCommittedUpdateDate, \\\n \"A new config variable default value update was added - raise \" \\\n \"cvar::IConfigVarUpdate::kLastCommittedUpdateDate to the same date in \" \\\n \"base/cvar.h to ensure coherence between different pull requests \" \\\n \"updating config variable defaults.\"); \\\n namespace cv { \\\n static const cvar::ConfigVarUpdateFromAny \\\n update_##name_##year_##month_##day_##utc_hour(cv_##name, year, month, \\\n day, utc_hour); \\\n }\n\n#define UPDATE_CVar(name, year, month, day, utc_hour, old_default_value, type) \\\n static_assert( \\\n cvar::MakeConfigVarUpdateDate(year, month, day, utc_hour) <= \\\n cvar::IConfigVarUpdate::kLastCommittedUpdateDate, \\\n \"A new config variable default value update was added - raise \" \\\n \"cvar::IConfigVarUpdate::kLastCommittedUpdateDate to the same date in \" \\\n \"base/cvar.h to ensure coherence between different pull requests \" \\\n \"updating config variable defaults.\"); \\\n namespace cv { \\\n static const cvar::ConfigVarUpdate \\\n update_##name_##year_##month_##day_##utc_hour(cv_##name, year, month, \\\n day, utc_hour, \\\n old_default_value); \\\n }\n\n#define UPDATE_from_bool(name, year, month, day, utc_hour, old_default_value) \\\n UPDATE_CVar(name, year, month, day, utc_hour, old_default_value, bool)\n\n#define UPDATE_from_int32(name, year, month, day, utc_hour, old_default_value) \\\n UPDATE_CVar(name, year, month, day, utc_hour, old_default_value, int32_t)\n\n#define UPDATE_from_uint32(name, year, month, day, utc_hour, \\\n old_default_value) \\\n UPDATE_CVar(name, year, month, day, utc_hour, old_default_value, uint32_t)\n\n#define UPDATE_from_uint64(name, year, month, day, utc_hour, \\\n old_default_value) \\\n UPDATE_CVar(name, year, month, day, utc_hour, old_default_value, uint64_t)\n\n#define UPDATE_from_double(name, year, month, day, utc_hour, \\\n old_default_value) \\\n UPDATE_CVar(name, year, month, day, utc_hour, old_default_value, double)\n\n#define UPDATE_from_string(name, year, month, day, utc_hour, \\\n old_default_value) \\\n UPDATE_CVar(name, year, month, day, utc_hour, old_default_value, std::string)\n\n#define UPDATE_from_path(name, year, month, day, utc_hour, old_default_value) \\\n UPDATE_CVar(name, year, month, day, utc_hour, old_default_value, \\\n std::filesystem::path)\n\n} // namespace cvar\n\n#endif // XENIA_CVAR_H_\n" }, { "path": "src/xenia/base/cvar_android.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/main_android.h\"\n\nnamespace cvar {\n\nvoid ParseLaunchArgumentsFromAndroidBundle(jobject bundle) {\n if (!ConfigVars) {\n return;\n }\n\n JNIEnv* jni_env = xe::GetAndroidThreadJniEnv();\n if (!jni_env) {\n return;\n }\n\n jclass bundle_class = jni_env->GetObjectClass(bundle);\n if (!bundle_class) {\n return;\n }\n bool bundle_methods_obtained = true;\n jmethodID bundle_get_boolean =\n jni_env->GetMethodID(bundle_class, \"getBoolean\", \"(Ljava/lang/String;)Z\");\n bundle_methods_obtained &= (bundle_get_boolean != nullptr);\n jmethodID bundle_get_double =\n jni_env->GetMethodID(bundle_class, \"getDouble\", \"(Ljava/lang/String;)D\");\n bundle_methods_obtained &= (bundle_get_double != nullptr);\n jmethodID bundle_get_int =\n jni_env->GetMethodID(bundle_class, \"getInt\", \"(Ljava/lang/String;)I\");\n bundle_methods_obtained &= (bundle_get_int != nullptr);\n jmethodID bundle_get_long =\n jni_env->GetMethodID(bundle_class, \"getLong\", \"(Ljava/lang/String;)J\");\n bundle_methods_obtained &= (bundle_get_long != nullptr);\n jmethodID bundle_get_string = jni_env->GetMethodID(\n bundle_class, \"getString\", \"(Ljava/lang/String;)Ljava/lang/String;\");\n bundle_methods_obtained &= (bundle_get_string != nullptr);\n jmethodID bundle_key_set_method_id =\n jni_env->GetMethodID(bundle_class, \"keySet\", \"()Ljava/util/Set;\");\n bundle_methods_obtained &= (bundle_key_set_method_id != nullptr);\n if (!bundle_methods_obtained) {\n jni_env->DeleteLocalRef(bundle_class);\n return;\n }\n\n jobject key_set = jni_env->CallObjectMethod(bundle, bundle_key_set_method_id);\n if (!key_set) {\n jni_env->DeleteLocalRef(bundle_class);\n return;\n }\n\n jclass set_class = jni_env->GetObjectClass(key_set);\n if (!set_class) {\n jni_env->DeleteLocalRef(key_set);\n jni_env->DeleteLocalRef(bundle_class);\n return;\n }\n bool set_methods_obtained = true;\n jmethodID set_iterator_method_id =\n jni_env->GetMethodID(set_class, \"iterator\", \"()Ljava/util/Iterator;\");\n set_methods_obtained &= (set_iterator_method_id != nullptr);\n if (!set_methods_obtained) {\n jni_env->DeleteLocalRef(set_class);\n jni_env->DeleteLocalRef(key_set);\n jni_env->DeleteLocalRef(bundle_class);\n return;\n }\n\n jobject key_set_iterator =\n jni_env->CallObjectMethod(key_set, set_iterator_method_id);\n if (!key_set_iterator) {\n jni_env->DeleteLocalRef(set_class);\n jni_env->DeleteLocalRef(key_set);\n jni_env->DeleteLocalRef(bundle_class);\n return;\n }\n\n jclass iterator_class = jni_env->GetObjectClass(key_set_iterator);\n if (!iterator_class) {\n jni_env->DeleteLocalRef(key_set_iterator);\n jni_env->DeleteLocalRef(set_class);\n jni_env->DeleteLocalRef(key_set);\n jni_env->DeleteLocalRef(bundle_class);\n return;\n }\n bool iterator_methods_obtained = true;\n jmethodID iterator_has_next =\n jni_env->GetMethodID(iterator_class, \"hasNext\", \"()Z\");\n iterator_methods_obtained &= (iterator_has_next != nullptr);\n jmethodID iterator_next =\n jni_env->GetMethodID(iterator_class, \"next\", \"()Ljava/lang/Object;\");\n iterator_methods_obtained &= (iterator_next != nullptr);\n if (!iterator_methods_obtained) {\n jni_env->DeleteLocalRef(iterator_class);\n jni_env->DeleteLocalRef(key_set_iterator);\n jni_env->DeleteLocalRef(set_class);\n jni_env->DeleteLocalRef(key_set);\n jni_env->DeleteLocalRef(bundle_class);\n return;\n }\n\n while (jni_env->CallBooleanMethod(key_set_iterator, iterator_has_next)) {\n jstring key = reinterpret_cast(\n jni_env->CallObjectMethod(key_set_iterator, iterator_next));\n if (!key) {\n continue;\n }\n const char* key_utf = jni_env->GetStringUTFChars(key, nullptr);\n if (!key_utf) {\n jni_env->DeleteLocalRef(key);\n continue;\n }\n auto cvar_it = ConfigVars->find(key_utf);\n jni_env->ReleaseStringUTFChars(key, key_utf);\n // key_utf can't be used from now on.\n if (cvar_it == ConfigVars->end()) {\n jni_env->DeleteLocalRef(key);\n continue;\n }\n IConfigVar* cvar = cvar_it->second;\n auto cvar_bool = dynamic_cast*>(cvar);\n if (cvar_bool) {\n cvar_bool->SetCommandLineValue(\n bool(jni_env->CallBooleanMethod(bundle, bundle_get_boolean, key)));\n jni_env->DeleteLocalRef(key);\n continue;\n }\n auto cvar_int32 = dynamic_cast*>(cvar);\n if (cvar_int32) {\n cvar_int32->SetCommandLineValue(\n jni_env->CallIntMethod(bundle, bundle_get_int, key));\n jni_env->DeleteLocalRef(key);\n continue;\n }\n auto cvar_uint32 = dynamic_cast*>(cvar);\n if (cvar_uint32) {\n cvar_uint32->SetCommandLineValue(\n uint32_t(jni_env->CallIntMethod(bundle, bundle_get_int, key)));\n jni_env->DeleteLocalRef(key);\n continue;\n }\n auto cvar_uint64 = dynamic_cast*>(cvar);\n if (cvar_uint64) {\n cvar_uint64->SetCommandLineValue(\n uint64_t(jni_env->CallLongMethod(bundle, bundle_get_long, key)));\n jni_env->DeleteLocalRef(key);\n continue;\n }\n auto cvar_double = dynamic_cast*>(cvar);\n if (cvar_double) {\n cvar_double->SetCommandLineValue(\n jni_env->CallDoubleMethod(bundle, bundle_get_double, key));\n jni_env->DeleteLocalRef(key);\n continue;\n }\n auto cvar_string = dynamic_cast*>(cvar);\n if (cvar_string) {\n jstring cvar_string_value = reinterpret_cast(\n jni_env->CallObjectMethod(bundle, bundle_get_string, key));\n if (cvar_string_value) {\n const char* cvar_string_value_utf =\n jni_env->GetStringUTFChars(cvar_string_value, nullptr);\n if (cvar_string_value_utf) {\n cvar_string->SetCommandLineValue(cvar_string_value_utf);\n jni_env->ReleaseStringUTFChars(cvar_string_value,\n cvar_string_value_utf);\n }\n jni_env->DeleteLocalRef(cvar_string_value);\n }\n jni_env->DeleteLocalRef(key);\n continue;\n }\n auto cvar_path = dynamic_cast*>(cvar);\n if (cvar_path) {\n jstring cvar_string_value = reinterpret_cast(\n jni_env->CallObjectMethod(bundle, bundle_get_string, key));\n if (cvar_string_value) {\n const char* cvar_string_value_utf =\n jni_env->GetStringUTFChars(cvar_string_value, nullptr);\n if (cvar_string_value_utf) {\n cvar_path->SetCommandLineValue(xe::to_path(cvar_string_value_utf));\n jni_env->ReleaseStringUTFChars(cvar_string_value,\n cvar_string_value_utf);\n }\n jni_env->DeleteLocalRef(cvar_string_value);\n }\n jni_env->DeleteLocalRef(key);\n continue;\n }\n assert_always(\"Unsupported type of cvar {}\", cvar->name().c_str());\n jni_env->DeleteLocalRef(key);\n }\n\n jni_env->DeleteLocalRef(iterator_class);\n jni_env->DeleteLocalRef(key_set_iterator);\n jni_env->DeleteLocalRef(set_class);\n jni_env->DeleteLocalRef(key_set);\n jni_env->DeleteLocalRef(bundle_class);\n}\n\n} // namespace cvar\n" }, { "path": "src/xenia/base/debug_visualizers.natvis", "content": "\n\n\n \n \n \n {(((value & 0xFF00000000000000) >> 56) |\n ((value & 0x00FF000000000000) >> 40) |\n ((value & 0x0000FF0000000000) >> 24) |\n ((value & 0x000000FF00000000) >> 8 ) |\n ((value & 0x00000000FF000000) << 8 ) |\n ((value & 0x0000000000FF0000) << 24) |\n ((value & 0x000000000000FF00) << 40) |\n ((value & 0x00000000000000FF) << 56))}\n \n \n \n \n {(((value & 0xFF00000000000000) >> 56) |\n ((value & 0x00FF000000000000) >> 40) |\n ((value & 0x0000FF0000000000) >> 24) |\n ((value & 0x000000FF00000000) >> 8 ) |\n ((value & 0x00000000FF000000) << 8 ) |\n ((value & 0x0000000000FF0000) << 24) |\n ((value & 0x000000000000FF00) << 40) |\n ((value & 0x00000000000000FF) << 56))}\n \n \n\n \n \n {(((value & 0xFF000000) >> 24) |\n ((value & 0x00FF0000) >> 8) |\n ((value & 0x0000FF00) << 8) |\n ((value & 0x000000FF) << 24))}\n \n \n \n \n {(((value & 0xFF000000) >> 24) |\n ((value & 0x00FF0000) >> 8) |\n ((value & 0x0000FF00) << 8) |\n ((value & 0x000000FF) << 24))}\n \n \n\n \n \n {(((value & 0xFF00) >> 8) |\n ((value & 0x00FF) << 8))}\n \n \n \n \n {(((value & 0xFF00) >> 8) |\n ((value & 0x00FF) << 8))}\n \n \n\n \n \n {value}\n \n \n \n \n {value}\n \n \n \n \n \n \n {($T1)((storage & ((((value_type)~0) >> (8 * sizeof(value_type) - $T3)) << $T2)) >> $T2)}\n \n \n\n\n" }, { "path": "src/xenia/base/debugging.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_DEBUGGING_H_\n#define XENIA_BASE_DEBUGGING_H_\n\n#include \n\n#include \"third_party/fmt/include/fmt/format.h\"\n\nnamespace xe {\nnamespace debugging {\n\n// Returns true if a debugger is attached to this process.\n// The state may change at any time (attach after launch, etc), so do not\n// cache this value. Determining if the debugger is attached is expensive,\n// though, so avoid calling it frequently.\nbool IsDebuggerAttached();\n\n// Breaks into the debugger if it is attached.\n// If no debugger is present, a signal will be raised.\nvoid Break();\n\nnamespace internal {\nvoid DebugPrint(const char* s);\n}\n\n// Prints a message to the attached debugger.\n// This bypasses the normal logging mechanism. If no debugger is attached it's\n// likely to no-op.\ntemplate \nvoid DebugPrint(fmt::string_view format, const Args&... args) {\n internal::DebugPrint(\n fmt::vformat(format, fmt::make_format_args(args...)).c_str());\n}\n\n} // namespace debugging\n} // namespace xe\n\n#endif // XENIA_BASE_DEBUGGING_H_\n" }, { "path": "src/xenia/base/debugging_mac.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/debugging.h\"\n\n#include \n#include \n\nnamespace xe {\nnamespace debugging {\n\nbool IsDebuggerAttached() {\n // https://developer.apple.com/library/mac/qa/qa1361/_index.html\n kinfo_proc info;\n info.kp_proc.p_flag = 0;\n int mib[] = {CTL_KERN, KERN_PROC, KERN_PROC_PID, getpid()};\n size_t size = sizeof(info);\n sysctl(mib, sizeof(mib) / sizeof(*mib), &info, &size, nullptr, 0);\n return (info.kp_proc.p_flag & P_TRACED) != 0;\n}\n\n// TODO(benvanik): find a more reliable way.\nvoid Break() {\n // __asm__(\"int $3\");\n __builtin_debugtrap();\n}\n\n} // namespace debugging\n} // namespace xe\n" }, { "path": "src/xenia/base/debugging_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2017 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/debugging.h\"\n\n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/string_buffer.h\"\n\nnamespace xe {\nnamespace debugging {\n\nbool IsDebuggerAttached() {\n std::ifstream proc_status_stream(\"/proc/self/status\");\n if (!proc_status_stream.is_open()) {\n return false;\n }\n std::string line;\n while (std::getline(proc_status_stream, line)) {\n std::istringstream line_stream(line);\n std::string key;\n line_stream >> key;\n if (key == \"TracerPid:\") {\n uint32_t tracer_pid;\n line_stream >> tracer_pid;\n return tracer_pid != 0;\n }\n }\n return false;\n}\n\nvoid Break() {\n static std::once_flag flag;\n std::call_once(flag, []() {\n // Install handler for sigtrap only once\n std::signal(SIGTRAP, [](int) {\n // Forward signal to default handler after being caught\n std::signal(SIGTRAP, SIG_DFL);\n });\n });\n std::raise(SIGTRAP);\n}\n\nnamespace internal {\nvoid DebugPrint(const char* s) { std::clog << s << std::endl; }\n} // namespace internal\n\n} // namespace debugging\n} // namespace xe\n" }, { "path": "src/xenia/base/debugging_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/debugging.h\"\n\n#include \"xenia/base/platform_win.h\"\n#include \"xenia/base/string_buffer.h\"\n\nnamespace xe {\nnamespace debugging {\n\nbool IsDebuggerAttached() { return IsDebuggerPresent() ? true : false; }\n\nvoid Break() { __debugbreak(); }\n\nnamespace internal {\nvoid DebugPrint(const char* s) { OutputDebugStringA(s); }\n} // namespace internal\n\n} // namespace debugging\n} // namespace xe\n" }, { "path": "src/xenia/base/delegate.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_DELEGATE_H_\n#define XENIA_BASE_DELEGATE_H_\n\n#include \n#include \n#include \n\nnamespace xe {\n\n// TODO(benvanik): go lockfree, and don't hold the lock while emitting.\n\ntemplate \nclass Delegate {\n public:\n typedef std::function Listener;\n\n void AddListener(Listener const& listener) {\n std::lock_guard guard(lock_);\n listeners_.push_back(listener);\n }\n\n void RemoveAllListeners() {\n std::lock_guard guard(lock_);\n listeners_.clear();\n }\n\n void operator()(Args... args) {\n std::lock_guard guard(lock_);\n for (auto& listener : listeners_) {\n listener(args...);\n }\n }\n\n private:\n std::mutex lock_;\n std::vector listeners_;\n};\n\ntemplate <>\nclass Delegate {\n public:\n typedef std::function Listener;\n\n void AddListener(Listener const& listener) {\n std::lock_guard guard(lock_);\n listeners_.push_back(listener);\n }\n\n void RemoveAllListeners() {\n std::lock_guard guard(lock_);\n listeners_.clear();\n }\n\n void operator()() {\n std::lock_guard guard(lock_);\n for (auto& listener : listeners_) {\n listener();\n }\n }\n\n private:\n std::mutex lock_;\n std::vector listeners_;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_DELEGATE_H_\n" }, { "path": "src/xenia/base/exception_handler.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/exception_handler.h\"\n\nnamespace xe {\n\n// Based on VIXL Instruction::IsLoad and IsStore.\n// https://github.com/Linaro/vixl/blob/d48909dd0ac62197edb75d26ed50927e4384a199/src/aarch64/instructions-aarch64.cc#L484\n//\n// Copyright 2015, VIXL authors\n// All rights reserved.\n//\n// Redistribution and use in source and binary forms, with or without\n// modification, are permitted provided that the following conditions are met:\n//\n// * Redistributions of source code must retain the above copyright notice,\n// this list of conditions and the following disclaimer.\n// * Redistributions in binary form must reproduce the above copyright notice,\n// this list of conditions and the following disclaimer in the documentation\n// and/or other materials provided with the distribution.\n// * Neither the name of ARM Limited nor the names of its contributors may be\n// used to endorse or promote products derived from this software without\n// specific prior written permission.\n//\n// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS \"AS IS\" AND\n// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED\n// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\n// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE\n// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL\n// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR\n// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER\n// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,\n// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\n// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\nbool IsArm64LoadPrefetchStore(uint32_t instruction, bool& is_store_out) {\n if ((instruction & kArm64LoadLiteralFMask) == kArm64LoadLiteralFixed) {\n return true;\n }\n if ((instruction & kArm64LoadStoreAnyFMask) != kArm64LoadStoreAnyFixed) {\n return false;\n }\n if ((instruction & kArm64LoadStorePairAnyFMask) ==\n kArm64LoadStorePairAnyFixed) {\n is_store_out = !(instruction & kArm64LoadStorePairLoadBit);\n return true;\n }\n switch (Arm64LoadStoreOp(instruction & kArm64LoadStoreMask)) {\n case Arm64LoadStoreOp::kLDRB_w:\n case Arm64LoadStoreOp::kLDRH_w:\n case Arm64LoadStoreOp::kLDR_w:\n case Arm64LoadStoreOp::kLDR_x:\n case Arm64LoadStoreOp::kLDRSB_x:\n case Arm64LoadStoreOp::kLDRSH_x:\n case Arm64LoadStoreOp::kLDRSW_x:\n case Arm64LoadStoreOp::kLDRSB_w:\n case Arm64LoadStoreOp::kLDRSH_w:\n case Arm64LoadStoreOp::kLDR_b:\n case Arm64LoadStoreOp::kLDR_h:\n case Arm64LoadStoreOp::kLDR_s:\n case Arm64LoadStoreOp::kLDR_d:\n case Arm64LoadStoreOp::kLDR_q:\n case Arm64LoadStoreOp::kPRFM:\n is_store_out = false;\n return true;\n case Arm64LoadStoreOp::kSTRB_w:\n case Arm64LoadStoreOp::kSTRH_w:\n case Arm64LoadStoreOp::kSTR_w:\n case Arm64LoadStoreOp::kSTR_x:\n case Arm64LoadStoreOp::kSTR_b:\n case Arm64LoadStoreOp::kSTR_h:\n case Arm64LoadStoreOp::kSTR_s:\n case Arm64LoadStoreOp::kSTR_d:\n case Arm64LoadStoreOp::kSTR_q:\n is_store_out = true;\n return true;\n default:\n return false;\n }\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/exception_handler.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_EXCEPTION_HANDLER_H_\n#define XENIA_BASE_EXCEPTION_HANDLER_H_\n\n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/host_thread_context.h\"\n\nnamespace xe {\n\n// AArch64 load and store decoding based on VIXL.\n// https://github.com/Linaro/vixl/blob/ae5957cd66517b3f31dbf37e9bf39db6594abfe3/src/aarch64/constants-aarch64.h\n//\n// Copyright 2015, VIXL authors\n// All rights reserved.\n//\n// Redistribution and use in source and binary forms, with or without\n// modification, are permitted provided that the following conditions are met:\n//\n// * Redistributions of source code must retain the above copyright notice,\n// this list of conditions and the following disclaimer.\n// * Redistributions in binary form must reproduce the above copyright notice,\n// this list of conditions and the following disclaimer in the documentation\n// and/or other materials provided with the distribution.\n// * Neither the name of ARM Limited nor the names of its contributors may be\n// used to endorse or promote products derived from this software without\n// specific prior written permission.\n//\n// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS \"AS IS\" AND\n// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED\n// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\n// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE\n// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL\n// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR\n// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER\n// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,\n// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\n// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n\n// `Instruction address + literal offset` loads.\n// This includes PRFM_lit.\nconstexpr uint32_t kArm64LoadLiteralFMask = UINT32_C(0x3B000000);\nconstexpr uint32_t kArm64LoadLiteralFixed = UINT32_C(0x18000000);\n\nconstexpr uint32_t kArm64LoadStoreAnyFMask = UINT32_C(0x0A000000);\nconstexpr uint32_t kArm64LoadStoreAnyFixed = UINT32_C(0x08000000);\n\nconstexpr uint32_t kArm64LoadStorePairAnyFMask = UINT32_C(0x3A000000);\nconstexpr uint32_t kArm64LoadStorePairAnyFixed = UINT32_C(0x28000000);\nconstexpr uint32_t kArm64LoadStorePairLoadBit = UINT32_C(1) << 22;\n\nconstexpr uint32_t kArm64LoadStoreMask = UINT32_C(0xC4C00000);\nenum class Arm64LoadStoreOp : uint32_t {\n kSTRB_w = UINT32_C(0x00000000),\n kSTRH_w = UINT32_C(0x40000000),\n kSTR_w = UINT32_C(0x80000000),\n kSTR_x = UINT32_C(0xC0000000),\n kLDRB_w = UINT32_C(0x00400000),\n kLDRH_w = UINT32_C(0x40400000),\n kLDR_w = UINT32_C(0x80400000),\n kLDR_x = UINT32_C(0xC0400000),\n kLDRSB_x = UINT32_C(0x00800000),\n kLDRSH_x = UINT32_C(0x40800000),\n kLDRSW_x = UINT32_C(0x80800000),\n kLDRSB_w = UINT32_C(0x00C00000),\n kLDRSH_w = UINT32_C(0x40C00000),\n kSTR_b = UINT32_C(0x04000000),\n kSTR_h = UINT32_C(0x44000000),\n kSTR_s = UINT32_C(0x84000000),\n kSTR_d = UINT32_C(0xC4000000),\n kSTR_q = UINT32_C(0x04800000),\n kLDR_b = UINT32_C(0x04400000),\n kLDR_h = UINT32_C(0x44400000),\n kLDR_s = UINT32_C(0x84400000),\n kLDR_d = UINT32_C(0xC4400000),\n kLDR_q = UINT32_C(0x04C00000),\n kPRFM = UINT32_C(0xC0800000),\n};\n\nconstexpr uint32_t kArm64LoadStoreOffsetFMask = UINT32_C(0x3B200C00);\nenum class Arm64LoadStoreOffsetFixed : uint32_t {\n kUnscaledOffset = UINT32_C(0x38000000),\n kPostIndex = UINT32_C(0x38000400),\n kPreIndex = UINT32_C(0x38000C00),\n kRegisterOffset = UINT32_C(0x38200800),\n};\n\nconstexpr uint32_t kArm64LoadStoreUnsignedOffsetFMask = UINT32_C(0x3B000000);\nconstexpr uint32_t kArm64LoadStoreUnsignedOffsetFixed = UINT32_C(0x39000000);\n\nbool IsArm64LoadPrefetchStore(uint32_t instruction, bool& is_store_out);\n\nclass Exception {\n public:\n enum class Code {\n kInvalidException = 0,\n kAccessViolation,\n kIllegalInstruction,\n };\n\n enum class AccessViolationOperation {\n kUnknown,\n kRead,\n kWrite,\n };\n\n void InitializeAccessViolation(HostThreadContext* thread_context,\n uint64_t fault_address,\n AccessViolationOperation operation) {\n code_ = Code::kAccessViolation;\n thread_context_ = thread_context;\n fault_address_ = fault_address;\n access_violation_operation_ = operation;\n }\n void InitializeIllegalInstruction(HostThreadContext* thread_context) {\n code_ = Code::kIllegalInstruction;\n thread_context_ = thread_context;\n }\n\n Code code() const { return code_; }\n\n // Returns the platform-specific thread context info.\n // Note that certain registers must be modified through Modify* proxy\n // functions rather than directly:\n // x86-64:\n // - General-purpose registers (r##, r8-r15).\n // - XMM registers.\n // AArch64:\n // - General-purpose registers (Xn), including FP and LR.\n // - SIMD and floating-point registers (Vn).\n HostThreadContext* thread_context() const { return thread_context_; }\n\n // Returns the program counter where the exception occurred.\n uint64_t pc() const {\n#if XE_ARCH_AMD64\n return thread_context_->rip;\n#elif XE_ARCH_ARM64\n return thread_context_->pc;\n#else\n assert_always();\n return 0;\n#endif // XE_ARCH\n }\n\n // Sets the program counter where execution will resume.\n void set_resume_pc(uint64_t pc) {\n#if XE_ARCH_AMD64\n thread_context_->rip = pc;\n#elif XE_ARCH_ARM64\n thread_context_->pc = pc;\n#else\n assert_always();\n#endif // XE_ARCH\n }\n\n#if XE_ARCH_AMD64\n // The index is relative to X64Register::kIntRegisterFirst.\n uint64_t& ModifyIntRegister(uint32_t index) {\n assert_true(index <= 15);\n modified_int_registers_ |= UINT16_C(1) << index;\n return thread_context_->int_registers[index];\n }\n uint16_t modified_int_registers() const { return modified_int_registers_; }\n vec128_t& ModifyXmmRegister(uint32_t index) {\n assert_true(index <= 15);\n modified_xmm_registers_ |= UINT16_C(1) << index;\n return thread_context_->xmm_registers[index];\n }\n uint16_t modified_xmm_registers() const { return modified_xmm_registers_; }\n#elif XE_ARCH_ARM64\n uint64_t& ModifyXRegister(uint32_t index) {\n assert_true(index <= 30);\n modified_x_registers_ |= UINT32_C(1) << index;\n return thread_context_->x[index];\n }\n uint32_t modified_x_registers() const { return modified_x_registers_; }\n vec128_t& ModifyVRegister(uint32_t index) {\n assert_true(index <= 31);\n modified_v_registers_ |= UINT32_C(1) << index;\n return thread_context_->v[index];\n }\n uint32_t modified_v_registers() const { return modified_v_registers_; }\n#endif // XE_ARCH\n\n // In case of AV, address that was read from/written to.\n uint64_t fault_address() const { return fault_address_; }\n\n // In case of AV, what kind of operation caused it.\n AccessViolationOperation access_violation_operation() const {\n return access_violation_operation_;\n }\n\n private:\n Code code_ = Code::kInvalidException;\n HostThreadContext* thread_context_ = nullptr;\n#if XE_ARCH_AMD64\n uint16_t modified_int_registers_ = 0;\n uint16_t modified_xmm_registers_ = 0;\n#elif XE_ARCH_ARM64\n uint32_t modified_x_registers_ = 0;\n uint32_t modified_v_registers_ = 0;\n#endif // XE_ARCH\n uint64_t fault_address_ = 0;\n AccessViolationOperation access_violation_operation_ =\n AccessViolationOperation::kUnknown;\n};\n\nclass ExceptionHandler {\n public:\n typedef bool (*Handler)(Exception* ex, void* data);\n\n // Installs an exception handler.\n // Handlers are called in the order they are installed.\n static void Install(Handler fn, void* data);\n\n // Uninstalls a previously-installed exception handler.\n static void Uninstall(Handler fn, void* data);\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_EXCEPTION_HANDLER_H_\n" }, { "path": "src/xenia/base/exception_handler_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/exception_handler.h\"\n\n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/host_thread_context.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\n\nbool signal_handlers_installed_ = false;\nstruct sigaction original_sigill_handler_;\nstruct sigaction original_sigsegv_handler_;\n\n// This can be as large as needed, but isn't often needed.\n// As we will be sometimes firing many exceptions we want to avoid having to\n// scan the table too much or invoke many custom handlers.\nconstexpr size_t kMaxHandlerCount = 8;\n\n// All custom handlers, left-aligned and null terminated.\n// Executed in order.\nstd::pair handlers_[kMaxHandlerCount];\n\nstatic void ExceptionHandlerCallback(int signal_number, siginfo_t* signal_info,\n void* signal_context) {\n mcontext_t& mcontext =\n reinterpret_cast(signal_context)->uc_mcontext;\n\n HostThreadContext thread_context;\n\n#if XE_ARCH_AMD64\n thread_context.rip = uint64_t(mcontext.gregs[REG_RIP]);\n thread_context.eflags = uint32_t(mcontext.gregs[REG_EFL]);\n // The REG_ order may be different than the register indices in the\n // instruction encoding.\n thread_context.rax = uint64_t(mcontext.gregs[REG_RAX]);\n thread_context.rcx = uint64_t(mcontext.gregs[REG_RCX]);\n thread_context.rdx = uint64_t(mcontext.gregs[REG_RDX]);\n thread_context.rbx = uint64_t(mcontext.gregs[REG_RBX]);\n thread_context.rsp = uint64_t(mcontext.gregs[REG_RSP]);\n thread_context.rbp = uint64_t(mcontext.gregs[REG_RBP]);\n thread_context.rsi = uint64_t(mcontext.gregs[REG_RSI]);\n thread_context.rdi = uint64_t(mcontext.gregs[REG_RDI]);\n thread_context.r8 = uint64_t(mcontext.gregs[REG_R8]);\n thread_context.r9 = uint64_t(mcontext.gregs[REG_R9]);\n thread_context.r10 = uint64_t(mcontext.gregs[REG_R10]);\n thread_context.r11 = uint64_t(mcontext.gregs[REG_R11]);\n thread_context.r12 = uint64_t(mcontext.gregs[REG_R12]);\n thread_context.r13 = uint64_t(mcontext.gregs[REG_R13]);\n thread_context.r14 = uint64_t(mcontext.gregs[REG_R14]);\n thread_context.r15 = uint64_t(mcontext.gregs[REG_R15]);\n std::memcpy(thread_context.xmm_registers, mcontext.fpregs->_xmm,\n sizeof(thread_context.xmm_registers));\n#elif XE_ARCH_ARM64\n std::memcpy(thread_context.x, mcontext.regs, sizeof(thread_context.x));\n thread_context.sp = mcontext.sp;\n thread_context.pc = mcontext.pc;\n thread_context.pstate = mcontext.pstate;\n struct fpsimd_context* mcontext_fpsimd = nullptr;\n struct esr_context* mcontext_esr = nullptr;\n for (struct _aarch64_ctx* mcontext_extension =\n reinterpret_cast(mcontext.__reserved);\n mcontext_extension->magic;\n mcontext_extension = reinterpret_cast(\n reinterpret_cast(mcontext_extension) +\n mcontext_extension->size)) {\n switch (mcontext_extension->magic) {\n case FPSIMD_MAGIC:\n mcontext_fpsimd =\n reinterpret_cast(mcontext_extension);\n break;\n case ESR_MAGIC:\n mcontext_esr =\n reinterpret_cast(mcontext_extension);\n break;\n default:\n break;\n }\n }\n assert_not_null(mcontext_fpsimd);\n if (mcontext_fpsimd) {\n thread_context.fpsr = mcontext_fpsimd->fpsr;\n thread_context.fpcr = mcontext_fpsimd->fpcr;\n std::memcpy(thread_context.v, mcontext_fpsimd->vregs,\n sizeof(thread_context.v));\n }\n#endif // XE_ARCH\n\n Exception ex;\n switch (signal_number) {\n case SIGILL:\n ex.InitializeIllegalInstruction(&thread_context);\n break;\n case SIGSEGV: {\n Exception::AccessViolationOperation access_violation_operation;\n#if XE_ARCH_AMD64\n // x86_pf_error_code::X86_PF_WRITE\n constexpr uint64_t kX86PageFaultErrorCodeWrite = UINT64_C(1) << 1;\n access_violation_operation =\n (uint64_t(mcontext.gregs[REG_ERR]) & kX86PageFaultErrorCodeWrite)\n ? Exception::AccessViolationOperation::kWrite\n : Exception::AccessViolationOperation::kRead;\n#elif XE_ARCH_ARM64\n // For a Data Abort (EC - ESR_EL1 bits 31:26 - 0b100100 from a lower\n // Exception Level, 0b100101 without a change in the Exception Level),\n // bit 6 is 0 for reading from a memory location, 1 for writing to a\n // memory location.\n if (mcontext_esr && ((mcontext_esr->esr >> 26) & 0b111110) == 0b100100) {\n access_violation_operation =\n (mcontext_esr->esr & (UINT64_C(1) << 6))\n ? Exception::AccessViolationOperation::kWrite\n : Exception::AccessViolationOperation::kRead;\n } else {\n // Determine the memory access direction based on which instruction has\n // requested it.\n // esr_context may be unavailable on certain hosts (for instance, on\n // Android, it was added only in NDK r16 - which is the first NDK\n // version to support the Android API level 27, while NDK r15 doesn't\n // have esr_context in its API 26 sigcontext.h).\n // On AArch64 (unlike on AArch32), the program counter is the address of\n // the currently executing instruction.\n bool instruction_is_store;\n if (IsArm64LoadPrefetchStore(\n *reinterpret_cast(mcontext.pc),\n instruction_is_store)) {\n access_violation_operation =\n instruction_is_store ? Exception::AccessViolationOperation::kWrite\n : Exception::AccessViolationOperation::kRead;\n } else {\n assert_always(\n \"No ESR in the exception thread context, or it's not a Data \"\n \"Abort, and the faulting instruction is not a known load, \"\n \"prefetch or store instruction\");\n access_violation_operation =\n Exception::AccessViolationOperation::kUnknown;\n }\n }\n#else\n access_violation_operation =\n Exception::AccessViolationOperation::kUnknown;\n#endif // XE_ARCH\n ex.InitializeAccessViolation(\n &thread_context, reinterpret_cast(signal_info->si_addr),\n access_violation_operation);\n } break;\n default:\n assert_unhandled_case(signal_number);\n }\n\n for (size_t i = 0; i < xe::countof(handlers_) && handlers_[i].first; ++i) {\n if (handlers_[i].first(&ex, handlers_[i].second)) {\n // Exception handled.\n#if XE_ARCH_AMD64\n mcontext.gregs[REG_RIP] = greg_t(thread_context.rip);\n mcontext.gregs[REG_EFL] = greg_t(thread_context.eflags);\n uint32_t modified_register_index;\n // The order must match the order in X64Register.\n static const size_t kIntRegisterMap[] = {\n REG_RAX, REG_RCX, REG_RDX, REG_RBX, REG_RSP, REG_RBP,\n REG_RSI, REG_RDI, REG_R8, REG_R9, REG_R10, REG_R11,\n REG_R12, REG_R13, REG_R14, REG_R15,\n };\n uint16_t modified_int_registers_remaining = ex.modified_int_registers();\n while (xe::bit_scan_forward(modified_int_registers_remaining,\n &modified_register_index)) {\n modified_int_registers_remaining &=\n ~(UINT16_C(1) << modified_register_index);\n mcontext.gregs[kIntRegisterMap[modified_register_index]] =\n thread_context.int_registers[modified_register_index];\n }\n uint16_t modified_xmm_registers_remaining = ex.modified_xmm_registers();\n while (xe::bit_scan_forward(modified_xmm_registers_remaining,\n &modified_register_index)) {\n modified_xmm_registers_remaining &=\n ~(UINT16_C(1) << modified_register_index);\n std::memcpy(&mcontext.fpregs->_xmm[modified_register_index],\n &thread_context.xmm_registers[modified_register_index],\n sizeof(vec128_t));\n }\n#elif XE_ARCH_ARM64\n uint32_t modified_register_index;\n uint32_t modified_x_registers_remaining = ex.modified_x_registers();\n while (xe::bit_scan_forward(modified_x_registers_remaining,\n &modified_register_index)) {\n modified_x_registers_remaining &=\n ~(UINT32_C(1) << modified_register_index);\n mcontext.regs[modified_register_index] =\n thread_context.x[modified_register_index];\n }\n mcontext.sp = thread_context.sp;\n mcontext.pc = thread_context.pc;\n mcontext.pstate = thread_context.pstate;\n if (mcontext_fpsimd) {\n mcontext_fpsimd->fpsr = thread_context.fpsr;\n mcontext_fpsimd->fpcr = thread_context.fpcr;\n uint32_t modified_v_registers_remaining = ex.modified_v_registers();\n while (xe::bit_scan_forward(modified_v_registers_remaining,\n &modified_register_index)) {\n modified_v_registers_remaining &=\n ~(UINT32_C(1) << modified_register_index);\n std::memcpy(&mcontext_fpsimd->vregs[modified_register_index],\n &thread_context.v[modified_register_index],\n sizeof(vec128_t));\n mcontext.regs[modified_register_index] =\n thread_context.x[modified_register_index];\n }\n }\n#endif // XE_ARCH\n return;\n }\n }\n}\n\nvoid ExceptionHandler::Install(Handler fn, void* data) {\n if (!signal_handlers_installed_) {\n struct sigaction signal_handler;\n\n std::memset(&signal_handler, 0, sizeof(signal_handler));\n signal_handler.sa_sigaction = ExceptionHandlerCallback;\n signal_handler.sa_flags = SA_SIGINFO;\n\n if (sigaction(SIGILL, &signal_handler, &original_sigill_handler_) != 0) {\n assert_always(\"Failed to install new SIGILL handler\");\n }\n if (sigaction(SIGSEGV, &signal_handler, &original_sigsegv_handler_) != 0) {\n assert_always(\"Failed to install new SIGSEGV handler\");\n }\n signal_handlers_installed_ = true;\n }\n\n for (size_t i = 0; i < xe::countof(handlers_); ++i) {\n if (!handlers_[i].first) {\n handlers_[i].first = fn;\n handlers_[i].second = data;\n return;\n }\n }\n assert_always(\"Too many exception handlers installed\");\n}\n\nvoid ExceptionHandler::Uninstall(Handler fn, void* data) {\n for (size_t i = 0; i < xe::countof(handlers_); ++i) {\n if (handlers_[i].first == fn && handlers_[i].second == data) {\n for (; i < xe::countof(handlers_) - 1; ++i) {\n handlers_[i] = handlers_[i + 1];\n }\n handlers_[i].first = nullptr;\n handlers_[i].second = nullptr;\n break;\n }\n }\n\n bool has_any = false;\n for (size_t i = 0; i < xe::countof(handlers_); ++i) {\n if (handlers_[i].first) {\n has_any = true;\n break;\n }\n }\n if (!has_any) {\n if (signal_handlers_installed_) {\n if (sigaction(SIGILL, &original_sigill_handler_, NULL) != 0) {\n assert_always(\"Failed to restore original SIGILL handler\");\n }\n if (sigaction(SIGSEGV, &original_sigsegv_handler_, NULL) != 0) {\n assert_always(\"Failed to restore original SIGSEGV handler\");\n }\n signal_handlers_installed_ = false;\n }\n }\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/exception_handler_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/exception_handler.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/platform_win.h\"\n\nnamespace xe {\n\n// Handle of the added VectoredExceptionHandler.\nvoid* veh_handle_ = nullptr;\n// Handle of the added VectoredContinueHandler.\nvoid* vch_handle_ = nullptr;\n\n// This can be as large as needed, but isn't often needed.\n// As we will be sometimes firing many exceptions we want to avoid having to\n// scan the table too much or invoke many custom handlers.\nconstexpr size_t kMaxHandlerCount = 8;\n\n// All custom handlers, left-aligned and null terminated.\n// Executed in order.\nstd::pair handlers_[kMaxHandlerCount];\n\nLONG CALLBACK ExceptionHandlerCallback(PEXCEPTION_POINTERS ex_info) {\n // Visual Studio SetThreadName.\n if (ex_info->ExceptionRecord->ExceptionCode == 0x406D1388) {\n return EXCEPTION_CONTINUE_SEARCH;\n }\n\n HostThreadContext thread_context;\n thread_context.rip = ex_info->ContextRecord->Rip;\n thread_context.eflags = ex_info->ContextRecord->EFlags;\n std::memcpy(thread_context.int_registers, &ex_info->ContextRecord->Rax,\n sizeof(thread_context.int_registers));\n std::memcpy(thread_context.xmm_registers, &ex_info->ContextRecord->Xmm0,\n sizeof(thread_context.xmm_registers));\n\n // https://msdn.microsoft.com/en-us/library/ms679331(v=vs.85).aspx\n // https://msdn.microsoft.com/en-us/library/aa363082(v=vs.85).aspx\n Exception ex;\n switch (ex_info->ExceptionRecord->ExceptionCode) {\n case STATUS_ILLEGAL_INSTRUCTION:\n ex.InitializeIllegalInstruction(&thread_context);\n break;\n case STATUS_ACCESS_VIOLATION: {\n Exception::AccessViolationOperation access_violation_operation;\n switch (ex_info->ExceptionRecord->ExceptionInformation[0]) {\n case 0:\n access_violation_operation =\n Exception::AccessViolationOperation::kRead;\n break;\n case 1:\n access_violation_operation =\n Exception::AccessViolationOperation::kWrite;\n break;\n default:\n access_violation_operation =\n Exception::AccessViolationOperation::kUnknown;\n break;\n }\n ex.InitializeAccessViolation(\n &thread_context, ex_info->ExceptionRecord->ExceptionInformation[1],\n access_violation_operation);\n } break;\n default:\n // Unknown/unhandled type.\n return EXCEPTION_CONTINUE_SEARCH;\n }\n\n for (size_t i = 0; i < xe::countof(handlers_) && handlers_[i].first; ++i) {\n if (handlers_[i].first(&ex, handlers_[i].second)) {\n // Exception handled.\n ex_info->ContextRecord->Rip = thread_context.rip;\n ex_info->ContextRecord->EFlags = thread_context.eflags;\n uint32_t modified_register_index;\n uint16_t modified_int_registers_remaining = ex.modified_int_registers();\n while (xe::bit_scan_forward(modified_int_registers_remaining,\n &modified_register_index)) {\n modified_int_registers_remaining &=\n ~(UINT16_C(1) << modified_register_index);\n (&ex_info->ContextRecord->Rax)[modified_register_index] =\n thread_context.int_registers[modified_register_index];\n }\n uint16_t modified_xmm_registers_remaining = ex.modified_xmm_registers();\n while (xe::bit_scan_forward(modified_xmm_registers_remaining,\n &modified_register_index)) {\n modified_xmm_registers_remaining &=\n ~(UINT16_C(1) << modified_register_index);\n std::memcpy(&ex_info->ContextRecord->Xmm0 + modified_register_index,\n &thread_context.xmm_registers[modified_register_index],\n sizeof(vec128_t));\n }\n return EXCEPTION_CONTINUE_EXECUTION;\n }\n }\n return EXCEPTION_CONTINUE_SEARCH;\n}\n\nvoid ExceptionHandler::Install(Handler fn, void* data) {\n if (!veh_handle_) {\n veh_handle_ = AddVectoredExceptionHandler(1, ExceptionHandlerCallback);\n\n if (IsDebuggerPresent()) {\n // TODO(benvanik): do we need a continue handler if a debugger is\n // attached?\n // vch_handle_ = AddVectoredContinueHandler(1, ExceptionHandlerCallback);\n }\n }\n\n for (size_t i = 0; i < xe::countof(handlers_); ++i) {\n if (!handlers_[i].first) {\n handlers_[i].first = fn;\n handlers_[i].second = data;\n return;\n }\n }\n assert_always(\"Too many exception handlers installed\");\n}\n\nvoid ExceptionHandler::Uninstall(Handler fn, void* data) {\n for (size_t i = 0; i < xe::countof(handlers_); ++i) {\n if (handlers_[i].first == fn && handlers_[i].second == data) {\n for (; i < xe::countof(handlers_) - 1; ++i) {\n handlers_[i] = handlers_[i + 1];\n }\n handlers_[i].first = nullptr;\n handlers_[i].second = nullptr;\n break;\n }\n }\n\n bool has_any = false;\n for (size_t i = 0; i < xe::countof(handlers_); ++i) {\n if (handlers_[i].first) {\n has_any = true;\n break;\n }\n }\n if (!has_any) {\n if (veh_handle_) {\n RemoveVectoredExceptionHandler(veh_handle_);\n veh_handle_ = nullptr;\n }\n if (vch_handle_) {\n RemoveVectoredContinueHandler(vch_handle_);\n vch_handle_ = nullptr;\n }\n }\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/filesystem.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/filesystem.h\"\n\n#include \n\nnamespace xe {\nnamespace filesystem {\n\nbool CreateParentFolder(const std::filesystem::path& path) {\n if (path.has_parent_path()) {\n auto parent_path = path.parent_path();\n if (!std::filesystem::exists(parent_path)) {\n return std::filesystem::create_directories(parent_path);\n }\n }\n return true;\n}\n\n} // namespace filesystem\n} // namespace xe\n" }, { "path": "src/xenia/base/filesystem.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_FILESYSTEM_H_\n#define XENIA_BASE_FILESYSTEM_H_\n\n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/string.h\"\n\nnamespace xe {\n\nstd::string path_to_utf8(const std::filesystem::path& path);\nstd::u16string path_to_utf16(const std::filesystem::path& path);\nstd::filesystem::path to_path(const std::string_view source);\nstd::filesystem::path to_path(const std::u16string_view source);\n\nnamespace filesystem {\n\n// Get executable path.\nstd::filesystem::path GetExecutablePath();\n\n// Get executable folder.\nstd::filesystem::path GetExecutableFolder();\n\n// Get user folder.\nstd::filesystem::path GetUserFolder();\n\n// Creates the parent folder of the specified path if needed.\n// This can be used to ensure the destination path for a new file exists before\n// attempting to create it.\nbool CreateParentFolder(const std::filesystem::path& path);\n\n// Creates an empty file at the given path, overwriting if it exists.\nbool CreateEmptyFile(const std::filesystem::path& path);\n\n// Opens the file at the given path with the specified mode.\n// This behaves like fopen and the returned handle can be used with stdio.\nFILE* OpenFile(const std::filesystem::path& path, const std::string_view mode);\n\n// Wrapper for the 64-bit version of fseek, returns true on success.\nbool Seek(FILE* file, int64_t offset, int origin);\n\n// Wrapper for the 64-bit version of ftell, returns a positive value on success.\nint64_t Tell(FILE* file);\n\n// Reduces the size of a stdio file opened for writing. The file pointer is\n// clamped. If this returns false, the size of the file and the file pointer are\n// undefined.\nbool TruncateStdioFile(FILE* file, uint64_t length);\n\nstruct FileAccess {\n // Implies kFileReadData.\n static const uint32_t kGenericRead = 0x80000000;\n // Implies kFileWriteData.\n static const uint32_t kGenericWrite = 0x40000000;\n static const uint32_t kGenericExecute = 0x20000000;\n static const uint32_t kGenericAll = 0x10000000;\n static const uint32_t kFileReadData = 0x00000001;\n static const uint32_t kFileWriteData = 0x00000002;\n static const uint32_t kFileAppendData = 0x00000004;\n};\n\nclass FileHandle {\n public:\n // Opens the file, failing if it doesn't exist.\n // The desired_access bitmask denotes the permissions on the file.\n static std::unique_ptr OpenExisting(\n const std::filesystem::path& path, uint32_t desired_access);\n\n virtual ~FileHandle() = default;\n\n const std::filesystem::path& path() const { return path_; }\n\n // Reads the requested number of bytes from the file starting at the given\n // offset. The total number of bytes read is returned only if the complete\n // read succeeds.\n virtual bool Read(size_t file_offset, void* buffer, size_t buffer_length,\n size_t* out_bytes_read) = 0;\n\n // Writes the given buffer to the file starting at the given offset.\n // The total number of bytes written is returned only if the complete\n // write succeeds.\n virtual bool Write(size_t file_offset, const void* buffer,\n size_t buffer_length, size_t* out_bytes_written) = 0;\n\n // Set length of the file in bytes.\n virtual bool SetLength(size_t length) = 0;\n\n // Flushes any pending write buffers to the underlying filesystem.\n virtual void Flush() = 0;\n\n protected:\n explicit FileHandle(const std::filesystem::path& path) : path_(path) {}\n\n std::filesystem::path path_;\n};\n\nstruct FileInfo {\n enum class Type {\n kFile,\n kDirectory,\n };\n Type type;\n std::filesystem::path name;\n std::filesystem::path path;\n size_t total_size;\n uint64_t create_timestamp;\n uint64_t access_timestamp;\n uint64_t write_timestamp;\n};\nbool GetInfo(const std::filesystem::path& path, FileInfo* out_info);\nstd::vector ListFiles(const std::filesystem::path& path);\n\n#if XE_PLATFORM_ANDROID\nvoid AndroidInitialize();\nvoid AndroidShutdown();\nbool IsAndroidContentUri(const std::string_view source);\nint OpenAndroidContentFileDescriptor(const std::string_view uri,\n const char* mode);\n#endif // XE_PLATFORM_ANDROID\n\n} // namespace filesystem\n} // namespace xe\n\n#endif // XENIA_BASE_FILESYSTEM_H_\n" }, { "path": "src/xenia/base/filesystem_android.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n#include \n\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/main_android.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/string.h\"\n\nnamespace xe {\nnamespace filesystem {\n\n// Using Android logging because the file system may need to be initialized\n// before logging.\n\nstatic jobject android_content_resolver_;\nstatic jclass android_content_resolver_class_;\nstatic jmethodID android_content_resolver_open_file_descriptor_;\n\nstatic jclass android_parcel_file_descriptor_class_;\nstatic jmethodID android_parcel_file_descriptor_detach_fd_;\n\nstatic jclass android_uri_class_;\nstatic jmethodID android_uri_parse_;\n\nstatic bool android_content_resolver_initialized_;\n\nstatic void AndroidShutdownContentResolver() {\n android_content_resolver_initialized_ = false;\n android_uri_parse_ = nullptr;\n android_parcel_file_descriptor_detach_fd_ = nullptr;\n android_content_resolver_open_file_descriptor_ = nullptr;\n JNIEnv* jni_env = GetAndroidThreadJniEnv();\n if (jni_env) {\n if (android_uri_class_) {\n jni_env->DeleteGlobalRef(android_uri_class_);\n }\n if (android_parcel_file_descriptor_class_) {\n jni_env->DeleteGlobalRef(\n reinterpret_cast(android_parcel_file_descriptor_class_));\n }\n if (android_content_resolver_class_) {\n jni_env->DeleteGlobalRef(\n reinterpret_cast(android_content_resolver_class_));\n }\n if (android_content_resolver_) {\n jni_env->DeleteGlobalRef(\n reinterpret_cast(android_content_resolver_));\n }\n }\n android_uri_class_ = nullptr;\n android_parcel_file_descriptor_class_ = nullptr;\n android_content_resolver_class_ = nullptr;\n android_content_resolver_ = nullptr;\n}\n\nstatic void AndroidInitializeContentResolver() {\n JNIEnv* jni_env = GetAndroidThreadJniEnv();\n if (!jni_env) {\n return;\n }\n jobject application_context = GetAndroidApplicationContext();\n if (!application_context) {\n return;\n }\n {\n jclass context_class = jni_env->GetObjectClass(application_context);\n if (!context_class) {\n __android_log_write(ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the context class\");\n AndroidShutdownContentResolver();\n return;\n }\n jmethodID context_get_content_resolver =\n jni_env->GetMethodID(context_class, \"getContentResolver\",\n \"()Landroid/content/ContentResolver;\");\n if (!context_get_content_resolver) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the getContentResolver method of the context\");\n jni_env->DeleteLocalRef(reinterpret_cast(context_class));\n AndroidShutdownContentResolver();\n return;\n }\n jobject content_resolver = jni_env->CallObjectMethod(\n application_context, context_get_content_resolver);\n jni_env->DeleteLocalRef(reinterpret_cast(context_class));\n if (!content_resolver) {\n __android_log_write(ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the content resolver\");\n AndroidShutdownContentResolver();\n return;\n }\n android_content_resolver_ = jni_env->NewGlobalRef(content_resolver);\n jni_env->DeleteLocalRef(content_resolver);\n }\n if (!android_content_resolver_) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to create a global reference to the content resolver\");\n AndroidShutdownContentResolver();\n return;\n }\n {\n jobject content_resolver_class =\n jni_env->GetObjectClass(android_content_resolver_);\n if (!content_resolver_class) {\n __android_log_write(ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the content resolver class\");\n AndroidShutdownContentResolver();\n return;\n }\n android_content_resolver_class_ =\n reinterpret_cast(jni_env->NewGlobalRef(\n reinterpret_cast(content_resolver_class)));\n jni_env->DeleteLocalRef(reinterpret_cast(content_resolver_class));\n }\n if (!android_content_resolver_class_) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to create a global reference to the content resolver class\");\n AndroidShutdownContentResolver();\n return;\n }\n android_content_resolver_open_file_descriptor_ = jni_env->GetMethodID(\n android_content_resolver_class_, \"openFileDescriptor\",\n \"(Landroid/net/Uri;Ljava/lang/String;)Landroid/os/ParcelFileDescriptor;\");\n if (!android_content_resolver_open_file_descriptor_) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the openFileDescriptor method of the content resolver\");\n AndroidShutdownContentResolver();\n return;\n }\n\n {\n jclass parcel_file_descriptor_class =\n jni_env->FindClass(\"android/os/ParcelFileDescriptor\");\n if (!parcel_file_descriptor_class) {\n __android_log_write(ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the parcel file descriptor class\");\n AndroidShutdownContentResolver();\n return;\n }\n android_parcel_file_descriptor_class_ =\n reinterpret_cast(jni_env->NewGlobalRef(\n reinterpret_cast(parcel_file_descriptor_class)));\n jni_env->DeleteLocalRef(\n reinterpret_cast(parcel_file_descriptor_class));\n }\n if (!android_parcel_file_descriptor_class_) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to create a global reference to the parcel file descriptor \"\n \"class\");\n AndroidShutdownContentResolver();\n return;\n }\n android_parcel_file_descriptor_detach_fd_ = jni_env->GetMethodID(\n android_parcel_file_descriptor_class_, \"detachFd\", \"()I\");\n if (!android_parcel_file_descriptor_detach_fd_) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the detachFd method of the parcel file descriptor\");\n AndroidShutdownContentResolver();\n return;\n }\n\n {\n jclass uri_class = jni_env->FindClass(\"android/net/Uri\");\n if (!uri_class) {\n __android_log_write(ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the URI class\");\n AndroidShutdownContentResolver();\n return;\n }\n android_uri_class_ = reinterpret_cast(\n jni_env->NewGlobalRef(reinterpret_cast(uri_class)));\n jni_env->DeleteLocalRef(reinterpret_cast(uri_class));\n }\n if (!android_uri_class_) {\n __android_log_write(ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to create a global reference to the URI class\");\n AndroidShutdownContentResolver();\n return;\n }\n android_uri_parse_ = jni_env->GetStaticMethodID(\n android_uri_class_, \"parse\", \"(Ljava/lang/String;)Landroid/net/Uri;\");\n if (!android_uri_parse_) {\n __android_log_write(ANDROID_LOG_ERROR, \"AndroidInitializeContentResolver\",\n \"Failed to get the parse method of the URI class\");\n AndroidShutdownContentResolver();\n return;\n }\n\n android_content_resolver_initialized_ = true;\n}\n\nvoid AndroidInitialize() { AndroidInitializeContentResolver(); }\n\nvoid AndroidShutdown() { AndroidShutdownContentResolver(); }\n\nbool IsAndroidContentUri(const std::string_view source) {\n // A URI schema is case-insensitive. Though just content: defines the schema,\n // still including // in the comparison to distinguish from a file with a name\n // starting from content: (as this is the main purpose of this code -\n // separating URIs from file paths) more clearly.\n static const char kContentSchema[] = \"content://\";\n constexpr size_t kContentSchemaLength = xe::countof(kContentSchema) - 1;\n return source.size() >= kContentSchemaLength &&\n !xe_strncasecmp(source.data(), kContentSchema, kContentSchemaLength);\n}\n\nint OpenAndroidContentFileDescriptor(const std::string_view uri,\n const char* mode) {\n if (!android_content_resolver_initialized_) {\n return -1;\n }\n JNIEnv* jni_env = GetAndroidThreadJniEnv();\n if (!jni_env) {\n return -1;\n }\n\n jobject uri_object;\n {\n jstring uri_string;\n {\n std::u16string uri_u16 = xe::to_utf16(uri);\n uri_string = jni_env->NewString(\n reinterpret_cast(uri_u16.data()), uri_u16.size());\n }\n if (!uri_string) {\n return -1;\n }\n uri_object = jni_env->CallStaticObjectMethod(\n android_uri_class_, android_uri_parse_, uri_string);\n jni_env->DeleteLocalRef(uri_string);\n }\n if (!uri_object) {\n return -1;\n }\n\n jstring mode_string = jni_env->NewStringUTF(mode);\n if (!mode_string) {\n jni_env->DeleteLocalRef(uri_object);\n return -1;\n }\n\n jobject parcel_file_descriptor = jni_env->CallObjectMethod(\n android_content_resolver_, android_content_resolver_open_file_descriptor_,\n uri_object, mode_string);\n jni_env->DeleteLocalRef(mode_string);\n jni_env->DeleteLocalRef(uri_object);\n if (jni_env->ExceptionCheck()) {\n jni_env->ExceptionClear();\n return -1;\n }\n if (!parcel_file_descriptor) {\n return -1;\n }\n int file_descriptor = jni_env->CallIntMethod(\n parcel_file_descriptor, android_parcel_file_descriptor_detach_fd_);\n jni_env->DeleteLocalRef(parcel_file_descriptor);\n\n return file_descriptor;\n}\n\n} // namespace filesystem\n} // namespace xe\n" }, { "path": "src/xenia/base/filesystem_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/string.h\"\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\nnamespace xe {\n\nstd::string path_to_utf8(const std::filesystem::path& path) {\n return path.string();\n}\n\nstd::u16string path_to_utf16(const std::filesystem::path& path) {\n return xe::to_utf16(path.string());\n}\n\nstd::filesystem::path to_path(const std::string_view source) { return source; }\n\nstd::filesystem::path to_path(const std::u16string_view source) {\n return xe::to_utf8(source);\n}\n\nnamespace filesystem {\n\nstd::filesystem::path GetExecutablePath() {\n char buff[FILENAME_MAX] = \"\";\n readlink(\"/proc/self/exe\", buff, FILENAME_MAX);\n std::string s(buff);\n return s;\n}\n\nstd::filesystem::path GetExecutableFolder() {\n return GetExecutablePath().parent_path();\n}\n\nstd::filesystem::path GetUserFolder() {\n // get preferred data home\n char* home = std::getenv(\"XDG_DATA_HOME\");\n if (home) {\n return std::string(home);\n }\n\n // if XDG_DATA_HOME not set, fallback to HOME directory\n home = std::getenv(\"HOME\");\n\n // if HOME not set, fall back to this\n if (home == NULL) {\n struct passwd pw1;\n struct passwd* pw;\n char buf[4096]; // could potentionally lower this\n getpwuid_r(getuid(), &pw1, buf, sizeof(buf), &pw);\n assert(&pw1 == pw); // sanity check\n home = pw->pw_dir;\n }\n\n return std::filesystem::path(home) / \".local\" / \"share\";\n}\n\nFILE* OpenFile(const std::filesystem::path& path, const std::string_view mode) {\n return fopen(path.c_str(), std::string(mode).c_str());\n}\n\nbool Seek(FILE* file, int64_t offset, int origin) {\n return fseeko64(file, off64_t(offset), origin) == 0;\n}\n\nint64_t Tell(FILE* file) { return int64_t(ftello64(file)); }\n\nbool TruncateStdioFile(FILE* file, uint64_t length) {\n if (fflush(file)) {\n return false;\n }\n int64_t position = Tell(file);\n if (position < 0) {\n return false;\n }\n if (ftruncate64(fileno(file), off64_t(length))) {\n return false;\n }\n if (uint64_t(position) > length) {\n if (!Seek(file, 0, SEEK_END)) {\n return false;\n }\n }\n return true;\n}\n\nstatic int removeCallback(const char* fpath, const struct stat* sb,\n int typeflag, struct FTW* ftwbuf) {\n int rv = remove(fpath);\n return rv;\n}\n\nstatic uint64_t convertUnixtimeToWinFiletime(time_t unixtime) {\n // Linux uses number of seconds since 1/1/1970, and Windows uses\n // number of nanoseconds since 1/1/1601\n // so we convert linux time to nanoseconds and then add the number of\n // nanoseconds from 1601 to 1970\n // see https://msdn.microsoft.com/en-us/library/ms724228\n uint64_t filetime = filetime = (unixtime * 10000000) + 116444736000000000;\n return filetime;\n}\n\nbool CreateEmptyFile(const std::filesystem::path& path) {\n int file = creat(path.c_str(), 0774);\n if (file >= 0) {\n close(file);\n return true;\n }\n return false;\n}\n\nclass PosixFileHandle : public FileHandle {\n public:\n PosixFileHandle(std::filesystem::path path, int handle)\n : FileHandle(std::move(path)), handle_(handle) {}\n ~PosixFileHandle() override {\n close(handle_);\n handle_ = -1;\n }\n bool Read(size_t file_offset, void* buffer, size_t buffer_length,\n size_t* out_bytes_read) override {\n ssize_t out = pread(handle_, buffer, buffer_length, file_offset);\n *out_bytes_read = out;\n return out >= 0 ? true : false;\n }\n bool Write(size_t file_offset, const void* buffer, size_t buffer_length,\n size_t* out_bytes_written) override {\n ssize_t out = pwrite(handle_, buffer, buffer_length, file_offset);\n *out_bytes_written = out;\n return out >= 0 ? true : false;\n }\n bool SetLength(size_t length) override {\n return ftruncate(handle_, length) >= 0 ? true : false;\n }\n void Flush() override { fsync(handle_); }\n\n private:\n int handle_ = -1;\n};\n\nstd::unique_ptr FileHandle::OpenExisting(\n const std::filesystem::path& path, uint32_t desired_access) {\n int open_access = 0;\n if (desired_access & FileAccess::kGenericRead) {\n open_access |= O_RDONLY;\n }\n if (desired_access & FileAccess::kGenericWrite) {\n open_access |= O_WRONLY;\n }\n if (desired_access & FileAccess::kGenericExecute) {\n open_access |= O_RDONLY;\n }\n if (desired_access & FileAccess::kGenericAll) {\n open_access |= O_RDWR;\n }\n if (desired_access & FileAccess::kFileReadData) {\n open_access |= O_RDONLY;\n }\n if (desired_access & FileAccess::kFileWriteData) {\n open_access |= O_WRONLY;\n }\n if (desired_access & FileAccess::kFileAppendData) {\n open_access |= O_APPEND;\n }\n int handle = open(path.c_str(), open_access);\n if (handle == -1) {\n // TODO(benvanik): pick correct response.\n return nullptr;\n }\n return std::make_unique(path, handle);\n}\n\nbool GetInfo(const std::filesystem::path& path, FileInfo* out_info) {\n struct stat st;\n if (stat(path.c_str(), &st) == 0) {\n if (S_ISDIR(st.st_mode)) {\n out_info->type = FileInfo::Type::kDirectory;\n } else {\n out_info->type = FileInfo::Type::kFile;\n }\n out_info->create_timestamp = convertUnixtimeToWinFiletime(st.st_ctime);\n out_info->access_timestamp = convertUnixtimeToWinFiletime(st.st_atime);\n out_info->write_timestamp = convertUnixtimeToWinFiletime(st.st_mtime);\n return true;\n }\n return false;\n}\n\nstd::vector ListFiles(const std::filesystem::path& path) {\n std::vector result;\n\n DIR* dir = opendir(path.c_str());\n if (!dir) {\n return result;\n }\n\n while (auto ent = readdir(dir)) {\n if (strcmp(ent->d_name, \".\") == 0 || strcmp(ent->d_name, \"..\") == 0) {\n continue;\n }\n\n FileInfo info;\n\n info.name = ent->d_name;\n struct stat st;\n stat((path / info.name).c_str(), &st);\n info.create_timestamp = convertUnixtimeToWinFiletime(st.st_ctime);\n info.access_timestamp = convertUnixtimeToWinFiletime(st.st_atime);\n info.write_timestamp = convertUnixtimeToWinFiletime(st.st_mtime);\n info.path = path;\n if (ent->d_type == DT_DIR) {\n info.type = FileInfo::Type::kDirectory;\n info.total_size = 0;\n } else {\n info.type = FileInfo::Type::kFile;\n info.total_size = st.st_size;\n }\n result.push_back(info);\n }\n closedir(dir);\n return result;\n}\n\n} // namespace filesystem\n} // namespace xe\n" }, { "path": "src/xenia/base/filesystem_wildcard.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/filesystem_wildcard.h\"\n\n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/string.h\"\n\nnamespace xe::filesystem {\n\nWildcardFlags WildcardFlags::FIRST(true, false, false);\nWildcardFlags WildcardFlags::LAST(false, true, false);\nWildcardFlags WildcardFlags::ANY(false, false, true);\nWildcardFlags WildcardFlags::FIRST_AND_LAST(true, true, false);\n\nWildcardFlags::WildcardFlags()\n : FromStart(false), ToEnd(false), ExactLength(false) {}\n\nWildcardFlags::WildcardFlags(bool start, bool end, bool exact_length)\n : FromStart(start), ToEnd(end), ExactLength(exact_length) {}\n\nWildcardRule::WildcardRule(const std::string_view match,\n const WildcardFlags& flags)\n : match_(utf8::lower_ascii(match)), rules_(flags) {}\n\nbool WildcardRule::Check(const std::string_view lower,\n std::string::size_type* offset) const {\n if (match_.empty()) {\n return true;\n }\n\n if ((lower.size() - *offset) < match_.size()) {\n return false;\n }\n\n if (rules_.ExactLength) {\n *offset += match_.size();\n return true;\n }\n\n std::string_view::size_type result(lower.find(match_, *offset));\n\n if (result != std::string_view::npos) {\n if (rules_.FromStart && result != *offset) {\n return false;\n }\n\n if (rules_.ToEnd && result != (lower.size() - match_.size())) {\n return false;\n }\n\n *offset = (result + match_.size());\n return true;\n }\n\n return false;\n}\n\nvoid WildcardEngine::PreparePattern(const std::string_view pattern) {\n rules_.clear();\n\n WildcardFlags flags(WildcardFlags::FIRST);\n size_t n = 0;\n size_t last = 0;\n while ((n = pattern.find_first_of(\"*?\", last)) != pattern.npos) {\n if (last != n) {\n std::string str_str(pattern.substr(last, n - last));\n rules_.push_back(WildcardRule(str_str, flags));\n }\n if (pattern[n] == '?') {\n auto end = pattern.find_first_not_of('?', n + 1);\n auto count = end == pattern.npos ? (pattern.size() - n) : (end - n);\n rules_.push_back(\n WildcardRule(pattern.substr(n, count), WildcardFlags::ANY));\n last = n + count;\n } else if (pattern[n] == '*') {\n last = n + 1;\n } else {\n assert_always();\n }\n flags = WildcardFlags();\n }\n if (last != pattern.size()) {\n std::string str_str(pattern.substr(last));\n rules_.push_back(WildcardRule(\n str_str, last ? WildcardFlags::LAST : WildcardFlags::FIRST_AND_LAST));\n }\n}\n\nvoid WildcardEngine::SetRule(const std::string_view pattern) {\n PreparePattern(pattern);\n}\n\nbool WildcardEngine::Match(const std::string_view str) const {\n std::string str_lc = utf8::lower_ascii(str);\n std::string::size_type offset(0);\n for (const auto& rule : rules_) {\n if (!(rule.Check(str_lc, &offset))) {\n return false;\n }\n }\n\n return true;\n}\n\n} // namespace xe::filesystem\n" }, { "path": "src/xenia/base/filesystem_wildcard.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_FILESYSTEM_WILDCARD_H_\n#define XENIA_BASE_FILESYSTEM_WILDCARD_H_\n\n#include \n#include \n#include \n#include \n\n#include \"xenia/base/string.h\"\n\nnamespace xe {\nnamespace filesystem {\n\nclass WildcardFlags {\n public:\n bool FromStart : 1, ToEnd : 1, ExactLength : 1;\n\n WildcardFlags();\n WildcardFlags(bool start, bool end, bool exact_length);\n\n static WildcardFlags FIRST;\n static WildcardFlags LAST;\n static WildcardFlags ANY;\n static WildcardFlags FIRST_AND_LAST;\n};\n\nclass WildcardRule {\n public:\n WildcardRule(const std::string_view match, const WildcardFlags& flags);\n bool Check(const std::string_view lower,\n std::string_view::size_type* offset) const;\n\n private:\n std::string match_;\n WildcardFlags rules_;\n};\n\nclass WildcardEngine {\n public:\n void SetRule(const std::string_view pattern);\n\n // Always ignoring case\n bool Match(const std::string_view str) const;\n\n private:\n std::vector rules_;\n void PreparePattern(const std::string_view pattern);\n};\n\n} // namespace filesystem\n} // namespace xe\n\n#endif // XENIA_BASE_FILESYSTEM_WILDCARD_H_\n" }, { "path": "src/xenia/base/filesystem_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n#include \n\n#undef CreateFile\n\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/platform_win.h\"\n#include \"xenia/base/string.h\"\n\nnamespace xe {\n\nstd::string path_to_utf8(const std::filesystem::path& path) {\n return xe::to_utf8(path.u16string());\n}\n\nstd::u16string path_to_utf16(const std::filesystem::path& path) {\n return path.u16string();\n}\n\nstd::filesystem::path to_path(const std::string_view source) {\n return xe::to_utf16(source);\n}\n\nstd::filesystem::path to_path(const std::u16string_view source) {\n return source;\n}\n\nnamespace filesystem {\n\nstd::filesystem::path GetExecutablePath() {\n wchar_t* path;\n auto error = _get_wpgmptr(&path);\n return !error ? std::filesystem::path(path) : std::filesystem::path();\n}\n\nstd::filesystem::path GetExecutableFolder() {\n return GetExecutablePath().parent_path();\n}\n\nstd::filesystem::path GetUserFolder() {\n std::filesystem::path result;\n PWSTR path;\n if (SUCCEEDED(SHGetKnownFolderPath(FOLDERID_Documents, KF_FLAG_DEFAULT,\n nullptr, &path))) {\n result.assign(path);\n CoTaskMemFree(path);\n }\n return result;\n}\n\nbool CreateEmptyFile(const std::filesystem::path& path) {\n auto handle = CreateFileW(path.c_str(), 0, 0, nullptr, CREATE_ALWAYS,\n FILE_ATTRIBUTE_NORMAL, nullptr);\n if (handle == INVALID_HANDLE_VALUE) {\n return false;\n }\n\n CloseHandle(handle);\n return true;\n}\n\nFILE* OpenFile(const std::filesystem::path& path, const std::string_view mode) {\n // Dumb, but OK.\n const auto wmode = xe::to_utf16(mode);\n return _wfopen(path.c_str(), reinterpret_cast(wmode.c_str()));\n}\n\nbool Seek(FILE* file, int64_t offset, int origin) {\n return _fseeki64(file, offset, origin) == 0;\n}\n\nint64_t Tell(FILE* file) { return _ftelli64(file); }\n\nbool TruncateStdioFile(FILE* file, uint64_t length) {\n // Flush is necessary - if not flushing, stream position may be out of sync.\n if (fflush(file)) {\n return false;\n }\n int64_t position = Tell(file);\n if (position < 0) {\n return false;\n }\n if (_chsize_s(_fileno(file), int64_t(length))) {\n return false;\n }\n if (uint64_t(position) > length) {\n if (!Seek(file, 0, SEEK_END)) {\n return false;\n }\n }\n return true;\n}\n\nclass Win32FileHandle : public FileHandle {\n public:\n Win32FileHandle(const std::filesystem::path& path, HANDLE handle)\n : FileHandle(path), handle_(handle) {}\n ~Win32FileHandle() override {\n CloseHandle(handle_);\n handle_ = nullptr;\n }\n bool Read(size_t file_offset, void* buffer, size_t buffer_length,\n size_t* out_bytes_read) override {\n *out_bytes_read = 0;\n OVERLAPPED overlapped;\n overlapped.Pointer = (PVOID)file_offset;\n overlapped.hEvent = NULL;\n DWORD bytes_read = 0;\n BOOL read = ReadFile(handle_, buffer, (DWORD)buffer_length, &bytes_read,\n &overlapped);\n if (read) {\n *out_bytes_read = bytes_read;\n return true;\n } else {\n if (GetLastError() == ERROR_NOACCESS) {\n XELOGW(\n \"Win32FileHandle::Read(..., {}, 0x{:X}, ...) returned \"\n \"ERROR_NOACCESS. Read-only memory?\",\n buffer, buffer_length);\n }\n return false;\n }\n }\n bool Write(size_t file_offset, const void* buffer, size_t buffer_length,\n size_t* out_bytes_written) override {\n *out_bytes_written = 0;\n OVERLAPPED overlapped;\n overlapped.Pointer = (PVOID)file_offset;\n overlapped.hEvent = NULL;\n DWORD bytes_written = 0;\n BOOL wrote = WriteFile(handle_, buffer, (DWORD)buffer_length,\n &bytes_written, &overlapped);\n if (wrote) {\n *out_bytes_written = bytes_written;\n return true;\n } else {\n return false;\n }\n }\n bool SetLength(size_t length) {\n LARGE_INTEGER position;\n position.QuadPart = length;\n if (!SetFilePointerEx(handle_, position, nullptr, SEEK_SET)) {\n return false;\n }\n if (!SetEndOfFile(handle_)) {\n return false;\n }\n return true;\n }\n void Flush() override { FlushFileBuffers(handle_); }\n\n private:\n HANDLE handle_ = nullptr;\n};\n\nstd::unique_ptr FileHandle::OpenExisting(\n const std::filesystem::path& path, uint32_t desired_access) {\n DWORD open_access = 0;\n if (desired_access & FileAccess::kGenericRead) {\n open_access |= GENERIC_READ;\n }\n if (desired_access & FileAccess::kGenericWrite) {\n open_access |= GENERIC_WRITE;\n }\n if (desired_access & FileAccess::kGenericExecute) {\n open_access |= GENERIC_EXECUTE;\n }\n if (desired_access & FileAccess::kGenericAll) {\n open_access |= GENERIC_READ | GENERIC_WRITE;\n }\n if (desired_access & FileAccess::kFileReadData) {\n open_access |= FILE_READ_DATA;\n }\n if (desired_access & FileAccess::kFileWriteData) {\n open_access |= FILE_WRITE_DATA;\n }\n if (desired_access & FileAccess::kFileAppendData) {\n open_access |= FILE_APPEND_DATA;\n }\n DWORD share_mode = FILE_SHARE_READ | FILE_SHARE_WRITE;\n // We assume we've already created the file in the caller.\n DWORD creation_disposition = OPEN_EXISTING;\n HANDLE handle = CreateFileW(\n path.c_str(), open_access, share_mode, nullptr, creation_disposition,\n FILE_ATTRIBUTE_NORMAL | FILE_FLAG_BACKUP_SEMANTICS, nullptr);\n if (handle == INVALID_HANDLE_VALUE) {\n // TODO(benvanik): pick correct response.\n return nullptr;\n }\n return std::make_unique(path, handle);\n}\n\n#define COMBINE_TIME(t) (((uint64_t)t.dwHighDateTime << 32) | t.dwLowDateTime)\n\nbool GetInfo(const std::filesystem::path& path, FileInfo* out_info) {\n std::memset(out_info, 0, sizeof(FileInfo));\n WIN32_FILE_ATTRIBUTE_DATA data = {0};\n if (!GetFileAttributesEx(path.c_str(), GetFileExInfoStandard, &data)) {\n return false;\n }\n if (data.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) {\n out_info->type = FileInfo::Type::kDirectory;\n out_info->total_size = 0;\n } else {\n out_info->type = FileInfo::Type::kFile;\n out_info->total_size =\n (data.nFileSizeHigh * (size_t(MAXDWORD) + 1)) + data.nFileSizeLow;\n }\n out_info->path = path.parent_path();\n out_info->name = path.filename();\n out_info->create_timestamp = COMBINE_TIME(data.ftCreationTime);\n out_info->access_timestamp = COMBINE_TIME(data.ftLastAccessTime);\n out_info->write_timestamp = COMBINE_TIME(data.ftLastWriteTime);\n return true;\n}\n\nstd::vector ListFiles(const std::filesystem::path& path) {\n std::vector result;\n\n WIN32_FIND_DATA ffd;\n HANDLE handle = FindFirstFileW((path / \"*\").c_str(), &ffd);\n if (handle == INVALID_HANDLE_VALUE) {\n return result;\n }\n do {\n if (std::wcscmp(ffd.cFileName, L\".\") == 0 ||\n std::wcscmp(ffd.cFileName, L\"..\") == 0) {\n continue;\n }\n FileInfo info;\n if (ffd.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) {\n info.type = FileInfo::Type::kDirectory;\n info.total_size = 0;\n } else {\n info.type = FileInfo::Type::kFile;\n info.total_size =\n (ffd.nFileSizeHigh * (size_t(MAXDWORD) + 1)) + ffd.nFileSizeLow;\n }\n info.path = path;\n info.name = ffd.cFileName;\n info.create_timestamp = COMBINE_TIME(ffd.ftCreationTime);\n info.access_timestamp = COMBINE_TIME(ffd.ftLastAccessTime);\n info.write_timestamp = COMBINE_TIME(ffd.ftLastWriteTime);\n result.push_back(info);\n } while (FindNextFile(handle, &ffd) != 0);\n FindClose(handle);\n\n return result;\n}\n\n} // namespace filesystem\n} // namespace xe\n" }, { "path": "src/xenia/base/fuzzy.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/fuzzy.h\"\n\n#include \n#include \n#include \n\n// TODO(gibbed): UTF8 support.\n\nnamespace xe {\n\nint fuzzy_match(const std::string_view pattern, const char* value) {\n // https://github.com/mattyork/fuzzy/blob/master/lib/fuzzy.js\n // TODO(benvanik): look at\n // https://github.com/atom/fuzzaldrin/tree/master/src This does not weight\n // complete substrings or prefixes right, which kind of sucks.\n size_t pattern_index = 0;\n size_t value_length = std::strlen(value);\n int total_score = 0;\n int local_score = 0;\n for (size_t i = 0; i < value_length; ++i) {\n if (std::tolower(value[i]) == std::tolower(pattern[pattern_index])) {\n ++pattern_index;\n local_score += 1 + local_score;\n } else {\n local_score = 0;\n }\n total_score += local_score;\n }\n return total_score;\n}\n\nstd::vector> fuzzy_filter(const std::string_view pattern,\n const void* const* entries,\n size_t entry_count,\n size_t string_offset) {\n std::vector> results;\n results.reserve(entry_count);\n for (size_t i = 0; i < entry_count; ++i) {\n auto entry_value =\n reinterpret_cast(entries[i]) + string_offset;\n int score = fuzzy_match(pattern, entry_value);\n results.emplace_back(i, score);\n }\n return results;\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/fuzzy.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_FUZZY_H_\n#define XENIA_BASE_FUZZY_H_\n\n#include \n#include \n\nnamespace xe {\n\n// Tests a match against a case-insensitive fuzzy filter.\n// Returns the score of the match or 0 if none.\nint fuzzy_match(const std::string_view pattern, const char* value);\n\n// Applies a case-insensitive fuzzy filter to the given entries and ranks\n// results.\n// Entries is a list of pointers to opaque structs, each of which contains a\n// char* string at the given offset.\n// Returns an unsorted list of {original index, score}.\nstd::vector> fuzzy_filter(const std::string_view pattern,\n const void* const* entries,\n size_t entry_count,\n size_t string_offset);\ntemplate \nstd::vector> fuzzy_filter(const std::string_view pattern,\n const std::vector& entries,\n size_t string_offset) {\n return fuzzy_filter(pattern, reinterpret_cast(entries.data()),\n entries.size(), string_offset);\n}\n\n} // namespace xe\n\n#endif // XENIA_BASE_FUZZY_H_\n" }, { "path": "src/xenia/base/hash.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_HASH_H_\n#define XENIA_BASE_HASH_H_\n\n#include \n\n#include \"xenia/base/xxhash.h\"\n\nnamespace xe {\nnamespace hash {\n\n// For use in unordered_sets and unordered_maps (primarily multisets and\n// multimaps, with manual collision resolution), where the hash is calculated\n// externally (for instance, as XXH3), possibly requiring context data rather\n// than a pure function to calculate the hash\ntemplate \nstruct IdentityHasher {\n size_t operator()(const Key& key) const { return static_cast(key); }\n};\n\ntemplate \nstruct XXHasher {\n size_t operator()(const Key& key) const {\n return static_cast(XXH3_64bits(&key, sizeof(key)));\n }\n};\n\n} // namespace hash\n} // namespace xe\n\n#endif // XENIA_BASE_HASH_H_\n" }, { "path": "src/xenia/base/host_thread_context.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/host_thread_context.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/string_util.h\"\n\nnamespace xe {\n\n// NOTE: this order matches 1:1 with the HostRegister enums.\nstatic const char* kRegisterNames[] = {\n#if XE_ARCH_AMD64\n \"rip\", \"eflags\", \"rax\", \"rcx\", \"rdx\", \"rbx\", \"rsp\",\n \"rbp\", \"rsi\", \"rdi\", \"r8\", \"r9\", \"r10\", \"r11\",\n \"r12\", \"r13\", \"r14\", \"r15\", \"xmm0\", \"xmm1\", \"xmm2\",\n \"xmm3\", \"xmm4\", \"xmm5\", \"xmm6\", \"xmm7\", \"xmm8\", \"xmm9\",\n \"xmm10\", \"xmm11\", \"xmm12\", \"xmm13\", \"xmm14\", \"xmm15\",\n#elif XE_ARCH_ARM64\n \"x0\", \"x1\", \"x2\", \"x3\", \"x4\", \"x5\", \"x6\", \"x7\", \"x8\", \"x9\",\n \"x10\", \"x11\", \"x12\", \"x13\", \"x14\", \"x15\", \"x16\", \"x17\", \"x18\", \"x19\",\n \"x20\", \"x21\", \"x22\", \"x23\", \"x24\", \"x25\", \"x26\", \"x27\", \"x28\", \"x29\",\n \"x30\", \"sp\", \"pc\", \"pstate\", \"fpsr\", \"fpcr\", \"v0\", \"v1\", \"v2\", \"v3\",\n \"v4\", \"v5\", \"v6\", \"v7\", \"v8\", \"v9\", \"v10\", \"v11\", \"v12\", \"v13\",\n \"v14\", \"v15\", \"v16\", \"v17\", \"v18\", \"v19\", \"v20\", \"v21\", \"v22\", \"v23\",\n \"v24\", \"v25\", \"v26\", \"v27\", \"v28\", \"v29\", \"v30\", \"v31\",\n#endif // XE_ARCH\n};\n\nconst char* HostThreadContext::GetRegisterName(HostRegister reg) {\n return kRegisterNames[int(reg)];\n}\n\nstd::string HostThreadContext::GetStringFromValue(HostRegister reg,\n bool hex) const {\n#if XE_ARCH_AMD64\n switch (reg) {\n case X64Register::kRip:\n return hex ? string_util::to_hex_string(rip) : std::to_string(rip);\n case X64Register::kEflags:\n return hex ? string_util::to_hex_string(eflags) : std::to_string(eflags);\n default:\n if (reg >= X64Register::kIntRegisterFirst &&\n reg <= X64Register::kIntRegisterLast) {\n auto value =\n int_registers[int(reg) - int(X64Register::kIntRegisterFirst)];\n return hex ? string_util::to_hex_string(value) : std::to_string(value);\n } else if (reg >= X64Register::kXmm0 && reg <= X64Register::kXmm15) {\n auto value = xmm_registers[int(reg) - int(X64Register::kXmm0)];\n return hex ? string_util::to_hex_string(value) : xe::to_string(value);\n } else {\n assert_unhandled_case(reg);\n return std::string();\n }\n }\n#elif XE_ARCH_ARM64\n switch (reg) {\n case Arm64Register::kSp:\n return hex ? string_util::to_hex_string(sp) : std::to_string(sp);\n case Arm64Register::kPc:\n return hex ? string_util::to_hex_string(pc) : std::to_string(pc);\n case Arm64Register::kPstate:\n return hex ? string_util::to_hex_string(pstate) : std::to_string(pstate);\n case Arm64Register::kFpsr:\n return hex ? string_util::to_hex_string(fpsr) : std::to_string(fpsr);\n case Arm64Register::kFpcr:\n return hex ? string_util::to_hex_string(fpcr) : std::to_string(fpcr);\n default:\n if (reg >= Arm64Register::kX0 && reg <= Arm64Register::kX30) {\n auto value = x[int(reg) - int(Arm64Register::kX0)];\n return hex ? string_util::to_hex_string(value) : std::to_string(value);\n } else if (reg >= Arm64Register::kV0 && reg <= Arm64Register::kV31) {\n auto value = v[int(reg) - int(Arm64Register::kV0)];\n return hex ? string_util::to_hex_string(value) : xe::to_string(value);\n } else {\n assert_unhandled_case(reg);\n return std::string();\n }\n }\n#else\n assert_always(\n \"HostThreadContext::GetStringFromValue not implemented for the target \"\n \"CPU architecture\");\n return std::string();\n#endif // XE_ARCH\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/host_thread_context.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_HOST_THREAD_CONTEXT_H_\n#define XENIA_BASE_HOST_THREAD_CONTEXT_H_\n\n#include \n#include \n\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/vec128.h\"\n\n#if XE_ARCH_AMD64\n#include \n#endif\n\nnamespace xe {\n\n// NOTE: The order of the registers in the enumerations must match the order in\n// the string table in host_thread_context.cc, as well as remapping tables in\n// exception handler implementations.\n\nenum class X64Register {\n kRip,\n kEflags,\n\n kIntRegisterFirst,\n // The order matches the indices in the instruction encoding, as well as the\n // Windows CONTEXT structure.\n kRax = kIntRegisterFirst,\n kRcx,\n kRdx,\n kRbx,\n kRsp,\n kRbp,\n kRsi,\n kRdi,\n kR8,\n kR9,\n kR10,\n kR11,\n kR12,\n kR13,\n kR14,\n kR15,\n kIntRegisterLast = kR15,\n\n kXmm0,\n kXmm1,\n kXmm2,\n kXmm3,\n kXmm4,\n kXmm5,\n kXmm6,\n kXmm7,\n kXmm8,\n kXmm9,\n kXmm10,\n kXmm11,\n kXmm12,\n kXmm13,\n kXmm14,\n kXmm15,\n};\n\nenum class Arm64Register {\n kX0,\n kX1,\n kX2,\n kX3,\n kX4,\n kX5,\n kX6,\n kX7,\n kX8,\n kX9,\n kX10,\n kX11,\n kX12,\n kX13,\n kX14,\n kX15,\n kX16,\n kX17,\n kX18,\n kX19,\n kX20,\n kX21,\n kX22,\n kX23,\n kX24,\n kX25,\n kX26,\n kX27,\n kX28,\n // FP (frame pointer).\n kX29,\n // LR (link register).\n kX30,\n kSp,\n kPc,\n kPstate,\n kFpsr,\n kFpcr,\n // The whole 128 bits of a Vn register are also known as Qn (quadword).\n kV0,\n kV1,\n kV2,\n kV3,\n kV4,\n kV5,\n kV6,\n kV7,\n kV8,\n kV9,\n kV10,\n kV11,\n kV12,\n kV13,\n kV14,\n kV15,\n kV16,\n kV17,\n kV18,\n kV19,\n kV20,\n kV21,\n kV22,\n kV23,\n kV24,\n kV25,\n kV26,\n kV27,\n kV28,\n kV29,\n kV30,\n kV31,\n};\n\n#if XE_ARCH_AMD64\nusing HostRegister = X64Register;\n#elif XE_ARCH_ARM64\nusing HostRegister = Arm64Register;\n#else\nenum class HostRegister {};\n#endif // XE_ARCH\n\nclass HostThreadContext {\n public:\n#if XE_ARCH_AMD64\n uint64_t rip;\n uint32_t eflags;\n union {\n struct {\n uint64_t rax;\n uint64_t rcx;\n uint64_t rdx;\n uint64_t rbx;\n uint64_t rsp;\n uint64_t rbp;\n uint64_t rsi;\n uint64_t rdi;\n uint64_t r8;\n uint64_t r9;\n uint64_t r10;\n uint64_t r11;\n uint64_t r12;\n uint64_t r13;\n uint64_t r14;\n uint64_t r15;\n };\n uint64_t int_registers[16];\n };\n union {\n struct {\n vec128_t xmm0;\n vec128_t xmm1;\n vec128_t xmm2;\n vec128_t xmm3;\n vec128_t xmm4;\n vec128_t xmm5;\n vec128_t xmm6;\n vec128_t xmm7;\n vec128_t xmm8;\n vec128_t xmm9;\n vec128_t xmm10;\n vec128_t xmm11;\n vec128_t xmm12;\n vec128_t xmm13;\n vec128_t xmm14;\n vec128_t xmm15;\n };\n vec128_t xmm_registers[16];\n };\n#elif XE_ARCH_ARM64\n uint64_t x[31];\n uint64_t sp;\n uint64_t pc;\n uint64_t pstate;\n uint32_t fpsr;\n uint32_t fpcr;\n vec128_t v[32];\n#endif // XE_ARCH\n\n static const char* GetRegisterName(HostRegister reg);\n std::string GetStringFromValue(HostRegister reg, bool hex) const;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_X64_CONTEXT_H_\n" }, { "path": "src/xenia/base/literals.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_LITERALS_H_\n#define XENIA_BASE_LITERALS_H_\n\n#include \n\nnamespace xe::literals {\n\nconstexpr size_t operator\"\"_KiB(unsigned long long int x) {\n return 1024ULL * x;\n}\n\nconstexpr size_t operator\"\"_MiB(unsigned long long int x) {\n return 1024_KiB * x;\n}\n\nconstexpr size_t operator\"\"_GiB(unsigned long long int x) {\n return 1024_MiB * x;\n}\n\nconstexpr size_t operator\"\"_TiB(unsigned long long int x) {\n return 1024_GiB * x;\n}\n\nconstexpr size_t operator\"\"_PiB(unsigned long long int x) {\n return 1024_TiB * x;\n}\n\n} // namespace xe::literals\n\n#endif // XENIA_BASE_LITERALS_H_\n" }, { "path": "src/xenia/base/logging.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/logging.h\"\n\n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"third_party/disruptorplus/include/disruptorplus/multi_threaded_claim_strategy.hpp\"\n#include \"third_party/disruptorplus/include/disruptorplus/ring_buffer.hpp\"\n#include \"third_party/disruptorplus/include/disruptorplus/sequence_barrier.hpp\"\n#include \"third_party/disruptorplus/include/disruptorplus/spin_wait_strategy.hpp\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/atomic.h\"\n#include \"xenia/base/console.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/debugging.h\"\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/literals.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/memory.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/ring_buffer.h\"\n#include \"xenia/base/string.h\"\n#include \"xenia/base/system.h\"\n#include \"xenia/base/threading.h\"\n\n// TODO(benvanik): generic API? logging_win.cc?\n#if XE_PLATFORM_ANDROID\n#include \n#elif XE_PLATFORM_WIN32\n// For MessageBox:\n#include \"xenia/base/platform_win.h\"\n#endif // XE_PLATFORM\n\n#include \"third_party/fmt/include/fmt/format.h\"\n\n#if XE_PLATFORM_ANDROID\nDEFINE_bool(log_to_logcat, true, \"Write log output to Android Logcat.\",\n \"Logging\");\n#else\nDEFINE_path(log_file, \"\", \"Logs are written to the given file\", \"Logging\");\nDEFINE_bool(log_to_stdout, true, \"Write log output to stdout\", \"Logging\");\nDEFINE_bool(log_to_debugprint, false, \"Dump the log to DebugPrint.\", \"Logging\");\n#endif // XE_PLATFORM_ANDROID\nDEFINE_bool(flush_log, true, \"Flush log file after each log line batch.\",\n \"Logging\");\nDEFINE_int32(\n log_level, 2,\n \"Maximum level to be logged. (0=error, 1=warning, 2=info, 3=debug)\",\n \"Logging\");\n\nnamespace dp = disruptorplus;\nusing namespace xe::literals;\n\nnamespace xe {\n\nclass Logger;\n\nLogger* logger_ = nullptr;\n\nstruct LogLine {\n size_t buffer_length;\n uint32_t thread_id;\n uint16_t _pad_0; // (2b) padding\n bool terminate;\n char prefix_char;\n};\n\nthread_local char thread_log_buffer_[64_KiB];\n\nFileLogSink::~FileLogSink() {\n if (file_) {\n fflush(file_);\n if (owns_file_) {\n fclose(file_);\n }\n }\n}\n\nvoid FileLogSink::Write(const char* buf, size_t size) {\n if (file_) {\n fwrite(buf, 1, size, file_);\n }\n}\n\nvoid FileLogSink::Flush() {\n if (file_) {\n fflush(file_);\n }\n}\n\nvoid DebugPrintLogSink::Write(const char* buf, size_t size) {\n debugging::DebugPrint(\"{}\", std::string_view(buf, size));\n}\n\n#if XE_PLATFORM_ANDROID\nclass AndroidLogSink final : public LogSink {\n public:\n explicit AndroidLogSink(const std::string_view tag) : tag_(tag) {}\n\n void Write(const char* buf, size_t size) override;\n void Flush() override;\n\n private:\n // May be called with an empty line_buffer_ to write an empty log message (if\n // the original message contains a blank line as a separator, for instance).\n void WriteLineBuffer();\n\n const std::string tag_;\n android_LogPriority current_priority_ = ANDROID_LOG_INFO;\n bool is_line_continuation_ = false;\n size_t line_buffer_used_ = 0;\n // LOGGER_ENTRY_MAX_PAYLOAD is defined as 4076 on older Android versions and\n // as 4068 on newer ones. An attempt to write more than that amount to the\n // kernel logger will result in a truncated log entry. 4000 is the commonly\n // used limit considered safe.\n // However, \"Log message text may be truncated to less than an\n // implementation-specific limit (1023 bytes)\" - android/log.h.\n char line_buffer_[1023 + 1];\n};\n\nvoid AndroidLogSink::Write(const char* buf, size_t size) {\n // A null character, if appears, is fine - it will just truncate a single\n // line, but since every message ends with a newline, and WriteLineBuffer is\n // done after every line, it won't have lasting effects.\n // Using memchr and memcpy as they are vectorized in Bionic.\n while (size) {\n const void* newline = std::memchr(buf, '\\n', size);\n size_t line_remaining =\n newline ? size_t(static_cast(newline) - buf) : size;\n while (line_remaining) {\n assert_true(line_buffer_used_ < xe::countof(line_buffer_));\n size_t append_count =\n std::min(line_remaining,\n xe::countof(line_buffer_) - size_t(1) - line_buffer_used_);\n std::memcpy(line_buffer_ + line_buffer_used_, buf, append_count);\n buf += append_count;\n size -= append_count;\n line_remaining -= append_count;\n line_buffer_used_ += append_count;\n if (line_buffer_used_ >= xe::countof(line_buffer_) - size_t(1)) {\n WriteLineBuffer();\n // The next WriteLineBuffer belongs to the same line.\n is_line_continuation_ = true;\n }\n }\n if (newline) {\n // If the end of the buffer was reached right before a newline character,\n // so line_buffer_used_ is 0, and is_line_continuation_ has been set, the\n // line break has already been done by Android itself - don't write an\n // empty message. However, if the message intentionally contains blank\n // lines, write them.\n if (line_buffer_used_ || !is_line_continuation_) {\n WriteLineBuffer();\n }\n is_line_continuation_ = false;\n ++buf;\n --size;\n }\n }\n}\n\nvoid AndroidLogSink::Flush() {\n if (line_buffer_used_) {\n WriteLineBuffer();\n }\n}\n\nvoid AndroidLogSink::WriteLineBuffer() {\n // If this is a new line, check if it's a new log message, and if it is,\n // update the priority based on the prefix.\n if (!is_line_continuation_ && line_buffer_used_ >= 3 &&\n line_buffer_[1] == '>' && line_buffer_[2] == ' ') {\n switch (line_buffer_[0]) {\n case 'd':\n current_priority_ = ANDROID_LOG_DEBUG;\n break;\n case 'w':\n current_priority_ = ANDROID_LOG_WARN;\n break;\n case '!':\n current_priority_ = ANDROID_LOG_ERROR;\n break;\n case 'x':\n current_priority_ = ANDROID_LOG_FATAL;\n break;\n default:\n current_priority_ = ANDROID_LOG_INFO;\n break;\n }\n }\n // Android skips blank lines, but if writing a blank line was requested\n // explicitly for formatting, write a non-newline character.\n if (!line_buffer_used_) {\n line_buffer_[line_buffer_used_++] = ' ';\n }\n // Terminate the text.\n assert_true(line_buffer_used_ < xe::countof(line_buffer_));\n line_buffer_[line_buffer_used_] = '\\0';\n // Write.\n __android_log_write(current_priority_, tag_.c_str(), line_buffer_);\n line_buffer_used_ = 0;\n}\n#endif // XE_PLATFORM_ANDROID\n\nclass Logger {\n public:\n explicit Logger(const std::string_view app_name)\n : wait_strategy_(),\n claim_strategy_(kBlockCount, wait_strategy_),\n consumed_(wait_strategy_) {\n claim_strategy_.add_claim_barrier(consumed_);\n\n write_thread_ =\n xe::threading::Thread::Create({}, [this]() { WriteThread(); });\n assert_not_null(write_thread_);\n write_thread_->set_name(\"Logging Writer\");\n }\n\n ~Logger() {\n AppendLine(0, '\\0', nullptr, 0, true); // append a terminator\n xe::threading::Wait(write_thread_.get(), true);\n }\n\n void AddLogSink(std::unique_ptr&& sink) {\n sinks_.push_back(std::move(sink));\n }\n\n private:\n static const size_t kBufferSize = 8_MiB;\n uint8_t buffer_[kBufferSize];\n\n static const size_t kBlockSize = 256;\n static const size_t kBlockCount = kBufferSize / kBlockSize;\n static const size_t kBlockIndexMask = kBlockCount - 1;\n\n static const size_t kClaimStrategyFootprint =\n sizeof(std::atomic[kBlockCount]);\n\n static size_t BlockOffset(dp::sequence_t sequence) {\n return (sequence & kBlockIndexMask) * kBlockSize;\n }\n\n static size_t BlockCount(size_t byte_size) {\n return (byte_size + (kBlockSize - 1)) / kBlockSize;\n }\n\n dp::spin_wait_strategy wait_strategy_;\n dp::multi_threaded_claim_strategy claim_strategy_;\n dp::sequence_barrier consumed_;\n\n std::vector> sinks_;\n\n std::unique_ptr write_thread_;\n\n void Write(const char* buf, size_t size) {\n for (const auto& sink : sinks_) {\n sink->Write(buf, size);\n }\n }\n\n void WriteThread() {\n RingBuffer rb(buffer_, kBufferSize);\n\n size_t idle_loops = 0;\n\n dp::sequence_t next_sequence = 0;\n dp::sequence_t last_sequence = -1;\n\n size_t desired_count = 1;\n while (true) {\n // We want one block to find out how many blocks we need or we know how\n // many blocks needed for at least one log line.\n auto next_range = dp::sequence_range(next_sequence, desired_count);\n\n claim_strategy_.wait_until_published(next_range.last(), last_sequence);\n\n size_t read_count = 0;\n auto available_range = next_range;\n auto available_count = available_range.size();\n\n rb.set_write_offset(BlockOffset(available_range.end()));\n\n bool terminate = false;\n for (size_t i = available_range.first(); i != available_range.end();) {\n rb.set_read_offset(BlockOffset(i));\n\n LogLine line;\n rb.Read(&line, sizeof(line));\n\n auto needed_count = BlockCount(sizeof(LogLine) + line.buffer_length);\n if (read_count + needed_count > available_count) {\n // More blocks are needed for a complete line.\n desired_count = needed_count;\n break;\n } else {\n // Enough blocks to read this log line, advance by that many.\n read_count += needed_count;\n i += needed_count;\n\n if (line.prefix_char) {\n char prefix[] = {\n line.prefix_char,\n '>',\n ' ',\n '?', // Thread ID gets placed here (8 chars).\n '?',\n '?',\n '?',\n '?',\n '?',\n '?',\n '?',\n ' ',\n 0,\n };\n fmt::format_to_n(prefix + 3, sizeof(prefix) - 3, \"{:08X}\",\n line.thread_id);\n Write(prefix, sizeof(prefix) - 1);\n }\n\n if (line.buffer_length) {\n // Get access to the line data - which may be split in the ring\n // buffer - and write it out in parts.\n auto line_range = rb.BeginRead(line.buffer_length);\n Write(reinterpret_cast(line_range.first),\n line_range.first_length);\n if (line_range.second_length) {\n Write(reinterpret_cast(line_range.second),\n line_range.second_length);\n }\n\n // Always ensure there is a newline.\n char last_char =\n line_range.second\n ? line_range.second[line_range.second_length - 1]\n : line_range.first[line_range.first_length - 1];\n if (last_char != '\\n') {\n const char suffix[1] = {'\\n'};\n Write(suffix, 1);\n }\n\n rb.EndRead(std::move(line_range));\n } else {\n // Always ensure there is a newline.\n const char suffix[1] = {'\\n'};\n Write(suffix, 1);\n }\n\n if (line.terminate) {\n terminate = true;\n break;\n }\n }\n }\n\n if (terminate) {\n break;\n }\n\n if (read_count) {\n // Advance by the number of blocks we read.\n auto read_range = dp::sequence_range(next_sequence, read_count);\n next_sequence = read_range.end();\n last_sequence = read_range.last();\n consumed_.publish(last_sequence);\n\n desired_count = 1;\n\n if (cvars::flush_log) {\n for (const auto& sink : sinks_) {\n sink->Flush();\n }\n }\n\n idle_loops = 0;\n } else {\n if (idle_loops >= 1000) {\n // Introduce a waiting period.\n xe::threading::Sleep(std::chrono::milliseconds(50));\n } else {\n idle_loops++;\n }\n }\n }\n }\n\n public:\n void AppendLine(uint32_t thread_id, const char prefix_char,\n const char* buffer_data, size_t buffer_length,\n bool terminate = false) {\n size_t count = BlockCount(sizeof(LogLine) + buffer_length);\n\n auto range = claim_strategy_.claim(count);\n assert_true(range.size() == count);\n\n RingBuffer rb(buffer_, kBufferSize);\n rb.set_write_offset(BlockOffset(range.first()));\n rb.set_read_offset(BlockOffset(range.end()));\n\n LogLine line = {};\n line.buffer_length = buffer_length;\n line.thread_id = thread_id;\n line.prefix_char = prefix_char;\n line.terminate = terminate;\n\n rb.Write(&line, sizeof(LogLine));\n if (buffer_length) {\n rb.Write(buffer_data, buffer_length);\n }\n\n claim_strategy_.publish(range);\n }\n};\n\nvoid InitializeLogging(const std::string_view app_name) {\n auto mem = memory::AlignedAlloc(0x10);\n logger_ = new (mem) Logger(app_name);\n\n#if XE_PLATFORM_ANDROID\n // TODO(Triang3l): Enable file logging, but not by default as logs may be\n // huge.\n if (cvars::log_to_logcat) {\n logger_->AddLogSink(std::make_unique(app_name));\n }\n#else\n FILE* log_file = nullptr;\n if (cvars::log_file.empty()) {\n // Default to app name.\n auto file_name = fmt::format(\"{}.log\", app_name);\n auto file_path = std::filesystem::path(file_name);\n xe::filesystem::CreateParentFolder(file_path);\n\n log_file = xe::filesystem::OpenFile(file_path, \"wt\");\n } else {\n xe::filesystem::CreateParentFolder(cvars::log_file);\n log_file = xe::filesystem::OpenFile(cvars::log_file, \"wt\");\n }\n logger_->AddLogSink(std::make_unique(log_file, true));\n\n if (cvars::log_to_stdout) {\n logger_->AddLogSink(std::make_unique(stdout, false));\n }\n\n if (cvars::log_to_debugprint) {\n logger_->AddLogSink(std::make_unique());\n }\n#endif // XE_PLATFORM_ANDROID\n}\n\nvoid ShutdownLogging() {\n Logger* logger = logger_;\n logger_ = nullptr;\n\n logger->~Logger();\n memory::AlignedFree(logger);\n}\n\nbool logging::ShouldLog(LogLevel log_level) {\n return logger_ != nullptr &&\n static_cast(log_level) <= cvars::log_level;\n}\n\nstd::pair logging::internal::GetThreadBuffer() {\n return {thread_log_buffer_, sizeof(thread_log_buffer_)};\n}\n\nvoid logging::internal::AppendLogLine(LogLevel log_level,\n const char prefix_char, size_t written) {\n if (!ShouldLog(log_level) || !written) {\n return;\n }\n logger_->AppendLine(xe::threading::current_thread_id(), prefix_char,\n thread_log_buffer_, written);\n}\n\nvoid logging::AppendLogLine(LogLevel log_level, const char prefix_char,\n const std::string_view str) {\n if (!ShouldLog(log_level) || !str.size()) {\n return;\n }\n logger_->AppendLine(xe::threading::current_thread_id(), prefix_char,\n str.data(), str.size());\n}\n\nvoid FatalError(const std::string_view str) {\n logging::AppendLogLine(LogLevel::Error, 'x', str);\n\n if (!xe::has_console_attached()) {\n ShowSimpleMessageBox(SimpleMessageBoxType::Error, str);\n }\n\n ShutdownLogging();\n\n#if XE_PLATFORM_ANDROID\n // Throw an error that can be reported to the developers via the store.\n std::abort();\n#else\n std::exit(EXIT_FAILURE);\n#endif // XE_PLATFORM_ANDROID\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/logging.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_LOGGING_H_\n#define XENIA_BASE_LOGGING_H_\n\n#include \n#include \n#include \n\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/string.h\"\n\nnamespace xe {\n\n#define XE_OPTION_ENABLE_LOGGING 1\n\n// Log level is a general indication of the importance of a given log line.\n//\n// While log levels are named, they are a rough correlation of what the log line\n// may be related to. These names should not be taken as fact as what a given\n// log line from any log level actually is.\nenum class LogLevel {\n Error = 0,\n Warning,\n Info,\n Debug,\n Trace,\n};\n\nclass LogSink {\n public:\n virtual ~LogSink() = default;\n\n virtual void Write(const char* buf, size_t size) = 0;\n virtual void Flush() = 0;\n};\n\nclass FileLogSink final : public LogSink {\n public:\n explicit FileLogSink(FILE* file, bool own_file)\n : file_(file), owns_file_(own_file) {}\n ~FileLogSink();\n\n void Write(const char* buf, size_t size) override;\n void Flush() override;\n\n private:\n FILE* file_;\n bool owns_file_;\n};\n\nclass DebugPrintLogSink final : public LogSink {\n public:\n DebugPrintLogSink() = default;\n\n void Write(const char* buf, size_t size) override;\n void Flush() override {}\n};\n\n// Initializes the logging system and any outputs requested.\n// Must be called on startup.\nvoid InitializeLogging(const std::string_view app_name);\nvoid ShutdownLogging();\n\nnamespace logging {\n\nconstexpr char kPrefixCharError = '!';\nconstexpr char kPrefixCharWarning = 'w';\nconstexpr char kPrefixCharInfo = 'i';\nconstexpr char kPrefixCharDebug = 'd';\n\nbool ShouldLog(LogLevel log_level);\n\nnamespace internal {\n\nstd::pair GetThreadBuffer();\n\nvoid AppendLogLine(LogLevel log_level, const char prefix_char, size_t written);\n\n} // namespace internal\n\n// Appends a line to the log with {fmt}-style formatting.\ntemplate \nvoid AppendLogLineFormat(LogLevel log_level, const char prefix_char,\n const char* format, const Args&... args) {\n if (!ShouldLog(log_level)) {\n return;\n }\n auto target = internal::GetThreadBuffer();\n auto result = fmt::format_to_n(target.first, target.second, format, args...);\n internal::AppendLogLine(log_level, prefix_char, result.size);\n}\n\n// Appends a line to the log.\nvoid AppendLogLine(LogLevel log_level, const char prefix_char,\n const std::string_view str);\n\n} // namespace logging\n\n// Logs a fatal error and aborts the program.\nvoid FatalError(const std::string_view str);\n\n} // namespace xe\n\n#if XE_OPTION_ENABLE_LOGGING\n\ntemplate \nvoid XELOGE(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(\n xe::LogLevel::Error, xe::logging::kPrefixCharError, format, args...);\n}\n\ntemplate \nvoid XELOGW(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(\n xe::LogLevel::Warning, xe::logging::kPrefixCharWarning, format, args...);\n}\n\ntemplate \nvoid XELOGI(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(\n xe::LogLevel::Info, xe::logging::kPrefixCharInfo, format, args...);\n}\n\ntemplate \nvoid XELOGD(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(\n xe::LogLevel::Debug, xe::logging::kPrefixCharDebug, format, args...);\n}\n\ntemplate \nvoid XELOGCPU(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(xe::LogLevel::Info, 'C', format, args...);\n}\n\ntemplate \nvoid XELOGAPU(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(xe::LogLevel::Info, 'A', format, args...);\n}\n\ntemplate \nvoid XELOGGPU(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(xe::LogLevel::Info, 'G', format, args...);\n}\n\ntemplate \nvoid XELOGKERNEL(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(xe::LogLevel::Info, 'K', format, args...);\n}\n\ntemplate \nvoid XELOGFS(const char* format, const Args&... args) {\n xe::logging::AppendLogLineFormat(xe::LogLevel::Info, 'F', format, args...);\n}\n\n#else\n\n#define __XELOGDUMMY \\\n do { \\\n } while (false)\n\n#define XELOGE(...) __XELOGDUMMY\n#define XELOGW(...) __XELOGDUMMY\n#define XELOGI(...) __XELOGDUMMY\n#define XELOGD(...) __XELOGDUMMY\n#define XELOGCPU(...) __XELOGDUMMY\n#define XELOGAPU(...) __XELOGDUMMY\n#define XELOGGPU(...) __XELOGDUMMY\n#define XELOGKERNEL(...) __XELOGDUMMY\n#define XELOGFS(...) __XELOGDUMMY\n\n#endif // ENABLE_LOGGING\n\n#endif // XENIA_BASE_LOGGING_H_\n" }, { "path": "src/xenia/base/main_android.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/main_android.h\"\n\n#include \n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/memory.h\"\n#include \"xenia/base/system.h\"\n#include \"xenia/base/threading.h\"\n\nnamespace xe {\n\nstatic size_t android_initializations_ = 0;\n\nstatic int32_t android_api_level_ = __ANDROID_API__;\n\nstatic JNIEnv* android_main_thread_jni_env_ = nullptr;\nstatic JavaVM* android_java_vm_ = nullptr;\nstatic pthread_key_t android_thread_jni_env_key_;\nstatic jobject android_application_context_ = nullptr;\n\nstatic void AndroidThreadJNIEnvDestructor(void* jni_env_pointer) {\n // The JNIEnv pointer for the main thread is taken externally, the lifetime of\n // the attachment is not managed by the key.\n JNIEnv* jni_env = static_cast(jni_env_pointer);\n if (jni_env && jni_env != android_main_thread_jni_env_) {\n android_java_vm_->DetachCurrentThread();\n }\n // Multiple iterations of destructor invocations can be done - clear.\n pthread_setspecific(android_thread_jni_env_key_, nullptr);\n}\n\nvoid InitializeAndroidAppFromMainThread(int32_t api_level,\n JNIEnv* main_thread_jni_env,\n jobject application_context,\n jobject launch_arguments_bundle) {\n if (android_initializations_++) {\n // Already initialized for another component in the process.\n return;\n }\n\n // Set the API level before everything else if available - may be needed by\n // subsystem initialization itself.\n android_api_level_ = api_level;\n\n android_main_thread_jni_env_ = main_thread_jni_env;\n if (main_thread_jni_env) {\n // In a Java VM, not just in a process that runs an executable - set up\n // the attachment of threads to the Java VM.\n if (main_thread_jni_env->GetJavaVM(&android_java_vm_) < 0) {\n // Logging has not been initialized yet.\n __android_log_write(\n ANDROID_LOG_ERROR, \"InitializeAndroidAppFromMainThread\",\n \"Failed to get the Java VM from the JNI environment of the main \"\n \"thread\");\n std::abort();\n }\n if (pthread_key_create(&android_thread_jni_env_key_,\n AndroidThreadJNIEnvDestructor)) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"InitializeAndroidAppFromMainThread\",\n \"Failed to create the thread-specific JNI environment key\");\n std::abort();\n }\n if (pthread_setspecific(android_thread_jni_env_key_, main_thread_jni_env)) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"InitializeAndroidAppFromMainThread\",\n \"Failed to set the thread-specific JNI environment pointer for the \"\n \"main thread\");\n std::abort();\n }\n if (application_context) {\n android_application_context_ =\n main_thread_jni_env->NewGlobalRef(application_context);\n if (!android_application_context_) {\n __android_log_write(\n ANDROID_LOG_ERROR, \"InitializeAndroidAppFromMainThread\",\n \"Failed to create a global reference to the application context \"\n \"object\");\n std::abort();\n }\n }\n }\n\n // Logging uses threading.\n xe::threading::AndroidInitialize();\n\n xe::filesystem::AndroidInitialize();\n\n // Initialize the cvars before logging.\n if (launch_arguments_bundle) {\n cvar::ParseLaunchArgumentsFromAndroidBundle(launch_arguments_bundle);\n }\n\n // Multiple apps can be launched within one process - don't pass the actual\n // app name.\n xe::InitializeLogging(\"xenia\");\n\n xe::memory::AndroidInitialize();\n\n if (android_application_context_) {\n if (!xe::InitializeAndroidSystemForApplicationContext()) {\n __android_log_write(ANDROID_LOG_ERROR,\n \"InitializeAndroidAppFromMainThread\",\n \"Failed to initialize system UI interaction\");\n std::abort();\n }\n }\n}\n\nvoid ShutdownAndroidAppFromMainThread() {\n assert_not_zero(android_initializations_);\n if (!android_initializations_) {\n return;\n }\n if (--android_initializations_) {\n // Other components are still running.\n return;\n }\n\n xe::ShutdownAndroidSystem();\n\n xe::memory::AndroidShutdown();\n\n xe::ShutdownLogging();\n\n xe::filesystem::AndroidShutdown();\n\n xe::threading::AndroidShutdown();\n\n if (android_application_context_) {\n android_main_thread_jni_env_->DeleteGlobalRef(android_application_context_);\n android_application_context_ = nullptr;\n }\n if (android_java_vm_) {\n android_java_vm_ = nullptr;\n pthread_key_delete(android_thread_jni_env_key_);\n }\n android_main_thread_jni_env_ = nullptr;\n\n android_api_level_ = __ANDROID_API__;\n}\n\nint32_t GetAndroidApiLevel() { return android_api_level_; }\n\nJNIEnv* GetAndroidThreadJniEnv() {\n if (!android_java_vm_) {\n return nullptr;\n }\n return static_cast(pthread_getspecific(android_thread_jni_env_key_));\n}\n\njobject GetAndroidApplicationContext() { return android_application_context_; }\n\n} // namespace xe\n" }, { "path": "src/xenia/base/main_android.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_MAIN_ANDROID_H_\n#define XENIA_BASE_MAIN_ANDROID_H_\n\n#include \n#include \n\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\n\n// In activities, these functions must be called in onCreate and onDestroy.\n//\n// Multiple components may run in the same process, and they will be\n// instantiated in the main thread, which is, for regular applications (the\n// system application exception doesn't apply to Xenia), the UI thread.\n//\n// For this reason, it's okay to call these functions multiple times if\n// different Xenia windowed applications run in one process - there is call\n// counting internally.\n//\n// In standalone console apps built with $(BUILD_EXECUTABLE), these functions\n// must be called in `main`, with a null main thread JNI environment.\nvoid InitializeAndroidAppFromMainThread(int32_t api_level,\n JNIEnv* main_thread_jni_env,\n jobject application_context,\n jobject launch_arguments_bundle);\nvoid ShutdownAndroidAppFromMainThread();\n\n// May be the minimum supported level if the initialization was done without a\n// configuration.\nint32_t GetAndroidApiLevel();\n\n// Useful notes about JNI usage on Android within Xenia:\n// - All static libraries defining JNI native functions must be linked to shared\n// libraries via LOCAL_WHOLE_STATIC_LIBRARIES.\n// - If method or field IDs are cached, a global reference to the class needs to\n// be held - it prevents the class from being unloaded by the class loaders\n// (in a way that would make the IDs invalid when it's reloaded).\n// - GetStringUTFChars (UTF-8) returns null-terminated strings, GetStringChars\n// (UTF-16) does not.\n\n// May return null if not in a Java VM process, or in case of a failure to\n// attach on a non-main thread.\nJNIEnv* GetAndroidThreadJniEnv();\n// Returns the global reference if in an application context, or null otherwise.\n// This is the application context, not the activity one, because multiple\n// activities may be running in one process.\njobject GetAndroidApplicationContext();\n\n} // namespace xe\n\n#endif // XENIA_BASE_MAIN_ANDROID_H_\n" }, { "path": "src/xenia/base/main_init_android.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n// Nothing. Stub.\n" }, { "path": "src/xenia/base/main_init_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n// Nothing. Stub.\n" }, { "path": "src/xenia/base/main_init_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/platform_win.h\"\n\n#include \n\n// Includes Windows headers, so it goes after platform_win.h.\n#include \"third_party/xbyak/xbyak/xbyak_util.h\"\n\nclass StartupAvxCheck {\n public:\n StartupAvxCheck() {\n Xbyak::util::Cpu cpu;\n if (cpu.has(Xbyak::util::Cpu::tAVX)) {\n return;\n }\n // TODO(gibbed): detect app type and printf instead, if needed?\n MessageBoxA(\n nullptr,\n \"Your CPU does not support AVX, which is required by Xenia. See the \"\n \"FAQ for system requirements at https://xenia.jp\",\n \"Xenia Error\", MB_OK | MB_ICONERROR | MB_SETFOREGROUND);\n ExitProcess(static_cast(-1));\n }\n};\n\n// This is a hack to get an instance of StartupAvxCheck\n// constructed before any initialization code,\n// where the AVX check then happens in the constructor.\n//\n// https://docs.microsoft.com/en-us/cpp/preprocessor/init-seg\n#pragma warning(suppress : 4073)\n#pragma init_seg(lib)\nstatic StartupAvxCheck gStartupAvxCheck;\n" }, { "path": "src/xenia/base/main_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/main_win.h\"\n#include \"xenia/base/platform_win.h\"\n#include \"xenia/base/string.h\"\n\n// Autogenerated by `xb premake`.\n#include \"build/version.h\"\n\n// For RequestWin32MMCSS.\n#include \n// For RequestWin32HighResolutionTimer.\n#include \n\nDEFINE_bool(win32_high_resolution_timer, true,\n \"Requests high-resolution timer from the NT kernel\", \"Win32\");\nDEFINE_bool(\n win32_mmcss, true,\n \"Opt in the Multimedia Class Scheduler Service (MMCSS) scheduling for \"\n \"prioritized access to CPU resources\",\n \"Win32\");\n\nnamespace xe {\n\nstatic void RequestWin32HighResolutionTimer() {\n HMODULE ntdll_module = GetModuleHandleW(L\"ntdll.dll\");\n if (!ntdll_module) {\n return;\n }\n\n // clang-format off\n NTSTATUS (NTAPI* nt_query_timer_resolution)(OUT PULONG MinimumResolution,\n OUT PULONG MaximumResolution,\n OUT PULONG CurrentResolution);\n NTSTATUS (NTAPI* nt_set_timer_resolution)(IN ULONG DesiredResolution,\n IN BOOLEAN SetResolution,\n OUT PULONG CurrentResolution);\n // clang-format on\n nt_query_timer_resolution =\n reinterpret_cast(\n GetProcAddress(ntdll_module, \"NtQueryTimerResolution\"));\n nt_set_timer_resolution = reinterpret_cast(\n GetProcAddress(ntdll_module, \"NtSetTimerResolution\"));\n if (!nt_query_timer_resolution || !nt_set_timer_resolution) {\n return;\n }\n\n ULONG minimum_resolution, maximum_resolution, current_resolution;\n nt_query_timer_resolution(&minimum_resolution, &maximum_resolution,\n ¤t_resolution);\n nt_set_timer_resolution(maximum_resolution, TRUE, ¤t_resolution);\n}\n\nstatic void RequestWin32MMCSS() {\n HMODULE dwmapi_module = LoadLibraryW(L\"dwmapi.dll\");\n if (!dwmapi_module) {\n return;\n }\n // clang-format off\n HRESULT (STDAPICALLTYPE* dwm_enable_mmcss)(BOOL fEnableMMCSS);\n // clang-format on\n dwm_enable_mmcss = reinterpret_cast(\n GetProcAddress(dwmapi_module, \"DwmEnableMMCSS\"));\n if (dwm_enable_mmcss) {\n dwm_enable_mmcss(TRUE);\n }\n FreeLibrary(dwmapi_module);\n}\n\nbool ParseWin32LaunchArguments(\n bool transparent_options, const std::string_view positional_usage,\n const std::vector& positional_options,\n std::vector* args_out) {\n auto command_line = GetCommandLineW();\n\n int wargc;\n wchar_t** wargv = CommandLineToArgvW(command_line, &wargc);\n if (!wargv) {\n return false;\n }\n\n // Convert all args to narrow, as cxxopts doesn't support wchar.\n int argc = wargc;\n char** argv = reinterpret_cast(alloca(sizeof(char*) * argc));\n for (int n = 0; n < argc; n++) {\n size_t len = std::wcstombs(nullptr, wargv[n], 0);\n argv[n] = reinterpret_cast(alloca(sizeof(char) * (len + 1)));\n std::wcstombs(argv[n], wargv[n], len + 1);\n }\n\n LocalFree(wargv);\n\n if (!transparent_options) {\n cvar::ParseLaunchArguments(argc, argv, positional_usage,\n positional_options);\n }\n\n if (args_out) {\n args_out->clear();\n for (int n = 0; n < argc; n++) {\n args_out->push_back(std::string(argv[n]));\n }\n }\n\n return true;\n}\n\nint InitializeWin32App(const std::string_view app_name) {\n // Initialize logging. Needs parsed FLAGS.\n xe::InitializeLogging(app_name);\n\n // Print version info.\n XELOGI(\n \"Build: \"\n#ifdef XE_BUILD_IS_PR\n \"PR#\" XE_BUILD_PR_NUMBER \" \" XE_BUILD_PR_REPO \" \" XE_BUILD_PR_BRANCH\n \"@\" XE_BUILD_PR_COMMIT_SHORT \" against \"\n#endif\n XE_BUILD_BRANCH \"@\" XE_BUILD_COMMIT_SHORT \" on \" XE_BUILD_DATE);\n\n // Request high-performance timing and scheduling.\n if (cvars::win32_high_resolution_timer) {\n RequestWin32HighResolutionTimer();\n }\n if (cvars::win32_mmcss) {\n RequestWin32MMCSS();\n }\n\n return 0;\n}\n\nvoid ShutdownWin32App() { xe::ShutdownLogging(); }\n\n} // namespace xe\n" }, { "path": "src/xenia/base/main_win.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_MAIN_WIN_H_\n#define XENIA_BASE_MAIN_WIN_H_\n\n#include \n#include \n\nnamespace xe {\n\n// Functions for calling in both windowed and console entry points.\nbool ParseWin32LaunchArguments(\n bool transparent_options, const std::string_view positional_usage,\n const std::vector& positional_options,\n std::vector* args_out);\n// InitializeWin32App uses cvars, call ParseWin32LaunchArguments before.\nint InitializeWin32App(const std::string_view app_name);\nvoid ShutdownWin32App();\n\n} // namespace xe\n\n#endif // XENIA_BASE_MAIN_WIN_H_\n" }, { "path": "src/xenia/base/mapped_memory.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_MAPPED_MEMORY_H_\n#define XENIA_BASE_MAPPED_MEMORY_H_\n\n#include \n#include \n#include \n\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\n\nclass MappedMemory {\n public:\n enum class Mode {\n kRead,\n kReadWrite,\n };\n\n static std::unique_ptr Open(const std::filesystem::path& path,\n Mode mode, size_t offset = 0,\n size_t length = 0);\n#if XE_PLATFORM_ANDROID\n static std::unique_ptr OpenForAndroidContentUri(\n const std::string_view uri, Mode mode, size_t offset = 0,\n size_t length = 0);\n#endif // XE_PLATFORM_ANDROID\n\n MappedMemory() : data_(nullptr), size_(0) {}\n MappedMemory(void* data, size_t size) : data_(data), size_(size) {}\n MappedMemory(const MappedMemory& mapped_memory) = delete;\n MappedMemory& operator=(const MappedMemory& mapped_memory) = delete;\n MappedMemory(MappedMemory&& mapped_memory) = delete;\n MappedMemory& operator=(MappedMemory&& mapped_memory) = delete;\n virtual ~MappedMemory() = default;\n\n // The mapping is still backed by the object the slice was created from, a\n // slice is not owning.\n std::unique_ptr Slice(size_t offset, size_t length) {\n return std::unique_ptr(\n new MappedMemory(data() + offset, length));\n }\n\n uint8_t* data() const { return reinterpret_cast(data_); }\n size_t size() const { return size_; }\n\n // Close, and optionally truncate file to size\n virtual void Close(uint64_t truncate_size = 0) {}\n virtual void Flush() {}\n\n // Changes the offset inside the file. This will update data() and size()!\n virtual bool Remap(size_t offset, size_t length) { return false; }\n\n protected:\n void* data_;\n size_t size_;\n};\n\nclass ChunkedMappedMemoryWriter {\n public:\n virtual ~ChunkedMappedMemoryWriter() = default;\n\n static std::unique_ptr Open(\n const std::filesystem::path& path, size_t chunk_size,\n bool low_address_space = false);\n\n virtual uint8_t* Allocate(size_t length) = 0;\n virtual void Flush() = 0;\n virtual void FlushNew() = 0;\n\n protected:\n ChunkedMappedMemoryWriter(const std::filesystem::path& path,\n size_t chunk_size, bool low_address_space)\n : path_(path),\n chunk_size_(chunk_size),\n low_address_space_(low_address_space) {}\n\n std::filesystem::path path_;\n size_t chunk_size_;\n bool low_address_space_;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_MAPPED_MEMORY_H_\n" }, { "path": "src/xenia/base/mapped_memory_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/mapped_memory.h\"\n\n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\n\nclass PosixMappedMemory : public MappedMemory {\n public:\n PosixMappedMemory(void* data, size_t size, int file_descriptor)\n : MappedMemory(data, size), file_descriptor_(file_descriptor) {}\n\n ~PosixMappedMemory() override {\n if (data_) {\n munmap(data_, size());\n }\n if (file_descriptor_ >= 0) {\n close(file_descriptor_);\n }\n }\n\n static std::unique_ptr WrapFileDescriptor(\n int file_descriptor, Mode mode, size_t offset = 0, size_t length = 0) {\n int protection = 0;\n switch (mode) {\n case Mode::kRead:\n protection |= PROT_READ;\n break;\n case Mode::kReadWrite:\n protection |= PROT_READ | PROT_WRITE;\n break;\n }\n\n size_t map_length = length;\n if (!length) {\n struct stat64 file_stat;\n if (fstat64(file_descriptor, &file_stat)) {\n close(file_descriptor);\n return nullptr;\n }\n map_length = size_t(file_stat.st_size);\n }\n\n void* data =\n mmap(0, map_length, protection, MAP_SHARED, file_descriptor, offset);\n if (!data) {\n close(file_descriptor);\n return nullptr;\n }\n\n return std::make_unique(data, map_length,\n file_descriptor);\n }\n\n void Close(uint64_t truncate_size) override {\n if (data_) {\n munmap(data_, size());\n data_ = nullptr;\n }\n if (file_descriptor_ >= 0) {\n if (truncate_size) {\n ftruncate64(file_descriptor_, off64_t(truncate_size));\n }\n close(file_descriptor_);\n file_descriptor_ = -1;\n }\n }\n\n void Flush() override { msync(data(), size(), MS_ASYNC); }\n\n private:\n int file_descriptor_;\n};\n\nstd::unique_ptr MappedMemory::Open(\n const std::filesystem::path& path, Mode mode, size_t offset,\n size_t length) {\n int open_flags = 0;\n switch (mode) {\n case Mode::kRead:\n open_flags |= O_RDONLY;\n break;\n case Mode::kReadWrite:\n open_flags |= O_RDWR;\n break;\n }\n int file_descriptor = open(path.c_str(), open_flags);\n if (file_descriptor < 0) {\n return nullptr;\n }\n return PosixMappedMemory::WrapFileDescriptor(file_descriptor, mode, offset,\n length);\n}\n\n#if XE_PLATFORM_ANDROID\nstd::unique_ptr MappedMemory::OpenForAndroidContentUri(\n const std::string_view uri, Mode mode, size_t offset, size_t length) {\n const char* open_mode = nullptr;\n switch (mode) {\n case Mode::kRead:\n open_mode = \"r\";\n break;\n case Mode::kReadWrite:\n open_mode = \"rw\";\n break;\n }\n int file_descriptor =\n xe::filesystem::OpenAndroidContentFileDescriptor(uri, open_mode);\n if (file_descriptor < 0) {\n return nullptr;\n }\n return PosixMappedMemory::WrapFileDescriptor(file_descriptor, mode, offset,\n length);\n}\n#endif // XE_PLATFORM_ANDROID\n\nstd::unique_ptr ChunkedMappedMemoryWriter::Open(\n const std::filesystem::path& path, size_t chunk_size,\n bool low_address_space) {\n // TODO(DrChat)\n return nullptr;\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/mapped_memory_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n#include \n\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/mapped_memory.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/memory.h\"\n#include \"xenia/base/platform_win.h\"\n\n#if WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP | \\\n WINAPI_PARTITION_SYSTEM | WINAPI_PARTITION_GAMES)\n#define XE_BASE_MAPPED_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n#endif\n\nnamespace xe {\n\nclass Win32MappedMemory : public MappedMemory {\n public:\n // CreateFile returns INVALID_HANDLE_VALUE in case of failure.\n static constexpr HANDLE kFileHandleInvalid = INVALID_HANDLE_VALUE;\n // CreateFileMapping returns nullptr in case of failure.\n static constexpr HANDLE kMappingHandleInvalid = nullptr;\n\n ~Win32MappedMemory() override {\n if (data_) {\n UnmapViewOfFile(data_);\n }\n if (mapping_handle != kMappingHandleInvalid) {\n CloseHandle(mapping_handle);\n }\n if (file_handle != kFileHandleInvalid) {\n CloseHandle(file_handle);\n }\n }\n\n void Close(uint64_t truncate_size) override {\n if (data_) {\n UnmapViewOfFile(data_);\n data_ = nullptr;\n }\n if (mapping_handle != kMappingHandleInvalid) {\n CloseHandle(mapping_handle);\n mapping_handle = kMappingHandleInvalid;\n }\n if (file_handle != kFileHandleInvalid) {\n if (truncate_size) {\n LONG distance_high = truncate_size >> 32;\n SetFilePointer(file_handle, truncate_size & 0xFFFFFFFF, &distance_high,\n FILE_BEGIN);\n SetEndOfFile(file_handle);\n }\n\n CloseHandle(file_handle);\n file_handle = kFileHandleInvalid;\n }\n }\n\n void Flush() override { FlushViewOfFile(data(), size()); }\n bool Remap(size_t offset, size_t length) override {\n size_t aligned_offset = offset & ~(memory::allocation_granularity() - 1);\n size_t aligned_length = length + (offset - aligned_offset);\n\n UnmapViewOfFile(data_);\n#ifdef XE_BASE_MAPPED_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n data_ = MapViewOfFile(mapping_handle, view_access_, aligned_offset >> 32,\n aligned_offset & 0xFFFFFFFF, aligned_length);\n#else\n data_ = MapViewOfFileFromApp(mapping_handle, ULONG(view_access_),\n ULONG64(aligned_offset), aligned_length);\n#endif\n if (!data_) {\n return false;\n }\n\n if (length) {\n size_ = aligned_length;\n } else {\n DWORD length_high;\n size_t map_length = GetFileSize(file_handle, &length_high);\n map_length |= static_cast(length_high) << 32;\n size_ = map_length - aligned_offset;\n }\n\n return true;\n }\n\n HANDLE file_handle = kFileHandleInvalid;\n HANDLE mapping_handle = kMappingHandleInvalid;\n DWORD view_access_ = 0;\n};\n\nstd::unique_ptr MappedMemory::Open(\n const std::filesystem::path& path, Mode mode, size_t offset,\n size_t length) {\n DWORD file_access = 0;\n DWORD file_share = 0;\n DWORD create_mode = 0;\n DWORD mapping_protect = 0;\n DWORD view_access = 0;\n switch (mode) {\n case Mode::kRead:\n file_access |= GENERIC_READ;\n file_share |= FILE_SHARE_READ;\n create_mode |= OPEN_EXISTING;\n mapping_protect |= PAGE_READONLY;\n view_access |= FILE_MAP_READ;\n break;\n case Mode::kReadWrite:\n file_access |= GENERIC_READ | GENERIC_WRITE;\n file_share |= 0;\n create_mode |= OPEN_EXISTING;\n mapping_protect |= PAGE_READWRITE;\n view_access |= FILE_MAP_READ | FILE_MAP_WRITE;\n break;\n }\n\n SYSTEM_INFO system_info;\n GetSystemInfo(&system_info);\n\n const size_t aligned_offset =\n offset & ~static_cast(system_info.dwAllocationGranularity - 1);\n const size_t aligned_length = length + (offset - aligned_offset);\n\n auto mm = std::make_unique();\n mm->view_access_ = view_access;\n\n mm->file_handle = CreateFile(path.c_str(), file_access, file_share, nullptr,\n create_mode, FILE_ATTRIBUTE_NORMAL, nullptr);\n if (mm->file_handle == Win32MappedMemory::kFileHandleInvalid) {\n return nullptr;\n }\n\n#ifdef XE_BASE_MAPPED_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n mm->mapping_handle = CreateFileMapping(\n mm->file_handle, nullptr, mapping_protect, DWORD(aligned_length >> 32),\n DWORD(aligned_length), nullptr);\n#else\n mm->mapping_handle =\n CreateFileMappingFromApp(mm->file_handle, nullptr, ULONG(mapping_protect),\n ULONG64(aligned_length), nullptr);\n#endif\n if (mm->mapping_handle == Win32MappedMemory::kMappingHandleInvalid) {\n return nullptr;\n }\n\n#ifdef XE_BASE_MAPPED_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n mm->data_ = reinterpret_cast(MapViewOfFile(\n mm->mapping_handle, view_access, DWORD(aligned_offset >> 32),\n DWORD(aligned_offset), aligned_length));\n#else\n mm->data_ = reinterpret_cast(\n MapViewOfFileFromApp(mm->mapping_handle, ULONG(view_access),\n ULONG64(aligned_offset), aligned_length));\n#endif\n if (!mm->data_) {\n return nullptr;\n }\n\n if (length) {\n mm->size_ = aligned_length;\n } else {\n DWORD length_high;\n size_t map_length = GetFileSize(mm->file_handle, &length_high);\n map_length |= static_cast(length_high) << 32;\n mm->size_ = map_length - aligned_offset;\n }\n\n return std::move(mm);\n}\n\nclass Win32ChunkedMappedMemoryWriter : public ChunkedMappedMemoryWriter {\n public:\n Win32ChunkedMappedMemoryWriter(const std::filesystem::path& path,\n size_t chunk_size, bool low_address_space)\n : ChunkedMappedMemoryWriter(path, chunk_size, low_address_space) {}\n\n ~Win32ChunkedMappedMemoryWriter() override {\n std::lock_guard lock(mutex_);\n chunks_.clear();\n }\n\n uint8_t* Allocate(size_t length) override {\n std::lock_guard lock(mutex_);\n if (!chunks_.empty()) {\n uint8_t* result = chunks_.back()->Allocate(length);\n if (result != nullptr) {\n return result;\n }\n }\n auto chunk = std::make_unique(chunk_size_);\n auto chunk_path =\n path_.replace_extension(fmt::format(\".{}\", chunks_.size()));\n if (!chunk->Open(chunk_path, low_address_space_)) {\n return nullptr;\n }\n uint8_t* result = chunk->Allocate(length);\n chunks_.push_back(std::move(chunk));\n return result;\n }\n\n void Flush() override {\n std::lock_guard lock(mutex_);\n for (auto& chunk : chunks_) {\n chunk->Flush();\n }\n }\n\n void FlushNew() override {\n std::lock_guard lock(mutex_);\n for (auto& chunk : chunks_) {\n chunk->FlushNew();\n }\n }\n\n private:\n class Chunk {\n public:\n explicit Chunk(size_t capacity)\n : file_handle_(Win32MappedMemory::kFileHandleInvalid),\n mapping_handle_(Win32MappedMemory::kMappingHandleInvalid),\n data_(nullptr),\n offset_(0),\n capacity_(capacity),\n last_flush_offset_(0) {}\n\n ~Chunk() {\n if (data_) {\n UnmapViewOfFile(data_);\n }\n if (mapping_handle_ != Win32MappedMemory::kMappingHandleInvalid) {\n CloseHandle(mapping_handle_);\n }\n if (file_handle_ != Win32MappedMemory::kFileHandleInvalid) {\n CloseHandle(file_handle_);\n }\n }\n\n bool Open(const std::filesystem::path& path, bool low_address_space) {\n DWORD file_access = GENERIC_READ | GENERIC_WRITE;\n DWORD file_share = FILE_SHARE_READ;\n DWORD create_mode = CREATE_ALWAYS;\n DWORD mapping_protect = PAGE_READWRITE;\n DWORD view_access = FILE_MAP_READ | FILE_MAP_WRITE;\n\n file_handle_ = CreateFile(path.c_str(), file_access, file_share, nullptr,\n create_mode, FILE_ATTRIBUTE_NORMAL, nullptr);\n if (file_handle_ == Win32MappedMemory::kFileHandleInvalid) {\n return false;\n }\n\n#ifdef XE_BASE_MAPPED_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n mapping_handle_ =\n CreateFileMapping(file_handle_, nullptr, mapping_protect,\n DWORD(capacity_ >> 32), DWORD(capacity_), nullptr);\n#else\n mapping_handle_ = CreateFileMappingFromApp(file_handle_, nullptr,\n ULONG(mapping_protect),\n ULONG64(capacity_), nullptr);\n#endif\n if (mapping_handle_ == Win32MappedMemory::kMappingHandleInvalid) {\n return false;\n }\n\n // If specified, ensure the allocation occurs in the lower 32-bit address\n // space.\n if (low_address_space) {\n bool successful = false;\n data_ = reinterpret_cast(0x10000000);\n#ifndef XE_BASE_MAPPED_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n HANDLE process = GetCurrentProcess();\n#endif\n for (int i = 0; i < 1000; ++i) {\n#ifdef XE_BASE_MAPPED_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n if (MapViewOfFileEx(mapping_handle_, view_access, 0, 0, capacity_,\n data_)) {\n successful = true;\n }\n#else\n // VirtualAlloc2FromApp and MapViewOfFile3FromApp were added in\n // 10.0.17134.0.\n // https://docs.microsoft.com/en-us/uwp/win32-and-com/win32-apis\n if (VirtualAlloc2FromApp(process, data_, capacity_,\n MEM_RESERVE | MEM_RESERVE_PLACEHOLDER,\n PAGE_NOACCESS, nullptr, 0)) {\n if (MapViewOfFile3FromApp(mapping_handle_, process, data_, 0,\n capacity_, MEM_REPLACE_PLACEHOLDER,\n ULONG(mapping_protect), nullptr, 0)) {\n successful = true;\n } else {\n VirtualFree(data_, capacity_, MEM_RELEASE);\n }\n }\n#endif\n if (successful) {\n break;\n }\n data_ += capacity_;\n if (!successful) {\n XELOGE(\"Unable to find space for mapping\");\n data_ = nullptr;\n return false;\n }\n }\n } else {\n#ifdef XE_BASE_MAPPED_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n data_ = reinterpret_cast(\n MapViewOfFile(mapping_handle_, view_access, 0, 0, capacity_));\n#else\n data_ = reinterpret_cast(MapViewOfFileFromApp(\n mapping_handle_, ULONG(view_access), 0, capacity_));\n#endif\n }\n if (!data_) {\n return false;\n }\n\n return true;\n }\n\n uint8_t* Allocate(size_t length) {\n if (capacity_ - offset_ < length) {\n return nullptr;\n }\n uint8_t* result = data_ + offset_;\n offset_ += length;\n return result;\n }\n\n void Flush() { FlushViewOfFile(data_, offset_); }\n\n void FlushNew() {\n FlushViewOfFile(data_ + last_flush_offset_, offset_ - last_flush_offset_);\n last_flush_offset_ = offset_;\n }\n\n private:\n HANDLE file_handle_;\n HANDLE mapping_handle_;\n uint8_t* data_;\n size_t offset_;\n size_t capacity_;\n size_t last_flush_offset_;\n };\n\n std::mutex mutex_;\n std::vector> chunks_;\n};\n\nstd::unique_ptr ChunkedMappedMemoryWriter::Open(\n const std::filesystem::path& path, size_t chunk_size,\n bool low_address_space) {\n SYSTEM_INFO system_info;\n GetSystemInfo(&system_info);\n size_t aligned_chunk_size =\n xe::round_up(chunk_size, system_info.dwAllocationGranularity);\n return std::make_unique(\n path, aligned_chunk_size, low_address_space);\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/math.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2019 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_MATH_H_\n#define XENIA_BASE_MATH_H_\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#if defined __has_include\n#if __has_include()\n#include \n#endif\n#endif\n#if __cpp_lib_bitops\n#include \n#endif\n\n#include \"xenia/base/platform.h\"\n\n#if XE_ARCH_AMD64\n#include \n#endif\n\nnamespace xe {\n\ntemplate \nconstexpr size_t countof(T (&arr)[N]) {\n return std::extent::value;\n}\n\ntemplate \nconstexpr bool is_pow2(T value) {\n return (value & (value - 1)) == 0;\n}\n\n// Rounds up the given value to the given alignment.\ntemplate \nconstexpr T align(T value, T alignment) {\n return (value + alignment - 1) & ~(alignment - 1);\n}\n\n// Rounds the given number up to the next highest multiple.\ntemplate \nconstexpr T round_up(T value, V multiple, bool force_non_zero = true) {\n if (force_non_zero && !value) {\n return multiple;\n }\n return (value + multiple - 1) / multiple * multiple;\n}\n\n// For NaN, returns min_value (or, if it's NaN too, max_value).\n// If either of the boundaries is zero, and if the value is at that boundary or\n// exceeds it, the result will have the sign of that boundary. If both\n// boundaries are zero, which sign is selected among the argument signs is not\n// explicitly defined.\ntemplate \nT clamp_float(T value, T min_value, T max_value) {\n float clamped_to_min = std::isgreater(value, min_value) ? value : min_value;\n return std::isless(clamped_to_min, max_value) ? clamped_to_min : max_value;\n}\n\n// Using the same conventions as in shading languages, returning 0 for NaN.\n// 0 is always returned as positive.\ntemplate \nT saturate(T value) {\n return clamp_float(value, static_cast(0.0f), static_cast(1.0f));\n}\n\n// Gets the next power of two value that is greater than or equal to the given\n// value.\ntemplate \nT next_pow2(T value) {\n value--;\n value |= value >> 1;\n value |= value >> 2;\n value |= value >> 4;\n value |= value >> 8;\n value |= value >> 16;\n value++;\n return value;\n}\n\n#if __cpp_lib_gcd_lcm\ntemplate \nconstexpr T greatest_common_divisor(T a, T b) {\n return std::gcd(a, b);\n}\n#else\ntemplate \nconstexpr T greatest_common_divisor(T a, T b) {\n // Use the Euclid algorithm to calculate the greatest common divisor\n while (b) {\n a = std::exchange(b, a % b);\n }\n return a;\n}\n#endif\n\ntemplate \nconstexpr void reduce_fraction(T& numerator, T& denominator) {\n auto gcd = greatest_common_divisor(numerator, denominator);\n numerator /= gcd;\n denominator /= gcd;\n}\n\ntemplate \nconstexpr void reduce_fraction(std::pair& fraction) {\n reduce_fraction(fraction.first, fraction.second);\n}\n\nconstexpr uint32_t make_bitmask(uint32_t a, uint32_t b) {\n return (static_cast(-1) >> (31 - b)) & ~((1u << a) - 1);\n}\n\nconstexpr uint32_t select_bits(uint32_t value, uint32_t a, uint32_t b) {\n return (value & make_bitmask(a, b)) >> a;\n}\n\n#if __cpp_lib_bitops\ntemplate \nconstexpr inline uint32_t bit_count(T v) {\n return static_cast(std::popcount(v));\n}\n#else\n#if XE_COMPILER_MSVC || XE_COMPILER_INTEL\ninline uint32_t bit_count(uint32_t v) { return __popcnt(v); }\ninline uint32_t bit_count(uint64_t v) {\n return static_cast(__popcnt64(v));\n}\n#elif XE_COMPILER_GCC || XE_COMPILER_CLANG\nstatic_assert(sizeof(unsigned int) == sizeof(uint32_t));\nstatic_assert(sizeof(unsigned long long) == sizeof(uint64_t));\ninline uint32_t bit_count(uint32_t v) { return __builtin_popcount(v); }\ninline uint32_t bit_count(uint64_t v) { return __builtin_popcountll(v); }\n#else\ninline uint32_t bit_count(uint32_t v) {\n v = v - ((v >> 1) & 0x55555555);\n v = (v & 0x33333333) + ((v >> 2) & 0x33333333);\n return ((v + (v >> 4) & 0xF0F0F0F) * 0x1010101) >> 24;\n}\n\ninline uint32_t bit_count(uint64_t v) {\n v = (v & 0x5555555555555555LU) + (v >> 1 & 0x5555555555555555LU);\n v = (v & 0x3333333333333333LU) + (v >> 2 & 0x3333333333333333LU);\n v = v + (v >> 4) & 0x0F0F0F0F0F0F0F0FLU;\n v = v + (v >> 8);\n v = v + (v >> 16);\n v = v + (v >> 32) & 0x0000007F;\n return static_cast(v);\n}\n#endif\n#endif\n\n// lzcnt instruction, typed for integers of all sizes.\n// The number of leading zero bits in the value parameter. If value is zero, the\n// return value is the size of the input operand (8, 16, 32, or 64). If the most\n// significant bit of value is one, the return value is zero.\n#if XE_PLATFORM_WIN32\n// TODO(benvanik): runtime magic so these point to an appropriate implementation\n// at runtime based on CPU features\n#if 0\ninline uint8_t lzcnt(uint8_t v) {\n return static_cast(__lzcnt16(v) - 8);\n}\ninline uint8_t lzcnt(uint16_t v) { return static_cast(__lzcnt16(v)); }\ninline uint8_t lzcnt(uint32_t v) { return static_cast(__lzcnt(v)); }\ninline uint8_t lzcnt(uint64_t v) { return static_cast(__lzcnt64(v)); }\n#else\ninline uint8_t lzcnt(uint8_t v) {\n unsigned long index;\n unsigned long mask = v;\n unsigned char is_nonzero = _BitScanReverse(&index, mask);\n return static_cast(is_nonzero ? int8_t(index) ^ 0x7 : 8);\n}\ninline uint8_t lzcnt(uint16_t v) {\n unsigned long index;\n unsigned long mask = v;\n unsigned char is_nonzero = _BitScanReverse(&index, mask);\n return static_cast(is_nonzero ? int8_t(index) ^ 0xF : 16);\n}\ninline uint8_t lzcnt(uint32_t v) {\n unsigned long index;\n unsigned long mask = v;\n unsigned char is_nonzero = _BitScanReverse(&index, mask);\n return static_cast(is_nonzero ? int8_t(index) ^ 0x1F : 32);\n}\ninline uint8_t lzcnt(uint64_t v) {\n unsigned long index;\n unsigned long long mask = v;\n unsigned char is_nonzero = _BitScanReverse64(&index, mask);\n return static_cast(is_nonzero ? int8_t(index) ^ 0x3F : 64);\n}\n#endif // LZCNT supported\n\ninline uint8_t tzcnt(uint8_t v) {\n unsigned long index;\n unsigned long mask = v;\n unsigned char is_nonzero = _BitScanForward(&index, mask);\n return static_cast(is_nonzero ? int8_t(index) : 8);\n}\n\ninline uint8_t tzcnt(uint16_t v) {\n unsigned long index;\n unsigned long mask = v;\n unsigned char is_nonzero = _BitScanForward(&index, mask);\n return static_cast(is_nonzero ? int8_t(index) : 16);\n}\n\ninline uint8_t tzcnt(uint32_t v) {\n unsigned long index;\n unsigned long mask = v;\n unsigned char is_nonzero = _BitScanForward(&index, mask);\n return static_cast(is_nonzero ? int8_t(index) : 32);\n}\n\ninline uint8_t tzcnt(uint64_t v) {\n unsigned long index;\n unsigned long long mask = v;\n unsigned char is_nonzero = _BitScanForward64(&index, mask);\n return static_cast(is_nonzero ? int8_t(index) : 64);\n}\n\n#else // XE_PLATFORM_WIN32\ninline uint8_t lzcnt(uint8_t v) {\n return v == 0 ? 8 : static_cast(__builtin_clz(v) - 24);\n}\ninline uint8_t lzcnt(uint16_t v) {\n return v == 0 ? 16 : static_cast(__builtin_clz(v) - 16);\n}\ninline uint8_t lzcnt(uint32_t v) {\n return v == 0 ? 32 : static_cast(__builtin_clz(v));\n}\ninline uint8_t lzcnt(uint64_t v) {\n return v == 0 ? 64 : static_cast(__builtin_clzll(v));\n}\n\ninline uint8_t tzcnt(uint8_t v) {\n return v == 0 ? 8 : static_cast(__builtin_ctz(v));\n}\ninline uint8_t tzcnt(uint16_t v) {\n return v == 0 ? 16 : static_cast(__builtin_ctz(v));\n}\ninline uint8_t tzcnt(uint32_t v) {\n return v == 0 ? 32 : static_cast(__builtin_ctz(v));\n}\ninline uint8_t tzcnt(uint64_t v) {\n return v == 0 ? 64 : static_cast(__builtin_ctzll(v));\n}\n#endif\ninline uint8_t lzcnt(int8_t v) { return lzcnt(static_cast(v)); }\ninline uint8_t lzcnt(int16_t v) { return lzcnt(static_cast(v)); }\ninline uint8_t lzcnt(int32_t v) { return lzcnt(static_cast(v)); }\ninline uint8_t lzcnt(int64_t v) { return lzcnt(static_cast(v)); }\ninline uint8_t tzcnt(int8_t v) { return tzcnt(static_cast(v)); }\ninline uint8_t tzcnt(int16_t v) { return tzcnt(static_cast(v)); }\ninline uint8_t tzcnt(int32_t v) { return tzcnt(static_cast(v)); }\ninline uint8_t tzcnt(int64_t v) { return tzcnt(static_cast(v)); }\n\n// BitScanForward (bsf).\n// Search the value from least significant bit (LSB) to the most significant bit\n// (MSB) for a set bit (1).\n// Returns false if no bits are set and the output index is invalid.\n#if XE_PLATFORM_WIN32\ninline bool bit_scan_forward(uint32_t v, uint32_t* out_first_set_index) {\n return _BitScanForward(reinterpret_cast(out_first_set_index),\n v) != 0;\n}\ninline bool bit_scan_forward(uint64_t v, uint32_t* out_first_set_index) {\n return _BitScanForward64(\n reinterpret_cast(out_first_set_index), v) != 0;\n}\n#else\ninline bool bit_scan_forward(uint32_t v, uint32_t* out_first_set_index) {\n int i = ffs(v);\n *out_first_set_index = i - 1;\n return i != 0;\n}\ninline bool bit_scan_forward(uint64_t v, uint32_t* out_first_set_index) {\n int i = __builtin_ffsll(v);\n *out_first_set_index = i - 1;\n return i != 0;\n}\n#endif // XE_PLATFORM_WIN32\ninline bool bit_scan_forward(int32_t v, uint32_t* out_first_set_index) {\n return bit_scan_forward(static_cast(v), out_first_set_index);\n}\ninline bool bit_scan_forward(int64_t v, uint32_t* out_first_set_index) {\n return bit_scan_forward(static_cast(v), out_first_set_index);\n}\n\ntemplate \ninline T log2_floor(T v) {\n return sizeof(T) * 8 - 1 - lzcnt(v);\n}\ntemplate \ninline T log2_ceil(T v) {\n return sizeof(T) * 8 - lzcnt(v - 1);\n}\n\ntemplate \ninline T rotate_left(T v, uint8_t sh) {\n return (T(v) << sh) | (T(v) >> ((sizeof(T) * 8) - sh));\n}\n#if XE_PLATFORM_WIN32\ntemplate <>\ninline uint8_t rotate_left(uint8_t v, uint8_t sh) {\n return _rotl8(v, sh);\n}\ntemplate <>\ninline uint16_t rotate_left(uint16_t v, uint8_t sh) {\n return _rotl16(v, sh);\n}\ntemplate <>\ninline uint32_t rotate_left(uint32_t v, uint8_t sh) {\n return _rotl(v, sh);\n}\ntemplate <>\ninline uint64_t rotate_left(uint64_t v, uint8_t sh) {\n return _rotl64(v, sh);\n}\n#endif // XE_PLATFORM_WIN32\n\n#if XE_ARCH_AMD64\n// Utilities for SSE values.\ntemplate \nfloat m128_f32(const __m128& v) {\n float ret;\n _mm_store_ss(&ret, _mm_shuffle_ps(v, v, _MM_SHUFFLE(N, N, N, N)));\n return ret;\n}\ntemplate \nint32_t m128_i32(const __m128& v) {\n union {\n float f;\n int32_t i;\n } ret;\n _mm_store_ss(&ret.f, _mm_shuffle_ps(v, v, _MM_SHUFFLE(N, N, N, N)));\n return ret.i;\n}\ntemplate \ndouble m128_f64(const __m128d& v) {\n double ret;\n _mm_store_sd(&ret, _mm_shuffle_pd(v, v, _MM_SHUFFLE2(N, N)));\n return ret;\n}\ntemplate \ndouble m128_f64(const __m128& v) {\n return m128_f64(_mm_castps_pd(v));\n}\ntemplate \nint64_t m128_i64(const __m128d& v) {\n union {\n double f;\n int64_t i;\n } ret;\n _mm_store_sd(&ret.f, _mm_shuffle_pd(v, v, _MM_SHUFFLE2(N, N)));\n return ret.i;\n}\ntemplate \nint64_t m128_i64(const __m128& v) {\n return m128_i64(_mm_castps_pd(v));\n}\n#endif\n\n// Similar to the C++ implementation of XMConvertFloatToHalf and\n// XMConvertHalfToFloat from DirectXMath 3.00 (pre-3.04, which switched from the\n// Xenos encoding to IEEE 754), with the extended range instead of infinity and\n// NaN, and optionally with denormalized numbers - as used in vpkd3d128 (no\n// denormals, rounding towards zero) and on the Xenos (GL_OES_texture_float\n// alternative encoding).\n\ninline uint16_t float_to_xenos_half(float value, bool preserve_denormal = false,\n bool round_to_nearest_even = false) {\n uint32_t integer_value = *reinterpret_cast(&value);\n uint32_t abs_value = integer_value & 0x7FFFFFFFu;\n uint32_t result;\n if (abs_value >= 0x47FFE000u) {\n // Saturate.\n result = 0x7FFFu;\n } else {\n if (abs_value < 0x38800000u) {\n // The number is too small to be represented as a normalized half.\n if (preserve_denormal) {\n uint32_t shift =\n std::min(uint32_t(113u - (abs_value >> 23u)), uint32_t(24u));\n result = (0x800000u | (abs_value & 0x7FFFFFu)) >> shift;\n } else {\n result = 0u;\n }\n } else {\n // Rebias the exponent to represent the value as a normalized half.\n result = abs_value + 0xC8000000u;\n }\n if (round_to_nearest_even) {\n result += 0xFFFu + ((result >> 13u) & 1u);\n }\n result = (result >> 13u) & 0x7FFFu;\n }\n return uint16_t(result | ((integer_value & 0x80000000u) >> 16u));\n}\n\ninline float xenos_half_to_float(uint16_t value,\n bool preserve_denormal = false) {\n uint32_t mantissa = value & 0x3FFu;\n uint32_t exponent = (value >> 10u) & 0x1Fu;\n if (!exponent) {\n if (!preserve_denormal) {\n mantissa = 0;\n } else if (mantissa) {\n // Normalize the value in the resulting float.\n // do { Exponent--; Mantissa <<= 1; } while ((Mantissa & 0x0400) == 0)\n uint32_t mantissa_lzcnt = xe::lzcnt(mantissa) - (32u - 11u);\n exponent = uint32_t(1 - int32_t(mantissa_lzcnt));\n mantissa = (mantissa << mantissa_lzcnt) & 0x3FFu;\n }\n if (!mantissa) {\n exponent = uint32_t(-112);\n }\n }\n uint32_t result = (uint32_t(value & 0x8000u) << 16u) |\n ((exponent + 112u) << 23u) | (mantissa << 13u);\n return *reinterpret_cast(&result);\n}\n\n// https://locklessinc.com/articles/sat_arithmetic/\ntemplate \ninline T sat_add(T a, T b) {\n using TU = typename std::make_unsigned::type;\n TU result = TU(a) + TU(b);\n if (std::is_unsigned::value) {\n result |=\n TU(-static_cast::type>(result < TU(a)));\n } else {\n TU overflowed =\n (TU(a) >> (sizeof(T) * 8 - 1)) + std::numeric_limits::max();\n if (T((overflowed ^ TU(b)) | ~(TU(b) ^ result)) >= 0) {\n result = overflowed;\n }\n }\n return T(result);\n}\ntemplate \ninline T sat_sub(T a, T b) {\n using TU = typename std::make_unsigned::type;\n TU result = TU(a) - TU(b);\n if (std::is_unsigned::value) {\n result &=\n TU(-static_cast::type>(result <= TU(a)));\n } else {\n TU overflowed =\n (TU(a) >> (sizeof(T) * 8 - 1)) + std::numeric_limits::max();\n if (T((overflowed ^ TU(b)) & (overflowed ^ result)) < 0) {\n result = overflowed;\n }\n }\n return T(result);\n}\n\n} // namespace xe\n\n#endif // XENIA_BASE_MATH_H_\n" }, { "path": "src/xenia/base/memory.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/memory.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/platform.h\"\n\n#if XE_ARCH_ARM64\n#include \n#endif\n\n#include \n\nDEFINE_bool(\n writable_executable_memory, true,\n \"Allow mapping memory with both write and execute access, for simulating \"\n \"behavior on platforms where that's not supported\",\n \"Memory\");\n\nnamespace xe {\nnamespace memory {\n\nbool IsWritableExecutableMemoryPreferred() {\n return IsWritableExecutableMemorySupported() &&\n cvars::writable_executable_memory;\n}\n\n} // namespace memory\n\n// TODO(benvanik): fancy AVX versions.\n// https://github.com/gnuradio/volk/blob/master/kernels/volk/volk_16u_byteswap.h\n// https://github.com/gnuradio/volk/blob/master/kernels/volk/volk_32u_byteswap.h\n// https://github.com/gnuradio/volk/blob/master/kernels/volk/volk_64u_byteswap.h\n// Original links:\n// https://gnuradio.org/redmine/projects/gnuradio/repository/revisions/cb32b70b79f430456208a2cd521d028e0ece5d5b/entry/volk/kernels/volk/volk_16u_byteswap.h\n// https://gnuradio.org/redmine/projects/gnuradio/repository/revisions/f2bc76cc65ffba51a141950f98e75364e49df874/entry/volk/kernels/volk/volk_32u_byteswap.h\n// https://gnuradio.org/redmine/projects/gnuradio/repository/revisions/2c4c371885c31222362f70a1cd714415d1398021/entry/volk/kernels/volk/volk_64u_byteswap.h\n\nvoid copy_128_aligned(void* dest, const void* src, size_t count) {\n std::memcpy(dest, src, count * 16);\n}\n\n#if XE_ARCH_AMD64\n\n// This works around a GCC bug\n// https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100801\n// TODO(Joel Linn): Remove this when fixed GCC versions are common place.\n#if XE_COMPILER_GNUC\n#define XE_WORKAROUND_CONSTANT_RETURN_IF(x) \\\n if (__builtin_constant_p(x) && (x)) return;\n#else\n#define XE_WORKAROUND_CONSTANT_RETURN_IF(x)\n#endif\nvoid copy_and_swap_16_aligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n assert_zero(reinterpret_cast(dest_ptr) & 0xF);\n assert_zero(reinterpret_cast(src_ptr) & 0xF);\n\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n __m128i shufmask =\n _mm_set_epi8(0x0E, 0x0F, 0x0C, 0x0D, 0x0A, 0x0B, 0x08, 0x09, 0x06, 0x07,\n 0x04, 0x05, 0x02, 0x03, 0x00, 0x01);\n\n size_t i = 0;\n for (i = 0; i + 8 <= count; i += 8) {\n __m128i input = _mm_load_si128(reinterpret_cast(&src[i]));\n __m128i output = _mm_shuffle_epi8(input, shufmask);\n _mm_store_si128(reinterpret_cast<__m128i*>(&dest[i]), output);\n }\n XE_WORKAROUND_CONSTANT_RETURN_IF(count % 8 == 0);\n for (; i < count; ++i) { // handle residual elements\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_16_unaligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n __m128i shufmask =\n _mm_set_epi8(0x0E, 0x0F, 0x0C, 0x0D, 0x0A, 0x0B, 0x08, 0x09, 0x06, 0x07,\n 0x04, 0x05, 0x02, 0x03, 0x00, 0x01);\n\n size_t i;\n for (i = 0; i + 8 <= count; i += 8) {\n __m128i input = _mm_loadu_si128(reinterpret_cast(&src[i]));\n __m128i output = _mm_shuffle_epi8(input, shufmask);\n _mm_storeu_si128(reinterpret_cast<__m128i*>(&dest[i]), output);\n }\n XE_WORKAROUND_CONSTANT_RETURN_IF(count % 8 == 0);\n for (; i < count; ++i) { // handle residual elements\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_32_aligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n assert_zero(reinterpret_cast(dest_ptr) & 0xF);\n assert_zero(reinterpret_cast(src_ptr) & 0xF);\n\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n __m128i shufmask =\n _mm_set_epi8(0x0C, 0x0D, 0x0E, 0x0F, 0x08, 0x09, 0x0A, 0x0B, 0x04, 0x05,\n 0x06, 0x07, 0x00, 0x01, 0x02, 0x03);\n\n size_t i;\n for (i = 0; i + 4 <= count; i += 4) {\n __m128i input = _mm_load_si128(reinterpret_cast(&src[i]));\n __m128i output = _mm_shuffle_epi8(input, shufmask);\n _mm_store_si128(reinterpret_cast<__m128i*>(&dest[i]), output);\n }\n XE_WORKAROUND_CONSTANT_RETURN_IF(count % 4 == 0);\n for (; i < count; ++i) { // handle residual elements\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_32_unaligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n __m128i shufmask =\n _mm_set_epi8(0x0C, 0x0D, 0x0E, 0x0F, 0x08, 0x09, 0x0A, 0x0B, 0x04, 0x05,\n 0x06, 0x07, 0x00, 0x01, 0x02, 0x03);\n\n size_t i;\n for (i = 0; i + 4 <= count; i += 4) {\n __m128i input = _mm_loadu_si128(reinterpret_cast(&src[i]));\n __m128i output = _mm_shuffle_epi8(input, shufmask);\n _mm_storeu_si128(reinterpret_cast<__m128i*>(&dest[i]), output);\n }\n XE_WORKAROUND_CONSTANT_RETURN_IF(count % 4 == 0);\n for (; i < count; ++i) { // handle residual elements\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_64_aligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n assert_zero(reinterpret_cast(dest_ptr) & 0xF);\n assert_zero(reinterpret_cast(src_ptr) & 0xF);\n\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n __m128i shufmask =\n _mm_set_epi8(0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0x00, 0x01,\n 0x02, 0x03, 0x04, 0x05, 0x06, 0x07);\n\n size_t i;\n for (i = 0; i + 2 <= count; i += 2) {\n __m128i input = _mm_load_si128(reinterpret_cast(&src[i]));\n __m128i output = _mm_shuffle_epi8(input, shufmask);\n _mm_store_si128(reinterpret_cast<__m128i*>(&dest[i]), output);\n }\n XE_WORKAROUND_CONSTANT_RETURN_IF(count % 2 == 0);\n for (; i < count; ++i) { // handle residual elements\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_64_unaligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n __m128i shufmask =\n _mm_set_epi8(0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0x00, 0x01,\n 0x02, 0x03, 0x04, 0x05, 0x06, 0x07);\n\n size_t i;\n for (i = 0; i + 2 <= count; i += 2) {\n __m128i input = _mm_loadu_si128(reinterpret_cast(&src[i]));\n __m128i output = _mm_shuffle_epi8(input, shufmask);\n _mm_storeu_si128(reinterpret_cast<__m128i*>(&dest[i]), output);\n }\n XE_WORKAROUND_CONSTANT_RETURN_IF(count % 2 == 0);\n for (; i < count; ++i) { // handle residual elements\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_16_in_32_aligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n size_t i;\n for (i = 0; i + 4 <= count; i += 4) {\n __m128i input = _mm_load_si128(reinterpret_cast(&src[i]));\n __m128i output =\n _mm_or_si128(_mm_slli_epi32(input, 16), _mm_srli_epi32(input, 16));\n _mm_store_si128(reinterpret_cast<__m128i*>(&dest[i]), output);\n }\n XE_WORKAROUND_CONSTANT_RETURN_IF(count % 4 == 0);\n for (; i < count; ++i) { // handle residual elements\n dest[i] = (src[i] >> 16) | (src[i] << 16);\n }\n}\n\nvoid copy_and_swap_16_in_32_unaligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n size_t i;\n for (i = 0; i + 4 <= count; i += 4) {\n __m128i input = _mm_loadu_si128(reinterpret_cast(&src[i]));\n __m128i output =\n _mm_or_si128(_mm_slli_epi32(input, 16), _mm_srli_epi32(input, 16));\n _mm_storeu_si128(reinterpret_cast<__m128i*>(&dest[i]), output);\n }\n XE_WORKAROUND_CONSTANT_RETURN_IF(count % 4 == 0);\n for (; i < count; ++i) { // handle residual elements\n dest[i] = (src[i] >> 16) | (src[i] << 16);\n }\n}\n\n#elif XE_ARCH_ARM64\n\n// Although NEON offers vector rev instructions (like vrev32q_u8), they are\n// slower in benchmarks. Also, using uint8x16xN_t wasn't any faster in the\n// benchmarks, hence we use just use one SIMD register to minimize residual\n// processing.\n\nvoid copy_and_swap_16_aligned(void* dst_ptr, const void* src_ptr,\n size_t count) {\n copy_and_swap_16_unaligned(dst_ptr, src_ptr, count);\n}\n\nvoid copy_and_swap_16_unaligned(void* dst_ptr, const void* src_ptr,\n size_t count) {\n auto dst = reinterpret_cast(dst_ptr);\n auto src = reinterpret_cast(src_ptr);\n\n const uint8x16_t tbl_idx =\n vcombine_u8(vcreate_u8(UINT64_C(0x0607040502030001)),\n vcreate_u8(UINT64_C(0x0E0F0C0D0A0B0809)));\n\n while (count >= 8) {\n uint8x16_t data = vld1q_u8(src);\n data = vqtbl1q_u8(data, tbl_idx);\n vst1q_u8(dst, data);\n\n count -= 8;\n // These pointer increments will be combined with the load/stores (ldr/str)\n // into single instructions (at least by clang)\n dst += 16;\n src += 16;\n }\n\n while (count > 0) {\n store_and_swap(dst, load(src));\n\n count--;\n dst += 2;\n src += 2;\n }\n}\n\nvoid copy_and_swap_32_aligned(void* dst, const void* src, size_t count) {\n copy_and_swap_32_unaligned(dst, src, count);\n}\n\nvoid copy_and_swap_32_unaligned(void* dst_ptr, const void* src_ptr,\n size_t count) {\n auto dst = reinterpret_cast(dst_ptr);\n auto src = reinterpret_cast(src_ptr);\n\n const uint8x16_t tbl_idx =\n vcombine_u8(vcreate_u8(UINT64_C(0x405060700010203)),\n vcreate_u8(UINT64_C(0x0C0D0E0F08090A0B)));\n\n while (count >= 4) {\n uint8x16_t data = vld1q_u8(src);\n data = vqtbl1q_u8(data, tbl_idx);\n vst1q_u8(dst, data);\n\n count -= 4;\n dst += 16;\n src += 16;\n }\n\n while (count > 0) {\n store_and_swap(dst, load(src));\n\n count--;\n dst += 4;\n src += 4;\n }\n}\n\nvoid copy_and_swap_64_aligned(void* dst, const void* src, size_t count) {\n copy_and_swap_64_unaligned(dst, src, count);\n}\n\nvoid copy_and_swap_64_unaligned(void* dst_ptr, const void* src_ptr,\n size_t count) {\n auto dst = reinterpret_cast(dst_ptr);\n auto src = reinterpret_cast(src_ptr);\n\n const uint8x16_t tbl_idx =\n vcombine_u8(vcreate_u8(UINT64_C(0x0001020304050607)),\n vcreate_u8(UINT64_C(0x08090A0B0C0D0E0F)));\n\n while (count >= 2) {\n uint8x16_t data = vld1q_u8(src);\n data = vqtbl1q_u8(data, tbl_idx);\n vst1q_u8(dst, data);\n\n count -= 2;\n dst += 16;\n src += 16;\n }\n\n while (count > 0) {\n store_and_swap(dst, load(src));\n\n count--;\n dst += 8;\n src += 8;\n }\n}\n\nvoid copy_and_swap_16_in_32_aligned(void* dst, const void* src, size_t count) {\n return copy_and_swap_16_in_32_unaligned(dst, src, count);\n}\n\nvoid copy_and_swap_16_in_32_unaligned(void* dst_ptr, const void* src_ptr,\n size_t count) {\n auto dst = reinterpret_cast(dst_ptr);\n auto src = reinterpret_cast(src_ptr);\n while (count > 0) {\n uint16_t word0 = *src++;\n uint16_t word1 = *src++;\n *dst++ = word1;\n *dst++ = word0;\n\n count--;\n }\n}\n\n#else\n\n// Generic routines.\nvoid copy_and_swap_16_aligned(void* dest, const void* src, size_t count) {\n return copy_and_swap_16_unaligned(dest, src, count);\n}\n\nvoid copy_and_swap_16_unaligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n for (size_t i = 0; i < count; ++i) {\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_32_aligned(void* dest, const void* src, size_t count) {\n return copy_and_swap_32_unaligned(dest, src, count);\n}\n\nvoid copy_and_swap_32_unaligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n for (size_t i = 0; i < count; ++i) {\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_64_aligned(void* dest, const void* src, size_t count) {\n return copy_and_swap_64_unaligned(dest, src, count);\n}\n\nvoid copy_and_swap_64_unaligned(void* dest_ptr, const void* src_ptr,\n size_t count) {\n auto dest = reinterpret_cast(dest_ptr);\n auto src = reinterpret_cast(src_ptr);\n for (size_t i = 0; i < count; ++i) {\n dest[i] = byte_swap(src[i]);\n }\n}\n\nvoid copy_and_swap_16_in_32_aligned(void* dest, const void* src, size_t count) {\n return copy_and_swap_16_in_32_unaligned(dest, src, count);\n}\n\nvoid copy_and_swap_16_in_32_unaligned(void* dst_ptr, const void* src_ptr,\n size_t count) {\n auto dst = reinterpret_cast(dst_ptr);\n auto src = reinterpret_cast(src_ptr);\n while (count > 0) {\n uint16_t word0 = *src++;\n uint16_t word1 = *src++;\n *dst++ = word1;\n *dst++ = word0;\n\n count--;\n }\n}\n\n#endif\n\n} // namespace xe\n" }, { "path": "src/xenia/base/memory.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_MEMORY_H_\n#define XENIA_BASE_MEMORY_H_\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/byte_order.h\"\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\nnamespace memory {\n\n// For variable declarations (not return values or `this` pointer).\n// Not propagated.\n#define XE_RESTRICT_VAR __restrict\n\n// Aliasing-safe bit reinterpretation.\n// For more complex cases such as non-trivially-copyable types, write copying\n// code respecting the requirements for them externally instead of using these\n// functions.\n\ntemplate \nvoid Reinterpret(Dst& XE_RESTRICT_VAR dst, const Src& XE_RESTRICT_VAR src) {\n static_assert(sizeof(Dst) == sizeof(Src));\n static_assert(std::is_trivially_copyable_v);\n static_assert(std::is_trivially_copyable_v);\n std::memcpy(&dst, &src, sizeof(Dst));\n}\n\ntemplate \nDst Reinterpret(const Src& XE_RESTRICT_VAR src) {\n Dst dst;\n Reinterpret(dst, src);\n return dst;\n}\n\n#if XE_PLATFORM_ANDROID\nvoid AndroidInitialize();\nvoid AndroidShutdown();\n#endif\n\n// Returns the native page size of the system, in bytes.\n// This should be ~4KiB.\nsize_t page_size();\n\n// Returns the allocation granularity of the system, in bytes.\n// This is likely 64KiB.\nsize_t allocation_granularity();\n\nenum class PageAccess {\n kNoAccess = 0,\n kReadOnly = 1 << 0,\n kReadWrite = kReadOnly | 1 << 1,\n kExecuteReadOnly = kReadOnly | 1 << 2,\n kExecuteReadWrite = kReadWrite | 1 << 2,\n};\n\nenum class AllocationType {\n kReserve = 1 << 0,\n kCommit = 1 << 1,\n kReserveCommit = kReserve | kCommit,\n};\n\nenum class DeallocationType {\n kRelease = 1 << 0,\n kDecommit = 1 << 1,\n};\n\n// Whether the host allows the pages to be allocated or mapped with\n// PageAccess::kExecuteReadWrite - if not, separate mappings backed by the same\n// memory-mapped file must be used to write to executable pages.\nbool IsWritableExecutableMemorySupported();\n\n// Whether PageAccess::kExecuteReadWrite is a supported and preferred way of\n// writing executable memory, useful for simulating how Xenia would work without\n// writable executable memory on a system with it.\nbool IsWritableExecutableMemoryPreferred();\n\n// Allocates a block of memory at the given page-aligned base address.\n// Fails if the memory is not available.\n// Specify nullptr for base_address to leave it up to the system.\nvoid* AllocFixed(void* base_address, size_t length,\n AllocationType allocation_type, PageAccess access);\n\n// Deallocates and/or releases the given block of memory.\n// When releasing memory length must be zero, as all pages in the region are\n// released.\nbool DeallocFixed(void* base_address, size_t length,\n DeallocationType deallocation_type);\n\n// Sets the access rights for the given block of memory and returns the previous\n// access rights. Both base_address and length will be adjusted to page_size().\nbool Protect(void* base_address, size_t length, PageAccess access,\n PageAccess* out_old_access = nullptr);\n\n// Queries a region of pages to get the access rights. This will modify the\n// length parameter to the length of pages with the same consecutive access\n// rights. The length will start from the first byte of the first page of\n// the region.\nbool QueryProtect(void* base_address, size_t& length, PageAccess& access_out);\n\n// Allocates a block of memory for a type with the given alignment.\n// The memory must be freed with AlignedFree.\ntemplate \ninline T* AlignedAlloc(size_t alignment) {\n#if XE_COMPILER_MSVC\n return reinterpret_cast(_aligned_malloc(sizeof(T), alignment));\n#else\n void* ptr = nullptr;\n if (posix_memalign(&ptr, alignment, sizeof(T))) {\n return nullptr;\n }\n return reinterpret_cast(ptr);\n#endif // XE_COMPILER_MSVC\n}\n\n// Frees memory previously allocated with AlignedAlloc.\ntemplate \nvoid AlignedFree(T* ptr) {\n#if XE_COMPILER_MSVC\n _aligned_free(ptr);\n#else\n free(ptr);\n#endif // XE_COMPILER_MSVC\n}\n\n#if XE_PLATFORM_WIN32\n// HANDLE.\ntypedef void* FileMappingHandle;\nconstexpr FileMappingHandle kFileMappingHandleInvalid = nullptr;\n#else\n// File descriptor.\ntypedef int FileMappingHandle;\nconstexpr FileMappingHandle kFileMappingHandleInvalid = -1;\n#endif\n\nFileMappingHandle CreateFileMappingHandle(const std::filesystem::path& path,\n size_t length, PageAccess access,\n bool commit);\nvoid CloseFileMappingHandle(FileMappingHandle handle,\n const std::filesystem::path& path);\nvoid* MapFileView(FileMappingHandle handle, void* base_address, size_t length,\n PageAccess access, size_t file_offset);\nbool UnmapFileView(FileMappingHandle handle, void* base_address, size_t length);\n\ninline size_t hash_combine(size_t seed) { return seed; }\n\ntemplate \nsize_t hash_combine(size_t seed, const T& v, const Ts&... vs) {\n std::hash hasher;\n seed ^= hasher(v) + 0x9E3779B9 + (seed << 6) + (seed >> 2);\n return hash_combine(seed, vs...);\n}\n\n} // namespace memory\n\n// TODO(benvanik): move into xe::memory::\n\ninline void* low_address(void* address) {\n return reinterpret_cast(uint64_t(address) & 0xFFFFFFFF);\n}\n\nvoid copy_128_aligned(void* dest, const void* src, size_t count);\n\nvoid copy_and_swap_16_aligned(void* dest, const void* src, size_t count);\nvoid copy_and_swap_16_unaligned(void* dest, const void* src, size_t count);\nvoid copy_and_swap_32_aligned(void* dest, const void* src, size_t count);\nvoid copy_and_swap_32_unaligned(void* dest, const void* src, size_t count);\nvoid copy_and_swap_64_aligned(void* dest, const void* src, size_t count);\nvoid copy_and_swap_64_unaligned(void* dest, const void* src, size_t count);\nvoid copy_and_swap_16_in_32_aligned(void* dest, const void* src, size_t count);\nvoid copy_and_swap_16_in_32_unaligned(void* dest, const void* src,\n size_t count);\n\ntemplate \nvoid copy_and_swap(T* dest, const T* src, size_t count) {\n bool is_aligned = reinterpret_cast(dest) % 32 == 0 &&\n reinterpret_cast(src) % 32 == 0;\n if (sizeof(T) == 1) {\n std::memcpy(dest, src, count);\n } else if (sizeof(T) == 2) {\n auto ps = reinterpret_cast(src);\n auto pd = reinterpret_cast(dest);\n if (is_aligned) {\n copy_and_swap_16_aligned(pd, ps, count);\n } else {\n copy_and_swap_16_unaligned(pd, ps, count);\n }\n } else if (sizeof(T) == 4) {\n auto ps = reinterpret_cast(src);\n auto pd = reinterpret_cast(dest);\n if (is_aligned) {\n copy_and_swap_32_aligned(pd, ps, count);\n } else {\n copy_and_swap_32_unaligned(pd, ps, count);\n }\n } else if (sizeof(T) == 8) {\n auto ps = reinterpret_cast(src);\n auto pd = reinterpret_cast(dest);\n if (is_aligned) {\n copy_and_swap_64_aligned(pd, ps, count);\n } else {\n copy_and_swap_64_unaligned(pd, ps, count);\n }\n } else {\n assert_always(\"Invalid xe::copy_and_swap size\");\n }\n}\n\ntemplate \nT load(const void* mem);\ntemplate <>\ninline int8_t load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline uint8_t load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline int16_t load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline uint16_t load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline int32_t load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline uint32_t load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline int64_t load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline uint64_t load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline float load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline double load(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate \ninline T load(const void* mem) {\n if (sizeof(T) == 1) {\n return static_cast(load(mem));\n } else if (sizeof(T) == 2) {\n return static_cast(load(mem));\n } else if (sizeof(T) == 4) {\n return static_cast(load(mem));\n } else if (sizeof(T) == 8) {\n return static_cast(load(mem));\n } else {\n assert_always(\"Invalid xe::load size\");\n }\n}\n\ntemplate \nT load_and_swap(const void* mem);\ntemplate <>\ninline int8_t load_and_swap(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline uint8_t load_and_swap(const void* mem) {\n return *reinterpret_cast(mem);\n}\ntemplate <>\ninline int16_t load_and_swap(const void* mem) {\n return byte_swap(*reinterpret_cast(mem));\n}\ntemplate <>\ninline uint16_t load_and_swap(const void* mem) {\n return byte_swap(*reinterpret_cast(mem));\n}\ntemplate <>\ninline int32_t load_and_swap(const void* mem) {\n return byte_swap(*reinterpret_cast(mem));\n}\ntemplate <>\ninline uint32_t load_and_swap(const void* mem) {\n return byte_swap(*reinterpret_cast(mem));\n}\ntemplate <>\ninline int64_t load_and_swap(const void* mem) {\n return byte_swap(*reinterpret_cast(mem));\n}\ntemplate <>\ninline uint64_t load_and_swap(const void* mem) {\n return byte_swap(*reinterpret_cast(mem));\n}\ntemplate <>\ninline float load_and_swap(const void* mem) {\n return byte_swap(*reinterpret_cast(mem));\n}\ntemplate <>\ninline double load_and_swap(const void* mem) {\n return byte_swap(*reinterpret_cast(mem));\n}\ntemplate <>\ninline std::string load_and_swap(const void* mem) {\n std::string value;\n for (int i = 0;; ++i) {\n auto c =\n xe::load_and_swap(reinterpret_cast(mem) + i);\n if (!c) {\n break;\n }\n value.push_back(static_cast(c));\n }\n return value;\n}\ntemplate <>\ninline std::u16string load_and_swap(const void* mem) {\n std::u16string value;\n for (int i = 0;; ++i) {\n auto c =\n xe::load_and_swap(reinterpret_cast(mem) + i);\n if (!c) {\n break;\n }\n value.push_back(static_cast(c));\n }\n return value;\n}\n\ntemplate \nvoid store(void* mem, const T& value);\ntemplate <>\ninline void store(void* mem, const int8_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const uint8_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const int16_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const uint16_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const int32_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const uint32_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const int64_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const uint64_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const float& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store(void* mem, const double& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate \nconstexpr inline void store(const void* mem, const T& value) {\n if constexpr (sizeof(T) == 1) {\n store(mem, static_cast(value));\n } else if constexpr (sizeof(T) == 2) {\n store(mem, static_cast(value));\n } else if constexpr (sizeof(T) == 4) {\n store(mem, static_cast(value));\n } else if constexpr (sizeof(T) == 8) {\n store(mem, static_cast(value));\n } else {\n static_assert(\"Invalid xe::store size\");\n }\n}\n\ntemplate \nvoid store_and_swap(void* mem, const T& value);\ntemplate <>\ninline void store_and_swap(void* mem, const int8_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store_and_swap(void* mem, const uint8_t& value) {\n *reinterpret_cast(mem) = value;\n}\ntemplate <>\ninline void store_and_swap(void* mem, const int16_t& value) {\n *reinterpret_cast(mem) = byte_swap(value);\n}\ntemplate <>\ninline void store_and_swap(void* mem, const uint16_t& value) {\n *reinterpret_cast(mem) = byte_swap(value);\n}\ntemplate <>\ninline void store_and_swap(void* mem, const int32_t& value) {\n *reinterpret_cast(mem) = byte_swap(value);\n}\ntemplate <>\ninline void store_and_swap(void* mem, const uint32_t& value) {\n *reinterpret_cast(mem) = byte_swap(value);\n}\ntemplate <>\ninline void store_and_swap(void* mem, const int64_t& value) {\n *reinterpret_cast(mem) = byte_swap(value);\n}\ntemplate <>\ninline void store_and_swap(void* mem, const uint64_t& value) {\n *reinterpret_cast(mem) = byte_swap(value);\n}\ntemplate <>\ninline void store_and_swap(void* mem, const float& value) {\n *reinterpret_cast(mem) = byte_swap(value);\n}\ntemplate <>\ninline void store_and_swap(void* mem, const double& value) {\n *reinterpret_cast(mem) = byte_swap(value);\n}\ntemplate <>\ninline void store_and_swap(void* mem,\n const std::string_view& value) {\n for (auto i = 0; i < value.size(); ++i) {\n xe::store_and_swap(reinterpret_cast(mem) + i, value[i]);\n }\n}\ntemplate <>\ninline void store_and_swap(void* mem, const std::string& value) {\n return store_and_swap(mem, value);\n}\ntemplate <>\ninline void store_and_swap(\n void* mem, const std::u16string_view& value) {\n for (auto i = 0; i < value.size(); ++i) {\n xe::store_and_swap(reinterpret_cast(mem) + i,\n value[i]);\n }\n}\ntemplate <>\ninline void store_and_swap(void* mem,\n const std::u16string& value) {\n return store_and_swap(mem, value);\n}\n\nusing fourcc_t = uint32_t;\n\n// Get FourCC in host byte order\n// make_fourcc('a', 'b', 'c', 'd') == 0x61626364\nconstexpr inline fourcc_t make_fourcc(char a, char b, char c, char d) {\n return fourcc_t((static_cast(a) << 24) |\n (static_cast(b) << 16) |\n (static_cast(c) << 8) | static_cast(d));\n}\n\n// Get FourCC in host byte order\n// This overload requires fourcc.length() == 4\n// make_fourcc(\"abcd\") == 'abcd' == 0x61626364 for most compilers\nconstexpr inline fourcc_t make_fourcc(const std::string_view fourcc) {\n if (fourcc.length() != 4) {\n throw std::runtime_error(\"Invalid fourcc length\");\n }\n return make_fourcc(fourcc[0], fourcc[1], fourcc[2], fourcc[3]);\n}\n\n} // namespace xe\n\n#endif // XENIA_BASE_MEMORY_H_\n" }, { "path": "src/xenia/base/memory_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/memory.h\"\n\n#include \n#include \n#include \n#include \n\n#include \"xenia/base/math.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/string.h\"\n\n#if XE_PLATFORM_ANDROID\n#include \n#include \n#include \n#include \n\n#include \"xenia/base/main_android.h\"\n#endif\n\nnamespace xe {\nnamespace memory {\n\n#if XE_PLATFORM_ANDROID\n// May be null if no dynamically loaded functions are required.\nstatic void* libandroid_;\n// API 26+.\nstatic int (*android_ASharedMemory_create_)(const char* name, size_t size);\n\nvoid AndroidInitialize() {\n if (xe::GetAndroidApiLevel() >= 26) {\n libandroid_ = dlopen(\"libandroid.so\", RTLD_NOW);\n assert_not_null(libandroid_);\n if (libandroid_) {\n android_ASharedMemory_create_ =\n reinterpret_cast(\n dlsym(libandroid_, \"ASharedMemory_create\"));\n assert_not_null(android_ASharedMemory_create_);\n }\n }\n}\n\nvoid AndroidShutdown() {\n android_ASharedMemory_create_ = nullptr;\n if (libandroid_) {\n dlclose(libandroid_);\n libandroid_ = nullptr;\n }\n}\n#endif\n\nsize_t page_size() { return getpagesize(); }\nsize_t allocation_granularity() { return page_size(); }\n\nuint32_t ToPosixProtectFlags(PageAccess access) {\n switch (access) {\n case PageAccess::kNoAccess:\n return PROT_NONE;\n case PageAccess::kReadOnly:\n return PROT_READ;\n case PageAccess::kReadWrite:\n return PROT_READ | PROT_WRITE;\n case PageAccess::kExecuteReadOnly:\n return PROT_READ | PROT_EXEC;\n case PageAccess::kExecuteReadWrite:\n return PROT_READ | PROT_WRITE | PROT_EXEC;\n default:\n assert_unhandled_case(access);\n return PROT_NONE;\n }\n}\n\nbool IsWritableExecutableMemorySupported() { return true; }\n\nvoid* AllocFixed(void* base_address, size_t length,\n AllocationType allocation_type, PageAccess access) {\n // mmap does not support reserve / commit, so ignore allocation_type.\n uint32_t prot = ToPosixProtectFlags(access);\n void* result = mmap(base_address, length, prot,\n MAP_PRIVATE | MAP_FIXED | MAP_ANONYMOUS, -1, 0);\n if (result == MAP_FAILED) {\n return nullptr;\n } else {\n return result;\n }\n}\n\nbool DeallocFixed(void* base_address, size_t length,\n DeallocationType deallocation_type) {\n return munmap(base_address, length) == 0;\n}\n\nbool Protect(void* base_address, size_t length, PageAccess access,\n PageAccess* out_old_access) {\n // Linux does not have a syscall to query memory permissions.\n assert_null(out_old_access);\n\n uint32_t prot = ToPosixProtectFlags(access);\n return mprotect(base_address, length, prot) == 0;\n}\n\nbool QueryProtect(void* base_address, size_t& length, PageAccess& access_out) {\n return false;\n}\n\nFileMappingHandle CreateFileMappingHandle(const std::filesystem::path& path,\n size_t length, PageAccess access,\n bool commit) {\n#if XE_PLATFORM_ANDROID\n // TODO(Triang3l): Check if memfd can be used instead on API 30+.\n if (android_ASharedMemory_create_) {\n int sharedmem_fd = android_ASharedMemory_create_(path.c_str(), length);\n return sharedmem_fd >= 0 ? sharedmem_fd : kFileMappingHandleInvalid;\n }\n\n // Use /dev/ashmem on API versions below 26, which added ASharedMemory.\n // /dev/ashmem was disabled on API 29 for apps targeting it.\n // https://chromium.googlesource.com/chromium/src/+/master/third_party/ashmem/ashmem-dev.c\n int ashmem_fd = open(\"/\" ASHMEM_NAME_DEF, O_RDWR);\n if (ashmem_fd < 0) {\n return kFileMappingHandleInvalid;\n }\n char ashmem_name[ASHMEM_NAME_LEN];\n strlcpy(ashmem_name, path.c_str(), xe::countof(ashmem_name));\n if (ioctl(ashmem_fd, ASHMEM_SET_NAME, ashmem_name) < 0 ||\n ioctl(ashmem_fd, ASHMEM_SET_SIZE, length) < 0) {\n close(ashmem_fd);\n return kFileMappingHandleInvalid;\n }\n return ashmem_fd;\n#else\n int oflag;\n switch (access) {\n case PageAccess::kNoAccess:\n oflag = 0;\n break;\n case PageAccess::kReadOnly:\n case PageAccess::kExecuteReadOnly:\n oflag = O_RDONLY;\n break;\n case PageAccess::kReadWrite:\n case PageAccess::kExecuteReadWrite:\n oflag = O_RDWR;\n break;\n default:\n assert_always();\n return kFileMappingHandleInvalid;\n }\n oflag |= O_CREAT;\n auto full_path = \"/\" / path;\n int ret = shm_open(full_path.c_str(), oflag, 0777);\n if (ret < 0) {\n return kFileMappingHandleInvalid;\n }\n ftruncate64(ret, length);\n return ret;\n#endif\n}\n\nvoid CloseFileMappingHandle(FileMappingHandle handle,\n const std::filesystem::path& path) {\n close(handle);\n#if !XE_PLATFORM_ANDROID\n auto full_path = \"/\" / path;\n shm_unlink(full_path.c_str());\n#endif\n}\n\nvoid* MapFileView(FileMappingHandle handle, void* base_address, size_t length,\n PageAccess access, size_t file_offset) {\n uint32_t prot = ToPosixProtectFlags(access);\n return mmap64(base_address, length, prot, MAP_PRIVATE | MAP_ANONYMOUS, handle,\n file_offset);\n}\n\nbool UnmapFileView(FileMappingHandle handle, void* base_address,\n size_t length) {\n return munmap(base_address, length) == 0;\n}\n\n} // namespace memory\n} // namespace xe\n" }, { "path": "src/xenia/base/memory_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/memory.h\"\n\n#include \"xenia/base/platform_win.h\"\n\n#if WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP | \\\n WINAPI_PARTITION_SYSTEM | WINAPI_PARTITION_GAMES)\n#define XE_BASE_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n#endif\n\nnamespace xe {\nnamespace memory {\n\nsize_t page_size() {\n static size_t value = 0;\n if (!value) {\n SYSTEM_INFO si;\n GetSystemInfo(&si);\n value = si.dwPageSize;\n }\n return value;\n}\n\nsize_t allocation_granularity() {\n static size_t value = 0;\n if (!value) {\n SYSTEM_INFO si;\n GetSystemInfo(&si);\n value = si.dwAllocationGranularity;\n }\n return value;\n}\n\nDWORD ToWin32ProtectFlags(PageAccess access) {\n switch (access) {\n case PageAccess::kNoAccess:\n return PAGE_NOACCESS;\n case PageAccess::kReadOnly:\n return PAGE_READONLY;\n case PageAccess::kReadWrite:\n return PAGE_READWRITE;\n case PageAccess::kExecuteReadOnly:\n return PAGE_EXECUTE_READ;\n case PageAccess::kExecuteReadWrite:\n return PAGE_EXECUTE_READWRITE;\n default:\n assert_unhandled_case(access);\n return PAGE_NOACCESS;\n }\n}\n\nPageAccess ToXeniaProtectFlags(DWORD access) {\n if (access & PAGE_GUARD) {\n // Strip the page guard flag for now...\n access &= ~PAGE_GUARD;\n }\n\n switch (access) {\n case PAGE_NOACCESS:\n return PageAccess::kNoAccess;\n case PAGE_READONLY:\n return PageAccess::kReadOnly;\n case PAGE_READWRITE:\n return PageAccess::kReadWrite;\n case PAGE_EXECUTE_READ:\n return PageAccess::kExecuteReadOnly;\n case PAGE_EXECUTE_READWRITE:\n return PageAccess::kExecuteReadWrite;\n default:\n return PageAccess::kNoAccess;\n }\n}\n\nbool IsWritableExecutableMemorySupported() {\n#ifdef XE_BASE_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n return true;\n#else\n // To test FromApp functions on desktop, undefine\n // XE_BASE_MEMORY_WIN_USE_DESKTOP_FUNCTIONS and link to WindowsApp.lib.\n return false;\n#endif\n}\n\nvoid* AllocFixed(void* base_address, size_t length,\n AllocationType allocation_type, PageAccess access) {\n DWORD alloc_type = 0;\n switch (allocation_type) {\n case AllocationType::kReserve:\n alloc_type = MEM_RESERVE;\n break;\n case AllocationType::kCommit:\n alloc_type = MEM_COMMIT;\n break;\n case AllocationType::kReserveCommit:\n alloc_type = MEM_RESERVE | MEM_COMMIT;\n break;\n default:\n assert_unhandled_case(allocation_type);\n break;\n }\n DWORD protect = ToWin32ProtectFlags(access);\n#ifdef XE_BASE_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n return VirtualAlloc(base_address, length, alloc_type, protect);\n#else\n return VirtualAllocFromApp(base_address, length, ULONG(alloc_type),\n ULONG(protect));\n#endif\n}\n\nbool DeallocFixed(void* base_address, size_t length,\n DeallocationType deallocation_type) {\n DWORD free_type = 0;\n switch (deallocation_type) {\n case DeallocationType::kRelease:\n free_type = MEM_RELEASE;\n length = 0;\n break;\n case DeallocationType::kDecommit:\n free_type = MEM_DECOMMIT;\n break;\n default:\n assert_unhandled_case(deallocation_type);\n break;\n }\n return VirtualFree(base_address, length, free_type) ? true : false;\n}\n\nbool Protect(void* base_address, size_t length, PageAccess access,\n PageAccess* out_old_access) {\n if (out_old_access) {\n *out_old_access = PageAccess::kNoAccess;\n }\n DWORD new_protect = ToWin32ProtectFlags(access);\n#ifdef XE_BASE_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n DWORD old_protect = 0;\n BOOL result = VirtualProtect(base_address, length, new_protect, &old_protect);\n#else\n ULONG old_protect = 0;\n BOOL result = VirtualProtectFromApp(base_address, length, ULONG(new_protect),\n &old_protect);\n#endif\n if (!result) {\n return false;\n }\n if (out_old_access) {\n *out_old_access = ToXeniaProtectFlags(DWORD(old_protect));\n }\n return true;\n}\n\nbool QueryProtect(void* base_address, size_t& length, PageAccess& access_out) {\n access_out = PageAccess::kNoAccess;\n\n MEMORY_BASIC_INFORMATION info;\n ZeroMemory(&info, sizeof(info));\n\n SIZE_T result = VirtualQuery(base_address, &info, sizeof(info));\n if (!result) {\n return false;\n }\n\n length = info.RegionSize;\n access_out = ToXeniaProtectFlags(info.Protect);\n return true;\n}\n\nFileMappingHandle CreateFileMappingHandle(const std::filesystem::path& path,\n size_t length, PageAccess access,\n bool commit) {\n DWORD protect =\n ToWin32ProtectFlags(access) | (commit ? SEC_COMMIT : SEC_RESERVE);\n auto full_path = \"Local\" / path;\n#ifdef XE_BASE_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n return CreateFileMappingW(INVALID_HANDLE_VALUE, nullptr, protect,\n static_cast(length >> 32),\n static_cast(length), full_path.c_str());\n#else\n return CreateFileMappingFromApp(INVALID_HANDLE_VALUE, nullptr, ULONG(protect),\n ULONG64(length), full_path.c_str());\n#endif\n}\n\nvoid CloseFileMappingHandle(FileMappingHandle handle,\n const std::filesystem::path& path) {\n CloseHandle(handle);\n}\n\nvoid* MapFileView(FileMappingHandle handle, void* base_address, size_t length,\n PageAccess access, size_t file_offset) {\n#ifdef XE_BASE_MEMORY_WIN_USE_DESKTOP_FUNCTIONS\n DWORD target_address_low = static_cast(file_offset);\n DWORD target_address_high = static_cast(file_offset >> 32);\n DWORD file_access = 0;\n switch (access) {\n case PageAccess::kReadOnly:\n file_access = FILE_MAP_READ;\n break;\n case PageAccess::kReadWrite:\n file_access = FILE_MAP_ALL_ACCESS;\n break;\n case PageAccess::kExecuteReadOnly:\n file_access = FILE_MAP_READ | FILE_MAP_EXECUTE;\n break;\n case PageAccess::kExecuteReadWrite:\n file_access = FILE_MAP_ALL_ACCESS | FILE_MAP_EXECUTE;\n break;\n case PageAccess::kNoAccess:\n default:\n assert_unhandled_case(access);\n return nullptr;\n }\n return MapViewOfFileEx(handle, file_access, target_address_high,\n target_address_low, length, base_address);\n#else\n // VirtualAlloc2FromApp and MapViewOfFile3FromApp were added in 10.0.17134.0.\n // https://docs.microsoft.com/en-us/uwp/win32-and-com/win32-apis\n HANDLE process = GetCurrentProcess();\n void* placeholder = VirtualAlloc2FromApp(\n process, base_address, length, MEM_RESERVE | MEM_RESERVE_PLACEHOLDER,\n PAGE_NOACCESS, nullptr, 0);\n if (!placeholder) {\n return nullptr;\n }\n void* mapping = MapViewOfFile3FromApp(\n handle, process, placeholder, ULONG64(file_offset), length,\n MEM_REPLACE_PLACEHOLDER, ULONG(ToWin32ProtectFlags(access)), nullptr, 0);\n if (!mapping) {\n VirtualFree(placeholder, length, MEM_RELEASE);\n return nullptr;\n }\n return mapping;\n#endif\n}\n\nbool UnmapFileView(FileMappingHandle handle, void* base_address,\n size_t length) {\n return UnmapViewOfFile(base_address) ? true : false;\n}\n\n} // namespace memory\n} // namespace xe\n" }, { "path": "src/xenia/base/mutex.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/mutex.h\"\n\nnamespace xe {\n\nstd::recursive_mutex& global_critical_region::mutex() {\n static std::recursive_mutex global_mutex;\n return global_mutex;\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/mutex.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_MUTEX_H_\n#define XENIA_BASE_MUTEX_H_\n\n#include \n\nnamespace xe {\n\n// The global critical region mutex singleton.\n// This must guard any operation that may suspend threads or be sensitive to\n// being suspended such as global table locks and such.\n// To prevent deadlocks this should be the first lock acquired and be held\n// for the entire duration of the critical region (longer than any other lock).\n//\n// As a general rule if some code can only be accessed from the guest you can\n// guard it with only the global critical region and be assured nothing else\n// will touch it. If it will be accessed from non-guest threads you may need\n// some additional protection.\n//\n// You can think of this as disabling interrupts in the guest. The thread in the\n// global critical region has exclusive access to the entire system and cannot\n// be preempted. This also means that all activity done while in the critical\n// region must be extremely fast (no IO!), as it has the chance to block any\n// other thread until its done.\n//\n// For example, in the following situation thread 1 will not be able to suspend\n// thread 0 until it has exited its critical region, preventing it from being\n// suspended while holding the table lock:\n// [thread 0]:\n// DoKernelStuff():\n// auto global_lock = global_critical_region_.Acquire();\n// std::lock_guard table_lock(table_mutex_);\n// table_->InsertStuff();\n// [thread 1]:\n// MySuspendThread():\n// auto global_lock = global_critical_region_.Acquire();\n// ::SuspendThread(thread0);\n//\n// To use the region it's strongly recommended that you keep an instance near\n// the data requiring it. This makes it clear to those reading that the data\n// is protected by the global critical region. For example:\n// class MyType {\n// // Implies my_list_ is protected:\n// xe::global_critical_region global_critical_region_;\n// std::list<...> my_list_;\n// };\nclass global_critical_region {\n public:\n static std::recursive_mutex& mutex();\n\n // Acquires a lock on the global critical section.\n // Use this when keeping an instance is not possible. Otherwise, prefer\n // to keep an instance of global_critical_region near the members requiring\n // it to keep things readable.\n static std::unique_lock AcquireDirect() {\n return std::unique_lock(mutex());\n }\n\n // Acquires a lock on the global critical section.\n inline std::unique_lock Acquire() {\n return std::unique_lock(mutex());\n }\n\n // Acquires a deferred lock on the global critical section.\n inline std::unique_lock AcquireDeferred() {\n return std::unique_lock(mutex(), std::defer_lock);\n }\n\n // Tries to acquire a lock on the glboal critical section.\n // Check owns_lock() to see if the lock was successfully acquired.\n inline std::unique_lock TryAcquire() {\n return std::unique_lock(mutex(), std::try_to_lock);\n }\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_MUTEX_H_\n" }, { "path": "src/xenia/base/platform.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_PLATFORM_H_\n#define XENIA_BASE_PLATFORM_H_\n\n// This file contains the main platform switches used by xenia as well as any\n// fixups required to normalize the environment. Everything in here should be\n// largely portable.\n// Platform-specific headers, like platform_win.h, are used to house any\n// super platform-specific stuff that implies code is not platform-agnostic.\n//\n// NOTE: ordering matters here as sometimes multiple flags are defined on\n// certain platforms.\n//\n// Great resource on predefined macros:\n// https://sourceforge.net/p/predef/wiki/OperatingSystems/\n// Original link: https://predef.sourceforge.net/preos.html\n\n#if defined(__APPLE__)\n#include \n#endif\n\n#if defined(TARGET_OS_MAC) && TARGET_OS_MAC\n#define XE_PLATFORM_MAC 1\n#elif defined(WIN32) || defined(_WIN32)\n#define XE_PLATFORM_WIN32 1\n#elif defined(__ANDROID__)\n#define XE_PLATFORM_ANDROID 1\n#define XE_PLATFORM_LINUX 1\n#elif defined(__gnu_linux__)\n#define XE_PLATFORM_GNU_LINUX 1\n#define XE_PLATFORM_LINUX 1\n#else\n#error Unsupported target OS.\n#endif\n\n#if defined(__clang__)\n#define XE_COMPILER_CLANG 1\n#elif defined(__GNUC__)\n#define XE_COMPILER_GNUC 1\n#elif defined(_MSC_VER)\n#define XE_COMPILER_MSVC 1\n#elif defined(__MINGW32)\n#define XE_COMPILER_MINGW32 1\n#elif defined(__INTEL_COMPILER)\n#define XE_COMPILER_INTEL 1\n#else\n#define XE_COMPILER_UNKNOWN 1\n#endif\n\n#if defined(_M_AMD64) || defined(__amd64__)\n#define XE_ARCH_AMD64 1\n#elif defined(_M_ARM64) || defined(__aarch64__)\n#define XE_ARCH_ARM64 1\n#elif defined(_M_IX86) || defined(__i386__) || defined(_M_ARM) || \\\n defined(__arm__)\n#error Xenia is not supported on 32-bit platforms.\n#elif defined(_M_PPC) || defined(__powerpc__)\n#define XE_ARCH_PPC 1\n#endif\n\n#if XE_PLATFORM_WIN32\n#define WIN32_LEAN_AND_MEAN\n#define NOMINMAX // Don't want windows.h including min/max macros.\n#endif // XE_PLATFORM_WIN32\n\n#if XE_PLATFORM_WIN32\n#include \n#elif XE_ARCH_AMD64\n#include \n#endif // XE_PLATFORM_WIN32\n\n#if XE_PLATFORM_MAC\n#include \n#endif // XE_PLATFORM_MAC\n\n#if XE_COMPILER_MSVC\n#define _XEPACKEDSCOPE(body) __pragma(pack(push, 1)) body __pragma(pack(pop));\n#else\n#define _XEPACKEDSCOPE(body) \\\n _Pragma(\"pack(push, 1)\") body; \\\n _Pragma(\"pack(pop)\");\n#endif // XE_PLATFORM_WIN32\n\n#define XEPACKEDSTRUCT(name, value) _XEPACKEDSCOPE(struct name value)\n#define XEPACKEDSTRUCTANONYMOUS(value) _XEPACKEDSCOPE(struct value)\n#define XEPACKEDUNION(name, value) _XEPACKEDSCOPE(union name value)\n\nnamespace xe {\n\n#if XE_PLATFORM_WIN32\nconst char kPathSeparator = '\\\\';\n#else\nconst char kPathSeparator = '/';\n#endif // XE_PLATFORM_WIN32\n\nconst char kGuestPathSeparator = '\\\\';\n\n} // namespace xe\n\n#endif // XENIA_BASE_PLATFORM_H_\n" }, { "path": "src/xenia/base/platform_linux.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_PLATFORM_LINUX_H_\n#define XENIA_BASE_PLATFORM_LINUX_H_\n\n// NOTE: if you're including this file it means you are explicitly depending\n// on Linux headers. Including this file outside of linux platform specific\n// source code will break portability\n#include \n\n#include \"xenia/base/platform.h\"\n\n#endif // XENIA_BASE_PLATFORM_LINUX_H_\n" }, { "path": "src/xenia/base/platform_win.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_PLATFORM_WIN_H_\n#define XENIA_BASE_PLATFORM_WIN_H_\n\n// NOTE: if you're including this file it means you are explicitly depending\n// on Windows-specific headers. This is bad for portability and should be\n// avoided!\n\n#include \"xenia/base/platform.h\"\n\n#ifndef WIN32_LEAN_AND_MEAN\n#define WIN32_LEAN_AND_MEAN\n#endif\n#define NOMINMAX\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#undef DeleteBitmap\n#undef DeleteFile\n#undef GetFirstChild\n\n#endif // XENIA_BASE_PLATFORM_WIN_H_\n" }, { "path": "src/xenia/base/premake5.lua", "content": "project_root = \"../../..\"\ninclude(project_root..\"/tools/build\")\n\nproject(\"xenia-base\")\n uuid(\"aeadaf22-2b20-4941-b05f-a802d5679c11\")\n kind(\"StaticLib\")\n language(\"C++\")\n links({\n \"fmt\",\n })\n defines({\n })\n local_platform_files()\n removefiles({\"console_app_main_*.cc\"})\n removefiles({\"main_init_*.cc\"})\n files({\n \"debug_visualizers.natvis\",\n })\n\ninclude(\"testing\")\n" }, { "path": "src/xenia/base/profiling.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n// NOTE: this must be included before microprofile as macro expansion needs\n// XELOGI.\n#include \"xenia/base/logging.h\"\n\n#include \"third_party/fmt/include/fmt/printf.h\"\n\n// NOTE: microprofile must be setup first, before profiling.h is included.\n#define MICROPROFILE_ENABLED 1\n#define MICROPROFILEUI_ENABLED 1\n#define MICROPROFILE_IMPL 1\n#define MICROPROFILEUI_IMPL 1\n#define MICROPROFILE_PER_THREAD_BUFFER_SIZE (1024 * 1024 * 10)\n#define MICROPROFILE_USE_THREAD_NAME_CALLBACK 1\n#define MICROPROFILE_WEBSERVER_MAXFRAMES 3\n#define MICROPROFILE_PRINTF(...) \\\n do { \\\n auto xenia_profiler_formatted = fmt::sprintf(__VA_ARGS__); \\\n XELOGI(\"{}\", xenia_profiler_formatted); \\\n } while (false);\n#define MICROPROFILE_WEBSERVER 0\n#define MICROPROFILE_DEBUG 0\n#define MICROPROFILE_MAX_THREADS 128\n#include \"third_party/microprofile/microprofile.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/ui/ui_event.h\"\n#include \"xenia/ui/virtual_key.h\"\n#include \"xenia/ui/window.h\"\n\n#if XE_OPTION_PROFILING\n#include \"third_party/microprofile/microprofileui.h\"\n#endif // XE_OPTION_PROFILING\n\n#if XE_OPTION_PROFILING_UI\n#include \"xenia/ui/microprofile_drawer.h\"\n#endif // XE_OPTION_PROFILING_UI\n\nDEFINE_bool(profiler_dpi_scaling, false,\n \"Apply window DPI scaling to the profiler.\", \"UI\");\nDEFINE_bool(show_profiler, false, \"Show profiling UI by default.\", \"UI\");\n\nnamespace xe {\n\n#if XE_OPTION_PROFILING\n\nProfiler::ProfilerWindowInputListener Profiler::input_listener_;\nsize_t Profiler::z_order_ = 0;\nui::Window* Profiler::window_ = nullptr;\n#if XE_OPTION_PROFILING_UI\nProfiler::ProfilerUIDrawer Profiler::ui_drawer_;\nui::Presenter* Profiler::presenter_ = nullptr;\nstd::unique_ptr Profiler::drawer_;\nbool Profiler::dpi_scaling_ = false;\n#endif // XE_OPTION_PROFILING_UI\n\nbool Profiler::is_enabled() { return true; }\n\nbool Profiler::is_visible() { return is_enabled() && MicroProfileIsDrawing(); }\n\nvoid Profiler::Initialize() {\n // Custom groups.\n MicroProfileSetEnableAllGroups(false);\n MicroProfileForceEnableGroup(\"apu\", MicroProfileTokenTypeCpu);\n MicroProfileForceEnableGroup(\"cpu\", MicroProfileTokenTypeCpu);\n MicroProfileForceEnableGroup(\"gpu\", MicroProfileTokenTypeCpu);\n MicroProfileForceEnableGroup(\"internal\", MicroProfileTokenTypeCpu);\n g_MicroProfile.nGroupMask = g_MicroProfile.nForceGroup;\n g_MicroProfile.nActiveGroup = g_MicroProfile.nActiveGroupWanted =\n g_MicroProfile.nGroupMask;\n\n // Custom timers: time, average.\n g_MicroProfile.nBars |= 0x1 | 0x2;\n g_MicroProfile.nActiveBars |= 0x1 | 0x2;\n\n#if XE_OPTION_PROFILING_UI\n dpi_scaling_ = cvars::profiler_dpi_scaling;\n MicroProfileInitUI();\n g_MicroProfileUI.bShowSpikes = true;\n g_MicroProfileUI.nOpacityBackground = 0x40u << 24;\n g_MicroProfileUI.nOpacityForeground = 0xc0u << 24;\n if (cvars::show_profiler) {\n MicroProfileSetDisplayMode(1);\n }\n#else\n MicroProfileSetForceEnable(true);\n MicroProfileSetEnableAllGroups(true);\n MicroProfileSetForceMetaCounters(false);\n#endif // XE_OPTION_PROFILING_UI\n}\n\nvoid Profiler::Dump() {\n#if XE_OPTION_PROFILING_UI\n MicroProfileDumpTimers();\n#endif // XE_OPTION_PROFILING_UI\n // MicroProfileDumpHtml(\"profile.html\");\n // MicroProfileDumpHtmlToFile();\n}\n\nvoid Profiler::Shutdown() {\n SetUserIO(0, nullptr, nullptr, nullptr);\n window_ = nullptr;\n MicroProfileShutdown();\n}\n\nuint32_t Profiler::GetColor(const char* str) {\n std::hash fn;\n size_t value = fn(str);\n return value & 0xFFFFFF;\n}\n\nvoid Profiler::ThreadEnter(const char* name) {\n MicroProfileOnThreadCreate(name);\n}\n\nvoid Profiler::ThreadExit() { MicroProfileOnThreadExit(); }\n\nvoid Profiler::ProfilerWindowInputListener::OnKeyDown(ui::KeyEvent& e) {\n // https://msdn.microsoft.com/en-us/library/windows/desktop/dd375731(v=vs.85).aspx\n bool handled = true;\n switch (e.virtual_key()) {\n case ui::VirtualKey::kOem3: // `\n MicroProfileTogglePause();\n break;\n#if XE_OPTION_PROFILING_UI\n case ui::VirtualKey::kTab:\n ToggleDisplay();\n break;\n case ui::VirtualKey::k1:\n MicroProfileModKey(1);\n break;\n#endif // XE_OPTION_PROFILING_UI\n default:\n handled = false;\n break;\n }\n if (handled) {\n e.set_handled(true);\n }\n PostInputEvent();\n}\n\nvoid Profiler::ProfilerWindowInputListener::OnKeyUp(ui::KeyEvent& e) {\n bool handled = true;\n switch (e.virtual_key()) {\n#if XE_OPTION_PROFILING_UI\n case ui::VirtualKey::k1:\n MicroProfileModKey(0);\n break;\n#endif // XE_OPTION_PROFILING_UI\n default:\n handled = false;\n break;\n }\n if (handled) {\n e.set_handled(true);\n }\n PostInputEvent();\n}\n\n#if XE_OPTION_PROFILING_UI\n\nvoid Profiler::ProfilerWindowInputListener::OnMouseDown(ui::MouseEvent& e) {\n Profiler::SetMousePosition(e.x(), e.y(), 0);\n MicroProfileMouseButton(e.button() == ui::MouseEvent::Button::kLeft,\n e.button() == ui::MouseEvent::Button::kRight);\n e.set_handled(true);\n PostInputEvent();\n}\n\nvoid Profiler::ProfilerWindowInputListener::OnMouseUp(ui::MouseEvent& e) {\n Profiler::SetMousePosition(e.x(), e.y(), 0);\n MicroProfileMouseButton(0, 0);\n e.set_handled(true);\n PostInputEvent();\n}\n\nvoid Profiler::ProfilerWindowInputListener::OnMouseMove(ui::MouseEvent& e) {\n Profiler::SetMousePosition(e.x(), e.y(), 0);\n e.set_handled(true);\n PostInputEvent();\n}\n\nvoid Profiler::ProfilerWindowInputListener::OnMouseWheel(ui::MouseEvent& e) {\n Profiler::SetMousePosition(e.x(), e.y(), e.scroll_y());\n e.set_handled(true);\n PostInputEvent();\n}\n\nvoid Profiler::TogglePause() { MicroProfileTogglePause(); }\n\n#else\n\nvoid Profiler::TogglePause() {}\n\n#endif // XE_OPTION_PROFILING_UI\n\nvoid Profiler::ToggleDisplay() {\n bool was_visible = is_visible();\n MicroProfileToggleDisplayMode();\n if (is_visible() != was_visible) {\n if (window_) {\n if (was_visible) {\n window_->RemoveInputListener(&input_listener_);\n } else {\n window_->AddInputListener(&input_listener_, z_order_);\n }\n }\n#if XE_OPTION_PROFILING_UI\n if (presenter_) {\n if (was_visible) {\n presenter_->RemoveUIDrawerFromUIThread(&ui_drawer_);\n } else {\n presenter_->AddUIDrawerFromUIThread(&ui_drawer_, z_order_);\n }\n }\n#endif // XE_OPTION_PROFILING_UI\n }\n}\n\nvoid Profiler::SetUserIO(size_t z_order, ui::Window* window,\n ui::Presenter* presenter,\n ui::ImmediateDrawer* immediate_drawer) {\n#if XE_OPTION_PROFILING_UI\n if (presenter_ && is_visible()) {\n presenter_->RemoveUIDrawerFromUIThread(&ui_drawer_);\n }\n drawer_.reset();\n presenter_ = nullptr;\n#endif // XE_OPTION_PROFILING_UI\n\n if (window_) {\n if (is_visible()) {\n window_->RemoveInputListener(&input_listener_);\n }\n window_ = nullptr;\n }\n\n if (!window) {\n return;\n }\n\n z_order_ = z_order;\n window_ = window;\n\n#if XE_OPTION_PROFILING_UI\n if (presenter && immediate_drawer) {\n presenter_ = presenter;\n drawer_ = std::make_unique(immediate_drawer);\n }\n#endif // XE_OPTION_PROFILING_UI\n\n if (is_visible()) {\n window_->AddInputListener(&input_listener_, z_order_);\n#if XE_OPTION_PROFILING_UI\n if (presenter_) {\n presenter_->AddUIDrawerFromUIThread(&ui_drawer_, z_order_);\n }\n#endif // XE_OPTION_PROFILING_UI\n }\n}\n\nvoid Profiler::Flip() {\n MicroProfileFlip();\n // This can be called from non-UI threads, so not trying to access the drawer\n // to trigger redraw here as it's owned and managed exclusively by the UI\n // thread. Relying on continuous painting currently.\n}\n\n#if XE_OPTION_PROFILING_UI\nvoid Profiler::ProfilerUIDrawer::Draw(ui::UIDrawContext& ui_draw_context) {\n if (!window_ || !presenter_ || !drawer_) {\n return;\n }\n SCOPE_profile_cpu_f(\"internal\");\n uint32_t coordinate_space_width = dpi_scaling_\n ? window_->GetActualLogicalWidth()\n : window_->GetActualPhysicalWidth();\n uint32_t coordinate_space_height = dpi_scaling_\n ? window_->GetActualLogicalHeight()\n : window_->GetActualPhysicalHeight();\n drawer_->Begin(ui_draw_context, coordinate_space_width,\n coordinate_space_height);\n MicroProfileDraw(coordinate_space_width, coordinate_space_height);\n drawer_->End();\n // Continuous repaint.\n if (is_visible()) {\n presenter_->RequestUIPaintFromUIThread();\n }\n}\n#endif // XE_OPTION_PROFILING_UI\n\n#if XE_OPTION_PROFILING_UI\nvoid Profiler::SetMousePosition(int32_t x, int32_t y, int32_t wheel_delta) {\n if (!window_) {\n return;\n }\n if (dpi_scaling_) {\n x = window_->PositionToLogical(x);\n y = window_->PositionToLogical(y);\n }\n MicroProfileMousePosition(uint32_t(std::max(int32_t(0), x)),\n uint32_t(std::max(int32_t(0), y)), wheel_delta);\n}\n#endif // XE_OPTION_PROFILING_UI\n\nvoid Profiler::PostInputEvent() {\n // The profiler can be hidden from within the profiler (Mode > Off).\n if (!is_visible()) {\n window_->RemoveInputListener(&input_listener_);\n#if XE_OPTION_PROFILING_UI\n if (presenter_) {\n presenter_->RemoveUIDrawerFromUIThread(&ui_drawer_);\n }\n#endif // XE_OPTION_PROFILING_UI\n return;\n }\n // Relying on continuous painting currently, no need to request drawing.\n}\n\n#else\n\nbool Profiler::is_enabled() { return false; }\nbool Profiler::is_visible() { return false; }\nvoid Profiler::Initialize() {}\nvoid Profiler::Dump() {}\nvoid Profiler::Shutdown() {}\nuint32_t Profiler::GetColor(const char* str) { return 0; }\nvoid Profiler::ThreadEnter(const char* name) {}\nvoid Profiler::ThreadExit() {}\nvoid Profiler::ToggleDisplay() {}\nvoid Profiler::TogglePause() {}\nvoid Profiler::SetUserIO(size_t z_order, ui::Window* window,\n ui::Presenter* presenter,\n ui::ImmediateDrawer* immediate_drawer) {}\nvoid Profiler::Flip() {}\n\n#endif // XE_OPTION_PROFILING\n\n} // namespace xe\n\nuint32_t MicroProfileGpuInsertTimeStamp() { return 0; }\n\nconst char* MicroProfileGetThreadName() { return \"TODO: get thread name!\"; }\n\n#if XE_OPTION_PROFILING\n#if XE_OPTION_PROFILING_UI\n\nvoid MicroProfileDrawBox(int nX, int nY, int nX1, int nY1, uint32_t nColor,\n MicroProfileBoxType type) {\n auto drawer = xe::Profiler::drawer();\n if (!drawer) {\n return;\n }\n drawer->DrawBox(nX, nY, nX1, nY1, nColor,\n static_cast(type));\n}\n\nvoid MicroProfileDrawLine2D(uint32_t nVertices, float* pVertices,\n uint32_t nColor) {\n auto drawer = xe::Profiler::drawer();\n if (!drawer) {\n return;\n }\n drawer->DrawLine2D(nVertices, pVertices, nColor);\n}\n\nvoid MicroProfileDrawText(int nX, int nY, uint32_t nColor, const char* pText,\n uint32_t nLen) {\n auto drawer = xe::Profiler::drawer();\n if (!drawer) {\n return;\n }\n drawer->DrawTextString(nX, nY, nColor, pText, nLen);\n}\n\n#endif // XE_OPTION_PROFILING_UI\n\n#endif // XE_OPTION_PROFILING\n" }, { "path": "src/xenia/base/profiling.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_PROFILING_H_\n#define XENIA_BASE_PROFILING_H_\n\n#include \n#include \n#include \n\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/string.h\"\n#include \"xenia/ui/ui_drawer.h\"\n#include \"xenia/ui/virtual_key.h\"\n#include \"xenia/ui/window_listener.h\"\n\n#if XE_PLATFORM_WIN32\n#define XE_OPTION_PROFILING 1\n#define XE_OPTION_PROFILING_UI 1\n#else\n#define XE_OPTION_PROFILING 0\n#endif // XE_PLATFORM_WIN32\n\n#if XE_OPTION_PROFILING\n// Pollutes the global namespace. Yuck.\n#define MICROPROFILE_MAX_THREADS 128\n#include \n#endif // XE_OPTION_PROFILING\n\nnamespace xe {\nnamespace ui {\nclass ImmediateDrawer;\nclass MicroprofileDrawer;\nclass Presenter;\nclass Window;\n} // namespace ui\n} // namespace xe\n\nnamespace xe {\n\n#if XE_OPTION_PROFILING\n\n// Defines a profiling scope for CPU tasks.\n// Use `SCOPE_profile_cpu(name)` to activate the scope.\n#define DEFINE_profile_cpu(name, group_name, scope_name) \\\n MICROPROFILE_DEFINE(name, group_name, scope_name, \\\n xe::Profiler::GetColor(scope_name))\n\n// Declares a previously defined profile scope. Use in a translation unit.\n#define DECLARE_profile_cpu(name) MICROPROFILE_DECLARE(name)\n\n// Defines a profiling scope for GPU tasks.\n// Use `COUNT_profile_gpu(name)` to activate the scope.\n#define DEFINE_profile_gpu(name, group_name, scope_name) \\\n MICROPROFILE_DEFINE_GPU(name, group_name, scope_name, \\\n xe::Profiler::GetColor(scope_name))\n\n// Declares a previously defined profile scope. Use in a translation unit.\n#define DECLARE_profile_gpu(name) MICROPROFILE_DECLARE_GPU(name)\n\n// Enters a previously defined CPU profiling scope, active for the duration\n// of the containing block.\n#define SCOPE_profile_cpu(name) MICROPROFILE_SCOPE(name)\n\n// Enters a CPU profiling scope, active for the duration of the containing\n// block. No previous definition required.\n#define SCOPE_profile_cpu_i(group_name, scope_name) \\\n MICROPROFILE_SCOPEI(group_name, scope_name, \\\n xe::Profiler::GetColor(scope_name))\n\n// Enters a CPU profiling scope by function name, active for the duration of\n// the containing block. No previous definition required.\n#define SCOPE_profile_cpu_f(group_name) \\\n MICROPROFILE_SCOPEI(group_name, __FUNCTION__, \\\n xe::Profiler::GetColor(__FUNCTION__))\n\n// Enters a previously defined GPU profiling scope, active for the duration\n// of the containing block.\n#define SCOPE_profile_gpu(name) MICROPROFILE_SCOPEGPU(name)\n\n// Enters a GPU profiling scope, active for the duration of the containing\n// block. No previous definition required.\n#define SCOPE_profile_gpu_i(group_name, scope_name) \\\n MICROPROFILE_SCOPEGPUI(group_name, scope_name, \\\n xe::Profiler::GetColor(scope_name))\n\n// Enters a GPU profiling scope by function name, active for the duration of\n// the containing block. No previous definition required.\n#define SCOPE_profile_gpu_f(group_name) \\\n MICROPROFILE_SCOPEGPUI(group_name, __FUNCTION__, \\\n xe::Profiler::GetColor(__FUNCTION__))\n\n// Adds a number to a counter\n#define COUNT_profile_add(name, count) MICROPROFILE_COUNTER_ADD(name, count)\n\n// Subtracts a number to a counter\n#define COUNT_profile_sub(name, count) MICROPROFILE_COUNTER_SUB(name, count)\n\n// Sets a counter's value\n#define COUNT_profile_set(name, count) MICROPROFILE_COUNTER_SET(name, count)\n\n// Tracks a CPU value counter.\n#define COUNT_profile_cpu(name, count) MICROPROFILE_META_CPU(name, count)\n\n// Tracks a GPU value counter.\n#define COUNT_profile_gpu(name, count) MICROPROFILE_META_GPU(name, count)\n\n#else\n\n#define DEFINE_profile_cpu(name, group_name, scope_name)\n#define DEFINE_profile_gpu(name, group_name, scope_name)\n#define DECLARE_profile_cpu(name)\n#define DECLARE_profile_gpu(name)\n#define SCOPE_profile_cpu(name) \\\n do { \\\n } while (false)\n#define SCOPE_profile_cpu_f(name) \\\n do { \\\n } while (false)\n#define SCOPE_profile_cpu_i(group_name, scope_name) \\\n do { \\\n } while (false)\n#define SCOPE_profile_gpu(name) \\\n do { \\\n } while (false)\n#define SCOPE_profile_gpu_f(name) \\\n do { \\\n } while (false)\n#define SCOPE_profile_gpu_i(group_name, scope_name) \\\n do { \\\n } while (false)\n#define COUNT_profile_add(name, count) \\\n do { \\\n } while (false)\n#define COUNT_profile_sub(name, count) \\\n do { \\\n } while (false)\n#define COUNT_profile_set(name, count) \\\n do { \\\n } while (false)\n#define COUNT_profile_cpu(name, count) \\\n do { \\\n } while (false)\n#define COUNT_profile_gpu(name, count) \\\n do { \\\n } while (false)\n\n#ifndef MICROPROFILE_TEXT_WIDTH\n#define MICROPROFILE_TEXT_WIDTH 1\n#define MICROPROFILE_TEXT_HEIGHT 1\n#endif // !MICROPROFILE_TEXT_WIDTH\n\n#endif // XE_OPTION_PROFILING\n\nclass Profiler {\n public:\n static bool is_enabled();\n static bool is_visible();\n\n // Initializes the profiler. Call at startup.\n static void Initialize();\n // Dumps data to stdout.\n static void Dump();\n // Cleans up profiling, releasing all memory.\n static void Shutdown();\n\n // Computes a color from the given string.\n static uint32_t GetColor(const char* str);\n\n // Activates the calling thread for profiling.\n // This must be called immediately after launching a thread.\n static void ThreadEnter(const char* name = nullptr);\n // Deactivates the calling thread for profiling.\n static void ThreadExit();\n\n static void ToggleDisplay();\n static void TogglePause();\n\n // Initializes input for the given window and drawing for the given presenter\n // and immediate drawer.\n static void SetUserIO(size_t z_order, ui::Window* window,\n ui::Presenter* presenter,\n ui::ImmediateDrawer* immediate_drawer);\n // Gets the current drawer, if any.\n static ui::MicroprofileDrawer* drawer() {\n#if XE_OPTION_PROFILING_UI\n return drawer_.get();\n#else\n return nullptr;\n#endif\n }\n // Presents the profiler to the bound display, if any.\n static void Present(ui::UIDrawContext& ui_draw_context);\n // Starts a new frame on the profiler\n static void Flip();\n\n private:\n#if XE_OPTION_PROFILING\n class ProfilerWindowInputListener final : public ui::WindowInputListener {\n public:\n void OnKeyDown(ui::KeyEvent& e) override;\n void OnKeyUp(ui::KeyEvent& e) override;\n#if XE_OPTION_PROFILING_UI\n void OnMouseDown(ui::MouseEvent& e) override;\n void OnMouseMove(ui::MouseEvent& e) override;\n void OnMouseUp(ui::MouseEvent& e) override;\n void OnMouseWheel(ui::MouseEvent& e) override;\n#endif // XE_OPTION_PROFILING_UI\n };\n // For now, no need for OnDpiChanged in a WindowListener because redrawing is\n // done continuously.\n\n#if XE_OPTION_PROFILING_UI\n class ProfilerUIDrawer final : public ui::UIDrawer {\n public:\n void Draw(ui::UIDrawContext& context) override;\n };\n#endif // XE_OPTION_PROFILING_UI\n\n#if XE_OPTION_PROFILING_UI\n static void SetMousePosition(int32_t x, int32_t y, int32_t wheel_delta);\n#endif // XE_OPTION_PROFILING_UI\n static void PostInputEvent();\n\n static ProfilerWindowInputListener input_listener_;\n static size_t z_order_;\n static ui::Window* window_;\n#if XE_OPTION_PROFILING_UI\n static ProfilerUIDrawer ui_drawer_;\n static ui::Presenter* presenter_;\n static std::unique_ptr drawer_;\n static bool dpi_scaling_;\n#endif // XE_OPTION_PROFILING_UI\n#endif // XE_OPTION_PROFILING\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_PROFILING_H_\n" }, { "path": "src/xenia/base/reset_scope.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_RESET_SCOPE_H_\n#define XENIA_BASE_RESET_SCOPE_H_\n\nnamespace xe {\n\ntemplate \nclass ResetScope {\n public:\n explicit ResetScope(T* value) : value_(value) {}\n ~ResetScope() {\n if (value_) {\n value_->Reset();\n }\n }\n\n private:\n T* value_;\n};\n\ntemplate \ninline ResetScope make_reset_scope(T* value) {\n return ResetScope(value);\n}\n\ntemplate \ninline ResetScope make_reset_scope(const std::unique_ptr& value) {\n return ResetScope(value.get());\n}\n\n} // namespace xe\n\n#endif // XENIA_BASE_RESET_SCOPE_H_\n" }, { "path": "src/xenia/base/ring_buffer.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/ring_buffer.h\"\n\n#include \n#include \n\nnamespace xe {\n\nRingBuffer::RingBuffer(uint8_t* buffer, size_t capacity)\n : buffer_(buffer), capacity_(capacity) {}\n\nvoid RingBuffer::AdvanceRead(size_t count) {\n if (read_offset_ + count < capacity_) {\n read_offset_ += count;\n } else {\n size_t left_half = capacity_ - read_offset_;\n size_t right_half = count - left_half;\n read_offset_ = right_half;\n }\n}\n\nvoid RingBuffer::AdvanceWrite(size_t count) {\n if (write_offset_ + count < capacity_) {\n write_offset_ += count;\n } else {\n size_t left_half = capacity_ - write_offset_;\n size_t right_half = count - left_half;\n write_offset_ = right_half;\n }\n}\n\nRingBuffer::ReadRange RingBuffer::BeginRead(size_t count) {\n count = std::min(count, capacity_);\n if (!count) {\n return {0};\n }\n if (read_offset_ + count < capacity_) {\n return {buffer_ + read_offset_, count, nullptr, 0};\n } else {\n size_t left_half = capacity_ - read_offset_;\n size_t right_half = count - left_half;\n return {buffer_ + read_offset_, left_half, buffer_, right_half};\n }\n}\n\nvoid RingBuffer::EndRead(ReadRange read_range) {\n if (read_range.second) {\n read_offset_ = read_range.second_length;\n } else {\n read_offset_ += read_range.first_length;\n }\n}\n\nsize_t RingBuffer::Read(uint8_t* buffer, size_t count) {\n count = std::min(count, capacity_);\n if (!count) {\n return 0;\n }\n\n // Sanity check: Make sure we don't read over the write offset.\n if (read_offset_ < write_offset_) {\n assert_true(read_offset_ + count <= write_offset_);\n } else if (read_offset_ + count >= capacity_) {\n size_t left_half = capacity_ - read_offset_;\n assert_true(count - left_half <= write_offset_);\n }\n\n if (read_offset_ + count < capacity_) {\n std::memcpy(buffer, buffer_ + read_offset_, count);\n read_offset_ += count;\n } else {\n size_t left_half = capacity_ - read_offset_;\n size_t right_half = count - left_half;\n std::memcpy(buffer, buffer_ + read_offset_, left_half);\n std::memcpy(buffer + left_half, buffer_, right_half);\n read_offset_ = right_half;\n }\n\n return count;\n}\n\nsize_t RingBuffer::Write(const uint8_t* buffer, size_t count) {\n count = std::min(count, capacity_);\n if (!count) {\n return 0;\n }\n\n // Sanity check: Make sure we don't write over the read offset.\n if (write_offset_ < read_offset_) {\n assert_true(write_offset_ + count <= read_offset_);\n } else if (write_offset_ + count >= capacity_) {\n size_t left_half = capacity_ - write_offset_;\n assert_true(count - left_half <= read_offset_);\n }\n\n if (write_offset_ + count < capacity_) {\n std::memcpy(buffer_ + write_offset_, buffer, count);\n write_offset_ += count;\n } else {\n size_t left_half = capacity_ - write_offset_;\n size_t right_half = count - left_half;\n std::memcpy(buffer_ + write_offset_, buffer, left_half);\n std::memcpy(buffer_, buffer + left_half, right_half);\n write_offset_ = right_half;\n }\n\n return count;\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/ring_buffer.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_RING_BUFFER_H_\n#define XENIA_BASE_RING_BUFFER_H_\n\n#include \n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/byte_order.h\"\n\nnamespace xe {\n\nclass RingBuffer {\n public:\n RingBuffer(uint8_t* buffer, size_t capacity);\n\n uint8_t* buffer() const { return buffer_; }\n size_t capacity() const { return capacity_; }\n bool empty() const { return read_offset_ == write_offset_; }\n\n size_t read_offset() const { return read_offset_; }\n uintptr_t read_ptr() const { return uintptr_t(buffer_) + read_offset_; }\n void set_read_offset(size_t offset) { read_offset_ = offset % capacity_; }\n size_t read_count() const {\n if (read_offset_ == write_offset_) {\n return 0;\n } else if (read_offset_ < write_offset_) {\n return write_offset_ - read_offset_;\n } else {\n return (capacity_ - read_offset_) + write_offset_;\n }\n }\n\n size_t write_offset() const { return write_offset_; }\n uintptr_t write_ptr() const { return uintptr_t(buffer_) + write_offset_; }\n void set_write_offset(size_t offset) { write_offset_ = offset % capacity_; }\n size_t write_count() const {\n if (read_offset_ == write_offset_) {\n return capacity_;\n } else if (write_offset_ < read_offset_) {\n return read_offset_ - write_offset_;\n } else {\n return (capacity_ - write_offset_) + read_offset_;\n }\n }\n\n void AdvanceRead(size_t count);\n void AdvanceWrite(size_t count);\n\n struct ReadRange {\n const uint8_t* first;\n size_t first_length;\n const uint8_t* second;\n size_t second_length;\n };\n ReadRange BeginRead(size_t count);\n void EndRead(ReadRange read_range);\n\n size_t Read(uint8_t* buffer, size_t count);\n template \n size_t Read(T* buffer, size_t count) {\n return Read(reinterpret_cast(buffer), count);\n }\n\n template \n T Read() {\n static_assert(std::is_fundamental::value,\n \"Immediate read only supports basic types!\");\n\n T imm;\n size_t read = Read(reinterpret_cast(&imm), sizeof(T));\n assert_true(read == sizeof(T));\n return imm;\n }\n\n template \n T ReadAndSwap() {\n static_assert(std::is_fundamental::value,\n \"Immediate read only supports basic types!\");\n\n T imm;\n size_t read = Read(reinterpret_cast(&imm), sizeof(T));\n assert_true(read == sizeof(T));\n imm = xe::byte_swap(imm);\n return imm;\n }\n\n size_t Write(const uint8_t* buffer, size_t count);\n template \n size_t Write(const T* buffer, size_t count) {\n return Write(reinterpret_cast(buffer), count);\n }\n\n template \n size_t Write(T& data) {\n return Write(reinterpret_cast(&data), sizeof(T));\n }\n\n private:\n uint8_t* buffer_ = nullptr;\n size_t capacity_ = 0;\n size_t read_offset_ = 0;\n size_t write_offset_ = 0;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_RING_BUFFER_H_\n" }, { "path": "src/xenia/base/socket.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_SOCKET_H_\n#define XENIA_BASE_SOCKET_H_\n\n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/threading.h\"\n\nnamespace xe {\n\n// An asynchronous bidirectional socket.\n// Sockets are wait-based and operate in a thread-free manner. Data may be\n// read from or written to a socket from any thread with the wait handle as\n// a mechanism to efficiently wait until new data is available.\n// Consumers should wait on the wait handle and when signalled poll to see\n// if the socket is still connected or new data is available. When reading\n// data Receive should be called until it returns 0 new bytes and then the\n// wait handle should be waited on until new data arrives.\nclass Socket {\n public:\n // Synchronously connects to the given hostname:port.\n static std::unique_ptr Connect(std::string hostname, uint16_t port);\n\n virtual ~Socket() = default;\n\n // TODO(benvanik): socket info? remote IP? etc?\n\n // Returns a wait handle that can be used to wait on incoming client data or\n // status updates. When this handle is signalled consumers should poll for\n // new data and react accordingly.\n virtual xe::threading::WaitHandle* wait_handle() = 0;\n\n // Returns true if the client is connected and can send/receive data.\n virtual bool is_connected() = 0;\n\n // Closes the socket.\n // This will signal the wait handle.\n virtual void Close() = 0;\n\n // Receives incoming data up to the given capacity and returns the total\n // number of bytes received.\n // 0 is returned if there is no data available. -1 is returned if the client\n // has been closed.\n virtual size_t Receive(void* buffer, size_t buffer_capacity) = 0;\n\n // Asynchronously sends a set of buffers.\n // The buffers are copied and may be modified immediately after the function\n // returns.\n // Returns false if the socket is disconnected or the data cannot be sent.\n virtual bool Send(const std::pair* buffers,\n size_t buffer_count) = 0;\n\n // Asynchronously sends a set of buffers.\n // The buffers are copied and may be modified immediately after the function\n // returns.\n // Returns false if the socket is disconnected or the data cannot be sent.\n bool Send(std::vector> buffers) {\n return Send(buffers.data(), buffers.size());\n }\n\n // Asynchronously sends a buffer of data.\n // The buffer is copied and may be modified immediately after the function\n // returns.\n // Returns false if the socket is disconnected or the data cannot be sent.\n bool Send(const void* buffer, size_t buffer_length) {\n auto buffer_list = std::make_pair(buffer, buffer_length);\n return Send(&buffer_list, 1);\n }\n\n // Asynchronously sends a buffer of data.\n // The buffer is copied and may be modified immediately after the function\n // returns.\n // Returns false if the socket is disconnected or the data cannot be sent.\n bool Send(const std::vector& buffer) {\n auto buffer_list = std::make_pair(buffer.data(), buffer.size());\n return Send(&buffer_list, 1);\n }\n\n // Asynchronously sends a string buffer.\n // Returns false if the socket is disconnected or the data cannot be sent.\n bool Send(const std::string_view value) {\n return Send(value.data(), value.size());\n }\n};\n\n// Runs a socket server on the specified local port.\n// The server accepts clients from a background thread until closed (by\n// deletion). Clients are not tied to their creating server and the server\n// may be closed while clients remain connected.\nclass SocketServer {\n public:\n // Creates a new socket server bound to the given local port.\n // The accept callback will be called from a random thread when a new client\n // connects.\n // Returns null if the server cannot be bound to the given port (in use, etc).\n static std::unique_ptr Create(\n uint16_t port,\n std::function client)> accept_callback);\n\n virtual ~SocketServer() = default;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_SOCKET_H_\n" }, { "path": "src/xenia/base/socket_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/socket.h\"\n\n#include \n#include \n\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/platform_win.h\"\n\n// winsock includes must come after platform_win.h:\n#include // NOLINT(build/include_order)\n#include // NOLINT(build/include_order)\n#include // NOLINT(build/include_order)\n\nnamespace xe {\n\nvoid InitializeWinsock() {\n static bool has_initialized = false;\n if (!has_initialized) {\n has_initialized = true;\n WSADATA wsa_data;\n WSAStartup(MAKEWORD(2, 2), &wsa_data);\n }\n}\n\nclass Win32Socket : public Socket {\n public:\n Win32Socket() = default;\n ~Win32Socket() override { Close(); }\n\n bool Connect(std::string hostname, uint16_t port) {\n addrinfo hints = {0};\n hints.ai_family = AF_UNSPEC;\n hints.ai_socktype = SOCK_STREAM;\n hints.ai_protocol = IPPROTO_TCP;\n\n // Resolve the server address and port.\n // May return multiple results, so attempt to connect to an address until\n // one succeeds.\n addrinfo* result = nullptr;\n auto port_string = std::to_string(port);\n int ret =\n getaddrinfo(hostname.c_str(), port_string.c_str(), &hints, &result);\n if (ret != 0) {\n XELOGE(\"getaddrinfo failed with error: {}\", ret);\n return false;\n }\n SOCKET try_socket = INVALID_SOCKET;\n for (addrinfo* ptr = result; ptr; ptr = ptr->ai_next) {\n // Create a SOCKET for connecting to server.\n try_socket = socket(ptr->ai_family, ptr->ai_socktype, ptr->ai_protocol);\n if (try_socket == INVALID_SOCKET) {\n XELOGE(\"socket failed with error: {}\", WSAGetLastError());\n freeaddrinfo(result);\n return false;\n }\n // Connect to server.\n ret =\n connect(try_socket, ptr->ai_addr, static_cast(ptr->ai_addrlen));\n if (ret == SOCKET_ERROR) {\n closesocket(try_socket);\n try_socket = INVALID_SOCKET;\n continue;\n }\n break;\n }\n freeaddrinfo(result);\n if (try_socket == INVALID_SOCKET) {\n XELOGE(\"Unable to connect to server\");\n return false;\n }\n\n // Piggyback to setup the socket.\n return Accept(try_socket);\n }\n\n bool Accept(SOCKET socket) {\n std::lock_guard lock(mutex_);\n\n socket_ = socket;\n\n // Create event and bind to the socket, waiting for read/close\n // notifications.\n // Set true to start so we'll force a query of the socket on first run.\n event_ = xe::threading::Event::CreateManualResetEvent(true);\n assert_not_null(event_);\n WSAEventSelect(socket_, event_->native_handle(), FD_READ | FD_CLOSE);\n\n // Keepalive for a looong time, as we may be paused by the debugger/etc.\n struct tcp_keepalive alive;\n alive.onoff = TRUE;\n alive.keepalivetime = 7200000;\n alive.keepaliveinterval = 6000;\n DWORD bytes_returned;\n WSAIoctl(socket_, SIO_KEEPALIVE_VALS, &alive, sizeof(alive), nullptr, 0,\n &bytes_returned, nullptr, nullptr);\n\n return true;\n }\n\n xe::threading::WaitHandle* wait_handle() override { return event_.get(); }\n\n bool is_connected() override {\n std::lock_guard lock(mutex_);\n return socket_ != INVALID_SOCKET;\n }\n\n void Close() override {\n std::lock_guard lock(mutex_);\n\n if (socket_ != INVALID_SOCKET) {\n SOCKET socket = socket_;\n socket_ = INVALID_SOCKET;\n shutdown(socket, SD_SEND);\n closesocket(socket);\n }\n\n if (event_) {\n // Set event so any future waits will immediately succeed.\n event_->Set();\n }\n }\n\n size_t Receive(void* buffer, size_t buffer_capacity) override {\n std::lock_guard lock(mutex_);\n\n if (socket_ == INVALID_SOCKET) {\n return -1;\n }\n int ret = recv(socket_, reinterpret_cast(buffer),\n static_cast(buffer_capacity), 0);\n if (ret == SOCKET_ERROR) {\n int e = WSAGetLastError();\n if (e == WSAEWOULDBLOCK) {\n // Ok - no more data to read.\n // Reset our event so we'll block next time.\n event_->Reset();\n return 0;\n }\n XELOGE(\"Socket send error: {}\", e);\n Close();\n return -1;\n } else if (ret == 0) {\n // Socket gracefully closed.\n Close();\n return -1;\n }\n return ret;\n }\n\n bool Send(const std::pair* buffers,\n size_t buffer_count) override {\n std::lock_guard lock(mutex_);\n\n if (socket_ == INVALID_SOCKET) {\n return false;\n }\n std::vector buffer_data_list;\n std::vector buffer_data_sent;\n for (size_t i = 0; i < buffer_count; ++i) {\n WSABUF buf;\n buf.len = ULONG(buffers[i].second);\n buf.buf = reinterpret_cast(const_cast(buffers[i].first));\n buffer_data_list.emplace_back(buf);\n buffer_data_sent.emplace_back(0);\n }\n int ret = WSASend(socket_, buffer_data_list.data(), DWORD(buffer_count),\n buffer_data_sent.data(), 0, nullptr, nullptr);\n if (ret == SOCKET_ERROR) {\n int e = WSAGetLastError();\n XELOGE(\"Socket send error: {}\", e);\n Close();\n return false;\n }\n return true;\n }\n\n private:\n std::recursive_mutex mutex_;\n SOCKET socket_ = INVALID_SOCKET;\n std::unique_ptr event_;\n};\n\nstd::unique_ptr Socket::Connect(std::string hostname, uint16_t port) {\n InitializeWinsock();\n\n auto socket = std::make_unique();\n if (!socket->Connect(std::move(hostname), port)) {\n return nullptr;\n }\n return std::unique_ptr(socket.release());\n}\n\nclass Win32SocketServer : public SocketServer {\n public:\n Win32SocketServer(\n std::function client)> accept_callback)\n : accept_callback_(std::move(accept_callback)) {}\n ~Win32SocketServer() override {\n if (socket_ != INVALID_SOCKET) {\n SOCKET socket = socket_;\n socket_ = INVALID_SOCKET;\n linger so_linger;\n so_linger.l_onoff = TRUE;\n so_linger.l_linger = 30;\n setsockopt(socket, SOL_SOCKET, SO_LINGER,\n reinterpret_cast(&so_linger), sizeof(so_linger));\n shutdown(socket, SD_SEND);\n closesocket(socket);\n }\n\n if (accept_thread_) {\n // Join thread, which should die because the socket is now invalid.\n xe::threading::Wait(accept_thread_.get(), true);\n accept_thread_.reset();\n }\n }\n\n bool Bind(uint16_t port) {\n socket_ = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);\n if (socket_ < 1) {\n XELOGE(\"Unable to create listen socket\");\n return false;\n }\n struct tcp_keepalive alive;\n alive.onoff = TRUE;\n alive.keepalivetime = 7200000;\n alive.keepaliveinterval = 6000;\n DWORD bytes_returned;\n WSAIoctl(socket_, SIO_KEEPALIVE_VALS, &alive, sizeof(alive), nullptr, 0,\n &bytes_returned, nullptr, nullptr);\n int opt_value = 1;\n setsockopt(socket_, SOL_SOCKET, SO_REUSEADDR,\n reinterpret_cast(&opt_value), sizeof(opt_value));\n opt_value = 1;\n setsockopt(socket_, IPPROTO_TCP, TCP_NODELAY,\n reinterpret_cast(&opt_value), sizeof(opt_value));\n\n sockaddr_in socket_addr = {0};\n socket_addr.sin_family = AF_INET;\n socket_addr.sin_addr.s_addr = htonl(INADDR_LOOPBACK);\n socket_addr.sin_port = htons(port);\n if (bind(socket_, reinterpret_cast(&socket_addr),\n sizeof(socket_addr)) == SOCKET_ERROR) {\n int e = WSAGetLastError();\n XELOGE(\"Unable to bind debug socket: {}\", e);\n return false;\n }\n if (listen(socket_, 5) == SOCKET_ERROR) {\n int e = WSAGetLastError();\n XELOGE(\"Unable to listen on accept socket {}\", e);\n return false;\n }\n\n accept_thread_ = xe::threading::Thread::Create({}, [this, port]() {\n xe::threading::set_name(std::string(\"xe::SocketServer localhost:\") +\n std::to_string(port));\n while (socket_ != INVALID_SOCKET) {\n sockaddr_in6 client_addr;\n int client_count = sizeof(client_addr);\n SOCKET client_socket = accept(\n socket_, reinterpret_cast(&client_addr), &client_count);\n if (client_socket == INVALID_SOCKET) {\n continue;\n }\n\n auto client = std::make_unique();\n if (!client->Accept(client_socket)) {\n XELOGE(\"Unable to accept socket; ignoring\");\n continue;\n }\n accept_callback_(std::move(client));\n }\n });\n assert_not_null(accept_thread_);\n\n return true;\n }\n\n private:\n std::function client)> accept_callback_;\n std::unique_ptr accept_thread_;\n\n SOCKET socket_ = INVALID_SOCKET;\n};\n\nstd::unique_ptr SocketServer::Create(\n uint16_t port,\n std::function client)> accept_callback) {\n InitializeWinsock();\n\n auto socket_server =\n std::make_unique(std::move(accept_callback));\n if (!socket_server->Bind(port)) {\n return nullptr;\n }\n return std::unique_ptr(socket_server.release());\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/string.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/string.h\"\n\n#include \n#include \n#include \n\n#include \"xenia/base/platform.h\"\n\n#if !XE_PLATFORM_WIN32\n#include \n#endif // !XE_PLATFORM_WIN32\n\n#define UTF_CPP_CPLUSPLUS 201703L\n#include \"third_party/utfcpp/source/utf8.h\"\n\nnamespace utfcpp = utf8;\n\nnamespace xe {\n\nint xe_strcasecmp(const char* string1, const char* string2) {\n#if XE_PLATFORM_WIN32\n return _stricmp(string1, string2);\n#else\n return strcasecmp(string1, string2);\n#endif // XE_PLATFORM_WIN32\n}\n\nint xe_strncasecmp(const char* string1, const char* string2, size_t count) {\n#if XE_PLATFORM_WIN32\n return _strnicmp(string1, string2, count);\n#else\n return strncasecmp(string1, string2, count);\n#endif // XE_PLATFORM_WIN32\n}\n\nchar* xe_strdup(const char* source) {\n#if XE_PLATFORM_WIN32\n return _strdup(source);\n#else\n return strdup(source);\n#endif // XE_PLATFORM_WIN32\n}\n\nstd::string to_utf8(const std::u16string_view source) {\n return utfcpp::utf16to8(source);\n}\n\nstd::u16string to_utf16(const std::string_view source) {\n return utfcpp::utf8to16(source);\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/string.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_STRING_H_\n#define XENIA_BASE_STRING_H_\n\n#include \n\n#include \"xenia/base/utf8.h\"\n\nnamespace xe {\n\nint xe_strcasecmp(const char* string1, const char* string2);\nint xe_strncasecmp(const char* string1, const char* string2, size_t count);\nchar* xe_strdup(const char* source);\n\nstd::string to_utf8(const std::u16string_view source);\nstd::u16string to_utf16(const std::string_view source);\n\n} // namespace xe\n\n#endif // XENIA_BASE_STRING_H_\n" }, { "path": "src/xenia/base/string_buffer.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/string_buffer.h\"\n\n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/literals.h\"\n#include \"xenia/base/math.h\"\n\nnamespace xe {\n\nusing namespace xe::literals;\n\nStringBuffer::StringBuffer(size_t initial_capacity) {\n buffer_capacity_ = std::max(initial_capacity, static_cast(16_KiB));\n buffer_ = reinterpret_cast(std::malloc(buffer_capacity_));\n assert_not_null(buffer_);\n buffer_[0] = 0;\n}\n\nStringBuffer::~StringBuffer() {\n free(buffer_);\n buffer_ = nullptr;\n}\n\nvoid StringBuffer::Reset() {\n buffer_offset_ = 0;\n buffer_[0] = 0;\n}\n\nvoid StringBuffer::Grow(size_t additional_length) {\n if (buffer_offset_ + additional_length <= buffer_capacity_) {\n return;\n }\n size_t old_capacity = buffer_capacity_;\n size_t new_capacity =\n std::max(xe::round_up(buffer_offset_ + additional_length, 16_KiB),\n old_capacity * 2);\n auto new_buffer = std::realloc(buffer_, new_capacity);\n assert_not_null(new_buffer);\n buffer_ = reinterpret_cast(new_buffer);\n buffer_capacity_ = new_capacity;\n}\n\nvoid StringBuffer::Append(char c) {\n AppendBytes(reinterpret_cast(&c), 1);\n}\n\nvoid StringBuffer::Append(char c, size_t count) {\n Grow(count + 1);\n std::memset(buffer_ + buffer_offset_, c, count);\n buffer_offset_ += count;\n buffer_[buffer_offset_] = 0;\n}\n\nvoid StringBuffer::Append(const char* value) {\n AppendBytes(reinterpret_cast(value), std::strlen(value));\n}\n\nvoid StringBuffer::Append(const std::string_view value) {\n AppendBytes(reinterpret_cast(value.data()), value.size());\n}\n\nvoid StringBuffer::AppendVarargs(const char* format, va_list args) {\n int result = vsnprintf(nullptr, 0, format, args);\n if (result <= 0) {\n return;\n }\n auto length = static_cast(result);\n Grow(length + 1);\n vsnprintf(buffer_ + buffer_offset_, buffer_capacity_, format, args);\n buffer_offset_ += length;\n buffer_[buffer_offset_] = 0;\n}\n\nvoid StringBuffer::AppendBytes(const uint8_t* buffer, size_t length) {\n Grow(length + 1);\n memcpy(buffer_ + buffer_offset_, buffer, length);\n buffer_offset_ += length;\n buffer_[buffer_offset_] = 0;\n}\n\nstd::string StringBuffer::to_string() {\n return std::string(buffer_, buffer_offset_);\n}\n\nstd::string_view StringBuffer::to_string_view() const {\n return std::string_view(buffer_, buffer_offset_);\n}\n\nstd::vector StringBuffer::to_bytes() const {\n std::vector bytes(buffer_offset_);\n std::memcpy(bytes.data(), buffer_, buffer_offset_);\n return bytes;\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/string_buffer.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_STRING_BUFFER_H_\n#define XENIA_BASE_STRING_BUFFER_H_\n\n#include \n#include \n#include \n\n#include \"third_party/fmt/include/fmt/format.h\"\n\nnamespace xe {\n\nclass StringBuffer {\n public:\n explicit StringBuffer(size_t initial_capacity = 0);\n ~StringBuffer();\n\n char* buffer() const { return buffer_; }\n size_t length() const { return buffer_offset_; }\n\n void Reset();\n\n void Append(char c);\n void Append(char c, size_t count);\n void Append(const char* value);\n void Append(const std::string_view value);\n\n template \n void AppendFormat(const char* format, const Args&... args) {\n auto s = fmt::format(format, args...);\n Append(s.c_str());\n }\n\n void AppendVarargs(const char* format, va_list args);\n void AppendBytes(const uint8_t* buffer, size_t length);\n\n std::string to_string();\n std::string_view to_string_view() const;\n std::vector to_bytes() const;\n\n private:\n void Grow(size_t additional_length);\n\n char* buffer_ = nullptr;\n size_t buffer_offset_ = 0;\n size_t buffer_capacity_ = 0;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_STRING_BUFFER_H_\n" }, { "path": "src/xenia/base/string_key.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_STRING_KEY_H_\n#define XENIA_BASE_STRING_KEY_H_\n\n#include \n#include \n\n#include \"utf8.h\"\n\nnamespace xe {\n\nnamespace internal {\n\nstruct string_key_base {\n private:\n std::variant value_;\n\n public:\n explicit string_key_base(const std::string_view value) : value_(value) {}\n explicit string_key_base(std::string value) : value_(std::move(value)) {}\n\n std::string_view view() const {\n return std::holds_alternative(value_)\n ? std::get(value_)\n : std::get(value_);\n }\n};\n\n} // namespace internal\n\nstruct string_key : internal::string_key_base {\n public:\n explicit string_key(const std::string_view value) : string_key_base(value) {}\n explicit string_key(std::string value) : string_key_base(value) {}\n\n static string_key create(const std::string_view value) {\n return string_key(std::string(value));\n }\n\n static string_key create(std::string value) { return string_key(value); }\n\n bool operator==(const string_key& other) const {\n return other.view() == view();\n }\n\n size_t hash() const { return utf8::hash_fnv1a(view()); }\n\n struct Hash {\n size_t operator()(const string_key& t) const { return t.hash(); }\n };\n};\n\nstruct string_key_case : internal::string_key_base {\n public:\n explicit string_key_case(const std::string_view value)\n : string_key_base(value) {}\n explicit string_key_case(std::string value) : string_key_base(value) {}\n\n static string_key_case create(const std::string_view value) {\n return string_key_case(std::string(value));\n }\n\n static string_key_case create(std::string value) {\n return string_key_case(value);\n }\n\n bool operator==(const string_key_case& other) const {\n return utf8::equal_case(other.view(), view());\n }\n\n size_t hash() const { return utf8::hash_fnv1a_case(view()); }\n\n struct Hash {\n size_t operator()(const string_key_case& t) const { return t.hash(); }\n };\n};\n\n} // namespace xe\n\nnamespace std {\ntemplate <>\nstruct hash {\n std::size_t operator()(const xe::string_key& t) const { return t.hash(); }\n};\n\ntemplate <>\nstruct hash {\n std::size_t operator()(const xe::string_key_case& t) const {\n return t.hash();\n }\n};\n\n}; // namespace std\n\n#endif // XENIA_BASE_STRING_KEY_H_\n" }, { "path": "src/xenia/base/string_util.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_STRING_UTIL_H_\n#define XENIA_BASE_STRING_UTIL_H_\n\n#include \n#include \n#include \n#include \n#include \n\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/memory.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/string.h\"\n#include \"xenia/base/vec128.h\"\n\n// TODO(gibbed): Clang and GCC don't have std::from_chars for floating point(!)\n// despite it being part of the C++17 standard. Check this in the future to see\n// if it's been resolved.\n\n#if XE_COMPILER_CLANG || XE_COMPILER_GNUC\n#include \n#endif\n\nnamespace xe {\nnamespace string_util {\n\nenum class Safety {\n IDontKnowWhatIAmDoing,\n IKnowWhatIAmDoing,\n};\n\ninline size_t copy_truncating(char* dest, const std::string_view source,\n size_t dest_buffer_count) {\n if (!dest_buffer_count) {\n return 0;\n }\n size_t chars_copied = std::min(source.size(), dest_buffer_count - size_t(1));\n std::memcpy(dest, source.data(), chars_copied);\n dest[chars_copied] = '\\0';\n return chars_copied;\n}\n\ninline size_t copy_truncating(char16_t* dest, const std::u16string_view source,\n size_t dest_buffer_count) {\n if (!dest_buffer_count) {\n return 0;\n }\n size_t chars_copied = std::min(source.size(), dest_buffer_count - size_t(1));\n std::memcpy(dest, source.data(), chars_copied * sizeof(char16_t));\n dest[chars_copied] = u'\\0';\n return chars_copied;\n}\n\ninline size_t copy_and_swap_truncating(char16_t* dest,\n const std::u16string_view source,\n size_t dest_buffer_count) {\n if (!dest_buffer_count) {\n return 0;\n }\n size_t chars_copied = std::min(source.size(), dest_buffer_count - size_t(1));\n xe::copy_and_swap(dest, source.data(), chars_copied);\n dest[chars_copied] = u'\\0';\n return chars_copied;\n}\n\ntemplate \ninline size_t copy_maybe_truncating(char* dest, const std::string_view source,\n size_t dest_buffer_count) {\n static_assert(safety == Safety::IKnowWhatIAmDoing);\n if (!dest_buffer_count) {\n return 0;\n }\n size_t chars_copied = std::min(source.size(), dest_buffer_count);\n std::memcpy(dest, source.data(), chars_copied);\n return chars_copied;\n}\n\ntemplate \ninline size_t copy_maybe_truncating(char16_t* dest,\n const std::u16string_view source,\n size_t dest_buffer_count) {\n static_assert(safety == Safety::IKnowWhatIAmDoing);\n if (!dest_buffer_count) {\n return 0;\n }\n size_t chars_copied = std::min(source.size(), dest_buffer_count);\n std::memcpy(dest, source.data(), chars_copied * sizeof(char16_t));\n return chars_copied;\n}\n\ntemplate \ninline size_t copy_and_swap_maybe_truncating(char16_t* dest,\n const std::u16string_view source,\n size_t dest_buffer_count) {\n static_assert(safety == Safety::IKnowWhatIAmDoing);\n if (!dest_buffer_count) {\n return 0;\n }\n size_t chars_copied = std::min(source.size(), dest_buffer_count);\n xe::copy_and_swap(dest, source.data(), chars_copied);\n return chars_copied;\n}\n\ninline std::string to_hex_string(uint32_t value) {\n return fmt::format(\"{:08X}\", value);\n}\n\ninline std::string to_hex_string(uint64_t value) {\n return fmt::format(\"{:016X}\", value);\n}\n\ninline std::string to_hex_string(float value) {\n static_assert(sizeof(uint32_t) == sizeof(value));\n uint32_t pun;\n std::memcpy(&pun, &value, sizeof(value));\n return to_hex_string(pun);\n}\n\ninline std::string to_hex_string(double value) {\n static_assert(sizeof(uint64_t) == sizeof(value));\n uint64_t pun;\n std::memcpy(&pun, &value, sizeof(value));\n return to_hex_string(pun);\n}\n\ninline std::string to_hex_string(const vec128_t& value) {\n return fmt::format(\"[{:08X} {:08X} {:08X} {:08X}]\", value.u32[0],\n value.u32[1], value.u32[2], value.u32[3]);\n}\n\ntemplate \ninline T from_string(const std::string_view value, bool force_hex = false) {\n // Missing implementation for converting type T from string\n throw;\n}\n\nnamespace internal {\n\ntemplate >\ninline T make_negative(T value) {\n if constexpr (std::is_unsigned_v) {\n value = static_cast(-static_cast(value));\n } else {\n value = -value;\n }\n return value;\n}\n\n// integral_from_string\ntemplate \ninline T ifs(const std::string_view value, bool force_hex) {\n int base = 10;\n std::string_view range = value;\n bool is_hex = force_hex;\n bool is_negative = false;\n if (utf8::starts_with(range, \"-\")) {\n is_negative = true;\n range = range.substr(1);\n }\n if (utf8::starts_with(range, \"0x\")) {\n is_hex = true;\n range = range.substr(2);\n }\n if (utf8::ends_with(range, \"h\")) {\n is_hex = true;\n range = range.substr(0, range.length() - 1);\n }\n T result;\n if (is_hex) {\n base = 16;\n }\n // TODO(gibbed): do something more with errors?\n auto [p, error] =\n std::from_chars(range.data(), range.data() + range.size(), result, base);\n if (error != std::errc()) {\n assert_always();\n return T();\n }\n if (is_negative) {\n result = make_negative(result);\n }\n return result;\n}\n\n// floating_point_from_string\ntemplate \ninline T fpfs(const std::string_view value, bool force_hex) {\n static_assert(sizeof(T) == sizeof(PUN));\n std::string_view range = value;\n bool is_hex = force_hex;\n bool is_negative = false;\n if (utf8::starts_with(range, \"-\")) {\n is_negative = true;\n range = range.substr(1);\n }\n if (utf8::starts_with(range, \"0x\")) {\n is_hex = true;\n range = range.substr(2);\n }\n if (utf8::ends_with(range, \"h\")) {\n is_hex = true;\n range = range.substr(0, range.length() - 1);\n }\n T result;\n if (is_hex) {\n PUN pun = from_string(range, true);\n if (is_negative) {\n pun = make_negative(pun);\n }\n std::memcpy(&result, &pun, sizeof(PUN));\n } else {\n#if XE_COMPILER_CLANG || XE_COMPILER_GNUC\n auto temp = std::string(range);\n result = std::strtod(temp.c_str(), nullptr);\n#else\n auto [p, error] = std::from_chars(range.data(), range.data() + range.size(),\n result, std::chars_format::general);\n // TODO(gibbed): do something more with errors?\n if (error != std::errc()) {\n assert_always();\n return T();\n }\n#endif\n if (is_negative) {\n result = -result;\n }\n }\n return result;\n}\n\n} // namespace internal\n\ntemplate <>\ninline bool from_string(const std::string_view value, bool force_hex) {\n return value == \"true\" || value == \"1\";\n}\n\ntemplate <>\ninline int8_t from_string(const std::string_view value,\n bool force_hex) {\n return internal::ifs(value, force_hex);\n}\n\ntemplate <>\ninline uint8_t from_string(const std::string_view value,\n bool force_hex) {\n return internal::ifs(value, force_hex);\n}\n\ntemplate <>\ninline int16_t from_string(const std::string_view value,\n bool force_hex) {\n return internal::ifs(value, force_hex);\n}\n\ntemplate <>\ninline uint16_t from_string(const std::string_view value,\n bool force_hex) {\n return internal::ifs(value, force_hex);\n}\n\ntemplate <>\ninline int32_t from_string(const std::string_view value,\n bool force_hex) {\n return internal::ifs(value, force_hex);\n}\n\ntemplate <>\ninline uint32_t from_string(const std::string_view value,\n bool force_hex) {\n return internal::ifs(value, force_hex);\n}\n\ntemplate <>\ninline int64_t from_string(const std::string_view value,\n bool force_hex) {\n return internal::ifs(value, force_hex);\n}\n\ntemplate <>\ninline uint64_t from_string(const std::string_view value,\n bool force_hex) {\n return internal::ifs(value, force_hex);\n}\n\ntemplate <>\ninline float from_string(const std::string_view value, bool force_hex) {\n return internal::fpfs(value, force_hex);\n}\n\ntemplate <>\ninline double from_string(const std::string_view value,\n bool force_hex) {\n return internal::fpfs(value, force_hex);\n}\n\ntemplate <>\ninline vec128_t from_string(const std::string_view value,\n bool force_hex) {\n if (!value.size()) {\n return vec128_t();\n }\n vec128_t v;\n#if XE_COMPILER_CLANG || XE_COMPILER_GNUC\n auto temp = std::string(value);\n auto p = temp.c_str();\n auto end = temp.c_str() + temp.size();\n#else\n auto p = value.data();\n auto end = value.data() + value.size();\n#endif\n bool is_hex = force_hex;\n if (p != end && *p == '[') {\n is_hex = true;\n ++p;\n } else if (p != end && *p == '(') {\n is_hex = false;\n ++p;\n } else {\n // Assume hex?\n is_hex = true;\n }\n if (p == end) {\n assert_always();\n return vec128_t();\n }\n if (is_hex) {\n for (size_t i = 0; i < 4; i++) {\n while (p != end && (*p == ' ' || *p == ',')) {\n ++p;\n }\n if (p == end) {\n assert_always();\n return vec128_t();\n }\n auto result = std::from_chars(p, end, v.u32[i], 16);\n if (result.ec != std::errc()) {\n assert_always();\n return vec128_t();\n }\n p = result.ptr;\n }\n } else {\n for (size_t i = 0; i < 4; i++) {\n while (p != end && (*p == ' ' || *p == ',')) {\n ++p;\n }\n if (p == end) {\n assert_always();\n return vec128_t();\n }\n#if XE_COMPILER_CLANG || XE_COMPILER_GNUC\n char* next_p;\n v.f32[i] = std::strtof(p, &next_p);\n p = next_p;\n#else\n auto result =\n std::from_chars(p, end, v.f32[i], std::chars_format::general);\n if (result.ec != std::errc()) {\n assert_always();\n return vec128_t();\n }\n p = result.ptr;\n#endif\n }\n }\n return v;\n}\n\n} // namespace string_util\n} // namespace xe\n\n#endif // XENIA_BASE_STRING_UTIL_H_\n" }, { "path": "src/xenia/base/system.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_SYSTEM_H_\n#define XENIA_BASE_SYSTEM_H_\n\n#include \n#include \n\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/string.h\"\n\nnamespace xe {\n\n#if XE_PLATFORM_ANDROID\nbool InitializeAndroidSystemForApplicationContext();\nvoid ShutdownAndroidSystem();\n#endif\n\n// The URL must include the protocol.\nvoid LaunchWebBrowser(const std::string_view url);\nvoid LaunchFileExplorer(const std::filesystem::path& path);\n\nenum class SimpleMessageBoxType {\n Help,\n Warning,\n Error,\n};\n\n// This is expected to block the caller until the message box is closed.\nvoid ShowSimpleMessageBox(SimpleMessageBoxType type, std::string_view message);\n\n} // namespace xe\n\n#endif // XENIA_BASE_SYSTEM_H_\n" }, { "path": "src/xenia/base/system_android.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/main_android.h\"\n#include \"xenia/base/system.h\"\n\nnamespace xe {\n\n// To store method and field IDs persistently, global references to the classes\n// are required to prevent the classes from being unloaded and reloaded,\n// potentially changing the IDs.\n\nstatic jclass android_system_application_context_class_ = nullptr;\nstatic jmethodID android_system_application_context_start_activity_ = nullptr;\n\nstatic jclass android_system_uri_class_ = nullptr;\nstatic jmethodID android_system_uri_parse_ = nullptr;\n\nstatic jclass android_system_intent_class_ = nullptr;\nstatic jmethodID android_system_intent_init_action_uri_ = nullptr;\nstatic jmethodID android_system_intent_add_flags_ = nullptr;\nstatic jobject android_system_intent_action_view_ = nullptr;\nstatic jint android_system_intent_flag_activity_new_task_;\n\nstatic bool android_system_initialized_ = false;\n\nbool InitializeAndroidSystemForApplicationContext() {\n assert_false(android_system_initialized_);\n\n JNIEnv* jni_env = GetAndroidThreadJniEnv();\n if (!jni_env) {\n return false;\n }\n jobject application_context = xe::GetAndroidApplicationContext();\n if (!application_context) {\n return false;\n }\n\n // Application context.\n {\n {\n jclass application_context_class_local_ref =\n jni_env->GetObjectClass(application_context);\n if (!application_context_class_local_ref) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to get the \"\n \"class of the application context\");\n ShutdownAndroidSystem();\n return false;\n }\n android_system_application_context_class_ =\n reinterpret_cast(jni_env->NewGlobalRef(\n reinterpret_cast(application_context_class_local_ref)));\n jni_env->DeleteLocalRef(application_context_class_local_ref);\n }\n if (!android_system_application_context_class_) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to create a \"\n \"global reference to the class of the application context\");\n ShutdownAndroidSystem();\n return false;\n }\n bool application_context_ids_obtained = true;\n application_context_ids_obtained &=\n (android_system_application_context_start_activity_ =\n jni_env->GetMethodID(android_system_application_context_class_,\n \"startActivity\",\n \"(Landroid/content/Intent;)V\")) != nullptr;\n if (!application_context_ids_obtained) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to get the \"\n \"application context class IDs\");\n ShutdownAndroidSystem();\n return false;\n }\n }\n\n // URI.\n {\n {\n jclass uri_class_local_ref = jni_env->FindClass(\"android/net/Uri\");\n if (!uri_class_local_ref) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to find the \"\n \"URI class\");\n ShutdownAndroidSystem();\n return false;\n }\n android_system_uri_class_ =\n reinterpret_cast(jni_env->NewGlobalRef(\n reinterpret_cast(uri_class_local_ref)));\n jni_env->DeleteLocalRef(uri_class_local_ref);\n }\n if (!android_system_uri_class_) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to create a \"\n \"global reference to the URI class\");\n ShutdownAndroidSystem();\n return false;\n }\n bool uri_ids_obtained = true;\n uri_ids_obtained &=\n (android_system_uri_parse_ = jni_env->GetStaticMethodID(\n android_system_uri_class_, \"parse\",\n \"(Ljava/lang/String;)Landroid/net/Uri;\")) != nullptr;\n if (!uri_ids_obtained) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to get the URI \"\n \"class IDs\");\n ShutdownAndroidSystem();\n return false;\n }\n }\n\n // Intent.\n {\n {\n jclass intent_class_local_ref =\n jni_env->FindClass(\"android/content/Intent\");\n if (!intent_class_local_ref) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to find the \"\n \"intent class\");\n ShutdownAndroidSystem();\n return false;\n }\n android_system_intent_class_ =\n reinterpret_cast(jni_env->NewGlobalRef(\n reinterpret_cast(intent_class_local_ref)));\n jni_env->DeleteLocalRef(intent_class_local_ref);\n }\n if (!android_system_intent_class_) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to create a \"\n \"global reference to the intent class\");\n ShutdownAndroidSystem();\n return false;\n }\n bool intent_ids_obtained = true;\n jfieldID intent_action_view_id;\n intent_ids_obtained &= (intent_action_view_id = jni_env->GetStaticFieldID(\n android_system_intent_class_, \"ACTION_VIEW\",\n \"Ljava/lang/String;\")) != nullptr;\n jfieldID intent_flag_activity_new_task_id;\n intent_ids_obtained &=\n (intent_flag_activity_new_task_id = jni_env->GetStaticFieldID(\n android_system_intent_class_, \"FLAG_ACTIVITY_NEW_TASK\", \"I\")) !=\n nullptr;\n intent_ids_obtained &=\n (android_system_intent_init_action_uri_ = jni_env->GetMethodID(\n android_system_intent_class_, \"\",\n \"(Ljava/lang/String;Landroid/net/Uri;)V\")) != nullptr;\n intent_ids_obtained &=\n (android_system_intent_add_flags_ =\n jni_env->GetMethodID(android_system_intent_class_, \"addFlags\",\n \"(I)Landroid/content/Intent;\")) != nullptr;\n if (!intent_ids_obtained) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to get the \"\n \"intent class IDs\");\n ShutdownAndroidSystem();\n return false;\n }\n {\n jobject intent_action_view_local_ref = jni_env->GetStaticObjectField(\n android_system_intent_class_, intent_action_view_id);\n if (!intent_action_view_local_ref) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to get the \"\n \"intent view action string\");\n ShutdownAndroidSystem();\n return false;\n }\n android_system_intent_action_view_ =\n jni_env->NewGlobalRef(intent_action_view_local_ref);\n jni_env->DeleteLocalRef(intent_action_view_local_ref);\n if (!android_system_intent_action_view_) {\n XELOGE(\n \"InitializeAndroidSystemForApplicationContext: Failed to create a \"\n \"global reference to the intent view action string\");\n ShutdownAndroidSystem();\n return false;\n }\n }\n android_system_intent_flag_activity_new_task_ = jni_env->GetStaticIntField(\n android_system_intent_class_, intent_flag_activity_new_task_id);\n }\n\n android_system_initialized_ = true;\n return true;\n}\n\nvoid ShutdownAndroidSystem() {\n // May be called from InitializeAndroidSystemForApplicationContext as well.\n android_system_initialized_ = false;\n android_system_intent_add_flags_ = nullptr;\n android_system_intent_init_action_uri_ = nullptr;\n android_system_uri_parse_ = nullptr;\n android_system_application_context_start_activity_ = nullptr;\n JNIEnv* jni_env = GetAndroidThreadJniEnv();\n if (jni_env) {\n if (android_system_intent_action_view_) {\n jni_env->DeleteGlobalRef(android_system_intent_action_view_);\n }\n if (android_system_intent_class_) {\n jni_env->DeleteGlobalRef(android_system_intent_class_);\n }\n if (android_system_uri_class_) {\n jni_env->DeleteGlobalRef(android_system_uri_class_);\n }\n if (android_system_application_context_class_) {\n jni_env->DeleteGlobalRef(android_system_application_context_class_);\n }\n }\n android_system_intent_action_view_ = nullptr;\n android_system_intent_class_ = nullptr;\n android_system_uri_class_ = nullptr;\n android_system_application_context_class_ = nullptr;\n}\n\nvoid LaunchWebBrowser(const std::string_view url) {\n if (!android_system_initialized_) {\n return;\n }\n JNIEnv* jni_env = GetAndroidThreadJniEnv();\n if (!jni_env) {\n return;\n }\n jobject application_context = GetAndroidApplicationContext();\n if (!application_context) {\n return;\n }\n\n jstring uri_string = jni_env->NewStringUTF(std::string(url).c_str());\n if (!uri_string) {\n XELOGE(\"LaunchWebBrowser: Failed to create the URI string\");\n return;\n }\n jobject uri = jni_env->CallStaticObjectMethod(\n android_system_uri_class_, android_system_uri_parse_, uri_string);\n jni_env->DeleteLocalRef(uri_string);\n if (!uri) {\n XELOGE(\"LaunchWebBrowser: Failed to parse the URI\");\n return;\n }\n jobject intent = jni_env->NewObject(android_system_intent_class_,\n android_system_intent_init_action_uri_,\n android_system_intent_action_view_, uri);\n jni_env->DeleteLocalRef(uri);\n if (!intent) {\n XELOGE(\"LaunchWebBrowser: Failed to create the intent\");\n return;\n }\n // Start a new task - the user may want to be able to switch between the\n // emulator and the newly opened web browser, without having to quit the web\n // browser to return to the emulator. Also, since the application context, not\n // the activity, is used, the new task flag is required.\n {\n jobject intent_add_flags_result_local_ref = jni_env->CallObjectMethod(\n intent, android_system_intent_add_flags_,\n android_system_intent_flag_activity_new_task_);\n if (intent_add_flags_result_local_ref) {\n jni_env->DeleteLocalRef(intent_add_flags_result_local_ref);\n }\n }\n jni_env->CallVoidMethod(application_context,\n android_system_application_context_start_activity_,\n intent);\n jni_env->DeleteLocalRef(intent);\n}\n\nvoid LaunchFileExplorer(const std::filesystem::path& path) { assert_always(); }\n\nvoid ShowSimpleMessageBox(SimpleMessageBoxType type, std::string_view message) {\n // TODO(Triang3l): Likely not needed much at all. ShowSimpleMessageBox is a\n // concept pretty unfriendly to platforms like Android because it's blocking,\n // and because it can be called from threads other than the UI thread. In the\n // normal execution flow, dialogs should preferably be asynchronous, and used\n // only in the UI thread. However, non-blocking messages may be good for error\n // reporting - investigate the usage of Toasts with respect to threads, and\n // aborting the process immediately after showing a Toast. For a Toast, the\n // Java VM for the calling thread is needed.\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/system_gnulinux.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n#include \n\n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/string.h\"\n#include \"xenia/base/system.h\"\n\n// Use headers in third party to not depend on system sdl headers for building\n#include \"third_party/SDL2/include/SDL.h\"\n\nnamespace xe {\n\nvoid LaunchWebBrowser(const std::string_view url) {\n auto cmd = std::string(\"xdg-open \");\n cmd.append(url);\n system(cmd.c_str());\n}\n\nvoid LaunchFileExplorer(const std::filesystem::path& path) { assert_always(); }\n\nvoid ShowSimpleMessageBox(SimpleMessageBoxType type, std::string_view message) {\n void* libsdl2 = dlopen(\"libSDL2.so\", RTLD_LAZY | RTLD_LOCAL);\n assert_not_null(libsdl2);\n if (libsdl2) {\n auto* pSDL_ShowSimpleMessageBox =\n reinterpret_cast(\n dlsym(libsdl2, \"SDL_ShowSimpleMessageBox\"));\n assert_not_null(pSDL_ShowSimpleMessageBox);\n if (pSDL_ShowSimpleMessageBox) {\n Uint32 flags;\n const char* title;\n char* message_copy = reinterpret_cast(alloca(message.size() + 1));\n std::memcpy(message_copy, message.data(), message.size());\n message_copy[message.size()] = '\\0';\n\n switch (type) {\n default:\n case SimpleMessageBoxType::Help:\n title = \"Xenia Help\";\n flags = SDL_MESSAGEBOX_INFORMATION;\n break;\n case SimpleMessageBoxType::Warning:\n title = \"Xenia Warning\";\n flags = SDL_MESSAGEBOX_WARNING;\n break;\n case SimpleMessageBoxType::Error:\n title = \"Xenia Error\";\n flags = SDL_MESSAGEBOX_ERROR;\n break;\n }\n pSDL_ShowSimpleMessageBox(flags, title, message_copy, NULL);\n }\n dlclose(libsdl2);\n }\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/system_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/platform_win.h\"\n#include \"xenia/base/string.h\"\n#include \"xenia/base/system.h\"\n\nnamespace xe {\n\nvoid LaunchWebBrowser(const std::string_view url) {\n auto wide_url = xe::to_utf16(url);\n ShellExecuteW(nullptr, L\"open\", reinterpret_cast(wide_url.c_str()),\n nullptr, nullptr, SW_SHOWNORMAL);\n}\n\nvoid LaunchFileExplorer(const std::filesystem::path& url) {\n ShellExecuteW(nullptr, L\"explore\", url.c_str(), nullptr, nullptr,\n SW_SHOWNORMAL);\n}\n\nvoid ShowSimpleMessageBox(SimpleMessageBoxType type,\n const std::string_view message) {\n const wchar_t* title;\n std::u16string wide_message = xe::to_utf16(message);\n DWORD type_flags = MB_OK | MB_APPLMODAL | MB_SETFOREGROUND;\n switch (type) {\n default:\n case SimpleMessageBoxType::Help:\n title = L\"Xenia Help\";\n type_flags |= MB_ICONINFORMATION;\n break;\n case SimpleMessageBoxType::Warning:\n title = L\"Xenia Warning\";\n type_flags |= MB_ICONWARNING;\n break;\n case SimpleMessageBoxType::Error:\n title = L\"Xenia Error\";\n type_flags |= MB_ICONERROR;\n break;\n }\n MessageBoxW(nullptr, reinterpret_cast(wide_message.c_str()), title,\n type_flags);\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/testing/chrono_test.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/chrono.h\"\n\n#define CATCH_CONFIG_ENABLE_CHRONO_STRINGMAKER\n#include \"third_party/catch/include/catch.hpp\"\n\nnamespace xe::base::test {\n\n// If the tests run fast, the time duration may be zero. So we artificially\n// raise it to allow for some error.\ntemplate \nauto dur_bound(std::chrono::duration<_Rep, _Period> dur) {\n using namespace std::chrono_literals;\n return std::max>>(\n dur, 1ms);\n}\n\nTEST_CASE(\"WinSystemClock <-> system_clock\", \"[clock_cast]\") {\n using namespace xe::chrono;\n using namespace date;\n using sys_clock = std::chrono::system_clock;\n\n // First check some assumptions made in chrono.h\n\n SECTION(\"C++20 unix epoch\") {\n auto epoch = date::year{1970} / date::month{1} / date::day{1};\n std::chrono::system_clock::time_point epoch_tp{\n static_cast(epoch)};\n REQUIRE(epoch_tp.time_since_epoch().count() == 0);\n }\n\n SECTION(\"Epoch delta\") {\n std::chrono::seconds sys_delta = WinSystemClock::unix_epoch_delta();\n REQUIRE(sys_delta.count() == INT64_C(-11644473600));\n }\n\n SECTION(\"1993/12/21\") {\n static constexpr sys_days sys(1993_y / dec / 21);\n\n static constexpr auto wsys = WinSystemClock::from_sys(sys);\n REQUIRE(wsys.time_since_epoch().count() == 124009056000000000);\n\n REQUIRE(clock_cast(wsys) == sys);\n REQUIRE(clock_cast(sys) == wsys);\n }\n\n SECTION(\"2100/3/1\") {\n static constexpr sys_days sys(2100_y / mar / 1);\n\n static constexpr auto wsys = WinSystemClock::from_sys(sys);\n REQUIRE(wsys.time_since_epoch().count() == 157520160000000000);\n\n REQUIRE(clock_cast(wsys) == sys);\n REQUIRE(clock_cast(sys) == wsys);\n }\n}\n\nTEST_CASE(\"WinSystemClock <-> XSystemClock\", \"[clock_cast]\") {\n using namespace std::chrono_literals;\n using namespace xe::chrono;\n using namespace date;\n using sys_clock = std::chrono::system_clock;\n\n SECTION(\"1993/12/21, clock_no_scaling = true\") {\n static constexpr sys_days sys(1993_y / dec / 21);\n cvars::clock_no_scaling = true;\n Clock::set_guest_time_scalar(1.0);\n\n static constexpr auto wsys = WinSystemClock::from_sys(sys);\n\n auto start = std::chrono::system_clock::now();\n\n auto xsys = date::clock_cast(wsys);\n auto wxsys = date::clock_cast(xsys);\n\n auto duration = dur_bound(std::chrono::system_clock::now() - start);\n\n auto error = wsys.time_since_epoch() - wxsys.time_since_epoch();\n REQUIRE(error < duration);\n REQUIRE(error > -duration);\n }\n\n SECTION(\"1993/12/21, clock_no_scaling = false\") {\n static constexpr sys_days sys(1993_y / dec / 21);\n cvars::clock_no_scaling = false;\n Clock::set_guest_time_scalar(1.0);\n\n static constexpr auto wsys = WinSystemClock::from_sys(sys);\n\n auto start = std::chrono::system_clock::now();\n\n auto xsys = date::clock_cast(wsys);\n auto wxsys = date::clock_cast(xsys);\n\n auto duration = dur_bound(std::chrono::system_clock::now() - start);\n\n auto error1 = wsys.time_since_epoch() - xsys.time_since_epoch();\n auto error2 = xsys.time_since_epoch() - wxsys.time_since_epoch();\n auto error3 = wsys - wxsys;\n\n // In AppVeyor, the difference often can be as large as roughly 16ms.\n REQUIRE(error1 < 20ms);\n REQUIRE(error1 > -20ms);\n REQUIRE(error2 < 20ms);\n REQUIRE(error2 > -20ms);\n REQUIRE(error3 < duration);\n REQUIRE(error3 > -duration);\n }\n}\n\n} // namespace xe::base::test\n\n// only make these available now\n#include \"xenia/base/chrono_steady_cast.h\"\n\nnamespace xe::base::test {\n\nTEST_CASE(\"WinSystemClock <-> steady_clock\", \"[clock_cast]\") {\n using namespace xe::chrono;\n using namespace date;\n using sty_clock = std::chrono::steady_clock;\n\n // steady conversion is mostly used to convert wait times\n\n SECTION(\"now\") {\n auto sty = sty_clock::now();\n\n // Because steady casts are imprecise, we need to allow some margin of error\n auto start = sty_clock::now();\n\n auto wsty = clock_cast(sty);\n auto sty2 = clock_cast(wsty);\n\n auto duration = dur_bound(sty_clock::now() - start).count();\n auto error = std::abs((sty2 - sty).count());\n REQUIRE(error <= duration);\n }\n}\n\n} // namespace xe::base::test\n" }, { "path": "src/xenia/base/testing/memory_test.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/memory.h\"\n\n#include \"third_party/catch/include/catch.hpp\"\n#include \"third_party/fmt/include/fmt/format.h\"\n\n#include \"xenia/base/clock.h\"\n\n#include \n\nnamespace xe {\nnamespace base {\nnamespace test {\n\nTEST_CASE(\"copy_128_aligned\", \"[copy_and_swap]\") {\n alignas(128) uint8_t src[256], dest[256];\n for (uint8_t i = 0; i < 255; ++i) {\n src[i] = 255 - i;\n }\n std::memset(dest, 0, sizeof(dest));\n copy_128_aligned(dest, src, 1);\n REQUIRE(std::memcmp(dest, src, 128));\n REQUIRE(dest[128] == 0);\n\n std::memset(dest, 0, sizeof(dest));\n copy_128_aligned(dest, src, 2);\n REQUIRE(std::memcmp(dest, src, 256));\n\n std::memset(dest, 0, sizeof(dest));\n copy_128_aligned(dest, src + 1, 1);\n REQUIRE(std::memcmp(dest, src + 1, 128));\n}\n\nTEST_CASE(\"copy_and_swap_16_aligned\", \"[copy_and_swap]\") {\n alignas(16) uint16_t a = 0x1111, b = 0xABCD;\n copy_and_swap_16_aligned(&a, &b, 1);\n REQUIRE(a == 0xCDAB);\n REQUIRE(b == 0xABCD);\n\n alignas(16) uint16_t c[] = {0x0000, 0x0000, 0x0000, 0x0000};\n alignas(16) uint16_t d[] = {0x0123, 0x4567, 0x89AB, 0xCDEF};\n copy_and_swap_16_aligned(c, d, 1);\n REQUIRE(c[0] == 0x2301);\n REQUIRE(c[1] == 0x0000);\n REQUIRE(c[2] == 0x0000);\n REQUIRE(c[3] == 0x0000);\n\n copy_and_swap_16_aligned(c, d, 3);\n REQUIRE(c[0] == 0x2301);\n REQUIRE(c[1] == 0x6745);\n REQUIRE(c[2] == 0xAB89);\n REQUIRE(c[3] == 0x0000);\n\n copy_and_swap_16_aligned(c, d, 4);\n REQUIRE(c[0] == 0x2301);\n REQUIRE(c[1] == 0x6745);\n REQUIRE(c[2] == 0xAB89);\n REQUIRE(c[3] == 0xEFCD);\n\n alignas(16) uint64_t e;\n copy_and_swap_16_aligned(&e, d, 4);\n REQUIRE(e == 0xEFCDAB8967452301);\n\n alignas(16) char f[85] = {0x00};\n alignas(16) char g[] =\n \"This is a 85 byte long string... \"\n \"It's supposed to be longer than standard alignment.\";\n copy_and_swap_16_aligned(f, g, 42);\n REQUIRE(std::strcmp(f,\n \"hTsii s a58b ty eolgns rtni.g..I 't susppsodet oebl \"\n \"noeg rhtnas atdnra dlagimnne.t\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_16_aligned(f, g + 16, 34);\n REQUIRE(std::strcmp(f,\n \" eolgns rtni.g..I 't susppsodet oebl \"\n \"noeg rhtnas atdnra dlagimnne.t\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_16_aligned(f, g + 32, 26);\n REQUIRE(std::strcmp(f,\n \"I 't susppsodet oebl \"\n \"noeg rhtnas atdnra dlagimnne.t\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_16_aligned(f, g + 64, 10);\n REQUIRE(std::strcmp(f, \"s atdnra dlagimnne.t\") == 0);\n}\n\nTEST_CASE(\"copy_and_swap_16_unaligned\", \"[copy_and_swap]\") {\n uint16_t a = 0x1111, b = 0xABCD;\n copy_and_swap_16_unaligned(&a, &b, 1);\n REQUIRE(a == 0xCDAB);\n REQUIRE(b == 0xABCD);\n\n uint16_t c[] = {0x0000, 0x0000, 0x0000, 0x0000};\n uint16_t d[] = {0x0123, 0x4567, 0x89AB, 0xCDEF};\n copy_and_swap_16_unaligned(c, d, 1);\n REQUIRE(c[0] == 0x2301);\n REQUIRE(c[1] == 0x0000);\n REQUIRE(c[2] == 0x0000);\n REQUIRE(c[3] == 0x0000);\n\n copy_and_swap_16_unaligned(c, d, 4);\n REQUIRE(c[0] == 0x2301);\n REQUIRE(c[1] == 0x6745);\n REQUIRE(c[2] == 0xAB89);\n REQUIRE(c[3] == 0xEFCD);\n\n {\n constexpr size_t count = 100;\n std::array src{};\n std::array dst{};\n for (size_t i = 0; i < src.size(); ++i) {\n src[i] = static_cast(i) + 1; // no zero in array\n }\n copy_and_swap_16_unaligned(dst.data(), src.data(), count);\n for (size_t i = 0; i < src.size(); i += 2) {\n // Check src is untouched\n REQUIRE(static_cast(src[i + 0]) == i + 1);\n REQUIRE(static_cast(src[i + 1]) == i + 2);\n // Check swapped bytes\n REQUIRE(static_cast(dst[i]) == static_cast(src[i + 1]));\n REQUIRE(static_cast(dst[i + 1]) == static_cast(src[i]));\n }\n }\n\n uint64_t e;\n copy_and_swap_16_unaligned(&e, d, 4);\n REQUIRE(e == 0xEFCDAB8967452301);\n\n char f[85] = {0x00};\n char g[] =\n \"This is a 85 byte long string... \"\n \"It's supposed to be longer than standard alignment.\";\n copy_and_swap_16_unaligned(f, g, 42);\n REQUIRE(std::strcmp(f,\n \"hTsii s a58b ty eolgns rtni.g..I 't susppsodet oebl \"\n \"noeg rhtnas atdnra dlagimnne.t\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_16_unaligned(f, g + 1, 41);\n REQUIRE(std::strcmp(f,\n \"ih ssia 8 5ybetl no gtsirgn.. .tIs's puopes dotb \"\n \"eolgnret ah ntsnaaddra ilngemtn\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_16_unaligned(f, g + 2, 41);\n REQUIRE(std::strcmp(f,\n \"sii s a58b ty eolgns rtni.g..I 't susppsodet oebl \"\n \"noeg rhtnas atdnra dlagimnne.t\") == 0);\n}\n\nTEST_CASE(\"copy_and_swap_32_aligned\", \"[copy_and_swap]\") {\n alignas(32) uint32_t a = 0x11111111, b = 0x89ABCDEF;\n copy_and_swap_32_aligned(&a, &b, 1);\n REQUIRE(a == 0xEFCDAB89);\n REQUIRE(b == 0x89ABCDEF);\n\n alignas(32) uint32_t c[] = {0x00000000, 0x00000000, 0x00000000, 0x00000000};\n alignas(32) uint32_t d[] = {0x01234567, 0x89ABCDEF, 0xE887EEED, 0xD8514199};\n copy_and_swap_32_aligned(c, d, 1);\n REQUIRE(c[0] == 0x67452301);\n REQUIRE(c[1] == 0x00000000);\n REQUIRE(c[2] == 0x00000000);\n REQUIRE(c[3] == 0x00000000);\n\n copy_and_swap_32_aligned(c, d, 3);\n REQUIRE(c[0] == 0x67452301);\n REQUIRE(c[1] == 0xEFCDAB89);\n REQUIRE(c[2] == 0xEDEE87E8);\n REQUIRE(c[3] == 0x00000000);\n\n copy_and_swap_32_aligned(c, d, 4);\n REQUIRE(c[0] == 0x67452301);\n REQUIRE(c[1] == 0xEFCDAB89);\n REQUIRE(c[2] == 0xEDEE87E8);\n REQUIRE(c[3] == 0x994151D8);\n\n alignas(32) uint64_t e;\n copy_and_swap_32_aligned(&e, d, 2);\n REQUIRE(e == 0xEFCDAB8967452301);\n\n alignas(32) char f[85] = {0x00};\n alignas(32) char g[] =\n \"This is a 85 byte long string... \"\n \"It's supposed to be longer than standard alignment.\";\n copy_and_swap_32_aligned(f, g, 21);\n REQUIRE(std::strcmp(f,\n \"sihT si 58 atyb ol es gnnirt...g'tI us ssoppt deeb \"\n \"onol regnahtats radnla dmngi.tne\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_32_aligned(f, g + 16, 17);\n REQUIRE(std::strcmp(f,\n \"ol es gnnirt...g'tI us ssoppt deeb \"\n \"onol regnahtats radnla dmngi.tne\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_32_aligned(f, g + 32, 13);\n REQUIRE(std::strcmp(f,\n \"'tI us ssoppt deeb \"\n \"onol regnahtats radnla dmngi.tne\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_32_aligned(f, g + 64, 5);\n REQUIRE(std::strcmp(f, \"ats radnla dmngi.tne\") == 0);\n}\n\nTEST_CASE(\"copy_and_swap_32_unaligned\", \"[copy_and_swap]\") {\n uint32_t a = 0x11111111, b = 0x89ABCDEF;\n copy_and_swap_32_unaligned(&a, &b, 1);\n REQUIRE(a == 0xEFCDAB89);\n REQUIRE(b == 0x89ABCDEF);\n\n uint32_t c[] = {0x00000000, 0x00000000, 0x00000000, 0x00000000};\n uint32_t d[] = {0x01234567, 0x89ABCDEF, 0xE887EEED, 0xD8514199};\n copy_and_swap_32_unaligned(c, d, 1);\n REQUIRE(c[0] == 0x67452301);\n REQUIRE(c[1] == 0x00000000);\n REQUIRE(c[2] == 0x00000000);\n REQUIRE(c[3] == 0x00000000);\n\n copy_and_swap_32_unaligned(c, d, 3);\n REQUIRE(c[0] == 0x67452301);\n REQUIRE(c[1] == 0xEFCDAB89);\n REQUIRE(c[2] == 0xEDEE87E8);\n REQUIRE(c[3] == 0x00000000);\n\n copy_and_swap_32_unaligned(c, d, 4);\n REQUIRE(c[0] == 0x67452301);\n REQUIRE(c[1] == 0xEFCDAB89);\n REQUIRE(c[2] == 0xEDEE87E8);\n REQUIRE(c[3] == 0x994151D8);\n\n {\n constexpr size_t count = 17;\n std::array src{};\n std::array dst{};\n for (size_t i = 0; i < src.size(); ++i) {\n src[i] = static_cast(i) + 1; // no zero in array\n }\n copy_and_swap_32_unaligned(dst.data(), src.data(), count);\n for (size_t i = 0; i < src.size(); i += 4) {\n // Check src is untouched\n REQUIRE(static_cast(src[i + 0]) == i + 1);\n REQUIRE(static_cast(src[i + 1]) == i + 2);\n REQUIRE(static_cast(src[i + 2]) == i + 3);\n REQUIRE(static_cast(src[i + 3]) == i + 4);\n // Check swapped bytes\n REQUIRE(static_cast(dst[i + 0]) ==\n static_cast(src[i + 3]));\n REQUIRE(static_cast(dst[i + 1]) ==\n static_cast(src[i + 2]));\n REQUIRE(static_cast(dst[i + 2]) ==\n static_cast(src[i + 1]));\n REQUIRE(static_cast(dst[i + 3]) ==\n static_cast(src[i + 0]));\n }\n }\n\n uint64_t e;\n copy_and_swap_32_unaligned(&e, d, 2);\n REQUIRE(e == 0xEFCDAB8967452301);\n\n char f[85] = {0x00};\n char g[] =\n \"This is a 85 byte long string... \"\n \"It's supposed to be longer than standard alignment.\";\n copy_and_swap_32_unaligned(f, g, 21);\n REQUIRE(std::strcmp(f,\n \"sihT si 58 atyb ol es gnnirt...g'tI us ssoppt deeb \"\n \"onol regnahtats radnla dmngi.tne\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_32_unaligned(f, g + 1, 20);\n REQUIRE(std::strcmp(f,\n \" siha si 58 etybnol ts ggnir ...s'tIpus esopot d eb \"\n \"gnolt re nahnatsdradila emng\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_32_unaligned(f, g + 2, 20);\n REQUIRE(std::strcmp(f,\n \"i si a sb 58 etygnolrts .gniI .. s'tppusdeso ot l \"\n \"ebegnoht rs nadnat dragilanemn\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_32_unaligned(f, g + 3, 20);\n REQUIRE(std::strcmp(f,\n \"si s8 a yb 5l et gnoirts..gntI .s s'oppu desb otol \"\n \"eregnaht ts nadnaa drngiltnem\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_32_unaligned(f, g + 4, 20);\n REQUIRE(std::strcmp(f,\n \" si 58 atyb ol es gnnirt...g'tI us ssoppt deeb onol \"\n \"regnahtats radnla dmngi.tne\") == 0);\n}\n\nTEST_CASE(\"copy_and_swap_64_aligned\", \"[copy_and_swap]\") {\n alignas(64) uint64_t a = 0x1111111111111111, b = 0x0123456789ABCDEF;\n copy_and_swap_64_aligned(&a, &b, 1);\n REQUIRE(a == 0xEFCDAB8967452301);\n REQUIRE(b == 0x0123456789ABCDEF);\n\n alignas(64) uint64_t c[] = {0x0000000000000000, 0x0000000000000000,\n 0x0000000000000000, 0x0000000000000000};\n alignas(64) uint64_t d[] = {0x0123456789ABCDEF, 0xE887EEEDD8514199,\n 0x21D4745A1D4A7706, 0xA4174FED675766E3};\n copy_and_swap_64_aligned(c, d, 1);\n REQUIRE(c[0] == 0xEFCDAB8967452301);\n REQUIRE(c[1] == 0x0000000000000000);\n REQUIRE(c[2] == 0x0000000000000000);\n REQUIRE(c[3] == 0x0000000000000000);\n\n copy_and_swap_64_aligned(c, d, 3);\n REQUIRE(c[0] == 0xEFCDAB8967452301);\n REQUIRE(c[1] == 0x994151D8EDEE87E8);\n REQUIRE(c[2] == 0x06774A1D5A74D421);\n REQUIRE(c[3] == 0x0000000000000000);\n\n copy_and_swap_64_aligned(c, d, 4);\n REQUIRE(c[0] == 0xEFCDAB8967452301);\n REQUIRE(c[1] == 0x994151D8EDEE87E8);\n REQUIRE(c[2] == 0x06774A1D5A74D421);\n REQUIRE(c[3] == 0xE3665767ED4F17A4);\n\n alignas(64) uint64_t e;\n copy_and_swap_64_aligned(&e, d, 1);\n REQUIRE(e == 0xEFCDAB8967452301);\n\n alignas(64) char f[85] = {0x00};\n alignas(64) char g[] =\n \"This is a 85 byte long string... \"\n \"It's supposed to be longer than standard alignment.\";\n copy_and_swap_64_aligned(f, g, 10);\n REQUIRE(std::strcmp(f,\n \" si sihTtyb 58 as gnol e...gnirtus s'tI t desoppnol eb \"\n \"onaht regradnats mngila d\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_aligned(f, g + 16, 8);\n REQUIRE(std::strcmp(f,\n \"s gnol e...gnirtus s'tI t desoppnol eb \"\n \"onaht regradnats mngila d\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_aligned(f, g + 32, 6);\n REQUIRE(std::strcmp(f,\n \"us s'tI t desoppnol eb \"\n \"onaht regradnats mngila d\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_aligned(f, g + 64, 2);\n REQUIRE(std::strcmp(f, \"radnats mngila d\") == 0);\n}\n\nTEST_CASE(\"copy_and_swap_64_unaligned\", \"[copy_and_swap]\") {\n uint64_t a = 0x1111111111111111, b = 0x0123456789ABCDEF;\n copy_and_swap_64_unaligned(&a, &b, 1);\n REQUIRE(a == 0xEFCDAB8967452301);\n REQUIRE(b == 0x0123456789ABCDEF);\n\n uint64_t c[] = {0x0000000000000000, 0x0000000000000000, 0x0000000000000000,\n 0x0000000000000000};\n uint64_t d[] = {0x0123456789ABCDEF, 0xE887EEEDD8514199, 0x21D4745A1D4A7706,\n 0xA4174FED675766E3};\n copy_and_swap_64_unaligned(c, d, 1);\n REQUIRE(c[0] == 0xEFCDAB8967452301);\n REQUIRE(c[1] == 0x0000000000000000);\n REQUIRE(c[2] == 0x0000000000000000);\n REQUIRE(c[3] == 0x0000000000000000);\n\n copy_and_swap_64_unaligned(c, d, 3);\n REQUIRE(c[0] == 0xEFCDAB8967452301);\n REQUIRE(c[1] == 0x994151D8EDEE87E8);\n REQUIRE(c[2] == 0x06774A1D5A74D421);\n REQUIRE(c[3] == 0x0000000000000000);\n\n copy_and_swap_64_unaligned(c, d, 4);\n REQUIRE(c[0] == 0xEFCDAB8967452301);\n REQUIRE(c[1] == 0x994151D8EDEE87E8);\n REQUIRE(c[2] == 0x06774A1D5A74D421);\n REQUIRE(c[3] == 0xE3665767ED4F17A4);\n\n uint64_t e;\n copy_and_swap_64_unaligned(&e, d, 1);\n REQUIRE(e == 0xEFCDAB8967452301);\n\n char f[85] = {0x00};\n char g[] =\n \"This is a 85 byte long string... \"\n \"It's supposed to be longer than standard alignment.\";\n copy_and_swap_64_unaligned(f, g, 10);\n REQUIRE(std::strcmp(f,\n \" si sihTtyb 58 as gnol e...gnirtus s'tI t desoppnol eb \"\n \"onaht regradnats mngila d\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_unaligned(f, g + 1, 10);\n REQUIRE(std::strcmp(f,\n \"a si sihetyb 58 ts gnol ...gnirpus s'tIot desopgnol \"\n \"eb naht redradnatsemngila \") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_unaligned(f, g + 2, 10);\n REQUIRE(std::strcmp(f,\n \" a si si etyb 58rts gnolI ...gnippus s't ot desoegnol \"\n \"ebs naht r dradnatnemngila\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_unaligned(f, g + 3, 10);\n REQUIRE(std::strcmp(f,\n \"8 a si sl etyb 5irts gnotI ...gnoppus s'b ot desregnol \"\n \"ets naht a dradnatnemngil\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_unaligned(f, g + 4, 10);\n REQUIRE(std::strcmp(f,\n \"58 a si ol etyb nirts gn'tI ...gsoppus seb ot de \"\n \"regnol ats nahtla dradn.tnemngi\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_unaligned(f, g + 5, 9);\n REQUIRE(std::strcmp(f,\n \" 58 a sinol etybgnirts gs'tI ...esoppus eb ot dt \"\n \"regnolnats nahila drad\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_unaligned(f, g + 6, 9);\n REQUIRE(std::strcmp(f,\n \"b 58 a sgnol ety.gnirts s'tI ..desoppusl eb ot ht \"\n \"regnodnats nagila dra\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_unaligned(f, g + 7, 9);\n REQUIRE(std::strcmp(f,\n \"yb 58 a gnol et..gnirtss s'tI . desoppuol eb otaht \"\n \"regnadnats nngila dr\") == 0);\n\n std::memset(f, 0, sizeof(f));\n copy_and_swap_64_unaligned(f, g + 8, 9);\n REQUIRE(std::strcmp(f,\n \"tyb 58 as gnol e...gnirtus s'tI t desoppnol eb onaht \"\n \"regradnats mngila d\") == 0);\n}\n\nTEST_CASE(\"copy_and_swap_16_in_32_aligned\", \"[copy_and_swap]\") {\n constexpr size_t count = 17;\n alignas(16) std::array src{};\n alignas(16) std::array dst{};\n\n // Check alignment (if this fails, adjust allocation)\n REQUIRE((reinterpret_cast(src.data()) & 0xF) == 0);\n REQUIRE((reinterpret_cast(dst.data()) & 0xF) == 0);\n\n for (size_t i = 0; i < src.size(); ++i) {\n src[i] = static_cast(i) + 1; // no zero in array\n }\n\n copy_and_swap_16_in_32_aligned(dst.data(), src.data(), count);\n\n for (size_t i = 0; i < src.size(); i += 4) {\n // Check src is untouched\n REQUIRE(static_cast(src[i + 0]) == i + 1);\n REQUIRE(static_cast(src[i + 1]) == i + 2);\n REQUIRE(static_cast(src[i + 2]) == i + 3);\n REQUIRE(static_cast(src[i + 3]) == i + 4);\n // Check swapped bytes\n REQUIRE(static_cast(dst[i + 0]) == static_cast(src[i + 2]));\n REQUIRE(static_cast(dst[i + 1]) == static_cast(src[i + 3]));\n REQUIRE(static_cast(dst[i + 2]) == static_cast(src[i + 0]));\n REQUIRE(static_cast(dst[i + 3]) == static_cast(src[i + 1]));\n }\n}\n\nTEST_CASE(\"copy_and_swap_16_in_32_unaligned\", \"[copy_and_swap]\") {\n constexpr size_t count = 17;\n std::array src{};\n std::array dst{};\n for (size_t i = 0; i < src.size(); ++i) {\n src[i] = static_cast(i) + 1; // no zero in array\n }\n\n copy_and_swap_16_in_32_unaligned(dst.data(), src.data(), count);\n\n for (size_t i = 0; i < src.size(); i += 4) {\n // Check src is untouched\n REQUIRE(static_cast(src[i + 0]) == i + 1);\n REQUIRE(static_cast(src[i + 1]) == i + 2);\n REQUIRE(static_cast(src[i + 2]) == i + 3);\n REQUIRE(static_cast(src[i + 3]) == i + 4);\n // Check swapped bytes\n REQUIRE(static_cast(dst[i + 0]) == static_cast(src[i + 2]));\n REQUIRE(static_cast(dst[i + 1]) == static_cast(src[i + 3]));\n REQUIRE(static_cast(dst[i + 2]) == static_cast(src[i + 0]));\n REQUIRE(static_cast(dst[i + 3]) == static_cast(src[i + 1]));\n }\n}\n\nTEST_CASE(\"create_and_close_file_mapping\", \"Virtual Memory Mapping\") {\n auto path = fmt::format(\"xenia_test_{}\", Clock::QueryHostTickCount());\n auto memory = xe::memory::CreateFileMappingHandle(\n path, 0x100, xe::memory::PageAccess::kReadWrite, true);\n REQUIRE(memory != xe::memory::kFileMappingHandleInvalid);\n xe::memory::CloseFileMappingHandle(memory, path);\n}\n\nTEST_CASE(\"map_view\", \"[virtual_memory_mapping]\") {\n auto path = fmt::format(\"xenia_test_{}\", Clock::QueryHostTickCount());\n const size_t length = 0x100;\n auto memory = xe::memory::CreateFileMappingHandle(\n path, length, xe::memory::PageAccess::kReadWrite, true);\n REQUIRE(memory != xe::memory::kFileMappingHandleInvalid);\n\n uintptr_t address = 0x100000000;\n auto view =\n xe::memory::MapFileView(memory, reinterpret_cast(address), length,\n xe::memory::PageAccess::kReadWrite, 0);\n REQUIRE(reinterpret_cast(view) == address);\n\n xe::memory::UnmapFileView(memory, reinterpret_cast(address), length);\n xe::memory::CloseFileMappingHandle(memory, path);\n}\n\nTEST_CASE(\"read_write_view\", \"[virtual_memory_mapping]\") {\n const size_t length = 0x100;\n auto path = fmt::format(\"xenia_test_{}\", Clock::QueryHostTickCount());\n auto memory = xe::memory::CreateFileMappingHandle(\n path, length, xe::memory::PageAccess::kReadWrite, true);\n REQUIRE(memory != xe::memory::kFileMappingHandleInvalid);\n\n uintptr_t address = 0x100000000;\n auto view =\n xe::memory::MapFileView(memory, reinterpret_cast(address), length,\n xe::memory::PageAccess::kReadWrite, 0);\n REQUIRE(reinterpret_cast(view) == address);\n\n for (uint32_t i = 0; i < length; i += sizeof(uint8_t)) {\n auto p_value = reinterpret_cast(address + i);\n *p_value = i;\n }\n for (uint32_t i = 0; i < length; i += sizeof(uint8_t)) {\n auto p_value = reinterpret_cast(address + i);\n uint8_t value = *p_value;\n REQUIRE(value == i);\n }\n\n xe::memory::UnmapFileView(memory, reinterpret_cast(address), length);\n xe::memory::CloseFileMappingHandle(memory, path);\n}\n\nTEST_CASE(\"make_fourcc\", \"[fourcc]\") {\n SECTION(\"'1234'\") {\n const uint32_t fourcc_host = 0x31323334;\n constexpr fourcc_t fourcc_1 = make_fourcc('1', '2', '3', '4');\n constexpr fourcc_t fourcc_2 = make_fourcc(\"1234\");\n REQUIRE(fourcc_1 == fourcc_host);\n REQUIRE(fourcc_2 == fourcc_host);\n REQUIRE(fourcc_1 == fourcc_2);\n REQUIRE(fourcc_2 == fourcc_1);\n }\n\n SECTION(\"'ABcd'\") {\n const uint32_t fourcc_host = 0x41426364;\n constexpr fourcc_t fourcc_1 = make_fourcc('A', 'B', 'c', 'd');\n constexpr fourcc_t fourcc_2 = make_fourcc(\"ABcd\");\n REQUIRE(fourcc_1 == fourcc_host);\n REQUIRE(fourcc_2 == fourcc_host);\n REQUIRE(fourcc_1 == fourcc_2);\n REQUIRE(fourcc_2 == fourcc_1);\n }\n\n SECTION(\"'XEN\\\\0'\") {\n const uint32_t fourcc_host = 0x58454E00;\n constexpr fourcc_t fourcc = make_fourcc('X', 'E', 'N', '\\0');\n REQUIRE(fourcc == fourcc_host);\n }\n\n SECTION(\"length()!=4\") {\n REQUIRE_THROWS(make_fourcc(\"AB\\0\\0\"));\n REQUIRE_THROWS(make_fourcc(\"AB\\0\\0AB\"));\n REQUIRE_THROWS(make_fourcc(\"ABCDEFGH\"));\n }\n}\n\n} // namespace test\n} // namespace base\n} // namespace xe\n" }, { "path": "src/xenia/base/testing/premake5.lua", "content": "project_root = \"../../../..\"\ninclude(project_root..\"/tools/build\")\n\ntest_suite(\"xenia-base-tests\", project_root, \".\", {\n links = {\n \"fmt\",\n \"xenia-base\",\n },\n})\n" }, { "path": "src/xenia/base/testing/threading_test.cc", "content": "/**\n******************************************************************************\n* Xenia : Xbox 360 Emulator Research Project *\n******************************************************************************\n* Copyright 2022 Ben Vanik. All rights reserved. *\n* Released under the BSD license - see LICENSE in the root for more details. *\n******************************************************************************\n*/\n\n#include \n\n#include \"xenia/base/threading.h\"\n\n#define CATCH_CONFIG_ENABLE_CHRONO_STRINGMAKER\n#include \"third_party/catch/include/catch.hpp\"\n\n#include \"third_party/disruptorplus/include/disruptorplus/spin_wait.hpp\"\n\nnamespace xe {\nnamespace base {\nnamespace test {\nusing namespace threading;\nusing namespace std::chrono_literals;\n\n// Helpers to wait on a predicate which do not depend on complex sync primitives\ntemplate \nbool spin_wait_until(\n const std::chrono::time_point& timeout_time,\n Predicate stop_waiting) {\n disruptorplus::spin_wait spinner;\n while (!stop_waiting()) {\n if (std::chrono::steady_clock::now() >= timeout_time) {\n return false;\n }\n spinner.spin_once();\n }\n return true;\n}\ntemplate \nbool spin_wait_for(const std::chrono::duration& rel_time,\n Predicate stop_waiting) {\n return spin_wait_until(std::chrono::steady_clock::now() + rel_time,\n stop_waiting);\n}\n\ntemplate \nvoid spin_wait(Predicate stop_waiting) {\n disruptorplus::spin_wait spinner;\n while (!stop_waiting()) {\n spinner.spin_once();\n }\n}\n\nTEST_CASE(\"Fence\") {\n std::unique_ptr pFence;\n std::unique_ptr pTimer;\n\n // Signal without wait\n pFence = std::make_unique();\n pFence->Signal();\n\n // Signal once and wait\n pFence = std::make_unique();\n pFence->Signal();\n pFence->Wait();\n\n // Signal twice and wait\n pFence = std::make_unique();\n pFence->Signal();\n pFence->Signal();\n pFence->Wait();\n\n // Signal and wait two times\n pFence = std::make_unique();\n pFence->Signal();\n pFence->Wait();\n pFence->Signal();\n pFence->Wait();\n\n // Test to synchronize multiple threads\n std::atomic started(0);\n std::atomic finished(0);\n pFence = std::make_unique();\n auto func = [&pFence, &started, &finished] {\n started.fetch_add(1);\n pFence->Wait();\n finished.fetch_add(1);\n };\n\n auto threads = std::array({\n std::thread(func),\n std::thread(func),\n std::thread(func),\n std::thread(func),\n std::thread(func),\n });\n\n REQUIRE(spin_wait_for(1s, [&] { return started == threads.size(); }));\n REQUIRE(finished.load() == 0);\n\n pFence->Signal();\n\n for (auto& t : threads) t.join();\n REQUIRE(finished.load() == threads.size());\n} // namespace test\n\nTEST_CASE(\"Get number of logical processors\") {\n auto count = std::thread::hardware_concurrency();\n REQUIRE(logical_processor_count() == count);\n REQUIRE(logical_processor_count() == count);\n REQUIRE(logical_processor_count() == count);\n}\n\nTEST_CASE(\"Enable process to set thread affinity\") {\n EnableAffinityConfiguration();\n}\n\nTEST_CASE(\"Yield Current Thread\", \"[maybe_yield]\") {\n // Run to see if there are any errors\n MaybeYield();\n}\n\nTEST_CASE(\"Sync with Memory Barrier\", \"[sync_memory]\") {\n // Run to see if there are any errors\n SyncMemory();\n}\n\nTEST_CASE(\"Sleep Current Thread\", \"[sleep]\") {\n auto wait_time = 50ms;\n auto start = std::chrono::steady_clock::now();\n Sleep(wait_time);\n auto duration = std::chrono::steady_clock::now() - start;\n REQUIRE(duration >= wait_time);\n}\n\nTEST_CASE(\"Sleep Current Thread in Alertable State\", \"[sleep]\") {\n auto wait_time = 50ms;\n auto start = std::chrono::steady_clock::now();\n auto result = threading::AlertableSleep(wait_time);\n auto duration = std::chrono::steady_clock::now() - start;\n REQUIRE(duration >= wait_time);\n REQUIRE(result == threading::SleepResult::kSuccess);\n\n // TODO(bwrsandman): Test a Thread to return kAlerted.\n // Need callback to call extended I/O function (ReadFileEx or WriteFileEx)\n}\n\nTEST_CASE(\"TlsHandle\") {\n // Test Allocate\n auto handle = threading::AllocateTlsHandle();\n\n // Test Free\n REQUIRE(threading::FreeTlsHandle(handle));\n REQUIRE(!threading::FreeTlsHandle(handle));\n REQUIRE(!threading::FreeTlsHandle(threading::kInvalidTlsHandle));\n\n // Test setting values\n handle = threading::AllocateTlsHandle();\n REQUIRE(threading::GetTlsValue(handle) == 0);\n uint32_t value = 0xDEADBEEF;\n threading::SetTlsValue(handle, reinterpret_cast(&value));\n auto p_received_value = threading::GetTlsValue(handle);\n REQUIRE(threading::GetTlsValue(handle) != 0);\n auto received_value = *reinterpret_cast(p_received_value);\n REQUIRE(received_value == value);\n\n uintptr_t non_thread_local_value = 0;\n auto thread = Thread::Create({}, [&non_thread_local_value, &handle] {\n non_thread_local_value = threading::GetTlsValue(handle);\n });\n REQUIRE(thread);\n\n auto result = Wait(thread.get(), false, 50ms);\n REQUIRE(result == WaitResult::kSuccess);\n REQUIRE(non_thread_local_value == 0);\n\n // Cleanup\n REQUIRE(threading::FreeTlsHandle(handle));\n}\n\nTEST_CASE(\"HighResolutionTimer\") {\n // The wait time is 500ms with an interval of 50ms\n // Smaller values are not as precise and fail the test\n const auto wait_time = 500ms;\n\n const Thread* timer_thread = nullptr;\n\n // Time the actual sleep duration\n {\n const auto interval = 50ms;\n std::atomic counter(0);\n auto start = std::chrono::steady_clock::now();\n auto cb = [&counter, &timer_thread] {\n if (counter == 0) {\n timer_thread = Thread::GetCurrentThread();\n } else {\n REQUIRE(Thread::GetCurrentThread() == timer_thread);\n }\n ++counter;\n };\n auto pTimer = HighResolutionTimer::CreateRepeating(interval, cb);\n Sleep(wait_time);\n pTimer.reset();\n auto duration = std::chrono::steady_clock::now() - start;\n\n // Should have run as many times as wait_time / timer_interval plus or\n // minus 1 due to imprecision of Sleep\n REQUIRE(duration.count() >= wait_time.count());\n auto ratio = static_cast(duration / interval);\n REQUIRE(counter >= ratio - 1);\n REQUIRE(counter <= ratio + 1);\n }\n\n // Test concurrent timers\n {\n const auto interval1 = 100ms;\n const auto interval2 = 200ms;\n std::atomic counter1(0);\n std::atomic counter2(0);\n auto start = std::chrono::steady_clock::now();\n auto cb1 = [&counter1, timer_thread] {\n ++counter1;\n REQUIRE(Thread::GetCurrentThread() == timer_thread);\n };\n auto cb2 = [&counter2, timer_thread] {\n ++counter2;\n REQUIRE(Thread::GetCurrentThread() == timer_thread);\n };\n auto pTimer1 = HighResolutionTimer::CreateRepeating(interval1, cb1);\n auto pTimer2 = HighResolutionTimer::CreateRepeating(interval2, cb2);\n Sleep(wait_time);\n pTimer1.reset();\n pTimer2.reset();\n auto duration = std::chrono::steady_clock::now() - start;\n\n // Should have run as many times as wait_time / timer_interval plus or\n // minus 1 due to imprecision of Sleep\n REQUIRE(duration.count() >= wait_time.count());\n auto ratio1 = static_cast(duration / interval1);\n auto ratio2 = static_cast(duration / interval2);\n REQUIRE(counter1 >= ratio1 - 1);\n REQUIRE(counter1 <= ratio1 + 1);\n REQUIRE(counter2 >= ratio2 - 1);\n REQUIRE(counter2 <= ratio2 + 1);\n }\n\n // Test many timers\n {\n const auto interval = 50ms;\n const size_t timer_count = 128;\n std::atomic counter(0);\n auto cb = [&counter, &timer_thread] {\n ++counter;\n REQUIRE(Thread::GetCurrentThread() == timer_thread);\n };\n std::vector> timers;\n auto start = std::chrono::steady_clock::now();\n for (size_t i = 0; i < timer_count; i++) {\n timers.emplace_back(HighResolutionTimer::CreateRepeating(interval, cb));\n }\n Sleep(wait_time);\n timers.clear();\n auto duration = std::chrono::steady_clock::now() - start;\n\n REQUIRE(duration.count() >= wait_time.count());\n auto ratio = static_cast(timer_count * duration / interval);\n REQUIRE(counter >= ratio - timer_count);\n REQUIRE(counter <= ratio + timer_count);\n }\n\n // Check timer order\n {\n constexpr size_t timer_count = 16;\n using pair_t = std::pair,\n std::chrono::high_resolution_clock::time_point>;\n std::array time_points{};\n auto start = std::chrono::steady_clock::now();\n auto gen_callback = [&timer_thread, &time_points](size_t i) {\n return [&timer_thread, &time_points, i]() {\n auto& pair = time_points[i];\n if (pair.first.fetch_add(1) == 1) {\n pair.second = std::chrono::high_resolution_clock::now();\n pair.first++;\n }\n REQUIRE(Thread::GetCurrentThread() == timer_thread);\n };\n };\n std::vector> timers;\n for (size_t i = 0; i < timer_count; i++) {\n timers.emplace_back(HighResolutionTimer::CreateRepeating(\n 10ms * (timer_count - i), gen_callback(timer_count - i - 1)));\n }\n REQUIRE(spin_wait_for(2s, [&] {\n return std::all_of(time_points.cbegin(), time_points.cend(),\n [](auto& pair) { return pair.first >= 3; });\n }));\n\n timers.clear();\n\n REQUIRE(std::is_sorted(\n time_points.cbegin(), time_points.cend(),\n [](auto& left, auto& right) { return left.second < right.second; }));\n }\n}\n\nTEST_CASE(\"Wait on Multiple Handles\", \"[wait]\") {\n auto mutant = Mutant::Create(true);\n REQUIRE(mutant);\n auto semaphore = Semaphore::Create(10, 10);\n REQUIRE(semaphore);\n auto event_ = Event::CreateManualResetEvent(false);\n REQUIRE(event_);\n auto thread = Thread::Create({}, [&mutant, &semaphore, &event_] {\n event_->Set();\n Wait(mutant.get(), false, 25ms);\n semaphore->Release(1, nullptr);\n Wait(mutant.get(), false, 25ms);\n mutant->Release();\n });\n\n std::vector handles = {\n mutant.get(),\n semaphore.get(),\n event_.get(),\n thread.get(),\n };\n\n auto any_result = WaitAny(handles, false, 100ms);\n REQUIRE(any_result.first == WaitResult::kSuccess);\n REQUIRE(any_result.second == 0);\n\n auto all_result = WaitAll(handles, false, 100ms);\n REQUIRE(all_result == WaitResult::kSuccess);\n}\n\nTEST_CASE(\"Signal and Wait\") {\n WaitResult result;\n auto mutant = Mutant::Create(true);\n REQUIRE(mutant);\n auto event_ = Event::CreateAutoResetEvent(false);\n REQUIRE(event_);\n auto thread = Thread::Create({}, [&mutant, &event_] {\n Wait(mutant.get(), false);\n event_->Set();\n });\n result = Wait(event_.get(), false, 50ms);\n REQUIRE(result == WaitResult::kTimeout);\n result = SignalAndWait(mutant.get(), event_.get(), false, 50ms);\n REQUIRE(result == WaitResult::kSuccess);\n result = Wait(thread.get(), false, 50ms);\n REQUIRE(result == WaitResult::kSuccess);\n}\n\nTEST_CASE(\"Wait on Event\", \"[event]\") {\n auto evt = Event::CreateAutoResetEvent(false);\n REQUIRE(evt);\n WaitResult result;\n\n // Call wait on unset Event\n result = Wait(evt.get(), false, 50ms);\n REQUIRE(result == WaitResult::kTimeout);\n\n // Call wait on set Event\n evt->Set();\n result = Wait(evt.get(), false, 50ms);\n REQUIRE(result == WaitResult::kSuccess);\n\n // Call wait on now consumed Event\n result = Wait(evt.get(), false, 50ms);\n REQUIRE(result == WaitResult::kTimeout);\n}\n\nTEST_CASE(\"Reset Event\", \"[event]\") {\n auto evt = Event::CreateAutoResetEvent(false);\n REQUIRE(evt);\n WaitResult result;\n\n // Call wait on reset Event\n evt->Set();\n evt->Reset();\n result = Wait(evt.get(), false, 50ms);\n REQUIRE(result == WaitResult::kTimeout);\n\n // Test resetting the unset event\n evt->Reset();\n result = Wait(evt.get(), false, 50ms);\n REQUIRE(result == WaitResult::kTimeout);\n\n // Test setting the reset event\n evt->Set();\n result = Wait(evt.get(), false, 50ms);\n REQUIRE(result == WaitResult::kSuccess);\n}\n\nTEST_CASE(\"Wait on Multiple Events\", \"[event]\") {\n auto events = std::array, 4>{\n Event::CreateAutoResetEvent(false),\n Event::CreateAutoResetEvent(false),\n Event::CreateAutoResetEvent(false),\n Event::CreateManualResetEvent(false),\n };\n for (auto& event : events) {\n REQUIRE(event.get() != nullptr);\n }\n\n std::atomic_uint threads_started(0);\n std::array order = {0};\n std::atomic_uint index(0);\n auto sign_in = [&order, &index](uint32_t id) {\n auto i = index.fetch_add(1, std::memory_order::memory_order_relaxed);\n order[i] = static_cast('0' + id);\n };\n\n auto threads = std::array{\n std::thread([&events, &sign_in, &threads_started] {\n set_name(\"1\");\n threads_started++;\n auto res = WaitAll({events[1].get(), events[3].get()}, false);\n REQUIRE(res == WaitResult::kSuccess);\n if (res == WaitResult::kSuccess) {\n sign_in(1);\n }\n }),\n std::thread([&events, &sign_in, &threads_started] {\n set_name(\"2\");\n threads_started++;\n auto res = WaitAny({events[0].get(), events[2].get()}, false);\n REQUIRE(res.first == WaitResult::kSuccess);\n if (res.first == WaitResult::kSuccess) {\n sign_in(2);\n }\n }),\n std::thread([&events, &sign_in, &threads_started] {\n set_name(\"3\");\n threads_started++;\n auto res =\n WaitAll({events[0].get(), events[2].get(), events[3].get()}, false);\n REQUIRE(res == WaitResult::kSuccess);\n if (res == WaitResult::kSuccess) {\n sign_in(3);\n }\n }),\n std::thread([&events, &sign_in, &threads_started] {\n set_name(\"4\");\n threads_started++;\n auto res = WaitAny({events[1].get(), events[3].get()}, false);\n REQUIRE(res.first == WaitResult::kSuccess);\n if (res.first == WaitResult::kSuccess) {\n sign_in(4);\n }\n }),\n };\n\n // wait for all threads starting up\n REQUIRE(spin_wait_for(1s, [&] { return threads_started == 4; }));\n events[3]->Set(); // Signals thread id=4 and stays on for 1 and 3\n REQUIRE(spin_wait_for(1s, [&] { return index == 1; }));\n events[1]->Set(); // Signals thread id=1\n REQUIRE(spin_wait_for(1s, [&] { return index == 2; }));\n events[0]->Set(); // Signals thread id=2\n REQUIRE(spin_wait_for(1s, [&] { return index == 3; }));\n events[2]->Set(); // Partial signals thread id=3\n events[0]->Set(); // Signals thread id=3\n\n for (auto& t : threads) {\n t.join();\n }\n\n REQUIRE(index == 4);\n\n INFO(order.data());\n REQUIRE(order[0] == '4');\n REQUIRE(order[1] == '1');\n REQUIRE(order[2] == '2');\n REQUIRE(order[3] == '3');\n}\n\nTEST_CASE(\"Wait on Semaphore\", \"[semaphore]\") {\n WaitResult result;\n std::unique_ptr sem;\n int previous_count = 0;\n\n // Wait on semaphore with no room\n sem = Semaphore::Create(0, 5);\n REQUIRE(sem);\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kTimeout);\n\n // Add room in semaphore\n REQUIRE(sem->Release(2, &previous_count));\n REQUIRE(previous_count == 0);\n REQUIRE(sem->Release(1, &previous_count));\n REQUIRE(previous_count == 2);\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kSuccess);\n REQUIRE(sem->Release(1, &previous_count));\n REQUIRE(previous_count == 2);\n\n // Set semaphore over maximum_count\n sem = Semaphore::Create(5, 5);\n REQUIRE(sem);\n previous_count = -1;\n REQUIRE_FALSE(sem->Release(1, &previous_count));\n REQUIRE(previous_count == -1);\n REQUIRE_FALSE(sem->Release(10, &previous_count));\n REQUIRE(previous_count == -1);\n sem = Semaphore::Create(0, 5);\n REQUIRE(sem);\n REQUIRE_FALSE(sem->Release(10, &previous_count));\n REQUIRE(previous_count == -1);\n REQUIRE_FALSE(sem->Release(10, &previous_count));\n REQUIRE(previous_count == -1);\n\n // Test invalid Release parameters\n REQUIRE_FALSE(sem->Release(0, &previous_count));\n REQUIRE(previous_count == -1);\n REQUIRE_FALSE(sem->Release(-1, &previous_count));\n REQUIRE(previous_count == -1);\n\n // Wait on fully available semaphore\n sem = Semaphore::Create(5, 5);\n REQUIRE(sem);\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kSuccess);\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kSuccess);\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kSuccess);\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kSuccess);\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kSuccess);\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kTimeout);\n\n // Semaphore between threads\n sem = Semaphore::Create(5, 5);\n REQUIRE(sem);\n // Occupy the semaphore with 5 threads\n std::atomic wait_count(0);\n std::atomic threads_terminate(false);\n auto func = [&sem, &wait_count, &threads_terminate] {\n auto res = Wait(sem.get(), false, 100ms);\n wait_count++;\n\n REQUIRE(spin_wait_for(2s, [&] { return threads_terminate.load(); }));\n\n REQUIRE(res == WaitResult::kSuccess);\n sem->Release(1, nullptr);\n };\n auto threads = std::array{\n std::thread(func), std::thread(func), std::thread(func),\n std::thread(func), std::thread(func),\n };\n // Wait for threads to finish semaphore calls\n REQUIRE(spin_wait_for(1s, [&] { return wait_count == 5; }));\n // Attempt to acquire full semaphore with current (6th) thread\n result = Wait(sem.get(), false, 20ms);\n REQUIRE(result == WaitResult::kTimeout);\n // Give threads time to release semaphore\n threads_terminate = true;\n for (auto& t : threads) {\n t.join();\n }\n result = Wait(sem.get(), false, 10ms);\n REQUIRE(result == WaitResult::kSuccess);\n sem->Release(1, &previous_count);\n REQUIRE(previous_count == 4);\n}\n\nTEST_CASE(\"Invalid semaphore parameters\", \"[semaphore]\") {\n std::unique_ptr sem;\n\n // Test invalid construction parameters\n // These are invalid according to documentation\n sem = Semaphore::Create(-1, 5);\n REQUIRE(sem == nullptr);\n sem = Semaphore::Create(10, 5);\n REQUIRE(sem == nullptr);\n sem = Semaphore::Create(0, 0);\n REQUIRE(sem == nullptr);\n sem = Semaphore::Create(0, -1);\n REQUIRE(sem == nullptr);\n sem = Semaphore::Create(-1, 0);\n REQUIRE(sem == nullptr);\n}\n\nTEST_CASE(\"Wait on Multiple Semaphores\", \"[semaphore]\") {\n WaitResult all_result;\n std::pair any_result;\n int previous_count;\n std::unique_ptr sem0, sem1;\n\n // Test Wait all which should fail\n sem0 = Semaphore::Create(0, 5);\n sem1 = Semaphore::Create(5, 5);\n REQUIRE(sem0);\n REQUIRE(sem1);\n all_result = WaitAll({sem0.get(), sem1.get()}, false, 10ms);\n REQUIRE(all_result == WaitResult::kTimeout);\n previous_count = -1;\n REQUIRE(sem0->Release(1, &previous_count));\n REQUIRE(previous_count == 0);\n previous_count = -1;\n REQUIRE_FALSE(sem1->Release(1, &previous_count));\n REQUIRE(previous_count == -1);\n\n // Test Wait all again which should succeed\n sem0 = Semaphore::Create(1, 5);\n sem1 = Semaphore::Create(5, 5);\n REQUIRE(sem0);\n REQUIRE(sem1);\n all_result = WaitAll({sem0.get(), sem1.get()}, false, 10ms);\n REQUIRE(all_result == WaitResult::kSuccess);\n previous_count = -1;\n REQUIRE(sem0->Release(1, &previous_count));\n REQUIRE(previous_count == 0);\n previous_count = -1;\n REQUIRE(sem1->Release(1, &previous_count));\n REQUIRE(previous_count == 4);\n\n // Test Wait Any which should fail\n sem0 = Semaphore::Create(0, 5);\n sem1 = Semaphore::Create(0, 5);\n REQUIRE(sem0);\n REQUIRE(sem1);\n any_result = WaitAny({sem0.get(), sem1.get()}, false, 10ms);\n REQUIRE(any_result.first == WaitResult::kTimeout);\n REQUIRE(any_result.second == 0);\n previous_count = -1;\n REQUIRE(sem0->Release(1, &previous_count));\n REQUIRE(previous_count == 0);\n previous_count = -1;\n REQUIRE(sem1->Release(1, &previous_count));\n REQUIRE(previous_count == 0);\n\n // Test Wait Any which should succeed\n sem0 = Semaphore::Create(0, 5);\n sem1 = Semaphore::Create(5, 5);\n REQUIRE(sem0);\n REQUIRE(sem1);\n any_result = WaitAny({sem0.get(), sem1.get()}, false, 10ms);\n REQUIRE(any_result.first == WaitResult::kSuccess);\n REQUIRE(any_result.second == 1);\n previous_count = -1;\n REQUIRE(sem0->Release(1, &previous_count));\n REQUIRE(previous_count == 0);\n previous_count = -1;\n REQUIRE(sem1->Release(1, &previous_count));\n REQUIRE(previous_count == 4);\n}\n\nTEST_CASE(\"Wait on Mutant\", \"[mutant]\") {\n WaitResult result;\n std::unique_ptr mut;\n\n // Release on initially owned mutant\n mut = Mutant::Create(true);\n REQUIRE(mut->Release());\n REQUIRE_FALSE(mut->Release());\n\n // Release on initially not-owned mutant\n mut = Mutant::Create(false);\n REQUIRE_FALSE(mut->Release());\n\n // Wait on initially owned mutant\n mut = Mutant::Create(true);\n result = Wait(mut.get(), false, 1ms);\n REQUIRE(result == WaitResult::kSuccess);\n REQUIRE(mut->Release());\n REQUIRE(mut->Release());\n REQUIRE_FALSE(mut->Release());\n\n // Wait on initially not owned mutant\n mut = Mutant::Create(false);\n result = Wait(mut.get(), false, 1ms);\n REQUIRE(result == WaitResult::kSuccess);\n REQUIRE(mut->Release());\n REQUIRE_FALSE(mut->Release());\n\n // Multiple waits (or locks)\n mut = Mutant::Create(false);\n for (int i = 0; i < 10; ++i) {\n result = Wait(mut.get(), false, 1ms);\n REQUIRE(result == WaitResult::kSuccess);\n }\n for (int i = 0; i < 10; ++i) {\n REQUIRE(mut->Release());\n }\n REQUIRE_FALSE(mut->Release());\n\n // Test mutants on other threads\n std::atomic step(0);\n auto thread1 = std::thread([&mut, &step] {\n mut = Mutant::Create(true);\n step++; // 1\n REQUIRE(spin_wait_for(2s, [&] { return step == 2; }));\n mut->Release();\n });\n REQUIRE(spin_wait_for(1s, [&] { return step == 1; }));\n REQUIRE_FALSE(mut->Release());\n result = Wait(mut.get(), false, 50ms);\n REQUIRE(result == WaitResult::kTimeout);\n step++; // 2\n thread1.join();\n result = Wait(mut.get(), false, 1ms);\n REQUIRE(result == WaitResult::kSuccess);\n REQUIRE(mut->Release());\n}\n\nTEST_CASE(\"Wait on Multiple Mutants\", \"[mutant]\") {\n WaitResult all_result;\n std::pair any_result;\n std::unique_ptr mut0, mut1;\n std::atomic step(0);\n\n // Test which should fail for WaitAll and WaitAny\n auto thread0 = std::thread([&mut0, &mut1, &step] {\n mut0 = Mutant::Create(true);\n mut1 = Mutant::Create(true);\n step++; // 1\n REQUIRE(spin_wait_for(2s, [&] { return step == 2; }));\n mut0->Release();\n mut1->Release();\n });\n REQUIRE(spin_wait_for(1s, [&] { return step == 1; }));\n all_result = WaitAll({mut0.get(), mut1.get()}, false, 10ms);\n REQUIRE(all_result == WaitResult::kTimeout);\n REQUIRE_FALSE(mut0->Release());\n REQUIRE_FALSE(mut1->Release());\n any_result = WaitAny({mut0.get(), mut1.get()}, false, 10ms);\n REQUIRE(any_result.first == WaitResult::kTimeout);\n REQUIRE(any_result.second == 0);\n REQUIRE_FALSE(mut0->Release());\n REQUIRE_FALSE(mut1->Release());\n step++; // 2\n thread0.join();\n\n // Test which should fail for WaitAll but not WaitAny\n step = 0;\n auto thread1 = std::thread([&mut0, &mut1, &step] {\n mut0 = Mutant::Create(true);\n mut1 = Mutant::Create(false);\n step++; // 1\n REQUIRE(spin_wait_for(2s, [&] { return step == 2; }));\n mut0->Release();\n });\n REQUIRE(spin_wait_for(1s, [&] { return step == 1; }));\n all_result = WaitAll({mut0.get(), mut1.get()}, false, 10ms);\n REQUIRE(all_result == WaitResult::kTimeout);\n REQUIRE_FALSE(mut0->Release());\n REQUIRE_FALSE(mut1->Release());\n any_result = WaitAny({mut0.get(), mut1.get()}, false, 10ms);\n REQUIRE(any_result.first == WaitResult::kSuccess);\n REQUIRE(any_result.second == 1);\n REQUIRE_FALSE(mut0->Release());\n REQUIRE(mut1->Release());\n step++; // 2\n thread1.join();\n\n // Test which should pass for WaitAll and WaitAny\n step = 0;\n auto thread2 = std::thread([&mut0, &mut1, &step] {\n mut0 = Mutant::Create(false);\n mut1 = Mutant::Create(false);\n step++; // 1\n REQUIRE(spin_wait_for(2s, [&] { return step == 2; }));\n });\n REQUIRE(spin_wait_for(1s, [&] { return step == 1; }));\n all_result = WaitAll({mut0.get(), mut1.get()}, false, 10ms);\n REQUIRE(all_result == WaitResult::kSuccess);\n REQUIRE(mut0->Release());\n REQUIRE(mut1->Release());\n any_result = WaitAny({mut0.get(), mut1.get()}, false, 10ms);\n REQUIRE(any_result.first == WaitResult::kSuccess);\n REQUIRE(any_result.second == 0);\n REQUIRE(mut0->Release());\n REQUIRE_FALSE(mut1->Release());\n step++; // 2\n thread2.join();\n}\n\nTEST_CASE(\"Wait on Timer\", \"[timer]\") {\n WaitResult result;\n std::unique_ptr timer;\n\n // Test Manual Reset\n timer = Timer::CreateManualResetTimer();\n REQUIRE(timer);\n result = Wait(timer.get(), false, 1ms);\n REQUIRE(result == WaitResult::kTimeout);\n REQUIRE(timer->SetOnceAfter(1ms)); // Signals it\n result = Wait(timer.get(), false, 20ms);\n REQUIRE(result == WaitResult::kSuccess);\n result = Wait(timer.get(), false, 1ms);\n REQUIRE(result == WaitResult::kSuccess); // Did not reset\n\n // Test Synchronization\n timer = Timer::CreateSynchronizationTimer();\n REQUIRE(timer);\n result = Wait(timer.get(), false, 1ms);\n REQUIRE(result == WaitResult::kTimeout);\n REQUIRE(timer->SetOnceAfter(1ms)); // Signals it\n result = Wait(timer.get(), false, 20ms);\n REQUIRE(result == WaitResult::kSuccess);\n result = Wait(timer.get(), false, 1ms);\n REQUIRE(result == WaitResult::kTimeout); // Did reset\n\n // Test long due time\n timer = Timer::CreateSynchronizationTimer();\n REQUIRE(timer->SetOnceAfter(10s));\n result = Wait(timer.get(), false, 10ms);\n REQUIRE(result == WaitResult::kTimeout);\n\n // Test Repeating\n REQUIRE(timer->SetRepeatingAfter(1ms, 10ms));\n for (int i = 0; i < 10; ++i) {\n result = Wait(timer.get(), false, 20ms);\n INFO(i);\n REQUIRE(result == WaitResult::kSuccess);\n }\n MaybeYield();\n Sleep(10ms); // Skip a few events\n for (int i = 0; i < 10; ++i) {\n result = Wait(timer.get(), false, 20ms);\n REQUIRE(result == WaitResult::kSuccess);\n }\n // Cancel it\n timer->Cancel();\n result = Wait(timer.get(), false, 20ms);\n REQUIRE(result == WaitResult::kTimeout);\n MaybeYield();\n Sleep(10ms); // Skip a few events\n result = Wait(timer.get(), false, 20ms);\n REQUIRE(result == WaitResult::kTimeout);\n // Cancel with SetOnce\n REQUIRE(timer->SetRepeatingAfter(1ms, 10ms));\n for (int i = 0; i < 10; ++i) {\n result = Wait(timer.get(), false, 20ms);\n REQUIRE(result == WaitResult::kSuccess);\n }\n REQUIRE(timer->SetOnceAfter(1ms));\n result = Wait(timer.get(), false, 20ms);\n REQUIRE(result == WaitResult::kSuccess); // Signal from Set Once\n result = Wait(timer.get(), false, 20ms);\n REQUIRE(result == WaitResult::kTimeout); // No more signals from repeating\n}\n\nTEST_CASE(\"Wait on Multiple Timers\", \"[timer]\") {\n WaitResult all_result;\n std::pair any_result;\n\n auto timer0 = Timer::CreateSynchronizationTimer();\n auto timer1 = Timer::CreateManualResetTimer();\n\n // None signaled\n all_result = WaitAll({timer0.get(), timer1.get()}, false, 1ms);\n REQUIRE(all_result == WaitResult::kTimeout);\n any_result = WaitAny({timer0.get(), timer1.get()}, false, 1ms);\n REQUIRE(any_result.first == WaitResult::kTimeout);\n REQUIRE(any_result.second == 0);\n\n // Some signaled\n REQUIRE(timer1->SetOnceAfter(1ms));\n all_result = WaitAll({timer0.get(), timer1.get()}, false, 100ms);\n REQUIRE(all_result == WaitResult::kTimeout);\n any_result = WaitAny({timer0.get(), timer1.get()}, false, 100ms);\n REQUIRE(any_result.first == WaitResult::kSuccess);\n REQUIRE(any_result.second == 1);\n\n // All signaled\n REQUIRE(timer0->SetOnceAfter(1ms));\n all_result = WaitAll({timer0.get(), timer1.get()}, false, 100ms);\n REQUIRE(all_result == WaitResult::kSuccess);\n REQUIRE(timer0->SetOnceAfter(1ms));\n Sleep(2ms);\n any_result = WaitAny({timer0.get(), timer1.get()}, false, 100ms);\n REQUIRE(any_result.first == WaitResult::kSuccess);\n REQUIRE(any_result.second == 0);\n\n // Check that timer0 reset\n any_result = WaitAny({timer0.get(), timer1.get()}, false, 100ms);\n REQUIRE(any_result.first == WaitResult::kSuccess);\n REQUIRE(any_result.second == 1);\n}\n\nTEST_CASE(\"Create and Trigger Timer Callbacks\", \"[timer]\") {\n // TODO(bwrsandman): Check which thread performs callback and timing of\n // callback\n REQUIRE(true);\n}\n\nTEST_CASE(\"Set and Test Current Thread ID\", \"[thread]\") {\n // System ID\n auto system_id = current_thread_system_id();\n REQUIRE(system_id > 0);\n\n // Thread ID\n auto thread_id = current_thread_id();\n REQUIRE(thread_id == system_id);\n\n // Set a new thread id\n const uint32_t new_thread_id = 0xDEADBEEF;\n set_current_thread_id(new_thread_id);\n REQUIRE(current_thread_id() == new_thread_id);\n\n // Set back original thread id of system\n set_current_thread_id(std::numeric_limits::max());\n REQUIRE(current_thread_id() == system_id);\n\n // TODO(bwrsandman): Test on Thread object\n}\n\nTEST_CASE(\"Set and Test Current Thread Name\", \"[thread]\") {\n auto current_thread = Thread::GetCurrentThread();\n REQUIRE(current_thread);\n auto old_thread_name = current_thread->name();\n\n std::string new_thread_name = \"Threading Test\";\n REQUIRE_NOTHROW(set_name(new_thread_name));\n\n // Restore the old catch.hpp thread name\n REQUIRE_NOTHROW(set_name(old_thread_name));\n}\n\nTEST_CASE(\"Create and Run Thread\", \"[thread]\") {\n std::unique_ptr thread;\n WaitResult result;\n Thread::CreationParameters params = {};\n std::atomic fence(0);\n auto func = [&fence] {\n REQUIRE(spin_wait_for(1s, [&] { return fence == 1; }));\n fence++;\n };\n\n SECTION(\"Create most basic case of thread\") {\n fence = 0;\n thread = Thread::Create(params, func);\n REQUIRE(thread->native_handle() != nullptr);\n REQUIRE_NOTHROW(thread->affinity_mask());\n REQUIRE(thread->name().empty());\n fence++;\n result = Wait(thread.get(), false, 1s);\n REQUIRE(result == WaitResult::kSuccess);\n }\n\n SECTION(\"Add thread name\") {\n fence = 0;\n std::string new_name = \"Test thread name\";\n thread = Thread::Create(params, func);\n auto name = thread->name();\n INFO(name.c_str());\n REQUIRE(name.empty());\n thread->set_name(new_name);\n REQUIRE(thread->name() == new_name);\n fence++;\n result = Wait(thread.get(), false, 1s);\n REQUIRE(result == WaitResult::kSuccess);\n }\n\n SECTION(\"Use Terminate to end an infinitely looping thread\") {\n thread = Thread::Create(params, [] {\n while (true) {\n Sleep(1ms);\n }\n });\n result = Wait(thread.get(), false, 1s);\n REQUIRE(result == WaitResult::kTimeout);\n thread->Terminate(-1);\n result = Wait(thread.get(), false, 1s);\n REQUIRE(result == WaitResult::kSuccess);\n }\n\n SECTION(\"Call Exit from inside an infinitely looping thread\") {\n thread = Thread::Create(params, [] {\n Thread::Exit(-1);\n FAIL(\"Function must not return\");\n while (true)\n ;\n });\n result = Wait(thread.get(), false, 1s);\n REQUIRE(result == WaitResult::kSuccess);\n }\n\n SECTION(\"Call timeout wait on self\") {\n result = Wait(Thread::GetCurrentThread(), false, 50ms);\n REQUIRE(result == WaitResult::kTimeout);\n }\n\n SECTION(\"16Mb stack size\") {\n params.stack_size = 16_MiB;\n thread = Thread::Create(params, [] {\n Thread::Exit(-1);\n FAIL(\"Function must not return\");\n while (true)\n ;\n });\n REQUIRE(thread != nullptr);\n result = Wait(thread.get(), false, 1s);\n REQUIRE(result == WaitResult::kSuccess);\n }\n\n // TODO(bwrsandman): Test with different priorities\n // TODO(bwrsandman): Test setting and getting thread affinity\n}\n\nTEST_CASE(\"Test Suspending Thread\", \"[thread]\") {\n std::unique_ptr thread;\n WaitResult result;\n Thread::CreationParameters params = {};\n auto func = [] { Sleep(20ms); };\n\n // Create initially suspended\n params.create_suspended = true;\n thread = threading::Thread::Create(params, func);\n result = threading::Wait(thread.get(), false, 50ms);\n REQUIRE(result == threading::WaitResult::kTimeout);\n thread->Resume();\n result = threading::Wait(thread.get(), false, 50ms);\n REQUIRE(result == threading::WaitResult::kSuccess);\n params.create_suspended = false;\n\n // Create and then suspend\n thread = threading::Thread::Create(params, func);\n thread->Suspend();\n result = threading::Wait(thread.get(), false, 50ms);\n REQUIRE(result == threading::WaitResult::kTimeout);\n thread->Resume();\n result = threading::Wait(thread.get(), false, 50ms);\n REQUIRE(result == threading::WaitResult::kSuccess);\n\n // Test recursive suspend\n thread = threading::Thread::Create(params, func);\n thread->Suspend();\n thread->Suspend();\n result = threading::Wait(thread.get(), false, 50ms);\n REQUIRE(result == threading::WaitResult::kTimeout);\n thread->Resume();\n result = threading::Wait(thread.get(), false, 50ms);\n REQUIRE(result == threading::WaitResult::kTimeout);\n thread->Resume();\n result = threading::Wait(thread.get(), false, 50ms);\n REQUIRE(result == threading::WaitResult::kSuccess);\n\n // Test suspend count\n uint32_t suspend_count = 0;\n thread = threading::Thread::Create(params, func);\n thread->Suspend(&suspend_count);\n REQUIRE(suspend_count == 0);\n thread->Suspend(&suspend_count);\n REQUIRE(suspend_count == 1);\n thread->Suspend(&suspend_count);\n REQUIRE(suspend_count == 2);\n thread->Resume(&suspend_count);\n REQUIRE(suspend_count == 3);\n thread->Resume(&suspend_count);\n REQUIRE(suspend_count == 2);\n thread->Resume(&suspend_count);\n REQUIRE(suspend_count == 1);\n thread->Suspend(&suspend_count);\n REQUIRE(suspend_count == 0);\n thread->Resume(&suspend_count);\n REQUIRE(suspend_count == 1);\n result = threading::Wait(thread.get(), false, 50ms);\n REQUIRE(result == threading::WaitResult::kSuccess);\n}\n\nTEST_CASE(\"Test Thread QueueUserCallback\", \"[thread]\") {\n std::unique_ptr thread;\n WaitResult result;\n Thread::CreationParameters params = {};\n std::atomic_int order;\n int is_modified;\n int has_finished;\n auto callback = [&is_modified, &order] {\n is_modified = std::atomic_fetch_add_explicit(\n &order, 1, std::memory_order::memory_order_relaxed);\n };\n\n // Without alertable\n order = 0;\n is_modified = -1;\n has_finished = -1;\n thread = Thread::Create(params, [&has_finished, &order] {\n // Not using Alertable so callback is not registered\n order++; // 1\n Sleep(90ms);\n order++; // 2\n has_finished = std::atomic_fetch_add_explicit(\n &order, 1, std::memory_order::memory_order_relaxed);\n });\n REQUIRE(!spin_wait_for(50ms, [&] { return order == 2; }));\n REQUIRE(is_modified == -1);\n thread->QueueUserCallback(callback);\n result = Wait(thread.get(), true, 200ms);\n REQUIRE(result == WaitResult::kSuccess);\n REQUIRE(is_modified == -1);\n REQUIRE(has_finished == 2);\n\n // With alertable\n order = 0;\n is_modified = -1;\n has_finished = -1;\n thread = Thread::Create(params, [&has_finished, &order] {\n // Using Alertable so callback is registered\n order++; // 1\n AlertableSleep(90ms);\n order++; // 3\n has_finished = std::atomic_fetch_add_explicit(\n &order, 1, std::memory_order::memory_order_relaxed);\n });\n REQUIRE(!spin_wait_for(50ms, [&] { return order == 2; }));\n REQUIRE(is_modified == -1);\n thread->QueueUserCallback(callback);\n result = Wait(thread.get(), true, 200ms);\n REQUIRE(result == WaitResult::kSuccess);\n REQUIRE(is_modified == 1);\n REQUIRE(has_finished == 3);\n\n // Test Exit command with QueueUserCallback\n order = 0;\n is_modified = -1;\n has_finished = -1;\n thread = Thread::Create(params, [&is_modified, &has_finished, &order] {\n is_modified = std::atomic_fetch_add_explicit(\n &order, 1, std::memory_order::memory_order_relaxed);\n // Using Alertable so callback is registered\n order++; // 2\n AlertableSleep(1s);\n FAIL(\"Thread should have been terminated during alertable sleep\");\n while (true)\n ;\n });\n REQUIRE(!spin_wait_for(100ms, [&] { return order == 3; })); // timeout\n thread->QueueUserCallback([] { Thread::Exit(0); });\n result = Wait(thread.get(), true, 500ms);\n REQUIRE(result == WaitResult::kSuccess);\n REQUIRE(is_modified == 0);\n REQUIRE(order == 2);\n\n // TODO(bwrsandman): Test alertable wait returning kUserCallback by using IO\n // callbacks.\n}\n\n} // namespace test\n} // namespace base\n} // namespace xe\n" }, { "path": "src/xenia/base/testing/utf8_test.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n#include \"xenia/base/string.h\"\n\n#include \"third_party/catch/include/catch.hpp\"\n\nnamespace xe::base::test {\n\n// TODO(gibbed): bit messy?\n// TODO(gibbed): predicate variant?\n\n#define TEST_EXAMPLE(func, left, right) REQUIRE(func(left) == right)\n\n#define TEST_EXAMPLES_1(func, language, results) \\\n TEST_EXAMPLE(func, examples::k##language##Values[0], results.language[0])\n#define TEST_EXAMPLES_2(func, language, results) \\\n TEST_EXAMPLE(func, examples::k##language##Values[0], results.language[0]); \\\n TEST_EXAMPLE(func, examples::k##language##Values[1], results.language[1])\n#define TEST_EXAMPLES_3(func, language, results) \\\n TEST_EXAMPLE(func, examples::k##language##Values[0], results.language[0]); \\\n TEST_EXAMPLE(func, examples::k##language##Values[1], results.language[1]); \\\n TEST_EXAMPLE(func, examples::k##language##Values[2], results.language[2])\n\nnamespace examples {\n\n// https://www.cl.cam.ac.uk/~mgk25/ucs/examples/quickbrown.txt\n\nconst size_t kDanishCount = 1;\nconst char* kDanishValues[kDanishCount] = {\n u8\"Quizdeltagerne spiste jordbær med fløde, mens cirkusklovnen Wolther \"\n u8\"spillede på xylofon.\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Danish(func, results) \\\n TEST_EXAMPLES_1(func, Danish, results)\n\nconst size_t kGermanCount = 3;\nconst char* kGermanValues[kGermanCount] = {\n u8\"Falsches Üben von Xylophonmusik quält jeden größeren Zwerg\",\n u8\"Zwölf Boxkämpfer jagten Eva quer über den Sylter Deich\",\n u8\"Heizölrückstoßabdämpfung\",\n};\n#define TEST_LANGUAGE_EXAMPLES_German(func, results) \\\n TEST_EXAMPLES_2(func, German, results)\n\nconst size_t kGreekCount = 2;\nconst char* kGreekValues[kGreekCount] = {\n u8\"Γαζέες καὶ μυρτιὲς δὲν θὰ βρῶ πιὰ στὸ χρυσαφὶ ξέφωτο\",\n u8\"Ξεσκεπάζω τὴν ψυχοφθόρα βδελυγμία\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Greek(func, results) \\\n TEST_EXAMPLES_2(func, Greek, results)\n\nconst size_t kEnglishCount = 1;\nconst char* kEnglishValues[kEnglishCount] = {\n u8\"The quick brown fox jumps over the lazy dog\",\n};\n#define TEST_LANGUAGE_EXAMPLES_English(func, results) \\\n TEST_EXAMPLES_1(func, English, results)\n\nconst size_t kSpanishCount = 1;\nconst char* kSpanishValues[kSpanishCount] = {\n u8\"El pingüino Wenceslao hizo kilómetros bajo exhaustiva lluvia y frío, \"\n u8\"añoraba a su querido cachorro.\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Spanish(func, results) \\\n TEST_EXAMPLES_1(func, Spanish, results)\n\nconst size_t kFrenchCount = 3;\nconst char* kFrenchValues[kFrenchCount] = {\n u8\"Portez ce vieux whisky au juge blond qui fume sur son île intérieure, à \"\n u8\"côté de l'alcôve ovoïde, où les bûches se consument dans l'âtre, ce qui \"\n u8\"lui permet de penser à la cænogenèse de l'être dont il est question \"\n u8\"dans la cause ambiguë entendue à Moÿ, dans un capharnaüm qui, \"\n u8\"pense-t-il, diminue çà et là la qualité de son œuvre.\",\n u8\"l'île exiguë\\n\"\n u8\"Où l'obèse jury mûr\\n\"\n u8\"Fête l'haï volapük,\\n\"\n u8\"Âne ex aéquo au whist,\\n\"\n u8\"Ôtez ce vœu déçu.\",\n u8\"Le cœur déçu mais l'âme plutôt naïve, Louÿs rêva de crapaüter en canoë \"\n u8\"au delà des îles, près du mälström où brûlent les novæ.\",\n};\n#define TEST_LANGUAGE_EXAMPLES_French(func, results) \\\n TEST_EXAMPLES_3(func, French, results)\n\nconst size_t kIrishGaelicCount = 1;\nconst char* kIrishGaelicValues[kIrishGaelicCount] = {\n u8\"D'fhuascail Íosa, Úrmhac na hÓighe Beannaithe, pór Éava agus Ádhaimh\",\n};\n#define TEST_LANGUAGE_EXAMPLES_IrishGaelic(func, results) \\\n TEST_EXAMPLES_1(func, IrishGaelic, results)\n\nconst size_t kHungarianCount = 1;\nconst char* kHungarianValues[kHungarianCount] = {\n u8\"Árvíztűrő tükörfúrógép\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Hungarian(func, results) \\\n TEST_EXAMPLES_1(func, Hungarian, results)\n\nconst size_t kIcelandicCount = 2;\nconst char* kIcelandicValues[kIcelandicCount] = {\n u8\"Kæmi ný öxi hér ykist þjófum nú bæði víl og ádrepa\",\n u8\"Sævör grét áðan því úlpan var ónýt\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Icelandic(func, results) \\\n TEST_EXAMPLES_2(func, Icelandic, results)\n\nconst size_t kJapaneseCount = 2;\nconst char* kJapaneseValues[kJapaneseCount] = {\n u8\"いろはにほへとちりぬるを\\n\"\n u8\"わかよたれそつねならむ\\n\"\n u8\"うゐのおくやまけふこえて\\n\"\n u8\"あさきゆめみしゑひもせす\\n\",\n u8\"イロハニホヘト チリヌルヲ ワカヨタレソ ツネナラム\\n\"\n u8\"ウヰノオクヤマ ケフコエテ アサキユメミシ ヱヒモセスン\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Japanese(func, results) \\\n TEST_EXAMPLES_2(func, Japanese, results)\n\nconst size_t kHebrewCount = 1;\nconst char* kHebrewValues[kHebrewCount] = {\n u8\"? דג סקרן שט בים מאוכזב ולפתע מצא לו חברה איך הקליטה\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Hebrew(func, results) \\\n TEST_EXAMPLES_1(func, Hebrew, results)\n\nconst size_t kPolishCount = 1;\nconst char* kPolishValues[kPolishCount] = {\n u8\"Pchnąć w tę łódź jeża lub ośm skrzyń fig\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Polish(func, results) \\\n TEST_EXAMPLES_1(func, Polish, results)\n\nconst size_t kRussianCount = 2;\nconst char* kRussianValues[kRussianCount] = {\n u8\"В чащах юга жил бы цитрус? Да, но фальшивый экземпляр!\",\n u8\"Съешь же ещё этих мягких французских булок да выпей чаю\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Russian(func, results) \\\n TEST_EXAMPLES_2(func, Russian, results)\n\nconst size_t kTurkishCount = 1;\nconst char* kTurkishValues[kTurkishCount] = {\n u8\"Pijamalı hasta, yağız şoföre çabucak güvendi.\",\n};\n#define TEST_LANGUAGE_EXAMPLES_Turkish(func, results) \\\n TEST_EXAMPLES_1(func, Turkish, results)\n\n#define TEST_LANGUAGE_EXAMPLES(func, results) \\\n TEST_LANGUAGE_EXAMPLES_Danish(func, results); \\\n TEST_LANGUAGE_EXAMPLES_German(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Greek(func, results); \\\n TEST_LANGUAGE_EXAMPLES_English(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Spanish(func, results); \\\n TEST_LANGUAGE_EXAMPLES_French(func, results); \\\n TEST_LANGUAGE_EXAMPLES_IrishGaelic(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Hungarian(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Icelandic(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Japanese(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Hebrew(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Polish(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Russian(func, results); \\\n TEST_LANGUAGE_EXAMPLES_Turkish(func, results)\n\n} // namespace examples\n\n#define TEST_EXAMPLE_RESULT(language) T language[examples::k##language##Count]\ntemplate \nstruct example_results {\n TEST_EXAMPLE_RESULT(Danish);\n TEST_EXAMPLE_RESULT(German);\n TEST_EXAMPLE_RESULT(Greek);\n TEST_EXAMPLE_RESULT(English);\n TEST_EXAMPLE_RESULT(Spanish);\n TEST_EXAMPLE_RESULT(French);\n TEST_EXAMPLE_RESULT(IrishGaelic);\n TEST_EXAMPLE_RESULT(Hungarian);\n TEST_EXAMPLE_RESULT(Icelandic);\n TEST_EXAMPLE_RESULT(Japanese);\n TEST_EXAMPLE_RESULT(Hebrew);\n TEST_EXAMPLE_RESULT(Polish);\n TEST_EXAMPLE_RESULT(Russian);\n TEST_EXAMPLE_RESULT(Turkish);\n};\n#undef TEST_EXAMPLE_RESULT\n\nTEST_CASE(\"UTF-8 Count\", \"[utf8]\") {\n example_results results = {};\n results.Danish[0] = 88;\n results.German[0] = 58;\n results.German[1] = 54;\n results.Greek[0] = 52;\n results.Greek[1] = 33;\n results.English[0] = 43;\n results.Spanish[0] = 99;\n results.French[0] = 327;\n results.French[1] = 93;\n results.French[2] = 126;\n results.IrishGaelic[0] = 68;\n results.Hungarian[0] = 22;\n results.Icelandic[0] = 50;\n results.Icelandic[1] = 34;\n results.Japanese[0] = 51;\n results.Japanese[1] = 55;\n results.Hebrew[0] = 52;\n results.Polish[0] = 40;\n results.Russian[0] = 54;\n results.Russian[1] = 55;\n results.Turkish[0] = 45;\n TEST_LANGUAGE_EXAMPLES(utf8::count, results);\n}\n\n// TODO(gibbed): lower_ascii\n// TODO(gibbed): upper_ascii\n// TODO(gibbed): hash_fnv1a\n// TODO(gibbed): hash_fnv1a_case\n\nTEST_CASE(\"UTF-8 Split\", \"[utf8]\") {\n std::vector parts;\n\n // Danish\n parts = utf8::split(examples::kDanishValues[0], u8\"æcå\");\n REQUIRE(parts.size() == 4);\n REQUIRE(parts[0] == u8\"Quizdeltagerne spiste jordb\");\n REQUIRE(parts[1] == u8\"r med fløde, mens \");\n REQUIRE(parts[2] == u8\"irkusklovnen Wolther spillede p\");\n REQUIRE(parts[3] == u8\" xylofon.\");\n\n // German\n parts = utf8::split(examples::kGermanValues[0], u8\"ßS\");\n REQUIRE(parts.size() == 2);\n REQUIRE(parts[0] == u8\"Falsches Üben von Xylophonmusik quält jeden grö\");\n REQUIRE(parts[1] == u8\"eren Zwerg\");\n parts = utf8::split(examples::kGermanValues[1], u8\"ßS\");\n REQUIRE(parts.size() == 2);\n REQUIRE(parts[0] == u8\"Zwölf Boxkämpfer jagten Eva quer über den \");\n REQUIRE(parts[1] == u8\"ylter Deich\");\n parts = utf8::split(examples::kGermanValues[2], u8\"ßS\");\n REQUIRE(parts.size() == 2);\n REQUIRE(parts[0] == u8\"Heizölrücksto\");\n REQUIRE(parts[1] == u8\"abdämpfung\");\n\n // Greek\n parts = utf8::split(examples::kGreekValues[0], u8\"πφ\");\n REQUIRE(parts.size() == 4);\n REQUIRE(parts[0] == u8\"Γαζέες καὶ μυρτιὲς δὲν θὰ βρῶ \");\n REQUIRE(parts[1] == u8\"ιὰ στὸ χρυσα\");\n REQUIRE(parts[2] == u8\"ὶ ξέ\");\n REQUIRE(parts[3] == u8\"ωτο\");\n parts = utf8::split(examples::kGreekValues[1], u8\"πφ\");\n REQUIRE(parts.size() == 3);\n REQUIRE(parts[0] == u8\"Ξεσκε\");\n REQUIRE(parts[1] == u8\"άζω τὴν ψυχο\");\n REQUIRE(parts[2] == u8\"θόρα βδελυγμία\");\n\n // English\n parts = utf8::split(examples::kEnglishValues[0], \"xy\");\n REQUIRE(parts.size() == 3);\n REQUIRE(parts[0] == u8\"The quick brown fo\");\n REQUIRE(parts[1] == u8\" jumps over the laz\");\n REQUIRE(parts[2] == u8\" dog\");\n\n // Spanish\n parts = utf8::split(examples::kSpanishValues[0], u8\"ójd\");\n REQUIRE(parts.size() == 4);\n REQUIRE(parts[0] == u8\"El pingüino Wenceslao hizo kil\");\n REQUIRE(parts[1] == u8\"metros ba\");\n REQUIRE(parts[2] == u8\"o exhaustiva lluvia y frío, añoraba a su queri\");\n REQUIRE(parts[3] == u8\"o cachorro.\");\n\n // TODO(gibbed): French\n // TODO(gibbed): Irish Gaelic\n // TODO(gibbed): Hungarian\n // TODO(gibbed): Icelandic\n // TODO(gibbed): Japanese\n // TODO(gibbed): Hebrew\n // TODO(gibbed): Polish\n // TODO(gibbed): Russian\n // TODO(gibbed): Thai\n // TODO(gibbed): Turkish\n}\n\nTEST_CASE(\"UTF-8 Equal Z\", \"[utf8]\") {\n REQUIRE(utf8::equal_z(u8\"foo\", u8\"foo\\0\"));\n REQUIRE_FALSE(utf8::equal_z(u8\"bar\", u8\"baz\\0\"));\n}\n\nTEST_CASE(\"UTF-8 Equal Case\", \"[utf8]\") {\n REQUIRE(utf8::equal_case(u8\"foo\", u8\"foo\\0\"));\n REQUIRE_FALSE(utf8::equal_case(u8\"bar\", u8\"baz\\0\"));\n}\n\nTEST_CASE(\"UTF-8 Equal Case Z\", \"[utf8]\") {\n REQUIRE(utf8::equal_case_z(u8\"foo\", u8\"foo\\0\"));\n REQUIRE_FALSE(utf8::equal_case_z(u8\"bar\", u8\"baz\\0\"));\n}\n\n// TODO(gibbed): find_any_of\n// TODO(gibbed): find_any_of_case\n// TODO(gibbed): find_first_of\n// TODO(gibbed): find_first_of_case\n// TODO(gibbed): starts_with\n// TODO(gibbed): starts_with_case\n// TODO(gibbed): ends_with\n// TODO(gibbed): ends_with_case\n// TODO(gibbed): split_path\n\n#define TEST_PATH(func, input, output) \\\n do { \\\n std::string input_value = input; \\\n std::string output_value = output; \\\n REQUIRE(func(input_value, '/') == output_value); \\\n std::replace(input_value.begin(), input_value.end(), '/', '\\\\'); \\\n std::replace(output_value.begin(), output_value.end(), '/', '\\\\'); \\\n REQUIRE(func(input_value, '\\\\') == output_value); \\\n } while (0)\n\n#define TEST_PATH_RAW(func, input, output) \\\n do { \\\n std::string output_value = output; \\\n REQUIRE(func(input, '/') == output_value); \\\n std::replace(output_value.begin(), output_value.end(), '/', '\\\\'); \\\n REQUIRE(func(input, '\\\\') == output_value); \\\n } while (0)\n\n#define TEST_PATHS(func, output, ...) \\\n do { \\\n std::vector input_values = {__VA_ARGS__}; \\\n std::string output_value = output; \\\n REQUIRE(func(input_values, '/') == output_value); \\\n for (auto it = input_values.begin(); it != input_values.end(); ++it) { \\\n std::replace((*it).begin(), (*it).end(), '/', '\\\\'); \\\n } \\\n std::replace(output_value.begin(), output_value.end(), '/', '\\\\'); \\\n REQUIRE(func(input_values, '\\\\') == output_value); \\\n } while (0)\n\nTEST_CASE(\"UTF-8 Join Paths\", \"[utf8]\") {\n TEST_PATHS(utf8::join_paths, u8\"\");\n TEST_PATHS(utf8::join_paths, u8\"foo\", u8\"foo\");\n TEST_PATHS(utf8::join_paths, u8\"foo/bar\", u8\"foo\", u8\"bar\");\n TEST_PATHS(utf8::join_paths, \"X:/foo/bar/baz/qux\", u8\"X:\", u8\"foo\", u8\"bar\",\n u8\"baz\", u8\"qux\");\n}\n\n// TODO(gibbed): join_guest_paths\n\nTEST_CASE(\"UTF-8 Fix Path Separators\", \"[utf8]\") {\n TEST_PATH_RAW(utf8::fix_path_separators, \"\", \"\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"\\\\\", \"/\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"/\", \"/\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"\\\\foo\", \"/foo\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"\\\\foo/\", \"/foo/\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"/foo\", \"/foo\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"\\\\foo/bar\\\\baz/qux\",\n \"/foo/bar/baz/qux\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"\\\\\\\\foo//bar\\\\\\\\baz//qux\",\n \"/foo/bar/baz/qux\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"foo\", \"foo\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"foo/\", \"foo/\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"foo/bar\\\\baz/qux\",\n \"foo/bar/baz/qux\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"foo//bar\\\\\\\\baz//qux\",\n \"foo/bar/baz/qux\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"X:\", \"X:\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"X:\\\\\", \"X:/\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"X:/\", \"X:/\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"X:\\\\foo\", \"X:/foo\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"X:\\\\foo/\", \"X:/foo/\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"X:/foo\", \"X:/foo\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"X:\\\\foo/bar\\\\baz/qux\",\n \"X:/foo/bar/baz/qux\");\n TEST_PATH_RAW(utf8::fix_path_separators, \"X:\\\\\\\\foo//bar\\\\\\\\baz//qux\",\n \"X:/foo/bar/baz/qux\");\n}\n\n// TODO(gibbed): fix_guest_path_separators\n\nTEST_CASE(\"UTF-8 Find Name From Path\", \"[utf8]\") {\n TEST_PATH(utf8::find_name_from_path, \"/\", \"\");\n TEST_PATH(utf8::find_name_from_path, \"//\", \"\");\n TEST_PATH(utf8::find_name_from_path, \"///\", \"\");\n TEST_PATH(utf8::find_name_from_path, \"C/\", \"C\");\n TEST_PATH(utf8::find_name_from_path, \"/C/\", \"C\");\n TEST_PATH(utf8::find_name_from_path, \"C/D/\", \"D\");\n TEST_PATH(utf8::find_name_from_path, \"/C/D/E/\", \"E\");\n TEST_PATH(utf8::find_name_from_path, \"foo/bar/D/\", \"D\");\n TEST_PATH(utf8::find_name_from_path, \"/foo/bar/E/qux/\", \"qux\");\n TEST_PATH(utf8::find_name_from_path, \"foo/bar/baz/qux/\", \"qux\");\n TEST_PATH(utf8::find_name_from_path, \"foo/bar/baz/qux//\", \"qux\");\n TEST_PATH(utf8::find_name_from_path, \"foo/bar/baz/qux///\", \"qux\");\n TEST_PATH(utf8::find_name_from_path, \"foo/bar/baz/qux.txt\", \"qux.txt\");\n TEST_PATH(utf8::find_name_from_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ/\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ//\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ///\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ.txt\",\n \"ほげほげ.txt\");\n TEST_PATH(utf8::find_name_from_path, \"/foo\", \"foo\");\n TEST_PATH(utf8::find_name_from_path, \"//foo\", \"foo\");\n TEST_PATH(utf8::find_name_from_path, \"///foo\", \"foo\");\n TEST_PATH(utf8::find_name_from_path, \"/foo/bar/baz/qux.txt\", \"qux.txt\");\n TEST_PATH(utf8::find_name_from_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ/\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ//\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ///\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ.txt\",\n \"ほげほげ.txt\");\n TEST_PATH(utf8::find_name_from_path, \"X:/foo/bar/baz/qux.txt\", \"qux.txt\");\n TEST_PATH(utf8::find_name_from_path, \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ/\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ//\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ///\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_name_from_path, \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ.txt\",\n \"ほげほげ.txt\");\n TEST_PATH(utf8::find_name_from_path, \"X:/ほげ/ぴよ/ふが/ほげら.ほげほげ\",\n \"ほげら.ほげほげ\");\n}\n\n// TODO(gibbed): find_name_from_guest_path\n\nTEST_CASE(\"UTF-8 Find Base Name From Path\", \"[utf8]\") {\n TEST_PATH(utf8::find_base_name_from_path, \"foo/bar/baz/qux.txt\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"foo/bar/baz/qux/\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"foo/bar/baz/qux//\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"foo/bar/baz/qux///\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"C/\", \"C\");\n TEST_PATH(utf8::find_base_name_from_path, \"/C/\", \"C\");\n TEST_PATH(utf8::find_base_name_from_path, \"C/D/\", \"D\");\n TEST_PATH(utf8::find_base_name_from_path, \"/C/D/E/\", \"E\");\n TEST_PATH(utf8::find_base_name_from_path, \"foo/bar/D/\", \"D\");\n TEST_PATH(utf8::find_base_name_from_path,\n \"ほげ/ぴよ/ふが/ほげら/ほげほげ.txt\", \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ/\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ//\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ///\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path, \"ほげ/ぴよ/ふが/ほげら.ほげほげ\",\n \"ほげら\");\n TEST_PATH(utf8::find_base_name_from_path, \"/foo/bar/baz/qux.txt\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"/foo/bar/baz/qux/\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"/foo/bar/baz/qux//\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"/foo/bar/baz/qux///\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path,\n \"/ほげ/ぴよ/ふが/ほげら/ほげほげ.txt\", \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ/\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ//\",\n \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path,\n \"/ほげ/ぴよ/ふが/ほげら/ほげほげ///\", \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path, \"/ほげ/ぴよ/ふが/ほげら.ほげほげ\",\n \"ほげら\");\n TEST_PATH(utf8::find_base_name_from_path, \"X:/foo/bar/baz/qux.txt\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"X:/foo/bar/baz/qux/\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"X:/foo/bar/baz/qux//\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path, \"X:/foo/bar/baz/qux///\", \"qux\");\n TEST_PATH(utf8::find_base_name_from_path,\n \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ.txt\", \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path,\n \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ/\", \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path,\n \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ//\", \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path,\n \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ///\", \"ほげほげ\");\n TEST_PATH(utf8::find_base_name_from_path, \"X:/ほげ/ぴよ/ふが/ほげら.ほげほげ\",\n \"ほげら\");\n}\n\n// TODO(gibbed): find_base_name_from_guest_path\n\nTEST_CASE(\"UTF-8 Find Base Path\", \"[utf8]\") {\n TEST_PATH(utf8::find_base_path, \"\", \"\");\n TEST_PATH(utf8::find_base_path, \"/\", \"\");\n TEST_PATH(utf8::find_base_path, \"//\", \"\");\n TEST_PATH(utf8::find_base_path, \"///\", \"\");\n TEST_PATH(utf8::find_base_path, \"/foo\", \"\");\n TEST_PATH(utf8::find_base_path, \"//foo\", \"\");\n TEST_PATH(utf8::find_base_path, \"///foo\", \"\");\n TEST_PATH(utf8::find_base_path, \"/foo/\", \"\");\n TEST_PATH(utf8::find_base_path, \"/foo//\", \"\");\n TEST_PATH(utf8::find_base_path, \"/foo///\", \"\");\n TEST_PATH(utf8::find_base_path, \"//foo/\", \"\");\n TEST_PATH(utf8::find_base_path, \"//foo//\", \"\");\n TEST_PATH(utf8::find_base_path, \"//foo///\", \"\");\n TEST_PATH(utf8::find_base_path, \"///foo/\", \"\");\n TEST_PATH(utf8::find_base_path, \"///foo//\", \"\");\n TEST_PATH(utf8::find_base_path, \"///foo///\", \"\");\n TEST_PATH(utf8::find_base_path, \"/foo/bar\", \"/foo\");\n TEST_PATH(utf8::find_base_path, \"/foo/bar/\", \"/foo\");\n TEST_PATH(utf8::find_base_path, \"/foo/bar//\", \"/foo\");\n TEST_PATH(utf8::find_base_path, \"/foo/bar///\", \"/foo\");\n TEST_PATH(utf8::find_base_path, \"/foo/bar/baz/qux\", \"/foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"/foo/bar/baz/qux/\", \"/foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"/foo/bar/baz/qux//\", \"/foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"/foo/bar/baz/qux///\", \"/foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ\",\n \"/ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ/\",\n \"/ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ//\",\n \"/ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"/ほげ/ぴよ/ふが/ほげら/ほげほげ///\",\n \"/ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"foo\", \"\");\n TEST_PATH(utf8::find_base_path, \"foo/\", \"\");\n TEST_PATH(utf8::find_base_path, \"foo//\", \"\");\n TEST_PATH(utf8::find_base_path, \"foo///\", \"\");\n TEST_PATH(utf8::find_base_path, \"foo/bar\", \"foo\");\n TEST_PATH(utf8::find_base_path, \"foo/bar/\", \"foo\");\n TEST_PATH(utf8::find_base_path, \"foo/bar//\", \"foo\");\n TEST_PATH(utf8::find_base_path, \"foo/bar///\", \"foo\");\n TEST_PATH(utf8::find_base_path, \"foo/bar/baz/qux\", \"foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"foo/bar/baz/qux/\", \"foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"foo/bar/baz/qux//\", \"foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"foo/bar/baz/qux///\", \"foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ\",\n \"ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ/\",\n \"ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ//\",\n \"ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"ほげ/ぴよ/ふが/ほげら/ほげほげ///\",\n \"ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"X:\", \"\");\n TEST_PATH(utf8::find_base_path, \"X:/\", \"\");\n TEST_PATH(utf8::find_base_path, \"X://\", \"\");\n TEST_PATH(utf8::find_base_path, \"X:///\", \"\");\n TEST_PATH(utf8::find_base_path, \"X:/foo\", \"X:\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/\", \"X:\");\n TEST_PATH(utf8::find_base_path, \"X:/foo//\", \"X:\");\n TEST_PATH(utf8::find_base_path, \"X:/foo///\", \"X:\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/bar\", \"X:/foo\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/bar/\", \"X:/foo\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/bar//\", \"X:/foo\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/bar///\", \"X:/foo\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/bar/baz/qux\", \"X:/foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/bar/baz/qux/\", \"X:/foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/bar/baz/qux//\", \"X:/foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"X:/foo/bar/baz/qux///\", \"X:/foo/bar/baz\");\n TEST_PATH(utf8::find_base_path, \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ\",\n \"X:/ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ/\",\n \"X:/ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ//\",\n \"X:/ほげ/ぴよ/ふが/ほげら\");\n TEST_PATH(utf8::find_base_path, \"X:/ほげ/ぴよ/ふが/ほげら/ほげほげ///\",\n \"X:/ほげ/ぴよ/ふが/ほげら\");\n}\n\n// TODO(gibbed): find_base_guest_path\n\nTEST_CASE(\"UTF-8 Canonicalize Path\", \"[utf8]\") {\n TEST_PATH(utf8::canonicalize_path, \"foo/bar/baz/qux\", \"foo/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/bar/baz/qux/\", \"foo/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/bar/baz/qux//\", \"foo/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/bar/baz/qux///\", \"foo/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/./baz/qux\", \"foo/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/./baz/qux/\", \"foo/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/../baz/qux\", \"baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/../baz/qux/\", \"baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/./baz/../qux\", \"foo/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/./baz/../qux/\", \"foo/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/./../baz/qux\", \"baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"foo/./../baz/qux/\", \"baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"./bar/baz/qux\", \"bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"./bar/baz/qux/\", \"bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"../bar/baz/qux\", \"bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"../bar/baz/qux/\", \"bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"ほげ/ぴよ/./ふが/../ほげら/ほげほげ\",\n \"ほげ/ぴよ/ほげら/ほげほげ\");\n TEST_PATH(utf8::canonicalize_path, \"ほげ/ぴよ/./ふが/../ほげら/ほげほげ/\",\n \"ほげ/ぴよ/ほげら/ほげほげ\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/bar/baz/qux\", \"/foo/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/bar/baz/qux/\", \"/foo/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/./baz/qux\", \"/foo/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/./baz/qux/\", \"/foo/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/../baz/qux\", \"/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/../baz/qux/\", \"/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/./baz/../qux\", \"/foo/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/./baz/../qux/\", \"/foo/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/./../baz/qux\", \"/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/foo/./../baz/qux/\", \"/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/./bar/baz/qux\", \"/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/./bar/baz/qux/\", \"/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/../bar/baz/qux\", \"/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/../bar/baz/qux/\", \"/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"/ほげ/ぴよ/./ふが/../ほげら/ほげほげ\",\n \"/ほげ/ぴよ/ほげら/ほげほげ\");\n TEST_PATH(utf8::canonicalize_path, \"/ほげ/ぴよ/./ふが/../ほげら/ほげほげ/\",\n \"/ほげ/ぴよ/ほげら/ほげほげ\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/bar/baz/qux\",\n \"X:/foo/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/bar/baz/qux/\",\n \"X:/foo/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/./baz/qux\", \"X:/foo/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/./baz/qux/\", \"X:/foo/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/../baz/qux\", \"X:/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/../baz/qux/\", \"X:/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/./baz/../qux\", \"X:/foo/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/./baz/../qux/\", \"X:/foo/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/./../baz/qux\", \"X:/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/foo/./../baz/qux/\", \"X:/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/./bar/baz/qux\", \"X:/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/./bar/baz/qux/\", \"X:/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/../bar/baz/qux\", \"X:/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/../bar/baz/qux/\", \"X:/bar/baz/qux\");\n TEST_PATH(utf8::canonicalize_path, \"X:/ほげ/ぴよ/./ふが/../ほげら/ほげほげ\",\n \"X:/ほげ/ぴよ/ほげら/ほげほげ\");\n TEST_PATH(utf8::canonicalize_path, \"X:/ほげ/ぴよ/./ふが/../ほげら/ほげほげ/\",\n \"X:/ほげ/ぴよ/ほげら/ほげほげ\");\n}\n\n// TODO(gibbed): canonicalize_guest_path\n\n} // namespace xe::base::test\n" }, { "path": "src/xenia/base/threading.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/threading.h\"\n\nnamespace xe {\nnamespace threading {\n\nuint32_t logical_processor_count() {\n static uint32_t value = 0;\n if (!value) {\n value = std::thread::hardware_concurrency();\n }\n return value;\n}\n\nthread_local uint32_t current_thread_id_ = UINT_MAX;\n\nuint32_t current_thread_id() {\n return current_thread_id_ == UINT_MAX ? current_thread_system_id()\n : current_thread_id_;\n}\n\nvoid set_current_thread_id(uint32_t id) { current_thread_id_ = id; }\n\n} // namespace threading\n} // namespace xe\n" }, { "path": "src/xenia/base/threading.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_THREADING_H_\n#define XENIA_BASE_THREADING_H_\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/chrono.h\"\n#include \"xenia/base/literals.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/threading_timer_queue.h\"\n\nnamespace xe {\nnamespace threading {\n\nusing namespace xe::literals;\n\n#if XE_PLATFORM_ANDROID\nvoid AndroidInitialize();\nvoid AndroidShutdown();\n#endif\n\n// This is more like an Event with self-reset when returning from Wait()\nclass Fence {\n public:\n Fence() : signal_state_(0) {}\n\n void Signal() {\n std::unique_lock lock(mutex_);\n signal_state_ |= SIGMASK_;\n cond_.notify_all();\n }\n\n // Wait for the Fence to be signaled. Clears the signal on return.\n void Wait() {\n std::unique_lock lock(mutex_);\n assert_true((signal_state_ & ~SIGMASK_) < (SIGMASK_ - 1) &&\n \"Too many threads?\");\n\n // keep local copy to minimize loads\n auto signal_state = ++signal_state_;\n for (; !(signal_state & SIGMASK_); signal_state = signal_state_) {\n cond_.wait(lock);\n }\n\n // We can't just clear the signal as other threads may not have read it yet\n assert_true((signal_state & ~SIGMASK_) > 0); // wait_count > 0\n if (signal_state == (1 | SIGMASK_)) { // wait_count == 1\n // Last one out turn off the lights\n signal_state_ = 0;\n } else {\n // Oops, another thread is still waiting, set the new count and keep the\n // signal.\n signal_state_ = --signal_state;\n }\n }\n\n private:\n using state_t_ = uint_fast32_t;\n static constexpr state_t_ SIGMASK_ = state_t_(1)\n << (sizeof(state_t_) * 8 - 1);\n\n std::mutex mutex_;\n std::condition_variable cond_;\n // Use the highest bit (sign bit) as the signal flag and the rest to count\n // waiting threads.\n volatile state_t_ signal_state_;\n};\n\n// Returns the total number of logical processors in the host system.\nuint32_t logical_processor_count();\n\n// Enables the current process to set thread affinity.\n// Must be called at startup before attempting to set thread affinity.\nvoid EnableAffinityConfiguration();\n\n// Gets a stable thread-specific ID, but may not be. Use for informative\n// purposes only.\nuint32_t current_thread_system_id();\n\n// Gets a stable thread-specific ID that defaults to the same value as\n// current_thread_system_id but may be overridden.\n// Guest threads often change this to the guest thread handle.\nuint32_t current_thread_id();\nvoid set_current_thread_id(uint32_t id);\n\n// Sets the current thread name.\nvoid set_name(const std::string_view name);\n\n// Yields the current thread to the scheduler. Maybe.\nvoid MaybeYield();\n\n// Memory barrier (request - may be ignored).\nvoid SyncMemory();\n\n// Sleeps the current thread for at least as long as the given duration.\nvoid Sleep(std::chrono::microseconds duration);\ntemplate \nvoid Sleep(std::chrono::duration duration) {\n Sleep(std::chrono::duration_cast(duration));\n}\n\nenum class SleepResult {\n kSuccess,\n kAlerted,\n};\n// Sleeps the current thread for at least as long as the given duration.\n// The thread is put in an alertable state and may wake to dispatch user\n// callbacks. If this happens the sleep returns early with\n// SleepResult::kAlerted.\nSleepResult AlertableSleep(std::chrono::microseconds duration);\ntemplate \nSleepResult AlertableSleep(std::chrono::duration duration) {\n return AlertableSleep(\n std::chrono::duration_cast(duration));\n}\n\ntypedef uint32_t TlsHandle;\nconstexpr TlsHandle kInvalidTlsHandle = UINT_MAX;\n\nTlsHandle AllocateTlsHandle();\nbool FreeTlsHandle(TlsHandle handle);\nuintptr_t GetTlsValue(TlsHandle handle);\nbool SetTlsValue(TlsHandle handle, uintptr_t value);\n\n// A high-resolution timer capable of firing at millisecond-precision. All\n// timers created in this way are executed in the same thread so callbacks must\n// be kept short or else all timers will be impacted. This is a simplified\n// wrapper around QueueTimerRecurring which automatically cancels the timer on\n// destruction.\nclass HighResolutionTimer {\n HighResolutionTimer(std::chrono::milliseconds interval,\n std::function callback) {\n assert_not_null(callback);\n wait_item_ = QueueTimerRecurring(\n [callback = std::move(callback)](void*) { callback(); }, nullptr,\n TimerQueueWaitItem::clock::now(), interval);\n }\n\n public:\n ~HighResolutionTimer() {\n if (auto wait_item = wait_item_.lock()) {\n wait_item->Disarm();\n }\n }\n\n // Creates a new repeating timer with the given period.\n // The given function will be called back as close to the given period as\n // possible.\n static std::unique_ptr CreateRepeating(\n std::chrono::milliseconds period, std::function callback) {\n return std::unique_ptr(\n new HighResolutionTimer(period, std::move(callback)));\n }\n\n private:\n std::weak_ptr wait_item_;\n};\n\n// Results for a WaitHandle operation.\nenum class WaitResult {\n // The state of the specified object is signaled.\n // In a WaitAny the tuple will contain the index of the wait handle that\n // caused the wait to be satisfied.\n kSuccess,\n // The wait was ended by one or more user-mode callbacks queued to the thread.\n // This will occur when is_alertable is set true.\n kUserCallback,\n // The time-out interval elapsed, and the object's state is nonsignaled.\n kTimeout,\n // The specified object is a mutex object that was not released by the thread\n // that owned the mutex object before the owning thread terminated. Ownership\n // of the mutex object is granted to the calling thread and the mutex is set\n // to nonsignaled.\n // In a WaitAny the tuple will contain the index of the wait handle that\n // caused the wait to be abandoned.\n kAbandoned,\n // The function has failed.\n kFailed,\n};\n\nclass WaitHandle {\n public:\n virtual ~WaitHandle() = default;\n\n // Returns the native handle of the object on the host system.\n // This value is platform specific.\n virtual void* native_handle() const = 0;\n\n protected:\n WaitHandle() = default;\n};\n\n// Waits until the wait handle is in the signaled state, an alert triggers and\n// a user callback is queued to the thread, or the timeout interval elapses.\n// If timeout is zero the call will return immediately instead of waiting and\n// if the timeout is max() the wait will not time out.\nWaitResult Wait(\n WaitHandle* wait_handle, bool is_alertable,\n std::chrono::milliseconds timeout = std::chrono::milliseconds::max());\n\n// Signals one object and waits on another object as a single operation.\n// Waits until the wait handle is in the signaled state, an alert triggers and\n// a user callback is queued to the thread, or the timeout interval elapses.\n// If timeout is zero the call will return immediately instead of waiting and\n// if the timeout is max() the wait will not time out.\nWaitResult SignalAndWait(\n WaitHandle* wait_handle_to_signal, WaitHandle* wait_handle_to_wait_on,\n bool is_alertable,\n std::chrono::milliseconds timeout = std::chrono::milliseconds::max());\n\nstd::pair WaitMultiple(\n WaitHandle* wait_handles[], size_t wait_handle_count, bool wait_all,\n bool is_alertable,\n std::chrono::milliseconds timeout = std::chrono::milliseconds::max());\n\n// Waits until all of the specified objects are in the signaled state, a\n// user callback is queued to the thread, or the time-out interval elapses.\n// If timeout is zero the call will return immediately instead of waiting and\n// if the timeout is max() the wait will not time out.\ninline WaitResult WaitAll(\n WaitHandle* wait_handles[], size_t wait_handle_count, bool is_alertable,\n std::chrono::milliseconds timeout = std::chrono::milliseconds::max()) {\n return WaitMultiple(wait_handles, wait_handle_count, true, is_alertable,\n timeout)\n .first;\n}\ninline WaitResult WaitAll(\n std::vector wait_handles, bool is_alertable,\n std::chrono::milliseconds timeout = std::chrono::milliseconds::max()) {\n return WaitAll(wait_handles.data(), wait_handles.size(), is_alertable,\n timeout);\n}\n\n// Waits until any of the specified objects are in the signaled state, a\n// user callback is queued to the thread, or the time-out interval elapses.\n// If timeout is zero the call will return immediately instead of waiting and\n// if the timeout is max() the wait will not time out.\n// The second argument of the return tuple indicates which wait handle caused\n// the wait to be satisfied or abandoned.\ninline std::pair WaitAny(\n WaitHandle* wait_handles[], size_t wait_handle_count, bool is_alertable,\n std::chrono::milliseconds timeout = std::chrono::milliseconds::max()) {\n return WaitMultiple(wait_handles, wait_handle_count, false, is_alertable,\n timeout);\n}\ninline std::pair WaitAny(\n std::vector wait_handles, bool is_alertable,\n std::chrono::milliseconds timeout = std::chrono::milliseconds::max()) {\n return WaitAny(wait_handles.data(), wait_handles.size(), is_alertable,\n timeout);\n}\n\n// Models a Win32-like event object.\n// https://msdn.microsoft.com/en-us/library/windows/desktop/ms682396(v=vs.85).aspx\nclass Event : public WaitHandle {\n public:\n // Creates a manual-reset event object, which requires the use of the\n // Reset() function to set the event state to nonsignaled.\n // If initial_state is true the event will start in the signaled state.\n static std::unique_ptr CreateManualResetEvent(bool initial_state);\n\n // Creates an auto-reset event object, and system automatically resets the\n // event state to nonsignaled after a single waiting thread has been released.\n // If initial_state is true the event will start in the signaled state.\n static std::unique_ptr CreateAutoResetEvent(bool initial_state);\n\n // Sets the specified event object to the signaled state.\n // If this is a manual reset event the event stays signaled until Reset() is\n // called. If this is an auto reset event until exactly one wait is satisfied.\n virtual void Set() = 0;\n\n // Sets the specified event object to the nonsignaled state.\n // Resetting an event that is already reset has no effect.\n virtual void Reset() = 0;\n\n // Sets the specified event object to the signaled state and then resets it to\n // the nonsignaled state after releasing the appropriate number of waiting\n // threads.\n virtual void Pulse() = 0;\n};\n\n// Models a Win32-like semaphore object.\n// https://msdn.microsoft.com/en-us/library/windows/desktop/ms682438(v=vs.85).aspx\nclass Semaphore : public WaitHandle {\n public:\n // Creates a new semaphore object.\n // The initial_count must be greater than or equal to zero and less than or\n // equal to maximum_count. The state of a semaphore is signaled when its count\n // is greater than zero and nonsignaled when it is zero. The count is\n // decreased by one whenever a wait function releases a thread that was\n // waiting for the semaphore. The count is increased by a specified amount by\n // calling the Release() function.\n static std::unique_ptr Create(int initial_count,\n int maximum_count);\n\n // Increases the count of the specified semaphore object by a specified\n // amount.\n // release_count must be greater than zero.\n // Returns false if adding release_count would set the semaphore over the\n // initially specified maximum_count.\n virtual bool Release(int release_count, int* out_previous_count) = 0;\n};\n\n// Models a Win32-like mutant (mutex) object.\n// https://msdn.microsoft.com/en-us/library/windows/desktop/ms682411(v=vs.85).aspx\nclass Mutant : public WaitHandle {\n public:\n // Creates a new mutant object, initially owned by the calling thread if\n // specified.\n static std::unique_ptr Create(bool initial_owner);\n\n // Releases ownership of the specified mutex object.\n // Returns false if the calling thread does not own the mutant object.\n virtual bool Release() = 0;\n};\n\n// Models a Win32-like timer object.\n// https://msdn.microsoft.com/en-us/library/windows/desktop/ms687012(v=vs.85).aspx\nclass Timer : public WaitHandle {\n public:\n // Make vtable entries for both so we can defer conversions and only do them\n // if really necessary (let the calling code what clock it prefers). Windows\n // kernel sync primitives will work with WinSystemClock while our own\n // implementation works with steady_clock.\n using WClock_ = xe::chrono::WinSystemClock;\n using GClock_ = std::chrono::steady_clock; // generic\n\n // Creates a timer whose state remains signaled until SetOnce() or\n // SetRepeating() is called to establish a new due time.\n static std::unique_ptr CreateManualResetTimer();\n\n // Creates a timer whose state remains signaled until a thread completes a\n // wait operation on the timer object.\n static std::unique_ptr CreateSynchronizationTimer();\n\n // Activates the specified waitable timer. When the due time arrives, the\n // timer is signaled and the thread that set the timer calls the optional\n // completion routine.\n // Returns true on success.\n virtual bool SetOnceAfter(xe::chrono::hundrednanoseconds rel_time,\n std::function opt_callback = nullptr) = 0;\n virtual bool SetOnceAt(WClock_::time_point due_time,\n std::function opt_callback = nullptr) = 0;\n virtual bool SetOnceAt(GClock_::time_point due_time,\n std::function opt_callback = nullptr) = 0;\n\n // Activates the specified waitable timer. When the due time arrives, the\n // timer is signaled and the thread that set the timer calls the optional\n // completion routine. A periodic timer automatically reactivates each time\n // the period elapses, until the timer is canceled or reset.\n // Returns true on success.\n virtual bool SetRepeatingAfter(\n xe::chrono::hundrednanoseconds rel_time, std::chrono::milliseconds period,\n std::function opt_callback = nullptr) = 0;\n virtual bool SetRepeatingAt(WClock_::time_point due_time,\n std::chrono::milliseconds period,\n std::function opt_callback = nullptr) = 0;\n virtual bool SetRepeatingAt(GClock_::time_point due_time,\n std::chrono::milliseconds period,\n std::function opt_callback = nullptr) = 0;\n\n // Stops the timer before it can be set to the signaled state and cancels\n // outstanding callbacks. Threads performing a wait operation on the timer\n // remain waiting until they time out or the timer is reactivated and its\n // state is set to signaled. If the timer is already in the signaled state, it\n // remains in that state.\n // Returns true on success.\n virtual bool Cancel() = 0;\n};\n\nstruct ThreadPriority {\n static const int32_t kLowest = -2;\n static const int32_t kBelowNormal = -1;\n static const int32_t kNormal = 0;\n static const int32_t kAboveNormal = 1;\n static const int32_t kHighest = 2;\n};\n\n// Models a Win32-like thread object.\n// https://msdn.microsoft.com/en-us/library/windows/desktop/ms682453(v=vs.85).aspx\nclass Thread : public WaitHandle {\n public:\n struct CreationParameters {\n size_t stack_size = 4_MiB;\n bool create_suspended = false;\n int32_t initial_priority = 0;\n };\n\n // Creates a thread with the given parameters and calls the start routine from\n // within that thread.\n static std::unique_ptr Create(CreationParameters params,\n std::function start_routine);\n static Thread* GetCurrentThread();\n\n // Ends the calling thread.\n // No destructors are called, and this function does not return.\n // The state of the thread object becomes signaled, releasing any other\n // threads that had been waiting for the thread to terminate.\n static void Exit(int exit_code);\n\n // Returns the ID of the thread.\n virtual uint32_t system_id() const = 0;\n\n // Returns the current name of the thread, if previously specified.\n const std::string& name() const { return name_; }\n\n // Sets the name of the thread, used in debugging and logging.\n virtual void set_name(std::string name) { name_ = std::move(name); }\n\n // Returns the current priority value for the thread.\n virtual int32_t priority() = 0;\n\n // Sets the priority value for the thread. This value, together with the\n // priority class of the thread's process, determines the thread's base\n // priority level. ThreadPriority contains useful constants.\n virtual void set_priority(int32_t new_priority) = 0;\n\n // Returns the current processor affinity mask for the thread.\n virtual uint64_t affinity_mask() = 0;\n\n // Sets a processor affinity mask for the thread.\n // A thread affinity mask is a bit vector in which each bit represents a\n // logical processor that a thread is allowed to run on. A thread affinity\n // mask must be a subset of the process affinity mask for the containing\n // process of a thread.\n virtual void set_affinity_mask(uint64_t new_affinity_mask) = 0;\n\n // Adds a user-mode asynchronous procedure call request to the thread queue.\n // When a user-mode APC is queued, the thread is not directed to call the APC\n // function unless it is in an alertable state. After the thread is in an\n // alertable state, the thread handles all pending APCs in first in, first out\n // (FIFO) order, and the wait operation returns WaitResult::kUserCallback.\n virtual void QueueUserCallback(std::function callback) = 0;\n\n // Decrements a thread's suspend count. When the suspend count is decremented\n // to zero, the execution of the thread is resumed.\n virtual bool Resume(uint32_t* out_previous_suspend_count = nullptr) = 0;\n\n // Suspends the specified thread.\n virtual bool Suspend(uint32_t* out_previous_suspend_count = nullptr) = 0;\n\n // Terminates the thread.\n // No destructors are called, and this function does not return.\n // The state of the thread object becomes signaled, releasing any other\n // threads that had been waiting for the thread to terminate.\n virtual void Terminate(int exit_code) = 0;\n\n protected:\n std::string name_;\n};\n\n} // namespace threading\n} // namespace xe\n\n#endif // XENIA_BASE_THREADING_H_\n" }, { "path": "src/xenia/base/threading_mac.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/threading.h\"\n\n#include \n#include \n#include \n#include \n\nnamespace xe {\nnamespace threading {\n\nuint64_t ticks() { return mach_absolute_time(); }\n\nuint32_t current_thread_id() {\n mach_port_t tid = pthread_mach_thread_np(pthread_self());\n return static_cast(tid);\n}\n\nvoid set_name(const std::string& name) { pthread_setname_np(name.c_str()); }\n\nvoid MaybeYield() { pthread_yield_np(); }\n\nvoid Sleep(std::chrono::microseconds duration) {\n timespec rqtp = {duration.count() / 1000000, duration.count() % 1000};\n nanosleep(&rqtp, nullptr);\n // TODO(benvanik): spin while rmtp >0?\n}\n\n} // namespace threading\n} // namespace xe\n" }, { "path": "src/xenia/base/threading_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/threading.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/chrono_steady_cast.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/threading_timer_queue.h\"\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#if XE_PLATFORM_ANDROID\n#include \n\n#include \"xenia/base/main_android.h\"\n#include \"xenia/base/string_util.h\"\n#endif\n\n#if XE_PLATFORM_LINUX\n// SIGEV_THREAD_ID in timer_create(...) is a Linux extension\n#define XE_HAS_SIGEV_THREAD_ID 1\n#ifdef __GLIBC__\n#define sigev_notify_thread_id _sigev_un._tid\n#endif\n#if __GLIBC__ == 2 && __GLIBC_MINOR__ < 30\n#define gettid() syscall(SYS_gettid)\n#endif\n#else\n#define XE_HAS_SIGEV_THREAD_ID 0\n#endif\n\nnamespace xe {\nnamespace threading {\n\n#if XE_PLATFORM_ANDROID\n// May be null if no dynamically loaded functions are required.\nstatic void* android_libc_;\n// API 26+.\nstatic int (*android_pthread_getname_np_)(pthread_t pthread, char* buf,\n size_t n);\n\nvoid AndroidInitialize() {\n if (xe::GetAndroidApiLevel() >= 26) {\n android_libc_ = dlopen(\"libc.so\", RTLD_NOW);\n assert_not_null(android_libc_);\n if (android_libc_) {\n android_pthread_getname_np_ =\n reinterpret_cast(\n dlsym(android_libc_, \"pthread_getname_np\"));\n assert_not_null(android_pthread_getname_np_);\n }\n }\n}\n\nvoid AndroidShutdown() {\n android_pthread_getname_np_ = nullptr;\n if (android_libc_) {\n dlclose(android_libc_);\n android_libc_ = nullptr;\n }\n}\n#endif\n\ntemplate \ninline timespec DurationToTimeSpec(\n std::chrono::duration<_Rep, _Period> duration) {\n auto nanoseconds =\n std::chrono::duration_cast(duration);\n auto div = ldiv(nanoseconds.count(), 1000000000L);\n return timespec{div.quot, div.rem};\n}\n\n// Thread interruption is done using user-defined signals\n// This implementation uses the SIGRTMAX - SIGRTMIN to signal to a thread\n// gdb tip, for SIG = SIGRTMIN + SignalType : handle SIG nostop\n// lldb tip, for SIG = SIGRTMIN + SignalType : process handle SIG -s false\nenum class SignalType {\n kThreadSuspend,\n kThreadUserCallback,\n#if XE_PLATFORM_ANDROID\n // pthread_cancel is not available on Android, using a signal handler for\n // simplified PTHREAD_CANCEL_ASYNCHRONOUS-like behavior - not disabling\n // cancellation currently, so should be enough.\n kThreadTerminate,\n#endif\n k_Count\n};\n\nint GetSystemSignal(SignalType num) {\n auto result = SIGRTMIN + static_cast(num);\n assert_true(result < SIGRTMAX);\n return result;\n}\n\nSignalType GetSystemSignalType(int num) {\n return static_cast(num - SIGRTMIN);\n}\n\nthread_local std::array(SignalType::k_Count)>\n signal_handler_installed = {};\n\nstatic void signal_handler(int signal, siginfo_t* info, void* context);\n\nvoid install_signal_handler(SignalType type) {\n if (signal_handler_installed[static_cast(type)]) return;\n struct sigaction action {};\n action.sa_flags = SA_SIGINFO;\n action.sa_sigaction = signal_handler;\n sigemptyset(&action.sa_mask);\n if (sigaction(GetSystemSignal(type), &action, nullptr) == -1)\n signal_handler_installed[static_cast(type)] = true;\n}\n\n// TODO(dougvj)\nvoid EnableAffinityConfiguration() {}\n\n// uint64_t ticks() { return mach_absolute_time(); }\n\nuint32_t current_thread_system_id() {\n return static_cast(syscall(SYS_gettid));\n}\n\nvoid MaybeYield() {\n sched_yield();\n __sync_synchronize();\n}\n\nvoid SyncMemory() { __sync_synchronize(); }\n\nvoid Sleep(std::chrono::microseconds duration) {\n timespec rqtp = DurationToTimeSpec(duration);\n timespec rmtp = {};\n auto p_rqtp = &rqtp;\n auto p_rmtp = &rmtp;\n int ret = 0;\n do {\n ret = nanosleep(p_rqtp, p_rmtp);\n // Swap requested for remaining in case of signal interruption\n // in which case, we start sleeping again for the remainder\n std::swap(p_rqtp, p_rmtp);\n } while (ret == -1 && errno == EINTR);\n}\n\n// TODO(bwrsandman) Implement by allowing alert interrupts from IO operations\nthread_local bool alertable_state_ = false;\nSleepResult AlertableSleep(std::chrono::microseconds duration) {\n alertable_state_ = true;\n Sleep(duration);\n alertable_state_ = false;\n return SleepResult::kSuccess;\n}\n\nTlsHandle AllocateTlsHandle() {\n auto key = static_cast(-1);\n auto res = pthread_key_create(&key, nullptr);\n assert_zero(res);\n assert_true(key != static_cast(-1));\n return static_cast(key);\n}\n\nbool FreeTlsHandle(TlsHandle handle) {\n return pthread_key_delete(static_cast(handle)) == 0;\n}\n\nuintptr_t GetTlsValue(TlsHandle handle) {\n return reinterpret_cast(\n pthread_getspecific(static_cast(handle)));\n}\n\nbool SetTlsValue(TlsHandle handle, uintptr_t value) {\n return pthread_setspecific(static_cast(handle),\n reinterpret_cast(value)) == 0;\n}\n\nclass PosixConditionBase {\n public:\n virtual bool Signal() = 0;\n\n WaitResult Wait(std::chrono::milliseconds timeout) {\n bool executed;\n auto predicate = [this] { return this->signaled(); };\n auto lock = std::unique_lock(mutex_);\n if (predicate()) {\n executed = true;\n } else {\n if (timeout == std::chrono::milliseconds::max()) {\n cond_.wait(lock, predicate);\n executed = true; // Did not time out;\n } else {\n executed = cond_.wait_for(lock, timeout, predicate);\n }\n }\n if (executed) {\n post_execution();\n return WaitResult::kSuccess;\n } else {\n return WaitResult::kTimeout;\n }\n }\n\n static std::pair WaitMultiple(\n std::vector&& handles, bool wait_all,\n std::chrono::milliseconds timeout) {\n assert_true(handles.size() > 0);\n\n // Construct a condition for all or any depending on wait_all\n std::function predicate;\n {\n using iter_t = std::vector::const_iterator;\n const auto predicate_inner = [](auto h) { return h->signaled(); };\n const auto operation =\n wait_all ? std::all_of\n : std::any_of;\n predicate = [&handles, operation, predicate_inner] {\n return operation(handles.cbegin(), handles.cend(), predicate_inner);\n };\n }\n\n // TODO(bwrsandman, Triang3l) This is controversial, see issue #1677\n // This will probably cause a deadlock on the next thread doing any waiting\n // if the thread is suspended between locking and waiting\n std::unique_lock lock(PosixConditionBase::mutex_);\n\n bool wait_success = true;\n // If the timeout is infinite, wait without timeout.\n // The predicate will be checked before beginning the wait\n if (timeout == std::chrono::milliseconds::max()) {\n PosixConditionBase::cond_.wait(lock, predicate);\n } else {\n // Wait with timeout.\n wait_success =\n PosixConditionBase::cond_.wait_for(lock, timeout, predicate);\n }\n if (wait_success) {\n auto first_signaled = std::numeric_limits::max();\n for (auto i = 0u; i < handles.size(); ++i) {\n if (handles[i]->signaled()) {\n if (first_signaled > i) {\n first_signaled = i;\n }\n handles[i]->post_execution();\n if (!wait_all) break;\n }\n }\n assert_true(std::numeric_limits::max() != first_signaled);\n return std::make_pair(WaitResult::kSuccess, first_signaled);\n } else {\n return std::make_pair(WaitResult::kTimeout, 0);\n }\n }\n\n virtual void* native_handle() const { return cond_.native_handle(); }\n\n protected:\n inline virtual bool signaled() const = 0;\n inline virtual void post_execution() = 0;\n static std::condition_variable cond_;\n static std::mutex mutex_;\n};\n\nstd::condition_variable PosixConditionBase::cond_;\nstd::mutex PosixConditionBase::mutex_;\n\n// There really is no native POSIX handle for a single wait/signal construct\n// pthreads is at a lower level with more handles for such a mechanism.\n// This simple wrapper class functions as our handle and uses conditional\n// variables for waits and signals.\ntemplate \nclass PosixCondition {};\n\ntemplate <>\nclass PosixCondition : public PosixConditionBase {\n public:\n PosixCondition(bool manual_reset, bool initial_state)\n : signal_(initial_state), manual_reset_(manual_reset) {}\n virtual ~PosixCondition() = default;\n\n bool Signal() override {\n auto lock = std::unique_lock(mutex_);\n signal_ = true;\n cond_.notify_all();\n return true;\n }\n\n void Reset() {\n auto lock = std::unique_lock(mutex_);\n signal_ = false;\n }\n\n private:\n inline bool signaled() const override { return signal_; }\n inline void post_execution() override {\n if (!manual_reset_) {\n signal_ = false;\n }\n }\n bool signal_;\n const bool manual_reset_;\n};\n\ntemplate <>\nclass PosixCondition : public PosixConditionBase {\n public:\n PosixCondition(uint32_t initial_count, uint32_t maximum_count)\n : count_(initial_count), maximum_count_(maximum_count) {}\n\n bool Signal() override { return Release(1, nullptr); }\n\n bool Release(uint32_t release_count, int* out_previous_count) {\n if (maximum_count_ - count_ >= release_count) {\n auto lock = std::unique_lock(mutex_);\n if (out_previous_count) *out_previous_count = count_;\n count_ += release_count;\n cond_.notify_all();\n return true;\n }\n return false;\n }\n\n private:\n inline bool signaled() const override { return count_ > 0; }\n inline void post_execution() override {\n count_--;\n cond_.notify_all();\n }\n uint32_t count_;\n const uint32_t maximum_count_;\n};\n\ntemplate <>\nclass PosixCondition : public PosixConditionBase {\n public:\n explicit PosixCondition(bool initial_owner) : count_(0) {\n if (initial_owner) {\n count_ = 1;\n owner_ = std::this_thread::get_id();\n }\n }\n\n bool Signal() override { return Release(); }\n\n bool Release() {\n if (owner_ == std::this_thread::get_id() && count_ > 0) {\n auto lock = std::unique_lock(mutex_);\n --count_;\n // Free to be acquired by another thread\n if (count_ == 0) {\n cond_.notify_all();\n }\n return true;\n }\n return false;\n }\n\n void* native_handle() const override { return mutex_.native_handle(); }\n\n private:\n inline bool signaled() const override {\n return count_ == 0 || owner_ == std::this_thread::get_id();\n }\n inline void post_execution() override {\n count_++;\n owner_ = std::this_thread::get_id();\n }\n uint32_t count_;\n std::thread::id owner_;\n};\n\ntemplate <>\nclass PosixCondition : public PosixConditionBase {\n public:\n explicit PosixCondition(bool manual_reset)\n : callback_(nullptr), signal_(false), manual_reset_(manual_reset) {}\n\n virtual ~PosixCondition() { Cancel(); }\n\n bool Signal() override {\n std::lock_guard lock(mutex_);\n signal_ = true;\n cond_.notify_all();\n return true;\n }\n\n void SetOnce(std::chrono::steady_clock::time_point due_time,\n std::function opt_callback) {\n Cancel();\n\n std::lock_guard lock(mutex_);\n\n callback_ = std::move(opt_callback);\n signal_ = false;\n wait_item_ = QueueTimerOnce(&CompletionRoutine, this, due_time);\n }\n\n void SetRepeating(std::chrono::steady_clock::time_point due_time,\n std::chrono::milliseconds period,\n std::function opt_callback) {\n Cancel();\n\n std::lock_guard lock(mutex_);\n\n callback_ = std::move(opt_callback);\n signal_ = false;\n wait_item_ =\n QueueTimerRecurring(&CompletionRoutine, this, due_time, period);\n }\n\n void Cancel() {\n if (auto wait_item = wait_item_.lock()) {\n wait_item->Disarm();\n }\n }\n\n void* native_handle() const override {\n assert_always();\n return nullptr;\n }\n\n private:\n static void CompletionRoutine(void* userdata) {\n assert_not_null(userdata);\n auto timer = reinterpret_cast*>(userdata);\n timer->Signal();\n // As the callback may reset the timer, store local.\n std::function callback;\n {\n std::lock_guard lock(timer->mutex_);\n callback = timer->callback_;\n }\n if (callback) {\n callback();\n }\n }\n\n private:\n inline bool signaled() const override { return signal_; }\n inline void post_execution() override {\n if (!manual_reset_) {\n signal_ = false;\n }\n }\n std::weak_ptr wait_item_;\n std::function callback_;\n volatile bool signal_;\n const bool manual_reset_;\n};\n\nstruct ThreadStartData {\n std::function start_routine;\n bool create_suspended;\n Thread* thread_obj;\n};\n\ntemplate <>\nclass PosixCondition : public PosixConditionBase {\n enum class State {\n kUninitialized,\n kRunning,\n kSuspended,\n kFinished,\n };\n\n public:\n PosixCondition()\n : thread_(0),\n signaled_(false),\n exit_code_(0),\n state_(State::kUninitialized),\n suspend_count_(0) {\n#if XE_PLATFORM_ANDROID\n android_pre_api_26_name_[0] = '\\0';\n#endif\n }\n bool Initialize(Thread::CreationParameters params,\n ThreadStartData* start_data) {\n start_data->create_suspended = params.create_suspended;\n pthread_attr_t attr;\n if (pthread_attr_init(&attr) != 0) return false;\n if (pthread_attr_setstacksize(&attr, params.stack_size) != 0) {\n pthread_attr_destroy(&attr);\n return false;\n }\n if (params.initial_priority != 0) {\n sched_param sched{};\n sched.sched_priority = params.initial_priority + 1;\n if (pthread_attr_setschedpolicy(&attr, SCHED_FIFO) != 0) {\n pthread_attr_destroy(&attr);\n return false;\n }\n if (pthread_attr_setschedparam(&attr, &sched) != 0) {\n pthread_attr_destroy(&attr);\n return false;\n }\n }\n if (pthread_create(&thread_, &attr, ThreadStartRoutine, start_data) != 0) {\n pthread_attr_destroy(&attr);\n return false;\n }\n pthread_attr_destroy(&attr);\n return true;\n }\n\n /// Constructor for existing thread. This should only happen once called by\n /// Thread::GetCurrentThread() on the main thread\n explicit PosixCondition(pthread_t thread)\n : thread_(thread),\n signaled_(false),\n exit_code_(0),\n state_(State::kRunning) {\n#if XE_PLATFORM_ANDROID\n android_pre_api_26_name_[0] = '\\0';\n#endif\n }\n\n virtual ~PosixCondition() {\n if (thread_ && !signaled_) {\n#if XE_PLATFORM_ANDROID\n if (pthread_kill(thread_,\n GetSystemSignal(SignalType::kThreadTerminate)) != 0) {\n assert_always();\n }\n#else\n if (pthread_cancel(thread_) != 0) {\n assert_always();\n }\n#endif\n if (pthread_join(thread_, nullptr) != 0) {\n assert_always();\n }\n }\n }\n\n bool Signal() override { return true; }\n\n std::string name() const {\n WaitStarted();\n auto result = std::array{'\\0'};\n std::unique_lock lock(state_mutex_);\n if (state_ != State::kUninitialized && state_ != State::kFinished) {\n#if XE_PLATFORM_ANDROID\n // pthread_getname_np was added in API 26 - below that, store the name in\n // this object, which may be only modified through Xenia threading, but\n // should be enough in most cases.\n if (android_pthread_getname_np_) {\n if (android_pthread_getname_np_(thread_, result.data(),\n result.size() - 1) != 0) {\n assert_always();\n }\n } else {\n std::lock_guard lock(android_pre_api_26_name_mutex_);\n std::strcpy(result.data(), android_pre_api_26_name_);\n }\n#else\n if (pthread_getname_np(thread_, result.data(), result.size() - 1) != 0) {\n assert_always();\n }\n#endif\n }\n return std::string(result.data());\n }\n\n void set_name(const std::string& name) {\n WaitStarted();\n std::unique_lock lock(state_mutex_);\n if (state_ != State::kUninitialized && state_ != State::kFinished) {\n pthread_setname_np(thread_, std::string(name).c_str());\n#if XE_PLATFORM_ANDROID\n SetAndroidPreApi26Name(name);\n#endif\n }\n }\n\n#if XE_PLATFORM_ANDROID\n void SetAndroidPreApi26Name(const std::string_view name) {\n if (android_pthread_getname_np_) {\n return;\n }\n std::lock_guard lock(android_pre_api_26_name_mutex_);\n xe::string_util::copy_truncating(android_pre_api_26_name_, name,\n xe::countof(android_pre_api_26_name_));\n }\n#endif\n\n uint32_t system_id() const { return static_cast(thread_); }\n\n uint64_t affinity_mask() {\n WaitStarted();\n cpu_set_t cpu_set;\n#if XE_PLATFORM_ANDROID\n if (sched_getaffinity(pthread_gettid_np(thread_), sizeof(cpu_set_t),\n &cpu_set) != 0) {\n assert_always();\n }\n#else\n if (pthread_getaffinity_np(thread_, sizeof(cpu_set_t), &cpu_set) != 0) {\n assert_always();\n }\n#endif\n uint64_t result = 0;\n auto cpu_count = std::min(CPU_SETSIZE, 64);\n for (auto i = 0u; i < cpu_count; i++) {\n auto set = CPU_ISSET(i, &cpu_set);\n result |= set << i;\n }\n return result;\n }\n\n void set_affinity_mask(uint64_t mask) {\n WaitStarted();\n cpu_set_t cpu_set;\n CPU_ZERO(&cpu_set);\n for (auto i = 0u; i < 64; i++) {\n if (mask & (1 << i)) {\n CPU_SET(i, &cpu_set);\n }\n }\n#if XE_PLATFORM_ANDROID\n if (sched_setaffinity(pthread_gettid_np(thread_), sizeof(cpu_set_t),\n &cpu_set) != 0) {\n assert_always();\n }\n#else\n if (pthread_setaffinity_np(thread_, sizeof(cpu_set_t), &cpu_set) != 0) {\n assert_always();\n }\n#endif\n }\n\n int priority() {\n WaitStarted();\n int policy;\n sched_param param{};\n int ret = pthread_getschedparam(thread_, &policy, ¶m);\n if (ret != 0) {\n return -1;\n }\n\n return param.sched_priority;\n }\n\n void set_priority(int new_priority) {\n WaitStarted();\n sched_param param{};\n param.sched_priority = new_priority;\n if (pthread_setschedparam(thread_, SCHED_FIFO, ¶m) != 0)\n assert_always();\n }\n\n void QueueUserCallback(std::function callback) {\n WaitStarted();\n std::unique_lock lock(callback_mutex_);\n user_callback_ = std::move(callback);\n sigval value{};\n value.sival_ptr = this;\n#if XE_PLATFORM_ANDROID\n sigqueue(pthread_gettid_np(thread_),\n GetSystemSignal(SignalType::kThreadUserCallback), value);\n#else\n pthread_sigqueue(thread_, GetSystemSignal(SignalType::kThreadUserCallback),\n value);\n#endif\n }\n\n void CallUserCallback() {\n std::unique_lock lock(callback_mutex_);\n user_callback_();\n }\n\n bool Resume(uint32_t* out_previous_suspend_count = nullptr) {\n if (out_previous_suspend_count) {\n *out_previous_suspend_count = 0;\n }\n WaitStarted();\n std::unique_lock lock(state_mutex_);\n if (state_ != State::kSuspended) return false;\n if (out_previous_suspend_count) {\n *out_previous_suspend_count = suspend_count_;\n }\n --suspend_count_;\n state_signal_.notify_all();\n return true;\n }\n\n bool Suspend(uint32_t* out_previous_suspend_count = nullptr) {\n if (out_previous_suspend_count) {\n *out_previous_suspend_count = 0;\n }\n WaitStarted();\n {\n if (out_previous_suspend_count) {\n *out_previous_suspend_count = suspend_count_;\n }\n state_ = State::kSuspended;\n ++suspend_count_;\n }\n int result =\n pthread_kill(thread_, GetSystemSignal(SignalType::kThreadSuspend));\n return result == 0;\n }\n\n void Terminate(int exit_code) {\n bool is_current_thread = pthread_self() == thread_;\n {\n std::unique_lock lock(state_mutex_);\n if (state_ == State::kFinished) {\n if (is_current_thread) {\n // This is really bad. Some thread must have called Terminate() on us\n // just before we decided to terminate ourselves\n assert_always();\n for (;;) {\n // Wait for pthread_cancel() to actually happen.\n }\n }\n return;\n }\n state_ = State::kFinished;\n }\n\n {\n std::lock_guard lock(mutex_);\n\n exit_code_ = exit_code;\n signaled_ = true;\n cond_.notify_all();\n }\n if (is_current_thread) {\n pthread_exit(reinterpret_cast(exit_code));\n } else {\n#ifdef XE_PLATFORM_ANDROID\n if (pthread_kill(thread_,\n GetSystemSignal(SignalType::kThreadTerminate)) != 0) {\n assert_always();\n }\n#else\n if (pthread_cancel(thread_) != 0) {\n assert_always();\n }\n#endif\n }\n }\n\n void WaitStarted() const {\n std::unique_lock lock(state_mutex_);\n state_signal_.wait(lock,\n [this] { return state_ != State::kUninitialized; });\n }\n\n /// Set state to suspended and wait until it reset by another thread\n void WaitSuspended() {\n std::unique_lock lock(state_mutex_);\n state_signal_.wait(lock, [this] { return suspend_count_ == 0; });\n state_ = State::kRunning;\n }\n\n void* native_handle() const override {\n return reinterpret_cast(thread_);\n }\n\n private:\n static void* ThreadStartRoutine(void* parameter);\n inline bool signaled() const override { return signaled_; }\n inline void post_execution() override {\n if (thread_) {\n pthread_join(thread_, nullptr);\n }\n }\n pthread_t thread_;\n bool signaled_;\n int exit_code_;\n volatile State state_;\n volatile uint32_t suspend_count_;\n mutable std::mutex state_mutex_;\n mutable std::mutex callback_mutex_;\n mutable std::condition_variable state_signal_;\n std::function user_callback_;\n#if XE_PLATFORM_ANDROID\n // Name accessible via name() on Android before API 26 which added\n // pthread_getname_np.\n mutable std::mutex android_pre_api_26_name_mutex_;\n char android_pre_api_26_name_[16];\n#endif\n};\n\nclass PosixWaitHandle {\n public:\n virtual PosixConditionBase& condition() = 0;\n};\n\n// This wraps a condition object as our handle because posix has no single\n// native handle for higher level concurrency constructs such as semaphores\ntemplate \nclass PosixConditionHandle : public T, public PosixWaitHandle {\n public:\n PosixConditionHandle() = default;\n explicit PosixConditionHandle(bool);\n explicit PosixConditionHandle(pthread_t thread);\n PosixConditionHandle(bool manual_reset, bool initial_state);\n PosixConditionHandle(uint32_t initial_count, uint32_t maximum_count);\n ~PosixConditionHandle() override = default;\n\n PosixCondition& condition() override { return handle_; }\n void* native_handle() const override { return handle_.native_handle(); }\n\n protected:\n PosixCondition handle_;\n friend PosixCondition;\n};\n\ntemplate <>\nPosixConditionHandle::PosixConditionHandle(uint32_t initial_count,\n uint32_t maximum_count)\n : handle_(initial_count, maximum_count) {}\n\ntemplate <>\nPosixConditionHandle::PosixConditionHandle(bool initial_owner)\n : handle_(initial_owner) {}\n\ntemplate <>\nPosixConditionHandle::PosixConditionHandle(bool manual_reset)\n : handle_(manual_reset) {}\n\ntemplate <>\nPosixConditionHandle::PosixConditionHandle(bool manual_reset,\n bool initial_state)\n : handle_(manual_reset, initial_state) {}\n\ntemplate <>\nPosixConditionHandle::PosixConditionHandle(pthread_t thread)\n : handle_(thread) {}\n\nWaitResult Wait(WaitHandle* wait_handle, bool is_alertable,\n std::chrono::milliseconds timeout) {\n auto posix_wait_handle = dynamic_cast(wait_handle);\n if (posix_wait_handle == nullptr) {\n return WaitResult::kFailed;\n }\n if (is_alertable) alertable_state_ = true;\n auto result = posix_wait_handle->condition().Wait(timeout);\n if (is_alertable) alertable_state_ = false;\n return result;\n}\n\nWaitResult SignalAndWait(WaitHandle* wait_handle_to_signal,\n WaitHandle* wait_handle_to_wait_on, bool is_alertable,\n std::chrono::milliseconds timeout) {\n auto result = WaitResult::kFailed;\n auto posix_wait_handle_to_signal =\n dynamic_cast(wait_handle_to_signal);\n auto posix_wait_handle_to_wait_on =\n dynamic_cast(wait_handle_to_wait_on);\n if (posix_wait_handle_to_signal == nullptr ||\n posix_wait_handle_to_wait_on == nullptr) {\n return WaitResult::kFailed;\n }\n if (is_alertable) alertable_state_ = true;\n if (posix_wait_handle_to_signal->condition().Signal()) {\n result = posix_wait_handle_to_wait_on->condition().Wait(timeout);\n }\n if (is_alertable) alertable_state_ = false;\n return result;\n}\n\nstd::pair WaitMultiple(WaitHandle* wait_handles[],\n size_t wait_handle_count,\n bool wait_all, bool is_alertable,\n std::chrono::milliseconds timeout) {\n std::vector conditions;\n conditions.reserve(wait_handle_count);\n for (size_t i = 0u; i < wait_handle_count; ++i) {\n auto handle = dynamic_cast(wait_handles[i]);\n if (handle == nullptr) {\n return std::make_pair(WaitResult::kFailed, 0);\n }\n conditions.push_back(&handle->condition());\n }\n if (is_alertable) alertable_state_ = true;\n auto result = PosixConditionBase::WaitMultiple(std::move(conditions),\n wait_all, timeout);\n if (is_alertable) alertable_state_ = false;\n return result;\n}\n\nclass PosixEvent : public PosixConditionHandle {\n public:\n PosixEvent(bool manual_reset, bool initial_state)\n : PosixConditionHandle(manual_reset, initial_state) {}\n ~PosixEvent() override = default;\n void Set() override { handle_.Signal(); }\n void Reset() override { handle_.Reset(); }\n void Pulse() override {\n using namespace std::chrono_literals;\n handle_.Signal();\n MaybeYield();\n Sleep(10us);\n handle_.Reset();\n }\n};\n\nstd::unique_ptr Event::CreateManualResetEvent(bool initial_state) {\n return std::make_unique(true, initial_state);\n}\n\nstd::unique_ptr Event::CreateAutoResetEvent(bool initial_state) {\n return std::make_unique(false, initial_state);\n}\n\nclass PosixSemaphore : public PosixConditionHandle {\n public:\n PosixSemaphore(int initial_count, int maximum_count)\n : PosixConditionHandle(static_cast(initial_count),\n static_cast(maximum_count)) {}\n ~PosixSemaphore() override = default;\n bool Release(int release_count, int* out_previous_count) override {\n if (release_count < 1) {\n return false;\n }\n return handle_.Release(static_cast(release_count),\n out_previous_count);\n }\n};\n\nstd::unique_ptr Semaphore::Create(int initial_count,\n int maximum_count) {\n if (initial_count < 0 || initial_count > maximum_count ||\n maximum_count <= 0) {\n return nullptr;\n }\n return std::make_unique(initial_count, maximum_count);\n}\n\nclass PosixMutant : public PosixConditionHandle {\n public:\n explicit PosixMutant(bool initial_owner)\n : PosixConditionHandle(initial_owner) {}\n ~PosixMutant() override = default;\n bool Release() override { return handle_.Release(); }\n};\n\nstd::unique_ptr Mutant::Create(bool initial_owner) {\n return std::make_unique(initial_owner);\n}\n\nclass PosixTimer : public PosixConditionHandle {\n using WClock_ = Timer::WClock_;\n using GClock_ = Timer::GClock_;\n\n public:\n explicit PosixTimer(bool manual_reset) : PosixConditionHandle(manual_reset) {}\n ~PosixTimer() override = default;\n\n bool SetOnceAfter(xe::chrono::hundrednanoseconds rel_time,\n std::function opt_callback = nullptr) override {\n return SetOnceAt(GClock_::now() + rel_time, std::move(opt_callback));\n }\n bool SetOnceAt(WClock_::time_point due_time,\n std::function opt_callback = nullptr) override {\n return SetOnceAt(date::clock_cast(due_time),\n std::move(opt_callback));\n };\n bool SetOnceAt(GClock_::time_point due_time,\n std::function opt_callback = nullptr) override {\n handle_.SetOnce(due_time, std::move(opt_callback));\n return true;\n }\n\n bool SetRepeatingAfter(\n xe::chrono::hundrednanoseconds rel_time, std::chrono::milliseconds period,\n std::function opt_callback = nullptr) override {\n return SetRepeatingAt(GClock_::now() + rel_time, period,\n std::move(opt_callback));\n }\n bool SetRepeatingAt(WClock_::time_point due_time,\n std::chrono::milliseconds period,\n std::function opt_callback = nullptr) override {\n return SetRepeatingAt(date::clock_cast(due_time), period,\n std::move(opt_callback));\n }\n bool SetRepeatingAt(GClock_::time_point due_time,\n std::chrono::milliseconds period,\n std::function opt_callback = nullptr) override {\n handle_.SetRepeating(due_time, period, std::move(opt_callback));\n return true;\n }\n bool Cancel() override {\n handle_.Cancel();\n return true;\n }\n};\n\nstd::unique_ptr Timer::CreateManualResetTimer() {\n return std::make_unique(true);\n}\n\nstd::unique_ptr Timer::CreateSynchronizationTimer() {\n return std::make_unique(false);\n}\n\nclass PosixThread : public PosixConditionHandle {\n public:\n PosixThread() = default;\n explicit PosixThread(pthread_t thread) : PosixConditionHandle(thread) {}\n ~PosixThread() override = default;\n\n bool Initialize(CreationParameters params,\n std::function start_routine) {\n auto start_data =\n new ThreadStartData({std::move(start_routine), false, this});\n return handle_.Initialize(params, start_data);\n }\n\n void set_name(std::string name) override {\n handle_.WaitStarted();\n Thread::set_name(name);\n if (name.length() > 15) {\n name = name.substr(0, 15);\n }\n handle_.set_name(name);\n }\n\n uint32_t system_id() const override { return handle_.system_id(); }\n\n uint64_t affinity_mask() override { return handle_.affinity_mask(); }\n void set_affinity_mask(uint64_t mask) override {\n handle_.set_affinity_mask(mask);\n }\n\n int priority() override { return handle_.priority(); }\n void set_priority(int new_priority) override {\n handle_.set_priority(new_priority);\n }\n\n void QueueUserCallback(std::function callback) override {\n handle_.QueueUserCallback(std::move(callback));\n }\n\n bool Resume(uint32_t* out_previous_suspend_count) override {\n return handle_.Resume(out_previous_suspend_count);\n }\n\n bool Suspend(uint32_t* out_previous_suspend_count) override {\n return handle_.Suspend(out_previous_suspend_count);\n }\n\n void Terminate(int exit_code) override { handle_.Terminate(exit_code); }\n\n void WaitSuspended() { handle_.WaitSuspended(); }\n};\n\nthread_local PosixThread* current_thread_ = nullptr;\n\nvoid* PosixCondition::ThreadStartRoutine(void* parameter) {\n#if !XE_PLATFORM_ANDROID\n if (pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, nullptr) != 0) {\n assert_always();\n }\n#endif\n threading::set_name(\"\");\n\n auto start_data = static_cast(parameter);\n assert_not_null(start_data);\n assert_not_null(start_data->thread_obj);\n\n auto thread = dynamic_cast(start_data->thread_obj);\n auto start_routine = std::move(start_data->start_routine);\n auto create_suspended = start_data->create_suspended;\n delete start_data;\n\n current_thread_ = thread;\n {\n std::unique_lock lock(thread->handle_.state_mutex_);\n thread->handle_.state_ =\n create_suspended ? State::kSuspended : State::kRunning;\n thread->handle_.state_signal_.notify_all();\n }\n\n if (create_suspended) {\n std::unique_lock lock(thread->handle_.state_mutex_);\n thread->handle_.suspend_count_ = 1;\n thread->handle_.state_signal_.wait(\n lock, [thread] { return thread->handle_.suspend_count_ == 0; });\n }\n\n start_routine();\n\n {\n std::unique_lock lock(thread->handle_.state_mutex_);\n thread->handle_.state_ = State::kFinished;\n }\n\n std::unique_lock lock(mutex_);\n thread->handle_.exit_code_ = 0;\n thread->handle_.signaled_ = true;\n cond_.notify_all();\n\n current_thread_ = nullptr;\n return nullptr;\n}\n\nstd::unique_ptr Thread::Create(CreationParameters params,\n std::function start_routine) {\n install_signal_handler(SignalType::kThreadSuspend);\n install_signal_handler(SignalType::kThreadUserCallback);\n#if XE_PLATFORM_ANDROID\n install_signal_handler(SignalType::kThreadTerminate);\n#endif\n auto thread = std::make_unique();\n if (!thread->Initialize(params, std::move(start_routine))) return nullptr;\n assert_not_null(thread);\n return thread;\n}\n\nThread* Thread::GetCurrentThread() {\n if (current_thread_) {\n return current_thread_;\n }\n\n // Should take this route only for threads not created by Thread::Create.\n // The only thread not created by Thread::Create should be the main thread.\n pthread_t handle = pthread_self();\n\n current_thread_ = new PosixThread(handle);\n // TODO(bwrsandman): Disabling deleting thread_local current thread to prevent\n // assert in destructor. Since this is thread local, the\n // \"memory leaking\" is controlled.\n // atexit([] { delete current_thread_; });\n\n return current_thread_;\n}\n\nvoid Thread::Exit(int exit_code) {\n if (current_thread_) {\n current_thread_->Terminate(exit_code);\n } else {\n // Should only happen with the main thread\n pthread_exit(reinterpret_cast(exit_code));\n }\n // Function must not return\n assert_always();\n}\n\nvoid set_name(const std::string_view name) {\n pthread_setname_np(pthread_self(), std::string(name).c_str());\n#if XE_PLATFORM_ANDROID\n if (!android_pthread_getname_np_ && current_thread_) {\n current_thread_->condition().SetAndroidPreApi26Name(name);\n }\n#endif\n}\n\nstatic void signal_handler(int signal, siginfo_t* info, void* /*context*/) {\n switch (GetSystemSignalType(signal)) {\n case SignalType::kThreadSuspend: {\n assert_not_null(current_thread_);\n current_thread_->WaitSuspended();\n } break;\n case SignalType::kThreadUserCallback: {\n assert_not_null(info->si_value.sival_ptr);\n auto p_thread =\n static_cast*>(info->si_value.sival_ptr);\n if (alertable_state_) {\n p_thread->CallUserCallback();\n }\n } break;\n#if XE_PLATFORM_ANDROID\n case SignalType::kThreadTerminate: {\n pthread_exit(reinterpret_cast(-1));\n } break;\n#endif\n default:\n assert_always();\n }\n}\n\n} // namespace threading\n} // namespace xe\n" }, { "path": "src/xenia/base/threading_timer_queue.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n\n#include \"third_party/disruptorplus/include/disruptorplus/multi_threaded_claim_strategy.hpp\"\n#include \"third_party/disruptorplus/include/disruptorplus/ring_buffer.hpp\"\n#include \"third_party/disruptorplus/include/disruptorplus/sequence_barrier.hpp\"\n#include \"third_party/disruptorplus/include/disruptorplus/spin_wait.hpp\"\n#include \"third_party/disruptorplus/include/disruptorplus/spin_wait_strategy.hpp\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/threading.h\"\n#include \"xenia/base/threading_timer_queue.h\"\n\nnamespace dp = disruptorplus;\n\nnamespace xe {\nnamespace threading {\n\nusing WaitItem = TimerQueueWaitItem;\n\nclass TimerQueue {\n public:\n using clock = WaitItem::clock;\n static_assert(clock::is_steady);\n\n public:\n TimerQueue()\n : buffer_(kWaitCount),\n wait_strategy_(),\n claim_strategy_(kWaitCount, wait_strategy_),\n consumed_(wait_strategy_),\n shutdown_(false) {\n claim_strategy_.add_claim_barrier(consumed_);\n dispatch_thread_ = std::thread(&TimerQueue::TimerThreadMain, this);\n }\n\n ~TimerQueue() {\n shutdown_.store(true, std::memory_order_release);\n\n // Kick dispatch thread to check shutdown flag\n auto wait_item = std::make_shared(nullptr, nullptr, this,\n clock::time_point::min(),\n clock::duration::zero());\n wait_item->Disarm();\n QueueTimer(std::move(wait_item));\n\n dispatch_thread_.join();\n }\n\n void TimerThreadMain() {\n dp::sequence_t next_sequence = 0;\n const auto comp = [](const std::shared_ptr& left,\n const std::shared_ptr& right) {\n return left->due_ < right->due_;\n };\n\n xe::threading::set_name(\"xe::threading::TimerQueue\");\n\n while (!shutdown_.load(std::memory_order_relaxed)) {\n {\n // Consume new wait items and add them to sorted wait queue\n dp::sequence_t available = claim_strategy_.wait_until_published(\n next_sequence, next_sequence - 1,\n wait_queue_.empty() ? clock::time_point::max()\n : wait_queue_.front()->due_);\n\n // Check for timeout\n if (available != next_sequence - 1) {\n std::forward_list> wait_items;\n do {\n wait_items.push_front(std::move(buffer_[next_sequence]));\n } while (next_sequence++ != available);\n\n consumed_.publish(available);\n\n wait_items.sort(comp);\n wait_queue_.merge(wait_items, comp);\n }\n }\n\n {\n // Check wait queue, invoke callbacks and reschedule\n std::forward_list> wait_items;\n while (!wait_queue_.empty() &&\n wait_queue_.front()->due_ <= clock::now()) {\n auto wait_item = std::move(wait_queue_.front());\n wait_queue_.pop_front();\n\n // Ensure that it isn't disarmed\n auto state = WaitItem::State::kIdle;\n if (wait_item->state_.compare_exchange_strong(\n state, WaitItem::State::kInCallback,\n std::memory_order_acq_rel)) {\n // Possibility to dispatch to a thread pool here\n assert_not_null(wait_item->callback_);\n wait_item->callback_(wait_item->userdata_);\n\n if (wait_item->interval_ != clock::duration::zero() &&\n wait_item->state_.load(std::memory_order_acquire) !=\n WaitItem::State::kInCallbackSelfDisarmed) {\n // Item is recurring and didn't self-disarm during callback:\n wait_item->due_ += wait_item->interval_;\n wait_item->state_.store(WaitItem::State::kIdle,\n std::memory_order_release);\n wait_items.push_front(std::move(wait_item));\n } else {\n wait_item->state_.store(WaitItem::State::kDisarmed,\n std::memory_order_release);\n }\n } else {\n // Specifically, kInCallback is illegal here\n assert_true(WaitItem::State::kDisarmed == state);\n }\n }\n wait_items.sort(comp);\n wait_queue_.merge(wait_items, comp);\n }\n }\n }\n\n std::weak_ptr QueueTimer(std::shared_ptr wait_item) {\n auto wait_item_weak = std::weak_ptr(wait_item);\n\n // Mitigate callback flooding\n wait_item->due_ =\n std::max(clock::now() - wait_item->interval_, wait_item->due_);\n\n auto sequence = claim_strategy_.claim_one();\n buffer_[sequence] = std::move(wait_item);\n claim_strategy_.publish(sequence);\n\n return wait_item_weak;\n }\n\n const std::thread& dispatch_thread() const { return dispatch_thread_; }\n\n private:\n // This ring buffer will be used to introduce timers queued by the public API\n static constexpr size_t kWaitCount = 512;\n dp::ring_buffer> buffer_;\n dp::spin_wait_strategy wait_strategy_;\n dp::multi_threaded_claim_strategy claim_strategy_;\n dp::sequence_barrier consumed_;\n\n // This is a _sorted_ (ascending due_) list of active timers managed by a\n // dedicated thread\n std::forward_list> wait_queue_;\n std::atomic_bool shutdown_;\n std::thread dispatch_thread_;\n};\n\nxe::threading::TimerQueue timer_queue_;\n\nvoid TimerQueueWaitItem::Disarm() {\n State state;\n\n // Special case for calling from a callback itself\n if (std::this_thread::get_id() == parent_queue_->dispatch_thread().get_id()) {\n state = State::kInCallback;\n if (state_.compare_exchange_strong(state, State::kInCallbackSelfDisarmed,\n std::memory_order_acq_rel)) {\n // If we are self disarming from the callback set this special state and\n // exit\n return;\n }\n // Normal case can handle the rest\n }\n\n dp::spin_wait spinner;\n state = State::kIdle;\n // Classes which hold WaitItems will often call Disarm() to cancel them during\n // destruction. This may lead to race conditions when the dispatch thread\n // executes a callback which accesses memory that is freed simultaneously due\n // to this. Therefore, we need to guarantee that no callbacks will be running\n // once Disarm() has returned.\n while (!state_.compare_exchange_weak(state, State::kDisarmed,\n std::memory_order_acq_rel)) {\n if (state == State::kDisarmed) {\n // Do not break for kInCallbackSelfDisarmed and keep spinning in order to\n // meet guarantees\n break;\n }\n state = State::kIdle;\n spinner.spin_once();\n }\n}\n\nstd::weak_ptr QueueTimerOnce(std::function callback,\n void* userdata,\n WaitItem::clock::time_point due) {\n return timer_queue_.QueueTimer(\n std::make_shared(std::move(callback), userdata, &timer_queue_,\n due, WaitItem::clock::duration::zero()));\n}\n\nstd::weak_ptr QueueTimerRecurring(\n std::function callback, void* userdata,\n WaitItem::clock::time_point due, WaitItem::clock::duration interval) {\n return timer_queue_.QueueTimer(std::make_shared(\n std::move(callback), userdata, &timer_queue_, due, interval));\n}\n\n} // namespace threading\n} // namespace xe\n" }, { "path": "src/xenia/base/threading_timer_queue.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_THREADING_TIMER_QUEUE_H_\n#define XENIA_BASE_THREADING_TIMER_QUEUE_H_\n\n#include \n#include \n#include \n#include \n\n// This is a platform independent implementation of a timer queue similar to\n// Windows CreateTimerQueueTimer with WT_EXECUTEINTIMERTHREAD.\n\nnamespace xe::threading {\n\nclass TimerQueue;\n\nstruct TimerQueueWaitItem {\n using clock = std::chrono::steady_clock;\n\n TimerQueueWaitItem(std::function callback, void* userdata,\n TimerQueue* parent_queue, clock::time_point due,\n clock::duration interval)\n : callback_(std::move(callback)),\n userdata_(userdata),\n parent_queue_(parent_queue),\n due_(due),\n interval_(interval),\n state_(State::kIdle) {}\n\n // Cancel the pending wait item. No callbacks will be running after this call.\n // The function blocks if a callback is running and returns only after the\n // callback has finished (except when called from the corresponding callback\n // itself, where it will mark the wait item for disarmament and return\n // immediately). Deadlocks are possible when a lock is held during disamament\n // and the corresponding callback is running concurrently, trying to acquire\n // said lock.\n void Disarm();\n\n friend TimerQueue;\n\n private:\n enum class State : uint_least8_t {\n kIdle = 0, // Waiting for the due time\n kInCallback, // Callback is being executed\n kInCallbackSelfDisarmed, // Callback is being executed and disarmed itself\n kDisarmed // Disarmed, waiting for destruction\n };\n static_assert(std::atomic::is_always_lock_free);\n\n std::function callback_;\n void* userdata_;\n TimerQueue* parent_queue_;\n clock::time_point due_;\n clock::duration interval_; // zero if not recurring\n std::atomic state_;\n};\n\nstd::weak_ptr QueueTimerOnce(\n std::function callback, void* userdata,\n TimerQueueWaitItem::clock::time_point due);\n\n// Callback is first executed at due, then again repeatedly after interval\n// passes (unless interval == 0). The first callback will be scheduled at\n// `max(now() - interval, due)` to mitigate callback flooding.\nstd::weak_ptr QueueTimerRecurring(\n std::function callback, void* userdata,\n TimerQueueWaitItem::clock::time_point due,\n TimerQueueWaitItem::clock::duration interval);\n} // namespace xe::threading\n\n#endif\n" }, { "path": "src/xenia/base/threading_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/chrono_steady_cast.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/platform_win.h\"\n#include \"xenia/base/threading.h\"\n#include \"xenia/base/threading_timer_queue.h\"\n\n#define LOG_LASTERROR() \\\n { XELOGI(\"Win32 Error 0x{:08X} in \" __FUNCTION__ \"(...)\", GetLastError()); }\n\ntypedef HANDLE (*SetThreadDescriptionFn)(HANDLE hThread,\n PCWSTR lpThreadDescription);\n\nnamespace xe {\nnamespace threading {\n\nvoid EnableAffinityConfiguration() {\n HANDLE process_handle = GetCurrentProcess();\n DWORD_PTR process_affinity_mask;\n DWORD_PTR system_affinity_mask;\n GetProcessAffinityMask(process_handle, &process_affinity_mask,\n &system_affinity_mask);\n SetProcessAffinityMask(process_handle, system_affinity_mask);\n}\n\nuint32_t current_thread_system_id() {\n return static_cast(GetCurrentThreadId());\n}\n\n// https://msdn.microsoft.com/en-us/library/xcb2z8hs.aspx\n#pragma pack(push, 8)\nstruct THREADNAME_INFO {\n DWORD dwType; // Must be 0x1000.\n LPCSTR szName; // Pointer to name (in user addr space).\n DWORD dwThreadID; // Thread ID (-1=caller thread).\n DWORD dwFlags; // Reserved for future use, must be zero.\n};\n#pragma pack(pop)\n\nvoid raise_thread_name_exception(HANDLE thread, const std::string& name) {\n if (!IsDebuggerPresent()) {\n return;\n }\n THREADNAME_INFO info;\n info.dwType = 0x1000;\n info.szName = name.c_str();\n info.dwThreadID = ::GetThreadId(thread);\n info.dwFlags = 0;\n __try {\n RaiseException(0x406D1388, 0, sizeof(info) / sizeof(ULONG_PTR),\n reinterpret_cast(&info));\n } __except (EXCEPTION_EXECUTE_HANDLER) { // NOLINT\n }\n}\n\nstatic void set_name(HANDLE thread, const std::string_view name) {\n auto kernel = GetModuleHandleW(L\"kernel32.dll\");\n if (kernel) {\n auto func =\n (SetThreadDescriptionFn)GetProcAddress(kernel, \"SetThreadDescription\");\n if (func) {\n auto u16name = xe::to_utf16(name);\n func(thread, reinterpret_cast(u16name.c_str()));\n }\n }\n raise_thread_name_exception(thread, std::string(name));\n}\n\nvoid set_name(const std::string_view name) {\n set_name(GetCurrentThread(), name);\n}\n\nvoid MaybeYield() {\n SwitchToThread();\n MemoryBarrier();\n}\n\nvoid SyncMemory() { MemoryBarrier(); }\n\nvoid Sleep(std::chrono::microseconds duration) {\n if (duration.count() < 100) {\n MaybeYield();\n } else {\n ::Sleep(static_cast(duration.count() / 1000));\n }\n}\n\nSleepResult AlertableSleep(std::chrono::microseconds duration) {\n if (SleepEx(static_cast(duration.count() / 1000), TRUE) ==\n WAIT_IO_COMPLETION) {\n return SleepResult::kAlerted;\n }\n return SleepResult::kSuccess;\n}\n\nTlsHandle AllocateTlsHandle() { return TlsAlloc(); }\n\nbool FreeTlsHandle(TlsHandle handle) { return TlsFree(handle) ? true : false; }\n\nuintptr_t GetTlsValue(TlsHandle handle) {\n return reinterpret_cast(TlsGetValue(handle));\n}\n\nbool SetTlsValue(TlsHandle handle, uintptr_t value) {\n return TlsSetValue(handle, reinterpret_cast(value)) ? true : false;\n}\n\ntemplate \nclass Win32Handle : public T {\n public:\n explicit Win32Handle(HANDLE handle) : handle_(handle) {\n assert_not_null(handle);\n }\n ~Win32Handle() override {\n CloseHandle(handle_);\n handle_ = nullptr;\n }\n\n protected:\n void* native_handle() const override { return handle_; }\n\n HANDLE handle_ = nullptr;\n};\n\nWaitResult Wait(WaitHandle* wait_handle, bool is_alertable,\n std::chrono::milliseconds timeout) {\n HANDLE handle = wait_handle->native_handle();\n DWORD result = WaitForSingleObjectEx(handle, DWORD(timeout.count()),\n is_alertable ? TRUE : FALSE);\n switch (result) {\n case WAIT_OBJECT_0:\n return WaitResult::kSuccess;\n case WAIT_ABANDONED:\n return WaitResult::kAbandoned;\n case WAIT_IO_COMPLETION:\n return WaitResult::kUserCallback;\n case WAIT_TIMEOUT:\n return WaitResult::kTimeout;\n default:\n case WAIT_FAILED:\n return WaitResult::kFailed;\n }\n}\n\nWaitResult SignalAndWait(WaitHandle* wait_handle_to_signal,\n WaitHandle* wait_handle_to_wait_on, bool is_alertable,\n std::chrono::milliseconds timeout) {\n HANDLE handle_to_signal = wait_handle_to_signal->native_handle();\n HANDLE handle_to_wait_on = wait_handle_to_wait_on->native_handle();\n DWORD result =\n SignalObjectAndWait(handle_to_signal, handle_to_wait_on,\n DWORD(timeout.count()), is_alertable ? TRUE : FALSE);\n switch (result) {\n case WAIT_OBJECT_0:\n return WaitResult::kSuccess;\n case WAIT_ABANDONED:\n return WaitResult::kAbandoned;\n case WAIT_IO_COMPLETION:\n return WaitResult::kUserCallback;\n case WAIT_TIMEOUT:\n return WaitResult::kTimeout;\n default:\n case WAIT_FAILED:\n return WaitResult::kFailed;\n }\n}\n\nstd::pair WaitMultiple(WaitHandle* wait_handles[],\n size_t wait_handle_count,\n bool wait_all, bool is_alertable,\n std::chrono::milliseconds timeout) {\n std::vector handles(wait_handle_count);\n for (size_t i = 0; i < wait_handle_count; ++i) {\n handles[i] = wait_handles[i]->native_handle();\n }\n DWORD result = WaitForMultipleObjectsEx(\n DWORD(handles.size()), handles.data(), wait_all ? TRUE : FALSE,\n DWORD(timeout.count()), is_alertable ? TRUE : FALSE);\n if (result >= WAIT_OBJECT_0 && result < WAIT_OBJECT_0 + handles.size()) {\n return std::pair(WaitResult::kSuccess,\n result - WAIT_OBJECT_0);\n } else if (result >= WAIT_ABANDONED_0 &&\n result < WAIT_ABANDONED_0 + handles.size()) {\n return std::pair(WaitResult::kAbandoned,\n result - WAIT_ABANDONED_0);\n }\n switch (result) {\n case WAIT_IO_COMPLETION:\n return std::pair(WaitResult::kUserCallback, 0);\n case WAIT_TIMEOUT:\n return std::pair(WaitResult::kTimeout, 0);\n default:\n case WAIT_FAILED:\n return std::pair(WaitResult::kFailed, 0);\n }\n}\n\nclass Win32Event : public Win32Handle {\n public:\n explicit Win32Event(HANDLE handle) : Win32Handle(handle) {}\n ~Win32Event() override = default;\n void Set() override { SetEvent(handle_); }\n void Reset() override { ResetEvent(handle_); }\n void Pulse() override { PulseEvent(handle_); }\n};\n\nstd::unique_ptr Event::CreateManualResetEvent(bool initial_state) {\n HANDLE handle =\n CreateEvent(nullptr, TRUE, initial_state ? TRUE : FALSE, nullptr);\n if (handle) {\n return std::make_unique(handle);\n } else {\n LOG_LASTERROR();\n return nullptr;\n }\n}\n\nstd::unique_ptr Event::CreateAutoResetEvent(bool initial_state) {\n HANDLE handle =\n CreateEvent(nullptr, FALSE, initial_state ? TRUE : FALSE, nullptr);\n if (handle) {\n return std::make_unique(handle);\n } else {\n LOG_LASTERROR();\n return nullptr;\n }\n}\n\nclass Win32Semaphore : public Win32Handle {\n public:\n explicit Win32Semaphore(HANDLE handle) : Win32Handle(handle) {}\n ~Win32Semaphore() override = default;\n bool Release(int release_count, int* out_previous_count) override {\n return ReleaseSemaphore(handle_, release_count,\n reinterpret_cast(out_previous_count))\n ? true\n : false;\n }\n};\n\nstd::unique_ptr Semaphore::Create(int initial_count,\n int maximum_count) {\n HANDLE handle =\n CreateSemaphore(nullptr, initial_count, maximum_count, nullptr);\n if (handle) {\n return std::make_unique(handle);\n } else {\n LOG_LASTERROR();\n return nullptr;\n }\n}\n\nclass Win32Mutant : public Win32Handle {\n public:\n explicit Win32Mutant(HANDLE handle) : Win32Handle(handle) {}\n ~Win32Mutant() = default;\n bool Release() override { return ReleaseMutex(handle_) ? true : false; }\n};\n\nstd::unique_ptr Mutant::Create(bool initial_owner) {\n HANDLE handle = CreateMutex(nullptr, initial_owner ? TRUE : FALSE, nullptr);\n if (handle) {\n return std::make_unique(handle);\n } else {\n LOG_LASTERROR();\n return nullptr;\n }\n}\n\nclass Win32Timer : public Win32Handle {\n using WClock_ = Timer::WClock_;\n using GClock_ = Timer::GClock_;\n\n public:\n explicit Win32Timer(HANDLE handle) : Win32Handle(handle) {}\n ~Win32Timer() = default;\n\n bool SetOnceAfter(xe::chrono::hundrednanoseconds rel_time,\n std::function opt_callback) override {\n return SetOnceAt(WClock_::now() + rel_time, std::move(opt_callback));\n }\n bool SetOnceAt(GClock_::time_point due_time,\n std::function opt_callback) override {\n return SetOnceAt(date::clock_cast(due_time),\n std::move(opt_callback));\n }\n bool SetOnceAt(WClock_::time_point due_time,\n std::function opt_callback) override {\n std::lock_guard lock(mutex_);\n callback_ = std::move(opt_callback);\n LARGE_INTEGER due_time_li;\n due_time_li.QuadPart = WClock_::to_file_time(due_time);\n auto completion_routine =\n callback_ ? reinterpret_cast(CompletionRoutine)\n : NULL;\n return SetWaitableTimer(handle_, &due_time_li, 0, completion_routine, this,\n FALSE)\n ? true\n : false;\n }\n\n bool SetRepeatingAfter(\n xe::chrono::hundrednanoseconds rel_time, std::chrono::milliseconds period,\n std::function opt_callback = nullptr) override {\n return SetRepeatingAt(WClock_::now() + rel_time, period,\n std::move(opt_callback));\n }\n bool SetRepeatingAt(GClock_::time_point due_time,\n std::chrono::milliseconds period,\n std::function opt_callback = nullptr) {\n return SetRepeatingAt(date::clock_cast(due_time), period,\n std::move(opt_callback));\n }\n bool SetRepeatingAt(WClock_::time_point due_time,\n std::chrono::milliseconds period,\n std::function opt_callback) override {\n std::lock_guard lock(mutex_);\n callback_ = std::move(opt_callback);\n LARGE_INTEGER due_time_li;\n due_time_li.QuadPart = WClock_::to_file_time(due_time);\n auto completion_routine =\n callback_ ? reinterpret_cast(CompletionRoutine)\n : NULL;\n return SetWaitableTimer(handle_, &due_time_li, int32_t(period.count()),\n completion_routine, this, FALSE)\n ? true\n : false;\n }\n\n bool Cancel() override {\n // Reset the callback immediately so that any completions don't call it.\n std::lock_guard lock(mutex_);\n callback_ = nullptr;\n return CancelWaitableTimer(handle_) ? true : false;\n }\n\n private:\n static void CompletionRoutine(Win32Timer* timer, DWORD timer_low,\n DWORD timer_high) {\n // As the callback may reset the timer, store local.\n std::function callback;\n {\n std::lock_guard lock(timer->mutex_);\n callback = timer->callback_;\n }\n callback();\n }\n\n std::mutex mutex_;\n std::function callback_;\n};\n\nstd::unique_ptr Timer::CreateManualResetTimer() {\n HANDLE handle = CreateWaitableTimer(NULL, TRUE, NULL);\n if (handle) {\n return std::make_unique(handle);\n } else {\n LOG_LASTERROR();\n return nullptr;\n }\n}\n\nstd::unique_ptr Timer::CreateSynchronizationTimer() {\n HANDLE handle = CreateWaitableTimer(NULL, FALSE, NULL);\n if (handle) {\n return std::make_unique(handle);\n } else {\n LOG_LASTERROR();\n return nullptr;\n }\n}\n\nclass Win32Thread : public Win32Handle {\n public:\n explicit Win32Thread(HANDLE handle) : Win32Handle(handle) {}\n ~Win32Thread() = default;\n\n void set_name(std::string name) override {\n xe::threading::set_name(handle_, name);\n Thread::set_name(name);\n }\n\n int32_t priority() override { return GetThreadPriority(handle_); }\n uint32_t system_id() const override { return GetThreadId(handle_); }\n\n void set_priority(int32_t new_priority) override {\n SetThreadPriority(handle_, new_priority);\n }\n\n uint64_t affinity_mask() override {\n uint64_t value = 0;\n SetThreadAffinityMask(handle_, reinterpret_cast(&value));\n return value;\n }\n\n void set_affinity_mask(uint64_t new_affinity_mask) override {\n SetThreadAffinityMask(handle_, new_affinity_mask);\n }\n\n struct ApcData {\n std::function callback;\n };\n static void NTAPI DispatchApc(ULONG_PTR parameter) {\n auto apc_data = reinterpret_cast(parameter);\n apc_data->callback();\n delete apc_data;\n }\n\n void QueueUserCallback(std::function callback) override {\n auto apc_data = new ApcData({std::move(callback)});\n QueueUserAPC(DispatchApc, handle_, reinterpret_cast(apc_data));\n }\n\n bool Resume(uint32_t* out_previous_suspend_count = nullptr) override {\n if (out_previous_suspend_count) {\n *out_previous_suspend_count = 0;\n }\n DWORD result = ResumeThread(handle_);\n if (result == UINT_MAX) {\n return false;\n }\n if (out_previous_suspend_count) {\n *out_previous_suspend_count = result;\n }\n return true;\n }\n\n bool Suspend(uint32_t* out_previous_suspend_count = nullptr) override {\n if (out_previous_suspend_count) {\n *out_previous_suspend_count = 0;\n }\n DWORD result = SuspendThread(handle_);\n if (result == UINT_MAX) {\n return false;\n }\n if (out_previous_suspend_count) {\n *out_previous_suspend_count = result;\n }\n return true;\n }\n\n void Terminate(int exit_code) override {\n TerminateThread(handle_, exit_code);\n }\n\n private:\n void AssertCallingThread() {\n assert_true(GetCurrentThreadId() == GetThreadId(handle_));\n }\n};\n\nthread_local std::unique_ptr current_thread_ = nullptr;\n\nstruct ThreadStartData {\n std::function start_routine;\n};\nDWORD WINAPI ThreadStartRoutine(LPVOID parameter) {\n current_thread_ = std::make_unique(::GetCurrentThread());\n\n auto start_data = reinterpret_cast(parameter);\n start_data->start_routine();\n delete start_data;\n return 0;\n}\n\nstd::unique_ptr Thread::Create(CreationParameters params,\n std::function start_routine) {\n auto start_data = new ThreadStartData({std::move(start_routine)});\n HANDLE handle =\n CreateThread(NULL, params.stack_size, ThreadStartRoutine, start_data,\n params.create_suspended ? CREATE_SUSPENDED : 0, NULL);\n if (handle) {\n return std::make_unique(handle);\n } else {\n LOG_LASTERROR();\n delete start_data;\n return nullptr;\n }\n}\n\nThread* Thread::GetCurrentThread() {\n if (current_thread_) {\n return current_thread_.get();\n }\n\n HANDLE handle = ::GetCurrentThread();\n if (handle == INVALID_HANDLE_VALUE) {\n return nullptr;\n }\n\n current_thread_ = std::make_unique(handle);\n return current_thread_.get();\n}\n\nvoid Thread::Exit(int exit_code) { ExitThread(exit_code); }\n\n} // namespace threading\n} // namespace xe\n" }, { "path": "src/xenia/base/type_pool.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_TYPE_POOL_H_\n#define XENIA_BASE_TYPE_POOL_H_\n\n#include \n#include \n\nnamespace xe {\n\ntemplate \nclass TypePool {\n public:\n ~TypePool() { Reset(); }\n\n void Reset() {\n std::lock_guard guard(lock_);\n for (auto it = list_.begin(); it != list_.end(); ++it) {\n T* value = *it;\n delete value;\n }\n list_.clear();\n }\n\n T* Allocate(A arg0) {\n T* result = 0;\n {\n std::lock_guard guard(lock_);\n if (list_.size()) {\n result = list_.back();\n list_.pop_back();\n }\n }\n if (!result) {\n result = new T(arg0);\n }\n return result;\n }\n\n void Release(T* value) {\n std::lock_guard guard(lock_);\n list_.push_back(value);\n }\n\n private:\n std::mutex lock_;\n std::vector list_;\n};\n\n} // namespace xe\n\n#endif // XENIA_BASE_TYPE_POOL_H_\n" }, { "path": "src/xenia/base/utf8.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/base/utf8.h\"\n\n#include \n#include \n#include \n#include \n#include \n\n#define UTF_CPP_CPLUSPLUS 201703L\n#include \"third_party/utfcpp/source/utf8.h\"\n\nnamespace utfcpp = utf8;\n\nusing citer = std::string_view::const_iterator;\nusing utf8_citer = utfcpp::iterator;\n\nnamespace xe::utf8 {\n\nstd::string to_string(char32_t c) {\n std::string result;\n utfcpp::append(c, result);\n return result;\n}\n\nuint32_t lower_ascii(const uint32_t c) {\n return c >= 'A' && c <= 'Z' ? c + 32 : c;\n}\n\nuint32_t upper_ascii(const uint32_t c) {\n return c >= 'A' && c <= 'Z' ? c + 32 : c;\n}\n\nbool equal_ascii_case(const uint32_t l, const uint32_t r) {\n return l == r || lower_ascii(l) == lower_ascii(r);\n}\n\nstd::pair make_citer(const std::string_view view) {\n return {utf8_citer(view.cbegin(), view.cbegin(), view.cend()),\n utf8_citer(view.cend(), view.cbegin(), view.cend())};\n}\n\nstd::pair make_citer(const utf8_citer begin,\n const utf8_citer end) {\n return {utf8_citer(begin.base(), begin.base(), end.base()),\n utf8_citer(end.base(), begin.base(), end.base())};\n}\n\nsize_t byte_length(utf8_citer begin, utf8_citer end) {\n return size_t(std::distance(begin.base(), end.base()));\n}\n\nsize_t count(const std::string_view view) {\n return size_t(utfcpp::distance(view.cbegin(), view.cend()));\n}\n\nstd::string lower_ascii(const std::string_view view) {\n auto [begin, end] = make_citer(view);\n std::string result;\n for (auto it = begin; it != end; ++it) {\n utfcpp::append(char32_t(lower_ascii(*it)), result);\n }\n return result;\n}\n\nstd::string upper_ascii(const std::string_view view) {\n auto [begin, end] = make_citer(view);\n std::string result;\n for (auto it = begin; it != end; ++it) {\n utfcpp::append(char32_t(upper_ascii(*it)), result);\n }\n return result;\n}\n\ntemplate \ninline size_t hash_fnv1a(const std::string_view view) {\n const size_t offset_basis = 0xCBF29CE484222325ull;\n const size_t prime = 0x00000100000001B3ull;\n auto work = [&prime](size_t hash, uint8_t byte_of_data) {\n hash ^= byte_of_data;\n hash *= prime;\n return hash;\n };\n auto hash = offset_basis;\n auto [begin, end] = make_citer(view);\n for (auto it = begin; it != end; ++it) {\n uint32_t c;\n if constexpr (LOWER) {\n c = lower_ascii(*it);\n } else {\n c = *it;\n }\n hash = work(hash, uint8_t((c >> 0) & 0xFF));\n hash = work(hash, uint8_t((c >> 8) & 0xFF));\n hash = work(hash, uint8_t((c >> 16) & 0xFF));\n hash = work(hash, uint8_t((c >> 24) & 0xFF));\n }\n return hash;\n}\n\nsize_t hash_fnv1a(const std::string_view view) {\n return hash_fnv1a(view);\n}\n\nsize_t hash_fnv1a_case(const std::string_view view) {\n return hash_fnv1a(view);\n}\n\n// TODO(gibbed): this is a separate inline function instead of inline within\n// split due to a Clang bug: reference to local binding 'needle_begin' declared\n// in enclosing function 'split'.\ninline utf8_citer find_needle(utf8_citer haystack_it, utf8_citer haystack_end,\n utf8_citer needle_begin, utf8_citer needle_end) {\n return std::find_if(haystack_it, haystack_end, [&](const auto& c) {\n for (auto needle = needle_begin; needle != needle_end; ++needle) {\n if (c == *needle) {\n return true;\n }\n }\n return false;\n });\n}\n\ninline utf8_citer find_needle_case(utf8_citer haystack_it,\n utf8_citer haystack_end,\n utf8_citer needle_begin,\n utf8_citer needle_end) {\n return std::find_if(haystack_it, haystack_end, [&](const auto& c) {\n for (auto needle = needle_begin; needle != needle_end; ++needle) {\n if (equal_ascii_case(c, *needle)) {\n return true;\n }\n }\n return false;\n });\n}\n\nstd::vector split(const std::string_view haystack,\n const std::string_view needles,\n bool remove_empty) {\n std::vector result;\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needles);\n\n auto it = haystack_begin;\n auto last = it;\n for (;;) {\n it = find_needle(it, haystack_end, needle_begin, needle_end);\n if (it == haystack_end) {\n break;\n }\n\n if (it != last) {\n auto offset = byte_length(haystack_begin, last);\n auto length = byte_length(haystack_begin, it) - offset;\n result.push_back(haystack.substr(offset, length));\n } else if (!remove_empty) {\n result.push_back(\"\");\n }\n\n ++it;\n last = it;\n }\n\n if (last != haystack_end) {\n auto offset = byte_length(haystack_begin, last);\n result.push_back(haystack.substr(offset));\n }\n\n return result;\n}\n\nbool equal_z(const std::string_view left, const std::string_view right) {\n if (!left.size()) {\n return !right.size();\n } else if (!right.size()) {\n return false;\n }\n auto [left_begin, left_end] = make_citer(left);\n auto [right_begin, right_end] = make_citer(right);\n auto left_it = left_begin;\n auto right_it = right_begin;\n for (; left_it != left_end && *left_it != 0 && right_it != right_end &&\n *right_it != 0;\n ++left_it, ++right_it) {\n if (*left_it != *right_it) {\n return false;\n }\n }\n return (left_it == left_end || *left_it == 0) &&\n (right_it == right_end || *right_it == 0);\n}\n\nbool equal_case(const std::string_view left, const std::string_view right) {\n if (!left.size()) {\n return !right.size();\n } else if (!right.size()) {\n return false;\n }\n auto [left_begin, left_end] = make_citer(left);\n auto [right_begin, right_end] = make_citer(right);\n return std::equal(left_begin, left_end, right_begin, right_end,\n equal_ascii_case);\n}\n\nbool equal_case_z(const std::string_view left, const std::string_view right) {\n if (!left.size()) {\n return !right.size();\n } else if (!right.size()) {\n return false;\n }\n auto [left_begin, left_end] = make_citer(left);\n auto [right_begin, right_end] = make_citer(right);\n auto left_it = left_begin;\n auto right_it = right_begin;\n for (; left_it != left_end && *left_it != 0 && right_it != right_end &&\n *right_it != 0;\n ++left_it, ++right_it) {\n if (!equal_ascii_case(*left_it, *right_it)) {\n return false;\n }\n }\n return (left_it == left_end || *left_it == 0) &&\n (right_it == right_end || *right_it == 0);\n}\n\nstd::string_view::size_type find_any_of(const std::string_view haystack,\n const std::string_view needles) {\n if (needles.empty()) {\n return std::string_view::size_type(0);\n } else if (haystack.empty()) {\n return std::string_view::npos;\n }\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needles);\n\n auto it = find_needle(haystack_begin, haystack_end, needle_begin, needle_end);\n if (it == haystack_end) {\n return std::string_view::npos;\n }\n\n return std::string_view::size_type(byte_length(haystack_begin, it));\n}\n\nstd::string_view::size_type find_any_of_case(const std::string_view haystack,\n const std::string_view needles) {\n if (needles.empty()) {\n return std::string_view::size_type(0);\n } else if (haystack.empty()) {\n return std::string_view::npos;\n }\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needles);\n\n auto it =\n find_needle_case(haystack_begin, haystack_end, needle_begin, needle_end);\n if (it == haystack_end) {\n return std::string_view::npos;\n }\n\n return std::string_view::size_type(byte_length(haystack_begin, it));\n}\n\nstd::string_view::size_type find_first_of(const std::string_view haystack,\n const std::string_view needle) {\n if (needle.empty()) {\n return std::string_view::size_type(0);\n } else if (haystack.empty()) {\n return std::string_view::npos;\n }\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needle);\n auto needle_count = count(needle);\n\n auto it = haystack_begin;\n for (; it != haystack_end; ++it) {\n it = std::find(it, haystack_end, *needle_begin);\n if (it == haystack_end) {\n return std::string_view::npos;\n }\n\n auto end = it;\n for (size_t i = 0; i < needle_count; ++i) {\n if (end == haystack_end) {\n // not enough room in target for search\n return std::string_view::npos;\n }\n ++end;\n }\n\n auto [sub_start, sub_end] = make_citer(it, end);\n if (std::equal(needle_begin, needle_end, sub_start, sub_end)) {\n return std::string_view::size_type(byte_length(haystack_begin, it));\n }\n }\n\n return std::string_view::npos;\n}\n\nstd::string_view::size_type find_first_of_case(const std::string_view haystack,\n const std::string_view needle) {\n if (needle.empty()) {\n return std::string_view::size_type(0);\n } else if (haystack.empty()) {\n return std::string_view::npos;\n }\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needle);\n auto needle_count = count(needle);\n auto nc = *needle_begin;\n\n auto it = haystack_begin;\n for (; it != haystack_end; ++it) {\n it = std::find_if(it, haystack_end, [&nc](const uint32_t& c) {\n return equal_ascii_case(nc, c);\n });\n if (it == haystack_end) {\n return std::string_view::npos;\n }\n\n auto end = it;\n for (size_t i = 0; i < needle_count; ++i) {\n if (end == haystack_end) {\n // not enough room in target for search\n return std::string_view::npos;\n }\n ++end;\n }\n\n auto [sub_start, sub_end] = make_citer(it, end);\n if (std::equal(needle_begin, needle_end, sub_start, sub_end,\n equal_ascii_case)) {\n return std::string_view::size_type(byte_length(haystack_begin, it));\n }\n }\n\n return std::string_view::npos;\n}\n\nbool starts_with(const std::string_view haystack, char32_t needle) {\n if (haystack.empty()) {\n return false;\n }\n auto [it, end] = make_citer(haystack);\n return *it == uint32_t(needle);\n}\n\nbool starts_with(const std::string_view haystack,\n const std::string_view needle) {\n if (needle.empty()) {\n return true;\n } else if (haystack.empty()) {\n return false;\n }\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needle);\n auto needle_count = count(needle);\n\n auto it = haystack_begin;\n auto end = it;\n for (size_t i = 0; i < needle_count; ++i) {\n if (end == haystack_end) {\n // not enough room in target for search\n return false;\n }\n ++end;\n }\n\n auto [sub_start, sub_end] = make_citer(it, end);\n return std::equal(needle_begin, needle_end, sub_start, sub_end);\n}\n\nbool starts_with_case(const std::string_view haystack,\n const std::string_view needle) {\n if (needle.empty()) {\n return true;\n } else if (haystack.empty()) {\n return false;\n }\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needle);\n auto needle_count = count(needle);\n\n auto it = haystack_begin;\n auto end = it;\n for (size_t i = 0; i < needle_count; ++i) {\n if (end == haystack_end) {\n // not enough room in target for search\n return false;\n }\n ++end;\n }\n\n auto [sub_start, sub_end] = make_citer(it, end);\n return std::equal(needle_begin, needle_end, sub_start, sub_end,\n equal_ascii_case);\n}\n\nbool ends_with(const std::string_view haystack, const std::string_view needle) {\n if (needle.empty()) {\n return true;\n } else if (haystack.empty()) {\n return false;\n }\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needle);\n auto needle_count = count(needle);\n\n auto it = haystack_end;\n auto end = it;\n --it;\n\n for (size_t i = 1; i < needle_count; ++i) {\n if (it == haystack_begin) {\n // not enough room in target for search\n return false;\n }\n --it;\n }\n\n auto [sub_start, sub_end] = make_citer(it, end);\n return std::equal(needle_begin, needle_end, sub_start, sub_end);\n}\n\nbool ends_with_case(const std::string_view haystack,\n const std::string_view needle) {\n if (needle.empty()) {\n return true;\n } else if (haystack.empty()) {\n return false;\n }\n\n auto [haystack_begin, haystack_end] = make_citer(haystack);\n auto [needle_begin, needle_end] = make_citer(needle);\n auto needle_count = count(needle);\n\n auto it = haystack_end;\n auto end = it;\n --it;\n\n for (size_t i = 0; i < needle_count; ++i) {\n if (it == haystack_begin) {\n // not enough room in target for search\n return false;\n }\n --it;\n }\n\n auto [sub_start, sub_end] = make_citer(it, end);\n return std::equal(needle_begin, needle_end, sub_start, sub_end,\n equal_ascii_case);\n}\n\nstd::vector split_path(const std::string_view path) {\n return split(path, u8\"\\\\/\", true);\n}\n\nstd::string join_paths(const std::string_view left_path,\n const std::string_view right_path, char32_t separator) {\n if (!left_path.length()) {\n return std::string(right_path);\n } else if (!right_path.length()) {\n return std::string(left_path);\n }\n\n utf8_citer it;\n std::tie(std::ignore, it) = make_citer(left_path);\n --it;\n\n std::string result = std::string(left_path);\n if (*it != static_cast(separator)) {\n utfcpp::append(separator, result);\n }\n return result + std::string(right_path);\n}\n\nstd::string join_paths(const std::vector& paths,\n char32_t separator) {\n std::string result;\n auto it = paths.cbegin();\n if (it != paths.cend()) {\n result = *it++;\n for (; it != paths.cend(); ++it) {\n result = join_paths(result, *it, separator);\n }\n }\n return result;\n}\n\nstd::string join_paths(const std::vector& paths,\n char32_t separator) {\n std::string result;\n auto it = paths.cbegin();\n if (it != paths.cend()) {\n result = *it++;\n for (; it != paths.cend(); ++it) {\n result = join_paths(result, *it, separator);\n }\n }\n return result;\n}\n\nstd::string fix_path_separators(const std::string_view path,\n char32_t new_separator) {\n if (path.empty()) {\n return std::string();\n }\n\n // Swap all separators to new_sep.\n const char32_t old_separator = new_separator == U'\\\\' ? U'/' : U'\\\\';\n\n auto [path_begin, path_end] = make_citer(path);\n\n std::string result;\n auto it = path_begin;\n auto last = it;\n\n auto is_separator = [old_separator, new_separator](char32_t c) {\n return c == uint32_t(old_separator) || c == uint32_t(new_separator);\n };\n\n // Begins with a separator\n if (is_separator(*it)) {\n utfcpp::append(new_separator, result);\n ++it;\n last = it;\n }\n\n for (;;) {\n it = std::find_if(it, path_end, is_separator);\n if (it == path_end) {\n break;\n }\n\n if (it != last) {\n auto offset = byte_length(path_begin, last);\n auto length = byte_length(path_begin, it) - offset;\n result += path.substr(offset, length);\n utfcpp::append(new_separator, result);\n }\n\n ++it;\n last = it;\n }\n\n if (last == path_begin) {\n return std::string(path);\n }\n\n if (last != path_end) {\n auto offset = byte_length(path_begin, last);\n result += path.substr(offset);\n }\n\n return result;\n}\n\nstd::string find_name_from_path(const std::string_view path,\n char32_t separator) {\n if (path.empty()) {\n return std::string();\n }\n\n auto [begin, end] = make_citer(path);\n\n auto it = end;\n --it;\n\n // skip trailing separators at the end of the path\n while (*it == uint32_t(separator)) {\n if (it == begin) {\n // path is all separators, name is empty\n return std::string();\n }\n --it;\n }\n\n // update end so it is before any trailing separators\n end = std::next(it);\n\n // skip non-separators\n while (*it != uint32_t(separator)) {\n if (it == begin) {\n break;\n }\n --it;\n }\n\n // if the iterator is on a separator, advance\n if (*it == uint32_t(separator)) {\n ++it;\n }\n\n auto offset = byte_length(begin, it);\n auto length = byte_length(it, end);\n return std::string(path.substr(offset, length));\n}\n\nstd::string find_base_name_from_path(const std::string_view path,\n char32_t separator) {\n auto name = find_name_from_path(path, separator);\n if (!name.length()) {\n return std::string();\n }\n\n auto [begin, end] = make_citer(name);\n\n auto it = end;\n --it;\n\n while (it != begin) {\n if (*it == uint32_t('.')) {\n break;\n }\n --it;\n }\n\n if (it == begin) {\n return name;\n }\n\n auto length = byte_length(it, end);\n auto offset = name.length() - length;\n return std::string(name.substr(0, offset));\n}\n\nstd::string find_base_path(const std::string_view path, char32_t separator) {\n if (path.empty()) {\n return std::string();\n }\n\n auto [begin, end] = make_citer(path);\n\n auto it = end;\n --it;\n\n // skip trailing separators at the end of the path\n while (*it == uint32_t(separator)) {\n if (it == begin) {\n return std::string();\n }\n --it;\n }\n\n // skip non-separators\n while (*it != uint32_t(separator)) {\n if (it == begin) {\n // there are no separators, base path is empty\n return std::string();\n }\n --it;\n }\n\n // skip trailing separators at the end of the base path\n while (*it == uint32_t(separator)) {\n if (it == begin) {\n // base path is all separators, base path is empty\n return std::string();\n }\n --it;\n }\n\n auto length = byte_length(begin, std::next(it));\n return std::string(path.substr(0, length));\n}\n\nstd::string canonicalize_path(const std::string_view path, char32_t separator) {\n if (path.empty()) {\n return std::string();\n }\n\n auto is_rooted = starts_with(path, separator);\n\n auto parts = split_path(path);\n for (auto it = parts.begin(); it != parts.end();) {\n const auto& part = *it;\n if (part == \".\") {\n // Potential marker for current directory.\n it = parts.erase(it);\n } else if (part == \"..\") {\n // Ensure we don't override the device name.\n if (it != parts.begin()) {\n auto prev = std::prev(it);\n if (!ends_with(*prev, \":\")) {\n it = parts.erase(prev);\n }\n }\n it = parts.erase(it);\n } else {\n ++it;\n }\n }\n\n return !is_rooted ? join_paths(parts, separator)\n : to_string(separator) + join_paths(parts, separator);\n}\n\n} // namespace xe::utf8\n" }, { "path": "src/xenia/base/utf8.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_UTF8_H_\n#define XENIA_BASE_UTF8_H_\n\n#include \n#include \n\n#include \"xenia/base/platform.h\"\n\nnamespace xe::utf8 {\n\nsize_t count(const std::string_view view);\n\nstd::string lower_ascii(const std::string_view view);\nstd::string upper_ascii(const std::string_view view);\n\nsize_t hash_fnv1a(const std::string_view view);\nsize_t hash_fnv1a_case(const std::string_view view);\n\n// Splits the given haystack on any delimiters (needles) and returns all parts.\nstd::vector split(const std::string_view haystack,\n const std::string_view needles,\n bool remove_empty = false);\n\nbool equal_z(const std::string_view left, const std::string_view right);\n\nbool equal_case(const std::string_view left, const std::string_view right);\n\nbool equal_case_z(const std::string_view left, const std::string_view right);\n\nstd::string_view::size_type find_any_of(const std::string_view haystack,\n const std::string_view needles);\n\nstd::string_view::size_type find_any_of_case(const std::string_view haystack,\n const std::string_view needles);\n\nstd::string_view::size_type find_first_of(const std::string_view haystack,\n const std::string_view needle);\n\n// find_first_of string, case insensitive.\nstd::string_view::size_type find_first_of_case(const std::string_view haystack,\n const std::string_view needle);\n\nbool starts_with(const std::string_view haystack,\n const std::string_view needle);\n\nbool starts_with_case(const std::string_view haystack,\n const std::string_view needle);\n\nbool ends_with(const std::string_view haystack, const std::string_view needle);\n\nbool ends_with_case(const std::string_view haystack,\n const std::string_view needle);\n\n// Splits the given path on any valid path separator and returns all parts.\nstd::vector split_path(const std::string_view path);\n\n// Joins two path segments with the given separator.\nstd::string join_paths(const std::string_view left_path,\n const std::string_view right_path,\n char32_t separator = kPathSeparator);\n\nstd::string join_paths(const std::vector& paths,\n char32_t separator = kPathSeparator);\n\nstd::string join_paths(const std::vector& paths,\n char32_t separator = kPathSeparator);\n\ninline std::string join_paths(\n std::initializer_list paths,\n char32_t separator = kPathSeparator) {\n std::string result;\n for (auto path : paths) {\n result = join_paths(result, path, separator);\n }\n return result;\n}\n\ninline std::string join_guest_paths(const std::string_view left_path,\n const std::string_view right_path) {\n return join_paths(left_path, right_path, kGuestPathSeparator);\n}\n\ninline std::string join_guest_paths(const std::vector& paths) {\n return join_paths(paths, kGuestPathSeparator);\n}\n\ninline std::string join_guest_paths(\n const std::vector& paths) {\n return join_paths(paths, kGuestPathSeparator);\n}\n\ninline std::string join_guest_paths(\n std::initializer_list paths) {\n return join_paths(paths, kGuestPathSeparator);\n}\n\n// Replaces all path separators with the given value and removes redundant\n// separators.\nstd::string fix_path_separators(const std::string_view path,\n char32_t new_separator = kPathSeparator);\n\ninline std::string fix_guest_path_separators(const std::string_view path) {\n return fix_path_separators(path, kGuestPathSeparator);\n}\n\n// Find the top directory name or filename from a path.\nstd::string find_name_from_path(const std::string_view path,\n char32_t separator = kPathSeparator);\n\ninline std::string find_name_from_guest_path(const std::string_view path) {\n return find_name_from_path(path, kGuestPathSeparator);\n}\n\nstd::string find_base_name_from_path(const std::string_view path,\n char32_t separator = kPathSeparator);\n\ninline std::string find_base_name_from_guest_path(const std::string_view path) {\n return find_base_name_from_path(path, kGuestPathSeparator);\n}\n\n// Get parent path of the given directory or filename.\nstd::string find_base_path(const std::string_view path,\n char32_t separator = kPathSeparator);\n\ninline std::string find_base_guest_path(const std::string_view path) {\n return find_base_path(path, kGuestPathSeparator);\n}\n\n// Canonicalizes a path, removing ..'s.\nstd::string canonicalize_path(const std::string_view path,\n char32_t separator = kPathSeparator);\n\ninline std::string canonicalize_guest_path(const std::string_view path) {\n return canonicalize_path(path, kGuestPathSeparator);\n}\n\n} // namespace xe::utf8\n\n#endif // XENIA_BASE_UTF8_H_\n" }, { "path": "src/xenia/base/vec128.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n#include \n\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/string_util.h\"\n#include \"xenia/base/vec128.h\"\n\nnamespace xe {\n\nstd::string to_string(const vec128_t& value) {\n return fmt::format(\"({}, {}, {}, {})\", value.x, value.y, value.z, value.w);\n}\n\nstd::ostream& operator<<(std::ostream& os, const vec128_t& value) {\n os << string_util::to_hex_string(value);\n return os;\n}\n\n} // namespace xe\n" }, { "path": "src/xenia/base/vec128.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_VEC128_H_\n#define XENIA_BASE_VEC128_H_\n\n#include \n#include \n\n#include \"xenia/base/math.h\"\n#include \"xenia/base/platform.h\"\n\nnamespace xe {\n\n// The first rule of vector programming is to only rely on exact positions\n// when absolutely required - prefer dumb loops to exact offsets.\n// Vectors in memory are laid out as in AVX registers on little endian\n// machines. Note that little endian is dumb, so the byte at index 0 in\n// the vector is is really byte 15 (or the high byte of short 7 or int 3).\n// Because of this, all byte access should be via the accessors instead of\n// the direct array.\n\n// Altivec big endian layout: AVX little endian layout:\n// +---------+---------+---------+ +---------+---------+---------+\n// | int32 0 | int16 0 | int8 0 | | int32 3 | int16 7 | int8 15 |\n// | | +---------+ | | +---------+\n// | | | int8 1 | | | | int8 14 |\n// | +---------+---------+ | +---------+---------+\n// | | int16 1 | int8 2 | | | int16 6 | int8 13 |\n// | | +---------+ | | +---------+\n// | | | int8 3 | | | | int8 12 |\n// +---------+---------+---------+ +---------+---------+---------+\n// | int32 1 | int16 2 | int8 4 | | int32 2 | int16 5 | int8 11 |\n// | | +---------+ | | +---------+\n// | | | int8 5 | | | | int8 10 |\n// | +---------+---------+ | +---------+---------+\n// | | int16 3 | int8 6 | | | int16 4 | int8 9 |\n// | | +---------+ | | +---------+\n// | | | int8 7 | | | | int8 8 |\n// +---------+---------+---------+ +---------+---------+---------+\n// | int32 2 | int16 4 | int8 8 | | int32 1 | int16 3 | int8 7 |\n// | | +---------+ | | +---------+\n// | | | int8 9 | | | | int8 6 |\n// | +---------+---------+ | +---------+---------+\n// | | int16 5 | int8 10 | | | int16 2 | int8 5 |\n// | | +---------+ | | +---------+\n// | | | int8 11 | | | | int8 4 |\n// +---------+---------+---------+ +---------+---------+---------+\n// | int32 3 | int16 6 | int8 12 | | int32 0 | int16 1 | int8 3 |\n// | | +---------+ | | +---------+\n// | | | int8 13 | | | | int8 2 |\n// | +---------+---------+ | +---------+---------+\n// | | int16 7 | int8 14 | | | int16 0 | int8 1 |\n// | | +---------+ | | +---------+\n// | | | int8 15 | | | | int8 0 |\n// +---------+---------+---------+ +---------+---------+---------+\n//\n// Logical order:\n// +-----+-----+-----+-----+ +-----+-----+-----+-----+\n// | X | Y | Z | W | | W | Z | Y | X |\n// +-----+-----+-----+-----+ +-----+-----+-----+-----+\n//\n// Mapping indices is easy:\n// int32[i ^ 0x3]\n// int16[i ^ 0x7]\n// int8[i ^ 0xF]\ntypedef struct alignas(16) vec128_s {\n union {\n struct {\n float x;\n float y;\n float z;\n float w;\n };\n struct {\n int32_t ix;\n int32_t iy;\n int32_t iz;\n int32_t iw;\n };\n struct {\n uint32_t ux;\n uint32_t uy;\n uint32_t uz;\n uint32_t uw;\n };\n float f32[4];\n double f64[2];\n int8_t i8[16];\n uint8_t u8[16];\n int16_t i16[8];\n uint16_t u16[8];\n int32_t i32[4];\n uint32_t u32[4];\n int64_t i64[2];\n uint64_t u64[2];\n struct {\n uint64_t low;\n uint64_t high;\n };\n };\n\n bool operator==(const vec128_s& b) const {\n return low == b.low && high == b.high;\n }\n bool operator!=(const vec128_s& b) const {\n return low != b.low || high != b.high;\n }\n vec128_s operator^(const vec128_s& b) const {\n vec128_s a = *this;\n a.high ^= b.high;\n a.low ^= b.low;\n return a;\n };\n vec128_s& operator^=(const vec128_s& b) {\n *this = *this ^ b;\n return *this;\n };\n vec128_s operator&(const vec128_s& b) const {\n vec128_s a = *this;\n a.high &= b.high;\n a.low &= b.low;\n return a;\n };\n vec128_s& operator&=(const vec128_s& b) {\n *this = *this & b;\n return *this;\n };\n vec128_s operator|(const vec128_s& b) const {\n vec128_s a = *this;\n a.high |= b.high;\n a.low |= b.low;\n return a;\n };\n vec128_s& operator|=(const vec128_s& b) {\n *this = *this | b;\n return *this;\n };\n} vec128_t;\n\nstatic inline vec128_t vec128i(uint32_t src) {\n vec128_t v;\n for (auto i = 0; i < 4; ++i) {\n v.u32[i] = src;\n }\n return v;\n}\nstatic inline vec128_t vec128i(uint32_t x, uint32_t y, uint32_t z, uint32_t w) {\n vec128_t v;\n v.u32[0] = x;\n v.u32[1] = y;\n v.u32[2] = z;\n v.u32[3] = w;\n return v;\n}\nstatic inline vec128_t vec128q(uint64_t src) {\n vec128_t v;\n for (auto i = 0; i < 2; ++i) {\n v.i64[i] = src;\n }\n return v;\n}\nstatic inline vec128_t vec128q(uint64_t x, uint64_t y) {\n vec128_t v;\n v.i64[0] = x;\n v.i64[1] = y;\n return v;\n}\nstatic inline vec128_t vec128d(double src) {\n vec128_t v;\n for (auto i = 0; i < 2; ++i) {\n v.f64[i] = src;\n }\n return v;\n}\nstatic inline vec128_t vec128d(double x, double y) {\n vec128_t v;\n v.f64[0] = x;\n v.f64[1] = y;\n return v;\n}\nstatic inline vec128_t vec128f(float src) {\n vec128_t v;\n for (auto i = 0; i < 4; ++i) {\n v.f32[i] = src;\n }\n return v;\n}\nstatic inline vec128_t vec128f(float x, float y, float z, float w) {\n vec128_t v;\n v.f32[0] = x;\n v.f32[1] = y;\n v.f32[2] = z;\n v.f32[3] = w;\n return v;\n}\nstatic inline vec128_t vec128s(uint16_t src) {\n vec128_t v;\n for (auto i = 0; i < 8; ++i) {\n v.u16[i] = src;\n }\n return v;\n}\nstatic inline vec128_t vec128s(uint16_t x0, uint16_t x1, uint16_t y0,\n uint16_t y1, uint16_t z0, uint16_t z1,\n uint16_t w0, uint16_t w1) {\n vec128_t v;\n v.u16[0] = x1;\n v.u16[1] = x0;\n v.u16[2] = y1;\n v.u16[3] = y0;\n v.u16[4] = z1;\n v.u16[5] = z0;\n v.u16[6] = w1;\n v.u16[7] = w0;\n return v;\n}\nstatic inline vec128_t vec128b(uint8_t src) {\n vec128_t v;\n for (auto i = 0; i < 16; ++i) {\n v.u8[i] = src;\n }\n return v;\n}\nstatic inline vec128_t vec128b(uint8_t x0, uint8_t x1, uint8_t x2, uint8_t x3,\n uint8_t y0, uint8_t y1, uint8_t y2, uint8_t y3,\n uint8_t z0, uint8_t z1, uint8_t z2, uint8_t z3,\n uint8_t w0, uint8_t w1, uint8_t w2, uint8_t w3) {\n vec128_t v;\n v.u8[0] = x3;\n v.u8[1] = x2;\n v.u8[2] = x1;\n v.u8[3] = x0;\n v.u8[4] = y3;\n v.u8[5] = y2;\n v.u8[6] = y1;\n v.u8[7] = y0;\n v.u8[8] = z3;\n v.u8[9] = z2;\n v.u8[10] = z1;\n v.u8[11] = z0;\n v.u8[12] = w3;\n v.u8[13] = w2;\n v.u8[14] = w1;\n v.u8[15] = w0;\n return v;\n}\n\n// TODO(gibbed): Figure out why clang doesn't line forward declarations of\n// inline functions.\n\nstd::string to_string(const vec128_t& value);\n\nstd::ostream& operator<<(std::ostream& os, const vec128_t& value);\n\n} // namespace xe\n\n#endif // XENIA_BASE_VEC128_H_\n" }, { "path": "src/xenia/base/xxhash.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_BASE_XXHASH_H_\n#define XENIA_BASE_XXHASH_H_\n\n#define XXH_INLINE_ALL\n\n// Can't use XXH_X86DISPATCH because XXH is calculated on multiple threads,\n// while the dispatch writes the result (multiple pointers without any\n// synchronization) to XXH_g_dispatch at the first call.\n\n#include \"third_party/xxhash/xxhash.h\"\n\n#endif // XENIA_BASE_XXHASH_H_\n" }, { "path": "src/xenia/config.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"config.h\"\n\n#include \"third_party/cpptoml/include/cpptoml.h\"\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/filesystem.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/string.h\"\n#include \"xenia/base/string_buffer.h\"\n\nstd::shared_ptr ParseFile(\n const std::filesystem::path& filename) {\n std::ifstream file(filename);\n if (!file.is_open()) {\n throw cpptoml::parse_exception(xe::path_to_utf8(filename) +\n \" could not be opened for parsing\");\n }\n // since cpptoml can't parse files with a UTF-8 BOM we need to skip them\n char bom[3];\n file.read(bom, sizeof(bom));\n if (file.fail() || bom[0] != '\\xEF' || bom[1] != '\\xBB' || bom[2] != '\\xBF') {\n file.clear();\n file.seekg(0);\n }\n\n cpptoml::parser p(file);\n return p.parse();\n}\n\nCmdVar(config, \"\", \"Specifies the target config to load.\");\n\nDEFINE_uint32(\n defaults_date, 0,\n \"Do not modify - internal version of the default values in the config, for \"\n \"seamless updates if default value of any option is changed.\",\n \"Config\");\n\nnamespace config {\nstd::string config_name = \"xenia.config.toml\";\nstd::filesystem::path config_folder;\nstd::filesystem::path config_path;\nstd::string game_config_suffix = \".config.toml\";\n\nstd::shared_ptr ParseConfig(\n const std::filesystem::path& config_path) {\n try {\n return ParseFile(config_path);\n } catch (cpptoml::parse_exception e) {\n xe::FatalError(fmt::format(\"Failed to parse config file '{}':\\n\\n{}\",\n xe::path_to_utf8(config_path), e.what()));\n return nullptr;\n }\n}\n\nvoid ReadConfig(const std::filesystem::path& file_path,\n bool update_if_no_version_stored) {\n if (!cvar::ConfigVars) {\n return;\n }\n const auto config = ParseConfig(file_path);\n // Loading an actual global config file that exists - if there's no\n // defaults_date in it, it's very old (before updating was added at all, thus\n // all defaults need to be updated).\n auto defaults_date_cvar =\n dynamic_cast*>(cv::cv_defaults_date);\n assert_not_null(defaults_date_cvar);\n defaults_date_cvar->SetConfigValue(0);\n for (auto& it : *cvar::ConfigVars) {\n auto config_var = static_cast(it.second);\n auto config_key = config_var->category() + \".\" + config_var->name();\n if (config->contains_qualified(config_key)) {\n config_var->LoadConfigValue(config->get_qualified(config_key));\n }\n }\n uint32_t config_defaults_date = defaults_date_cvar->GetTypedConfigValue();\n if (update_if_no_version_stored || config_defaults_date) {\n cvar::IConfigVarUpdate::ApplyUpdates(config_defaults_date);\n }\n XELOGI(\"Loaded config: {}\", xe::path_to_utf8(file_path));\n}\n\nvoid ReadGameConfig(const std::filesystem::path& file_path) {\n if (!cvar::ConfigVars) {\n return;\n }\n const auto config = ParseConfig(file_path);\n for (auto& it : *cvar::ConfigVars) {\n auto config_var = static_cast(it.second);\n auto config_key = config_var->category() + \".\" + config_var->name();\n if (config->contains_qualified(config_key)) {\n config_var->LoadGameConfigValue(config->get_qualified(config_key));\n }\n }\n XELOGI(\"Loaded game config: {}\", xe::path_to_utf8(file_path));\n}\n\nvoid SaveConfig() {\n if (config_path.empty()) {\n return;\n }\n\n // All cvar defaults have been updated on loading - store the current date.\n auto defaults_date_cvar =\n dynamic_cast*>(cv::cv_defaults_date);\n assert_not_null(defaults_date_cvar);\n defaults_date_cvar->SetConfigValue(\n cvar::IConfigVarUpdate::GetLastUpdateDate());\n\n std::vector vars;\n if (cvar::ConfigVars) {\n for (const auto& s : *cvar::ConfigVars) {\n vars.push_back(s.second);\n }\n }\n std::sort(vars.begin(), vars.end(), [](auto a, auto b) {\n if (a->category() < b->category()) return true;\n if (a->category() > b->category()) return false;\n if (a->name() < b->name()) return true;\n return false;\n });\n\n // we use our own write logic because cpptoml doesn't\n // allow us to specify comments :(\n std::string last_category;\n bool last_multiline_description = false;\n xe::StringBuffer sb;\n for (auto config_var : vars) {\n if (config_var->is_transient()) {\n continue;\n }\n\n if (last_category != config_var->category()) {\n if (!last_category.empty()) {\n sb.Append('\\n', 2);\n }\n last_category = config_var->category();\n last_multiline_description = false;\n sb.AppendFormat(\"[{}]\\n\", last_category);\n } else if (last_multiline_description) {\n last_multiline_description = false;\n sb.Append('\\n');\n }\n\n auto value = config_var->config_value();\n size_t line_count;\n if (xe::utf8::find_any_of(value, \"\\n\") == std::string_view::npos) {\n auto line = fmt::format(\"{} = {}\", config_var->name(),\n config_var->config_value());\n sb.Append(line);\n line_count = xe::utf8::count(line);\n } else {\n auto lines = xe::utf8::split(value, \"\\n\");\n auto first = lines.cbegin();\n sb.AppendFormat(\"{} = {}\\n\", config_var->name(), *first);\n auto last = std::prev(lines.cend());\n for (auto it = std::next(first); it != last; ++it) {\n sb.Append(*it);\n sb.Append('\\n');\n }\n sb.Append(*last);\n line_count = xe::utf8::count(*last);\n }\n\n const size_t value_alignment = 50;\n const auto& description = config_var->description();\n if (!description.empty()) {\n if (line_count < value_alignment) {\n sb.Append(' ', value_alignment - line_count);\n }\n if (xe::utf8::find_any_of(description, \"\\n\") == std::string_view::npos) {\n sb.AppendFormat(\"\\t# {}\\n\", config_var->description());\n } else {\n auto lines = xe::utf8::split(description, \"\\n\");\n auto first = lines.cbegin();\n sb.Append(\"\\t# \");\n sb.Append(*first);\n sb.Append('\\n');\n for (auto it = std::next(first); it != lines.cend(); ++it) {\n sb.Append(' ', value_alignment);\n sb.Append(\"\\t# \");\n sb.Append(*it);\n sb.Append('\\n');\n }\n last_multiline_description = true;\n }\n }\n }\n\n // save the config file\n xe::filesystem::CreateParentFolder(config_path);\n\n auto handle = xe::filesystem::OpenFile(config_path, \"wb\");\n if (!handle) {\n XELOGE(\"Failed to open '{}' for writing.\", xe::path_to_utf8(config_path));\n } else {\n fwrite(sb.buffer(), 1, sb.length(), handle);\n fclose(handle);\n }\n}\n\nvoid SetupConfig(const std::filesystem::path& config_folder) {\n config::config_folder = config_folder;\n // check if the user specified a specific config to load\n if (!cvars::config.empty()) {\n config_path = xe::to_path(cvars::config);\n if (std::filesystem::exists(config_path)) {\n // An external config file may contain only explicit overrides - in this\n // case, it will likely not contain the defaults version; don't update\n // from the version 0 in this case. Or, it may be a full config - in this\n // case, if it's recent enough (created at least in 2021), it will contain\n // the version number - updates the defaults in it.\n ReadConfig(config_path, false);\n return;\n }\n }\n // if the user specified a --config argument, but the file doesn't exist,\n // let's also load the default config\n if (!config_folder.empty()) {\n config_path = config_folder / config_name;\n if (std::filesystem::exists(config_path)) {\n ReadConfig(config_path, true);\n }\n // Re-save the loaded config to present the most up-to-date list of\n // parameters to the user, if new options were added, descriptions were\n // updated, or default values were changed.\n SaveConfig();\n }\n}\n\nvoid LoadGameConfig(const std::string_view title_id) {\n const auto game_config_folder = config_folder / \"config\";\n const auto game_config_path =\n game_config_folder / (std::string(title_id) + game_config_suffix);\n if (std::filesystem::exists(game_config_path)) {\n ReadGameConfig(game_config_path);\n }\n}\n\n} // namespace config\n" }, { "path": "src/xenia/config.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CONFIG_H_\n#define XENIA_CONFIG_H_\n\n#include \n\nnamespace config {\nvoid SetupConfig(const std::filesystem::path& config_folder);\nvoid LoadGameConfig(const std::string_view title_id);\nvoid SaveConfig();\n} // namespace config\n\n#endif // XENIA_CONFIG_H_\n" }, { "path": "src/xenia/cpp.hint", "content": "// Hint files help the Visual Studio IDE interpret Visual C++ identifiers\n// such as names of functions and macros.\n// For more information see https://go.microsoft.com/fwlink/?linkid=865984\n\n#define DECLARE_XAM_EXPORT_(name, category, tags)\n#define DECLARE_XAM_EXPORT1(name, category, tag)\n#define DECLARE_XAM_EXPORT2(name, category, tag1, tag2)\n\n#define DECLARE_XBDM_EXPORT_(name, category, tags)\n#define DECLARE_XBDM_EXPORT1(name, category, tag)\n\n#define DECLARE_XBOXKRNL_EXPORT_(name, category, tags)\n#define DECLARE_XBOXKRNL_EXPORT1(name, category, tag)\n#define DECLARE_XBOXKRNL_EXPORT2(name, category, tag1, tag2)\n#define DECLARE_XBOXKRNL_EXPORT3(name, category, tag1, tag2, tag3)\n#define DECLARE_XBOXKRNL_EXPORT4(name, category, tag1, tag2, tag3, tag4)\n" }, { "path": "src/xenia/cpu/backend/assembler.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/assembler.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\n\nAssembler::Assembler(Backend* backend) : backend_(backend) {}\n\nAssembler::~Assembler() { Reset(); }\n\nbool Assembler::Initialize() { return true; }\n\nvoid Assembler::Reset() {}\n\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/assembler.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_ASSEMBLER_H_\n#define XENIA_CPU_BACKEND_ASSEMBLER_H_\n\n#include \n\nnamespace xe {\nnamespace cpu {\nclass FunctionDebugInfo;\nclass GuestFunction;\nnamespace hir {\nclass HIRBuilder;\n} // namespace hir\n} // namespace cpu\n} // namespace xe\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\n\nclass Backend;\n\nclass Assembler {\n public:\n explicit Assembler(Backend* backend);\n virtual ~Assembler();\n\n virtual bool Initialize();\n\n virtual void Reset();\n\n virtual bool Assemble(GuestFunction* function, hir::HIRBuilder* builder,\n uint32_t debug_info_flags,\n std::unique_ptr debug_info) = 0;\n\n protected:\n Backend* backend_;\n};\n\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_ASSEMBLER_H_\n" }, { "path": "src/xenia/cpu/backend/backend.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/backend.h\"\n\n#include \n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\n\nBackend::Backend() { std::memset(&machine_info_, 0, sizeof(machine_info_)); }\nBackend::~Backend() = default;\n\nbool Backend::Initialize(Processor* processor) {\n processor_ = processor;\n return true;\n}\n\nvoid* Backend::AllocThreadData() { return nullptr; }\n\nvoid Backend::FreeThreadData(void* thread_data) {}\n\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/backend.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_BACKEND_H_\n#define XENIA_CPU_BACKEND_BACKEND_H_\n\n#include \n\n#include \"xenia/cpu/backend/machine_info.h\"\n#include \"xenia/cpu/thread_debug_info.h\"\n\nnamespace xe {\nnamespace cpu {\nclass Breakpoint;\nclass Function;\nclass GuestFunction;\nclass Module;\nclass Processor;\n} // namespace cpu\n} // namespace xe\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\n\nclass Assembler;\nclass CodeCache;\n\nclass Backend {\n public:\n explicit Backend();\n virtual ~Backend();\n\n Processor* processor() const { return processor_; }\n const MachineInfo* machine_info() const { return &machine_info_; }\n CodeCache* code_cache() const { return code_cache_; }\n\n virtual bool Initialize(Processor* processor);\n\n virtual void* AllocThreadData();\n virtual void FreeThreadData(void* thread_data);\n\n virtual void CommitExecutableRange(uint32_t guest_low,\n uint32_t guest_high) = 0;\n\n virtual std::unique_ptr CreateAssembler() = 0;\n\n virtual std::unique_ptr CreateGuestFunction(\n Module* module, uint32_t address) = 0;\n\n // Calculates the next host instruction based on the current thread state and\n // current PC. This will look for branches and other control flow\n // instructions.\n virtual uint64_t CalculateNextHostInstruction(ThreadDebugInfo* thread_info,\n uint64_t current_pc) = 0;\n\n virtual void InstallBreakpoint(Breakpoint* breakpoint) {}\n virtual void InstallBreakpoint(Breakpoint* breakpoint, Function* fn) {}\n virtual void UninstallBreakpoint(Breakpoint* breakpoint) {}\n\n protected:\n Processor* processor_ = nullptr;\n MachineInfo machine_info_;\n CodeCache* code_cache_ = nullptr;\n};\n\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_BACKEND_H_\n" }, { "path": "src/xenia/cpu/backend/code_cache.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_CODE_CACHE_H_\n#define XENIA_CPU_BACKEND_CODE_CACHE_H_\n\n#include \n#include \n#include \n\n#include \"xenia/cpu/function.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\n\nclass CodeCache {\n public:\n CodeCache() = default;\n virtual ~CodeCache() = default;\n\n virtual const std::filesystem::path& file_name() const = 0;\n virtual uintptr_t execute_base_address() const = 0;\n virtual size_t total_size() const = 0;\n\n // Finds a function based on the given host PC (that may be within a\n // function).\n virtual GuestFunction* LookupFunction(uint64_t host_pc) = 0;\n\n // Finds platform-specific function unwind info for the given host PC.\n virtual void* LookupUnwindInfo(uint64_t host_pc) = 0;\n};\n\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_CODE_CACHE_H_\n" }, { "path": "src/xenia/cpu/backend/machine_info.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_MACHINE_INFO_H_\n#define XENIA_CPU_BACKEND_MACHINE_INFO_H_\n\n#include \n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\n\nstruct MachineInfo {\n bool supports_extended_load_store;\n\n struct RegisterSet {\n enum Types {\n INT_TYPES = (1 << 1),\n FLOAT_TYPES = (1 << 2),\n VEC_TYPES = (1 << 3),\n };\n uint8_t id;\n char name[4];\n uint32_t types;\n uint32_t count;\n } register_sets[8];\n};\n\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_MACHINE_INFO_H_\n" }, { "path": "src/xenia/cpu/backend/null_backend.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/null_backend.h\"\n\n#include \"xenia/cpu/backend/assembler.h\"\n#include \"xenia/cpu/function.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\n\nvoid NullBackend::CommitExecutableRange(uint32_t guest_low,\n uint32_t guest_high) {}\n\nstd::unique_ptr NullBackend::CreateAssembler() { return nullptr; }\n\nstd::unique_ptr NullBackend::CreateGuestFunction(\n Module* module, uint32_t address) {\n return nullptr;\n}\n\nuint64_t NullBackend::CalculateNextHostInstruction(ThreadDebugInfo* thread_info,\n uint64_t current_pc) {\n return current_pc;\n}\n\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/null_backend.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_NULL_BACKEND_H_\n#define XENIA_CPU_BACKEND_NULL_BACKEND_H_\n\n#include \"xenia/cpu/backend/backend.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\n\nclass NullBackend : public Backend {\n public:\n void CommitExecutableRange(uint32_t guest_low, uint32_t guest_high) override;\n\n std::unique_ptr CreateAssembler() override;\n\n std::unique_ptr CreateGuestFunction(Module* module,\n uint32_t address) override;\n\n uint64_t CalculateNextHostInstruction(ThreadDebugInfo* thread_info,\n uint64_t current_pc) override;\n};\n\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_NULL_BACKEND_H_\n" }, { "path": "src/xenia/cpu/backend/x64/premake5.lua", "content": "project_root = \"../../../../..\"\ninclude(project_root..\"/tools/build\")\n\ngroup(\"src\")\nproject(\"xenia-cpu-backend-x64\")\n uuid(\"7d8d5dce-4696-4197-952a-09506f725afe\")\n kind(\"StaticLib\")\n language(\"C++\")\n links({\n \"capstone\",\n \"fmt\",\n \"xenia-base\",\n \"xenia-cpu\",\n })\n defines({\n \"CAPSTONE_X86_ATT_DISABLE\",\n \"CAPSTONE_DIET_NO\",\n \"CAPSTONE_X86_REDUCE_NO\",\n \"CAPSTONE_HAS_X86\",\n \"CAPSTONE_USE_SYS_DYN_MEM\",\n \"XBYAK_NO_OP_NAMES\",\n \"XBYAK_ENABLE_OMITTED_OPERAND\",\n })\n -- Enable VTune, if it's installed.\n if os.isdir(project_root..\"/third_party/vtune\") then\n defines { \"ENABLE_VTUNE=1\" }\n end\n\n includedirs({\n project_root..\"/third_party/capstone/include\",\n })\n local_platform_files()\n" }, { "path": "src/xenia/cpu/backend/x64/x64_assembler.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_assembler.h\"\n\n#include \n\n#include \"third_party/capstone/include/capstone/capstone.h\"\n#include \"third_party/capstone/include/capstone/x86.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/base/reset_scope.h\"\n#include \"xenia/base/string.h\"\n#include \"xenia/cpu/backend/x64/x64_backend.h\"\n#include \"xenia/cpu/backend/x64/x64_code_cache.h\"\n#include \"xenia/cpu/backend/x64/x64_emitter.h\"\n#include \"xenia/cpu/backend/x64/x64_function.h\"\n#include \"xenia/cpu/cpu_flags.h\"\n#include \"xenia/cpu/hir/hir_builder.h\"\n#include \"xenia/cpu/hir/label.h\"\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nusing xe::cpu::hir::HIRBuilder;\n\nX64Assembler::X64Assembler(X64Backend* backend)\n : Assembler(backend), x64_backend_(backend), capstone_handle_(0) {\n if (cs_open(CS_ARCH_X86, CS_MODE_64, &capstone_handle_) != CS_ERR_OK) {\n assert_always(\"Failed to initialize capstone\");\n }\n cs_option(capstone_handle_, CS_OPT_SYNTAX, CS_OPT_SYNTAX_INTEL);\n cs_option(capstone_handle_, CS_OPT_DETAIL, CS_OPT_OFF);\n}\n\nX64Assembler::~X64Assembler() {\n // Emitter must be freed before the allocator.\n emitter_.reset();\n allocator_.reset();\n\n if (capstone_handle_) {\n cs_close(&capstone_handle_);\n }\n}\n\nbool X64Assembler::Initialize() {\n if (!Assembler::Initialize()) {\n return false;\n }\n\n allocator_.reset(new XbyakAllocator());\n emitter_.reset(new X64Emitter(x64_backend_, allocator_.get()));\n\n return true;\n}\n\nvoid X64Assembler::Reset() {\n string_buffer_.Reset();\n Assembler::Reset();\n}\n\nbool X64Assembler::Assemble(GuestFunction* function, HIRBuilder* builder,\n uint32_t debug_info_flags,\n std::unique_ptr debug_info) {\n SCOPE_profile_cpu_f(\"cpu\");\n\n // Reset when we leave.\n xe::make_reset_scope(this);\n\n // Lower HIR -> x64.\n void* machine_code = nullptr;\n size_t code_size = 0;\n if (!emitter_->Emit(function, builder, debug_info_flags, debug_info.get(),\n &machine_code, &code_size, &function->source_map())) {\n return false;\n }\n\n // Stash generated machine code.\n if (debug_info_flags & DebugInfoFlags::kDebugInfoDisasmMachineCode) {\n DumpMachineCode(machine_code, code_size, function->source_map(),\n &string_buffer_);\n debug_info->set_machine_code_disasm(xe_strdup(string_buffer_.buffer()));\n string_buffer_.Reset();\n }\n\n function->set_debug_info(std::move(debug_info));\n static_cast(function)->Setup(\n reinterpret_cast(machine_code), code_size);\n\n // Install into indirection table.\n uint64_t host_address = reinterpret_cast(machine_code);\n assert_true((host_address >> 32) == 0);\n reinterpret_cast(backend_->code_cache())\n ->AddIndirection(function->address(),\n static_cast(host_address));\n\n return true;\n}\n\nvoid X64Assembler::DumpMachineCode(\n void* machine_code, size_t code_size,\n const std::vector& source_map, StringBuffer* str) {\n if (source_map.empty()) {\n return;\n }\n auto source_map_index = 0;\n uint32_t next_code_offset = source_map[0].code_offset;\n\n const uint8_t* code_ptr = reinterpret_cast(machine_code);\n size_t remaining_code_size = code_size;\n uint64_t address = uint64_t(machine_code);\n cs_insn insn = {0};\n while (remaining_code_size &&\n cs_disasm_iter(capstone_handle_, &code_ptr, &remaining_code_size,\n &address, &insn)) {\n // Look up source offset.\n auto code_offset =\n uint32_t(code_ptr - reinterpret_cast(machine_code));\n if (code_offset >= next_code_offset &&\n source_map_index < source_map.size()) {\n auto& source_map_entry = source_map[source_map_index];\n str->AppendFormat(\"{:08X} \", source_map_entry.guest_address);\n ++source_map_index;\n next_code_offset = source_map_index < source_map.size()\n ? source_map[source_map_index].code_offset\n : UINT_MAX;\n } else {\n str->Append(\" \");\n }\n\n str->AppendFormat(\"{:08X} {:<6} {}\\n\", uint32_t(insn.address),\n insn.mnemonic, insn.op_str);\n }\n}\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_assembler.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_ASSEMBLER_H_\n#define XENIA_CPU_BACKEND_X64_X64_ASSEMBLER_H_\n\n#include \n#include \n\n#include \"xenia/base/string_buffer.h\"\n#include \"xenia/cpu/backend/assembler.h\"\n#include \"xenia/cpu/function.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nclass X64Backend;\nclass X64Emitter;\nclass XbyakAllocator;\n\nclass X64Assembler : public Assembler {\n public:\n explicit X64Assembler(X64Backend* backend);\n ~X64Assembler() override;\n\n bool Initialize() override;\n\n void Reset() override;\n\n bool Assemble(GuestFunction* function, hir::HIRBuilder* builder,\n uint32_t debug_info_flags,\n std::unique_ptr debug_info) override;\n\n private:\n void DumpMachineCode(void* machine_code, size_t code_size,\n const std::vector& source_map,\n StringBuffer* str);\n\n private:\n X64Backend* x64_backend_;\n std::unique_ptr emitter_;\n std::unique_ptr allocator_;\n uintptr_t capstone_handle_;\n\n StringBuffer string_buffer_;\n};\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_ASSEMBLER_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_backend.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2019 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_backend.h\"\n\n#include \n\n#include \"third_party/capstone/include/capstone/capstone.h\"\n#include \"third_party/capstone/include/capstone/x86.h\"\n\n#include \"xenia/base/exception_handler.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/cpu/backend/x64/x64_assembler.h\"\n#include \"xenia/cpu/backend/x64/x64_code_cache.h\"\n#include \"xenia/cpu/backend/x64/x64_emitter.h\"\n#include \"xenia/cpu/backend/x64/x64_function.h\"\n#include \"xenia/cpu/backend/x64/x64_sequences.h\"\n#include \"xenia/cpu/backend/x64/x64_stack_layout.h\"\n#include \"xenia/cpu/breakpoint.h\"\n#include \"xenia/cpu/processor.h\"\n#include \"xenia/cpu/stack_walker.h\"\n\nDEFINE_int32(x64_extension_mask, -1,\n \"Allow the detection and utilization of specific instruction set \"\n \"features.\\n\"\n \" 0 = x86_64 + AVX1\\n\"\n \" 1 = AVX2\\n\"\n \" 2 = FMA\\n\"\n \" 4 = LZCNT\\n\"\n \" 8 = BMI1\\n\"\n \" 16 = BMI2\\n\"\n \" 32 = F16C\\n\"\n \" 64 = Movbe\\n\"\n \" 128 = GFNI\\n\"\n \" 256 = AVX512F\\n\"\n \" 512 = AVX512VL\\n\"\n \" 1024 = AVX512BW\\n\"\n \" 2048 = AVX512DQ\\n\"\n \" 4096 = AVX512VBMI\\n\"\n \" -1 = Detect and utilize all possible processor features\\n\",\n \"x64\");\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nclass X64ThunkEmitter : public X64Emitter {\n public:\n X64ThunkEmitter(X64Backend* backend, XbyakAllocator* allocator);\n ~X64ThunkEmitter() override;\n HostToGuestThunk EmitHostToGuestThunk();\n GuestToHostThunk EmitGuestToHostThunk();\n ResolveFunctionThunk EmitResolveFunctionThunk();\n\n private:\n // The following four functions provide save/load functionality for registers.\n // They assume at least StackLayout::THUNK_STACK_SIZE bytes have been\n // allocated on the stack.\n void EmitSaveVolatileRegs();\n void EmitLoadVolatileRegs();\n void EmitSaveNonvolatileRegs();\n void EmitLoadNonvolatileRegs();\n};\n\nX64Backend::X64Backend() : Backend(), code_cache_(nullptr) {\n if (cs_open(CS_ARCH_X86, CS_MODE_64, &capstone_handle_) != CS_ERR_OK) {\n assert_always(\"Failed to initialize capstone\");\n }\n cs_option(capstone_handle_, CS_OPT_SYNTAX, CS_OPT_SYNTAX_INTEL);\n cs_option(capstone_handle_, CS_OPT_DETAIL, CS_OPT_ON);\n cs_option(capstone_handle_, CS_OPT_SKIPDATA, CS_OPT_OFF);\n}\n\nX64Backend::~X64Backend() {\n if (capstone_handle_) {\n cs_close(&capstone_handle_);\n }\n\n X64Emitter::FreeConstData(emitter_data_);\n ExceptionHandler::Uninstall(&ExceptionCallbackThunk, this);\n}\n\nbool X64Backend::Initialize(Processor* processor) {\n if (!Backend::Initialize(processor)) {\n return false;\n }\n\n Xbyak::util::Cpu cpu;\n if (!cpu.has(Xbyak::util::Cpu::tAVX)) {\n XELOGE(\"This CPU does not support AVX. The emulator will now crash.\");\n return false;\n }\n\n // Need movbe to do advanced LOAD/STORE tricks.\n if (cvars::x64_extension_mask & kX64EmitMovbe) {\n machine_info_.supports_extended_load_store =\n cpu.has(Xbyak::util::Cpu::tMOVBE);\n } else {\n machine_info_.supports_extended_load_store = false;\n }\n\n auto& gprs = machine_info_.register_sets[0];\n gprs.id = 0;\n std::strcpy(gprs.name, \"gpr\");\n gprs.types = MachineInfo::RegisterSet::INT_TYPES;\n gprs.count = X64Emitter::GPR_COUNT;\n\n auto& xmms = machine_info_.register_sets[1];\n xmms.id = 1;\n std::strcpy(xmms.name, \"xmm\");\n xmms.types = MachineInfo::RegisterSet::FLOAT_TYPES |\n MachineInfo::RegisterSet::VEC_TYPES;\n xmms.count = X64Emitter::XMM_COUNT;\n\n code_cache_ = X64CodeCache::Create();\n Backend::code_cache_ = code_cache_.get();\n if (!code_cache_->Initialize()) {\n return false;\n }\n\n // Generate thunks used to transition between jitted code and host code.\n XbyakAllocator allocator;\n X64ThunkEmitter thunk_emitter(this, &allocator);\n host_to_guest_thunk_ = thunk_emitter.EmitHostToGuestThunk();\n guest_to_host_thunk_ = thunk_emitter.EmitGuestToHostThunk();\n resolve_function_thunk_ = thunk_emitter.EmitResolveFunctionThunk();\n\n // Set the code cache to use the ResolveFunction thunk for default\n // indirections.\n assert_zero(uint64_t(resolve_function_thunk_) & 0xFFFFFFFF00000000ull);\n code_cache_->set_indirection_default(\n uint32_t(uint64_t(resolve_function_thunk_)));\n\n // Allocate some special indirections.\n code_cache_->CommitExecutableRange(0x9FFF0000, 0x9FFFFFFF);\n\n // Allocate emitter constant data.\n emitter_data_ = X64Emitter::PlaceConstData();\n\n // Setup exception callback\n ExceptionHandler::Install(&ExceptionCallbackThunk, this);\n\n return true;\n}\n\nvoid X64Backend::CommitExecutableRange(uint32_t guest_low,\n uint32_t guest_high) {\n code_cache_->CommitExecutableRange(guest_low, guest_high);\n}\n\nstd::unique_ptr X64Backend::CreateAssembler() {\n return std::make_unique(this);\n}\n\nstd::unique_ptr X64Backend::CreateGuestFunction(\n Module* module, uint32_t address) {\n return std::make_unique(module, address);\n}\n\nuint64_t ReadCapstoneReg(HostThreadContext* context, x86_reg reg) {\n switch (reg) {\n case X86_REG_RAX:\n return context->rax;\n case X86_REG_RCX:\n return context->rcx;\n case X86_REG_RDX:\n return context->rdx;\n case X86_REG_RBX:\n return context->rbx;\n case X86_REG_RSP:\n return context->rsp;\n case X86_REG_RBP:\n return context->rbp;\n case X86_REG_RSI:\n return context->rsi;\n case X86_REG_RDI:\n return context->rdi;\n case X86_REG_R8:\n return context->r8;\n case X86_REG_R9:\n return context->r9;\n case X86_REG_R10:\n return context->r10;\n case X86_REG_R11:\n return context->r11;\n case X86_REG_R12:\n return context->r12;\n case X86_REG_R13:\n return context->r13;\n case X86_REG_R14:\n return context->r14;\n case X86_REG_R15:\n return context->r15;\n default:\n assert_unhandled_case(reg);\n return 0;\n }\n}\n\n#define X86_EFLAGS_CF 0x00000001 // Carry Flag\n#define X86_EFLAGS_PF 0x00000004 // Parity Flag\n#define X86_EFLAGS_ZF 0x00000040 // Zero Flag\n#define X86_EFLAGS_SF 0x00000080 // Sign Flag\n#define X86_EFLAGS_OF 0x00000800 // Overflow Flag\nbool TestCapstoneEflags(uint32_t eflags, uint32_t insn) {\n // https://www.felixcloutier.com/x86/Jcc.html\n switch (insn) {\n case X86_INS_JAE:\n // CF=0 && ZF=0\n return ((eflags & X86_EFLAGS_CF) == 0) && ((eflags & X86_EFLAGS_ZF) == 0);\n case X86_INS_JA:\n // CF=0\n return (eflags & X86_EFLAGS_CF) == 0;\n case X86_INS_JBE:\n // CF=1 || ZF=1\n return ((eflags & X86_EFLAGS_CF) == X86_EFLAGS_CF) ||\n ((eflags & X86_EFLAGS_ZF) == X86_EFLAGS_ZF);\n case X86_INS_JB:\n // CF=1\n return (eflags & X86_EFLAGS_CF) == X86_EFLAGS_CF;\n case X86_INS_JE:\n // ZF=1\n return (eflags & X86_EFLAGS_ZF) == X86_EFLAGS_ZF;\n case X86_INS_JGE:\n // SF=OF\n return (eflags & X86_EFLAGS_SF) == (eflags & X86_EFLAGS_OF);\n case X86_INS_JG:\n // ZF=0 && SF=OF\n return ((eflags & X86_EFLAGS_ZF) == 0) &&\n ((eflags & X86_EFLAGS_SF) == (eflags & X86_EFLAGS_OF));\n case X86_INS_JLE:\n // ZF=1 || SF!=OF\n return ((eflags & X86_EFLAGS_ZF) == X86_EFLAGS_ZF) ||\n ((eflags & X86_EFLAGS_SF) != X86_EFLAGS_OF);\n case X86_INS_JL:\n // SF!=OF\n return (eflags & X86_EFLAGS_SF) != (eflags & X86_EFLAGS_OF);\n case X86_INS_JNE:\n // ZF=0\n return (eflags & X86_EFLAGS_ZF) == 0;\n case X86_INS_JNO:\n // OF=0\n return (eflags & X86_EFLAGS_OF) == 0;\n case X86_INS_JNP:\n // PF=0\n return (eflags & X86_EFLAGS_PF) == 0;\n case X86_INS_JNS:\n // SF=0\n return (eflags & X86_EFLAGS_SF) == 0;\n case X86_INS_JO:\n // OF=1\n return (eflags & X86_EFLAGS_OF) == X86_EFLAGS_OF;\n case X86_INS_JP:\n // PF=1\n return (eflags & X86_EFLAGS_PF) == X86_EFLAGS_PF;\n case X86_INS_JS:\n // SF=1\n return (eflags & X86_EFLAGS_SF) == X86_EFLAGS_SF;\n default:\n assert_unhandled_case(insn);\n return false;\n }\n}\n\nuint64_t X64Backend::CalculateNextHostInstruction(ThreadDebugInfo* thread_info,\n uint64_t current_pc) {\n auto machine_code_ptr = reinterpret_cast(current_pc);\n size_t remaining_machine_code_size = 64;\n uint64_t host_address = current_pc;\n cs_insn insn = {0};\n cs_detail all_detail = {0};\n insn.detail = &all_detail;\n cs_disasm_iter(capstone_handle_, &machine_code_ptr,\n &remaining_machine_code_size, &host_address, &insn);\n auto& detail = all_detail.x86;\n switch (insn.id) {\n default:\n // Not a branching instruction - just move over it.\n return current_pc + insn.size;\n case X86_INS_CALL: {\n assert_true(detail.op_count == 1);\n assert_true(detail.operands[0].type == X86_OP_REG);\n uint64_t target_pc =\n ReadCapstoneReg(&thread_info->host_context, detail.operands[0].reg);\n return target_pc;\n } break;\n case X86_INS_RET: {\n assert_zero(detail.op_count);\n auto stack_ptr =\n reinterpret_cast(thread_info->host_context.rsp);\n uint64_t target_pc = stack_ptr[0];\n return target_pc;\n } break;\n case X86_INS_JMP: {\n assert_true(detail.op_count == 1);\n if (detail.operands[0].type == X86_OP_IMM) {\n uint64_t target_pc = static_cast(detail.operands[0].imm);\n return target_pc;\n } else if (detail.operands[0].type == X86_OP_REG) {\n uint64_t target_pc =\n ReadCapstoneReg(&thread_info->host_context, detail.operands[0].reg);\n return target_pc;\n } else {\n // TODO(benvanik): find some more uses of this.\n assert_always(\"jmp branch emulation not yet implemented\");\n return current_pc + insn.size;\n }\n } break;\n case X86_INS_JCXZ:\n case X86_INS_JECXZ:\n case X86_INS_JRCXZ:\n assert_always(\"j*cxz branch emulation not yet implemented\");\n return current_pc + insn.size;\n case X86_INS_JAE:\n case X86_INS_JA:\n case X86_INS_JBE:\n case X86_INS_JB:\n case X86_INS_JE:\n case X86_INS_JGE:\n case X86_INS_JG:\n case X86_INS_JLE:\n case X86_INS_JL:\n case X86_INS_JNE:\n case X86_INS_JNO:\n case X86_INS_JNP:\n case X86_INS_JNS:\n case X86_INS_JO:\n case X86_INS_JP:\n case X86_INS_JS: {\n assert_true(detail.op_count == 1);\n assert_true(detail.operands[0].type == X86_OP_IMM);\n uint64_t target_pc = static_cast(detail.operands[0].imm);\n bool test_passed =\n TestCapstoneEflags(thread_info->host_context.eflags, insn.id);\n if (test_passed) {\n return target_pc;\n } else {\n return current_pc + insn.size;\n }\n } break;\n }\n}\n\nvoid X64Backend::InstallBreakpoint(Breakpoint* breakpoint) {\n breakpoint->ForEachHostAddress([breakpoint](uint64_t host_address) {\n auto ptr = reinterpret_cast(host_address);\n auto original_bytes = xe::load_and_swap(ptr);\n assert_true(original_bytes != 0x0F0B);\n xe::store_and_swap(ptr, 0x0F0B);\n breakpoint->backend_data().emplace_back(host_address, original_bytes);\n });\n}\n\nvoid X64Backend::InstallBreakpoint(Breakpoint* breakpoint, Function* fn) {\n assert_true(breakpoint->address_type() == Breakpoint::AddressType::kGuest);\n assert_true(fn->is_guest());\n auto guest_function = reinterpret_cast(fn);\n auto host_address =\n guest_function->MapGuestAddressToMachineCode(breakpoint->guest_address());\n if (!host_address) {\n assert_always();\n return;\n }\n\n // Assume we haven't already installed a breakpoint in this spot.\n auto ptr = reinterpret_cast(host_address);\n auto original_bytes = xe::load_and_swap(ptr);\n assert_true(original_bytes != 0x0F0B);\n xe::store_and_swap(ptr, 0x0F0B);\n breakpoint->backend_data().emplace_back(host_address, original_bytes);\n}\n\nvoid X64Backend::UninstallBreakpoint(Breakpoint* breakpoint) {\n for (auto& pair : breakpoint->backend_data()) {\n auto ptr = reinterpret_cast(pair.first);\n auto instruction_bytes = xe::load_and_swap(ptr);\n assert_true(instruction_bytes == 0x0F0B);\n xe::store_and_swap(ptr, static_cast(pair.second));\n }\n breakpoint->backend_data().clear();\n}\n\nbool X64Backend::ExceptionCallbackThunk(Exception* ex, void* data) {\n auto backend = reinterpret_cast(data);\n return backend->ExceptionCallback(ex);\n}\n\nbool X64Backend::ExceptionCallback(Exception* ex) {\n if (ex->code() != Exception::Code::kIllegalInstruction) {\n // We only care about illegal instructions. Other things will be handled by\n // other handlers (probably). If nothing else picks it up we'll be called\n // with OnUnhandledException to do real crash handling.\n return false;\n }\n\n // Verify an expected illegal instruction.\n auto instruction_bytes =\n xe::load_and_swap(reinterpret_cast(ex->pc()));\n if (instruction_bytes != 0x0F0B) {\n // Not our ud2 - not us.\n return false;\n }\n\n // Let the processor handle things.\n return processor()->OnThreadBreakpointHit(ex);\n}\n\nX64ThunkEmitter::X64ThunkEmitter(X64Backend* backend, XbyakAllocator* allocator)\n : X64Emitter(backend, allocator) {}\n\nX64ThunkEmitter::~X64ThunkEmitter() {}\n\nHostToGuestThunk X64ThunkEmitter::EmitHostToGuestThunk() {\n // rcx = target\n // rdx = arg0 (context)\n // r8 = arg1 (guest return address)\n\n struct _code_offsets {\n size_t prolog;\n size_t prolog_stack_alloc;\n size_t body;\n size_t epilog;\n size_t tail;\n } code_offsets = {};\n\n const size_t stack_size = StackLayout::THUNK_STACK_SIZE;\n\n code_offsets.prolog = getSize();\n\n // rsp + 0 = return address\n mov(qword[rsp + 8 * 3], r8);\n mov(qword[rsp + 8 * 2], rdx);\n mov(qword[rsp + 8 * 1], rcx);\n sub(rsp, stack_size);\n\n code_offsets.prolog_stack_alloc = getSize();\n code_offsets.body = getSize();\n\n // Save nonvolatile registers.\n EmitSaveNonvolatileRegs();\n\n mov(rax, rcx);\n mov(rsi, rdx); // context\n mov(rcx, r8); // return address\n call(rax);\n\n EmitLoadNonvolatileRegs();\n\n code_offsets.epilog = getSize();\n\n add(rsp, stack_size);\n mov(rcx, qword[rsp + 8 * 1]);\n mov(rdx, qword[rsp + 8 * 2]);\n mov(r8, qword[rsp + 8 * 3]);\n ret();\n\n code_offsets.tail = getSize();\n\n assert_zero(code_offsets.prolog);\n EmitFunctionInfo func_info = {};\n func_info.code_size.total = getSize();\n func_info.code_size.prolog = code_offsets.body - code_offsets.prolog;\n func_info.code_size.body = code_offsets.epilog - code_offsets.body;\n func_info.code_size.epilog = code_offsets.tail - code_offsets.epilog;\n func_info.code_size.tail = getSize() - code_offsets.tail;\n func_info.prolog_stack_alloc_offset =\n code_offsets.prolog_stack_alloc - code_offsets.prolog;\n func_info.stack_size = stack_size;\n\n void* fn = Emplace(func_info);\n return (HostToGuestThunk)fn;\n}\n\nGuestToHostThunk X64ThunkEmitter::EmitGuestToHostThunk() {\n // rcx = target function\n // rdx = arg0\n // r8 = arg1\n // r9 = arg2\n\n struct _code_offsets {\n size_t prolog;\n size_t prolog_stack_alloc;\n size_t body;\n size_t epilog;\n size_t tail;\n } code_offsets = {};\n\n const size_t stack_size = StackLayout::THUNK_STACK_SIZE;\n\n code_offsets.prolog = getSize();\n\n // rsp + 0 = return address\n sub(rsp, stack_size);\n\n code_offsets.prolog_stack_alloc = getSize();\n code_offsets.body = getSize();\n\n // Save off volatile registers.\n EmitSaveVolatileRegs();\n\n mov(rax, rcx); // function\n mov(rcx, GetContextReg()); // context\n call(rax);\n\n EmitLoadVolatileRegs();\n\n code_offsets.epilog = getSize();\n\n add(rsp, stack_size);\n ret();\n\n code_offsets.tail = getSize();\n\n assert_zero(code_offsets.prolog);\n EmitFunctionInfo func_info = {};\n func_info.code_size.total = getSize();\n func_info.code_size.prolog = code_offsets.body - code_offsets.prolog;\n func_info.code_size.body = code_offsets.epilog - code_offsets.body;\n func_info.code_size.epilog = code_offsets.tail - code_offsets.epilog;\n func_info.code_size.tail = getSize() - code_offsets.tail;\n func_info.prolog_stack_alloc_offset =\n code_offsets.prolog_stack_alloc - code_offsets.prolog;\n func_info.stack_size = stack_size;\n\n void* fn = Emplace(func_info);\n return (GuestToHostThunk)fn;\n}\n\n// X64Emitter handles actually resolving functions.\nuint64_t ResolveFunction(void* raw_context, uint64_t target_address);\n\nResolveFunctionThunk X64ThunkEmitter::EmitResolveFunctionThunk() {\n // ebx = target PPC address\n // rcx = context\n\n struct _code_offsets {\n size_t prolog;\n size_t prolog_stack_alloc;\n size_t body;\n size_t epilog;\n size_t tail;\n } code_offsets = {};\n\n const size_t stack_size = StackLayout::THUNK_STACK_SIZE;\n\n code_offsets.prolog = getSize();\n\n // rsp + 0 = return address\n sub(rsp, stack_size);\n\n code_offsets.prolog_stack_alloc = getSize();\n code_offsets.body = getSize();\n\n // Save volatile registers\n EmitSaveVolatileRegs();\n\n mov(rcx, rsi); // context\n mov(rdx, rbx);\n mov(rax, reinterpret_cast(&ResolveFunction));\n call(rax);\n\n EmitLoadVolatileRegs();\n\n code_offsets.epilog = getSize();\n\n add(rsp, stack_size);\n jmp(rax);\n\n code_offsets.tail = getSize();\n\n assert_zero(code_offsets.prolog);\n EmitFunctionInfo func_info = {};\n func_info.code_size.total = getSize();\n func_info.code_size.prolog = code_offsets.body - code_offsets.prolog;\n func_info.code_size.body = code_offsets.epilog - code_offsets.body;\n func_info.code_size.epilog = code_offsets.tail - code_offsets.epilog;\n func_info.code_size.tail = getSize() - code_offsets.tail;\n func_info.prolog_stack_alloc_offset =\n code_offsets.prolog_stack_alloc - code_offsets.prolog;\n func_info.stack_size = stack_size;\n\n void* fn = Emplace(func_info);\n return (ResolveFunctionThunk)fn;\n}\n\nvoid X64ThunkEmitter::EmitSaveVolatileRegs() {\n // Save off volatile registers.\n // mov(qword[rsp + offsetof(StackLayout::Thunk, r[0])], rax);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[1])], rcx);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[2])], rdx);\n#if XE_PLATFORM_LINUX\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[3])], rsi);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[4])], rdi);\n#endif\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[5])], r8);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[6])], r9);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[7])], r10);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[8])], r11);\n\n // vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[0])], xmm0);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[1])], xmm1);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[2])], xmm2);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[3])], xmm3);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[4])], xmm4);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[5])], xmm5);\n}\n\nvoid X64ThunkEmitter::EmitLoadVolatileRegs() {\n // mov(rax, qword[rsp + offsetof(StackLayout::Thunk, r[0])]);\n mov(rcx, qword[rsp + offsetof(StackLayout::Thunk, r[1])]);\n mov(rdx, qword[rsp + offsetof(StackLayout::Thunk, r[2])]);\n#if XE_PLATFORM_LINUX\n mov(rsi, qword[rsp + offsetof(StackLayout::Thunk, r[3])]);\n mov(rdi, qword[rsp + offsetof(StackLayout::Thunk, r[4])]);\n#endif\n mov(r8, qword[rsp + offsetof(StackLayout::Thunk, r[5])]);\n mov(r9, qword[rsp + offsetof(StackLayout::Thunk, r[6])]);\n mov(r10, qword[rsp + offsetof(StackLayout::Thunk, r[7])]);\n mov(r11, qword[rsp + offsetof(StackLayout::Thunk, r[8])]);\n\n // vmovaps(xmm0, qword[rsp + offsetof(StackLayout::Thunk, xmm[0])]);\n vmovaps(xmm1, qword[rsp + offsetof(StackLayout::Thunk, xmm[1])]);\n vmovaps(xmm2, qword[rsp + offsetof(StackLayout::Thunk, xmm[2])]);\n vmovaps(xmm3, qword[rsp + offsetof(StackLayout::Thunk, xmm[3])]);\n vmovaps(xmm4, qword[rsp + offsetof(StackLayout::Thunk, xmm[4])]);\n vmovaps(xmm5, qword[rsp + offsetof(StackLayout::Thunk, xmm[5])]);\n}\n\nvoid X64ThunkEmitter::EmitSaveNonvolatileRegs() {\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[0])], rbx);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[1])], rbp);\n#if XE_PLATFORM_WIN32\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[2])], rcx);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[3])], rsi);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[4])], rdi);\n#endif\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[5])], r12);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[6])], r13);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[7])], r14);\n mov(qword[rsp + offsetof(StackLayout::Thunk, r[8])], r15);\n\n // SysV does not have nonvolatile XMM registers.\n#if XE_PLATFORM_WIN32\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[0])], xmm6);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[1])], xmm7);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[2])], xmm8);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[3])], xmm9);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[4])], xmm10);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[5])], xmm11);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[6])], xmm12);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[7])], xmm13);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[8])], xmm14);\n vmovaps(qword[rsp + offsetof(StackLayout::Thunk, xmm[9])], xmm15);\n#endif\n}\n\nvoid X64ThunkEmitter::EmitLoadNonvolatileRegs() {\n mov(rbx, qword[rsp + offsetof(StackLayout::Thunk, r[0])]);\n mov(rbp, qword[rsp + offsetof(StackLayout::Thunk, r[1])]);\n#if XE_PLATFORM_WIN32\n mov(rcx, qword[rsp + offsetof(StackLayout::Thunk, r[2])]);\n mov(rsi, qword[rsp + offsetof(StackLayout::Thunk, r[3])]);\n mov(rdi, qword[rsp + offsetof(StackLayout::Thunk, r[4])]);\n#endif\n mov(r12, qword[rsp + offsetof(StackLayout::Thunk, r[5])]);\n mov(r13, qword[rsp + offsetof(StackLayout::Thunk, r[6])]);\n mov(r14, qword[rsp + offsetof(StackLayout::Thunk, r[7])]);\n mov(r15, qword[rsp + offsetof(StackLayout::Thunk, r[8])]);\n\n#if XE_PLATFORM_WIN32\n vmovaps(xmm6, qword[rsp + offsetof(StackLayout::Thunk, xmm[0])]);\n vmovaps(xmm7, qword[rsp + offsetof(StackLayout::Thunk, xmm[1])]);\n vmovaps(xmm8, qword[rsp + offsetof(StackLayout::Thunk, xmm[2])]);\n vmovaps(xmm9, qword[rsp + offsetof(StackLayout::Thunk, xmm[3])]);\n vmovaps(xmm10, qword[rsp + offsetof(StackLayout::Thunk, xmm[4])]);\n vmovaps(xmm11, qword[rsp + offsetof(StackLayout::Thunk, xmm[5])]);\n vmovaps(xmm12, qword[rsp + offsetof(StackLayout::Thunk, xmm[6])]);\n vmovaps(xmm13, qword[rsp + offsetof(StackLayout::Thunk, xmm[7])]);\n vmovaps(xmm14, qword[rsp + offsetof(StackLayout::Thunk, xmm[8])]);\n vmovaps(xmm15, qword[rsp + offsetof(StackLayout::Thunk, xmm[9])]);\n#endif\n}\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_backend.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_BACKEND_H_\n#define XENIA_CPU_BACKEND_X64_X64_BACKEND_H_\n\n#include \n\n#include \"xenia/base/cvar.h\"\n#include \"xenia/cpu/backend/backend.h\"\n\nDECLARE_int32(x64_extension_mask);\n\nnamespace xe {\nclass Exception;\n} // namespace xe\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nclass X64CodeCache;\n\ntypedef void* (*HostToGuestThunk)(void* target, void* arg0, void* arg1);\ntypedef void* (*GuestToHostThunk)(void* target, void* arg0, void* arg1);\ntypedef void (*ResolveFunctionThunk)();\n\nclass X64Backend : public Backend {\n public:\n static const uint32_t kForceReturnAddress = 0x9FFF0000u;\n\n explicit X64Backend();\n ~X64Backend() override;\n\n X64CodeCache* code_cache() const { return code_cache_.get(); }\n uintptr_t emitter_data() const { return emitter_data_; }\n\n // Call a generated function, saving all stack parameters.\n HostToGuestThunk host_to_guest_thunk() const { return host_to_guest_thunk_; }\n // Function that guest code can call to transition into host code.\n GuestToHostThunk guest_to_host_thunk() const { return guest_to_host_thunk_; }\n // Function that thunks to the ResolveFunction in X64Emitter.\n ResolveFunctionThunk resolve_function_thunk() const {\n return resolve_function_thunk_;\n }\n\n bool Initialize(Processor* processor) override;\n\n void CommitExecutableRange(uint32_t guest_low, uint32_t guest_high) override;\n\n std::unique_ptr CreateAssembler() override;\n\n std::unique_ptr CreateGuestFunction(Module* module,\n uint32_t address) override;\n\n uint64_t CalculateNextHostInstruction(ThreadDebugInfo* thread_info,\n uint64_t current_pc) override;\n\n void InstallBreakpoint(Breakpoint* breakpoint) override;\n void InstallBreakpoint(Breakpoint* breakpoint, Function* fn) override;\n void UninstallBreakpoint(Breakpoint* breakpoint) override;\n\n private:\n static bool ExceptionCallbackThunk(Exception* ex, void* data);\n bool ExceptionCallback(Exception* ex);\n\n uintptr_t capstone_handle_ = 0;\n\n std::unique_ptr code_cache_;\n uintptr_t emitter_data_ = 0;\n\n HostToGuestThunk host_to_guest_thunk_;\n GuestToHostThunk guest_to_host_thunk_;\n ResolveFunctionThunk resolve_function_thunk_;\n};\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_BACKEND_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_code_cache.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_code_cache.h\"\n\n#include \n#include \n\n#if ENABLE_VTUNE\n#include \"third_party/vtune/include/jitprofiling.h\"\n#pragma comment(lib, \"../third_party/vtune/lib64/jitprofiling.lib\")\n#endif\n\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/clock.h\"\n#include \"xenia/base/literals.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/memory.h\"\n#include \"xenia/cpu/function.h\"\n#include \"xenia/cpu/module.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nusing namespace xe::literals;\n\nX64CodeCache::X64CodeCache() = default;\n\nX64CodeCache::~X64CodeCache() {\n if (indirection_table_base_) {\n xe::memory::DeallocFixed(indirection_table_base_, 0,\n xe::memory::DeallocationType::kRelease);\n }\n\n // Unmap all views and close mapping.\n if (mapping_ != xe::memory::kFileMappingHandleInvalid) {\n if (generated_code_write_base_ &&\n generated_code_write_base_ != generated_code_execute_base_) {\n xe::memory::UnmapFileView(mapping_, generated_code_write_base_,\n kGeneratedCodeSize);\n }\n if (generated_code_execute_base_) {\n xe::memory::UnmapFileView(mapping_, generated_code_execute_base_,\n kGeneratedCodeSize);\n }\n xe::memory::CloseFileMappingHandle(mapping_, file_name_);\n mapping_ = xe::memory::kFileMappingHandleInvalid;\n }\n}\n\nbool X64CodeCache::Initialize() {\n indirection_table_base_ = reinterpret_cast(xe::memory::AllocFixed(\n reinterpret_cast(kIndirectionTableBase), kIndirectionTableSize,\n xe::memory::AllocationType::kReserve,\n xe::memory::PageAccess::kReadWrite));\n if (!indirection_table_base_) {\n XELOGE(\"Unable to allocate code cache indirection table\");\n XELOGE(\n \"This is likely because the {:X}-{:X} range is in use by some other \"\n \"system DLL\",\n static_cast(kIndirectionTableBase),\n kIndirectionTableBase + kIndirectionTableSize);\n }\n\n // Create mmap file. This allows us to share the code cache with the debugger.\n file_name_ = fmt::format(\"xenia_code_cache_{}\", Clock::QueryHostTickCount());\n mapping_ = xe::memory::CreateFileMappingHandle(\n file_name_, kGeneratedCodeSize, xe::memory::PageAccess::kExecuteReadWrite,\n false);\n if (mapping_ == xe::memory::kFileMappingHandleInvalid) {\n XELOGE(\"Unable to create code cache mmap\");\n return false;\n }\n\n // Map generated code region into the file. Pages are committed as required.\n if (xe::memory::IsWritableExecutableMemoryPreferred()) {\n generated_code_execute_base_ =\n reinterpret_cast(xe::memory::MapFileView(\n mapping_, reinterpret_cast(kGeneratedCodeExecuteBase),\n kGeneratedCodeSize, xe::memory::PageAccess::kExecuteReadWrite, 0));\n generated_code_write_base_ = generated_code_execute_base_;\n if (!generated_code_execute_base_ || !generated_code_write_base_) {\n XELOGE(\"Unable to allocate code cache generated code storage\");\n XELOGE(\n \"This is likely because the {:X}-{:X} range is in use by some other \"\n \"system DLL\",\n uint64_t(kGeneratedCodeExecuteBase),\n uint64_t(kGeneratedCodeExecuteBase + kGeneratedCodeSize));\n return false;\n }\n } else {\n generated_code_execute_base_ =\n reinterpret_cast(xe::memory::MapFileView(\n mapping_, reinterpret_cast(kGeneratedCodeExecuteBase),\n kGeneratedCodeSize, xe::memory::PageAccess::kExecuteReadOnly, 0));\n generated_code_write_base_ =\n reinterpret_cast(xe::memory::MapFileView(\n mapping_, reinterpret_cast(kGeneratedCodeWriteBase),\n kGeneratedCodeSize, xe::memory::PageAccess::kReadWrite, 0));\n if (!generated_code_execute_base_ || !generated_code_write_base_) {\n XELOGE(\"Unable to allocate code cache generated code storage\");\n XELOGE(\n \"This is likely because the {:X}-{:X} and {:X}-{:X} ranges are in \"\n \"use by some other system DLL\",\n uint64_t(kGeneratedCodeExecuteBase),\n uint64_t(kGeneratedCodeExecuteBase + kGeneratedCodeSize),\n uint64_t(kGeneratedCodeWriteBase),\n uint64_t(kGeneratedCodeWriteBase + kGeneratedCodeSize));\n return false;\n }\n }\n\n // Preallocate the function map to a large, reasonable size.\n generated_code_map_.reserve(kMaximumFunctionCount);\n\n return true;\n}\n\nvoid X64CodeCache::set_indirection_default(uint32_t default_value) {\n indirection_default_value_ = default_value;\n}\n\nvoid X64CodeCache::AddIndirection(uint32_t guest_address,\n uint32_t host_address) {\n if (!indirection_table_base_) {\n return;\n }\n\n uint32_t* indirection_slot = reinterpret_cast(\n indirection_table_base_ + (guest_address - kIndirectionTableBase));\n *indirection_slot = host_address;\n}\n\nvoid X64CodeCache::CommitExecutableRange(uint32_t guest_low,\n uint32_t guest_high) {\n if (!indirection_table_base_) {\n return;\n }\n\n // Commit the memory.\n xe::memory::AllocFixed(\n indirection_table_base_ + (guest_low - kIndirectionTableBase),\n guest_high - guest_low, xe::memory::AllocationType::kCommit,\n xe::memory::PageAccess::kReadWrite);\n\n // Fill memory with the default value.\n uint32_t* p = reinterpret_cast(indirection_table_base_);\n for (uint32_t address = guest_low; address < guest_high; ++address) {\n p[(address - kIndirectionTableBase) / 4] = indirection_default_value_;\n }\n}\n\nvoid X64CodeCache::PlaceHostCode(uint32_t guest_address, void* machine_code,\n const EmitFunctionInfo& func_info,\n void*& code_execute_address_out,\n void*& code_write_address_out) {\n // Same for now. We may use different pools or whatnot later on, like when\n // we only want to place guest code in a serialized cache on disk.\n PlaceGuestCode(guest_address, machine_code, func_info, nullptr,\n code_execute_address_out, code_write_address_out);\n}\n\nvoid X64CodeCache::PlaceGuestCode(uint32_t guest_address, void* machine_code,\n const EmitFunctionInfo& func_info,\n GuestFunction* function_info,\n void*& code_execute_address_out,\n void*& code_write_address_out) {\n // Hold a lock while we bump the pointers up. This is important as the\n // unwind table requires entries AND code to be sorted in order.\n size_t low_mark;\n size_t high_mark;\n uint8_t* code_execute_address;\n UnwindReservation unwind_reservation;\n {\n auto global_lock = global_critical_region_.Acquire();\n\n low_mark = generated_code_offset_;\n\n // Reserve code.\n // Always move the code to land on 16b alignment.\n code_execute_address =\n generated_code_execute_base_ + generated_code_offset_;\n code_execute_address_out = code_execute_address;\n uint8_t* code_write_address =\n generated_code_write_base_ + generated_code_offset_;\n code_write_address_out = code_write_address;\n generated_code_offset_ += xe::round_up(func_info.code_size.total, 16);\n\n auto tail_write_address =\n generated_code_write_base_ + generated_code_offset_;\n\n // Reserve unwind info.\n // We go on the high size of the unwind info as we don't know how big we\n // need it, and a few extra bytes of padding isn't the worst thing.\n unwind_reservation = RequestUnwindReservation(generated_code_write_base_ +\n generated_code_offset_);\n generated_code_offset_ += xe::round_up(unwind_reservation.data_size, 16);\n\n auto end_write_address =\n generated_code_write_base_ + generated_code_offset_;\n\n high_mark = generated_code_offset_;\n\n // Store in map. It is maintained in sorted order of host PC dependent on\n // us also being append-only.\n generated_code_map_.emplace_back(\n (uint64_t(code_execute_address - generated_code_execute_base_) << 32) |\n generated_code_offset_,\n function_info);\n\n // TODO(DrChat): The following code doesn't really need to be under the\n // global lock except for PlaceCode (but it depends on the previous code\n // already being ran)\n\n // If we are going above the high water mark of committed memory, commit\n // some more. It's ok if multiple threads do this, as redundant commits\n // aren't harmful.\n size_t old_commit_mark, new_commit_mark;\n do {\n old_commit_mark = generated_code_commit_mark_;\n if (high_mark <= old_commit_mark) break;\n\n new_commit_mark = old_commit_mark + 16_MiB;\n if (generated_code_execute_base_ == generated_code_write_base_) {\n xe::memory::AllocFixed(generated_code_execute_base_, new_commit_mark,\n xe::memory::AllocationType::kCommit,\n xe::memory::PageAccess::kExecuteReadWrite);\n } else {\n xe::memory::AllocFixed(generated_code_execute_base_, new_commit_mark,\n xe::memory::AllocationType::kCommit,\n xe::memory::PageAccess::kExecuteReadOnly);\n xe::memory::AllocFixed(generated_code_write_base_, new_commit_mark,\n xe::memory::AllocationType::kCommit,\n xe::memory::PageAccess::kReadWrite);\n }\n } while (generated_code_commit_mark_.compare_exchange_weak(\n old_commit_mark, new_commit_mark));\n\n // Copy code.\n std::memcpy(code_write_address, machine_code, func_info.code_size.total);\n\n // Fill unused slots with 0xCC\n std::memset(tail_write_address, 0xCC,\n static_cast(end_write_address - tail_write_address));\n\n // Notify subclasses of placed code.\n PlaceCode(guest_address, machine_code, func_info, code_execute_address,\n unwind_reservation);\n }\n\n#if ENABLE_VTUNE\n if (iJIT_IsProfilingActive() == iJIT_SAMPLING_ON) {\n std::string method_name;\n if (function_info && function_info->name().size() != 0) {\n method_name = function_info->name();\n } else {\n method_name = xe::format_string(\"sub_%.8X\", guest_address);\n }\n\n iJIT_Method_Load_V2 method = {0};\n method.method_id = iJIT_GetNewMethodID();\n method.method_load_address = code_execute_address;\n method.method_size = uint32_t(code_size);\n method.method_name = const_cast(method_name.data());\n method.module_name = function_info\n ? (char*)function_info->module()->name().c_str()\n : nullptr;\n iJIT_NotifyEvent(iJVM_EVENT_TYPE_METHOD_LOAD_FINISHED_V2, (void*)&method);\n }\n#endif\n\n // Now that everything is ready, fix up the indirection table.\n // Note that we do support code that doesn't have an indirection fixup, so\n // ignore those when we see them.\n if (guest_address && indirection_table_base_) {\n uint32_t* indirection_slot = reinterpret_cast(\n indirection_table_base_ + (guest_address - kIndirectionTableBase));\n *indirection_slot =\n uint32_t(reinterpret_cast(code_execute_address));\n }\n}\n\nuint32_t X64CodeCache::PlaceData(const void* data, size_t length) {\n // Hold a lock while we bump the pointers up.\n size_t high_mark;\n uint8_t* data_address = nullptr;\n {\n auto global_lock = global_critical_region_.Acquire();\n\n // Reserve code.\n // Always move the code to land on 16b alignment.\n data_address = generated_code_write_base_ + generated_code_offset_;\n generated_code_offset_ += xe::round_up(length, 16);\n\n high_mark = generated_code_offset_;\n }\n\n // If we are going above the high water mark of committed memory, commit some\n // more. It's ok if multiple threads do this, as redundant commits aren't\n // harmful.\n size_t old_commit_mark, new_commit_mark;\n do {\n old_commit_mark = generated_code_commit_mark_;\n if (high_mark <= old_commit_mark) break;\n\n new_commit_mark = old_commit_mark + 16_MiB;\n if (generated_code_execute_base_ == generated_code_write_base_) {\n xe::memory::AllocFixed(generated_code_execute_base_, new_commit_mark,\n xe::memory::AllocationType::kCommit,\n xe::memory::PageAccess::kExecuteReadWrite);\n } else {\n xe::memory::AllocFixed(generated_code_execute_base_, new_commit_mark,\n xe::memory::AllocationType::kCommit,\n xe::memory::PageAccess::kExecuteReadOnly);\n xe::memory::AllocFixed(generated_code_write_base_, new_commit_mark,\n xe::memory::AllocationType::kCommit,\n xe::memory::PageAccess::kReadWrite);\n }\n } while (generated_code_commit_mark_.compare_exchange_weak(old_commit_mark,\n new_commit_mark));\n\n // Copy code.\n std::memcpy(data_address, data, length);\n\n return uint32_t(uintptr_t(data_address));\n}\n\nGuestFunction* X64CodeCache::LookupFunction(uint64_t host_pc) {\n uint32_t key = uint32_t(host_pc - kGeneratedCodeExecuteBase);\n void* fn_entry = std::bsearch(\n &key, generated_code_map_.data(), generated_code_map_.size() + 1,\n sizeof(std::pair),\n [](const void* key_ptr, const void* element_ptr) {\n auto key = *reinterpret_cast(key_ptr);\n auto element =\n reinterpret_cast*>(\n element_ptr);\n if (key < (element->first >> 32)) {\n return -1;\n } else if (key > uint32_t(element->first)) {\n return 1;\n } else {\n return 0;\n }\n });\n if (fn_entry) {\n return reinterpret_cast*>(\n fn_entry)\n ->second;\n } else {\n return nullptr;\n }\n}\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_code_cache.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_CODE_CACHE_H_\n#define XENIA_CPU_BACKEND_X64_X64_CODE_CACHE_H_\n\n#include \n#include \n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/memory.h\"\n#include \"xenia/base/mutex.h\"\n#include \"xenia/cpu/backend/code_cache.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nstruct EmitFunctionInfo {\n struct _code_size {\n size_t prolog;\n size_t body;\n size_t epilog;\n size_t tail;\n size_t total;\n } code_size;\n size_t prolog_stack_alloc_offset; // offset of instruction after stack alloc\n size_t stack_size;\n};\n\nclass X64CodeCache : public CodeCache {\n public:\n ~X64CodeCache() override;\n\n static std::unique_ptr Create();\n\n virtual bool Initialize();\n\n const std::filesystem::path& file_name() const override { return file_name_; }\n uintptr_t execute_base_address() const override {\n return kGeneratedCodeExecuteBase;\n }\n size_t total_size() const override { return kGeneratedCodeSize; }\n\n // TODO(benvanik): ELF serialization/etc\n // TODO(benvanik): keep track of code blocks\n // TODO(benvanik): padding/guards/etc\n\n bool has_indirection_table() { return indirection_table_base_ != nullptr; }\n void set_indirection_default(uint32_t default_value);\n void AddIndirection(uint32_t guest_address, uint32_t host_address);\n\n void CommitExecutableRange(uint32_t guest_low, uint32_t guest_high);\n\n void PlaceHostCode(uint32_t guest_address, void* machine_code,\n const EmitFunctionInfo& func_info,\n void*& code_execute_address_out,\n void*& code_write_address_out);\n void PlaceGuestCode(uint32_t guest_address, void* machine_code,\n const EmitFunctionInfo& func_info,\n GuestFunction* function_info,\n void*& code_execute_address_out,\n void*& code_write_address_out);\n uint32_t PlaceData(const void* data, size_t length);\n\n GuestFunction* LookupFunction(uint64_t host_pc) override;\n\n protected:\n // All executable code falls within 0x80000000 to 0x9FFFFFFF, so we can\n // only map enough for lookups within that range.\n static const size_t kIndirectionTableSize = 0x1FFFFFFF;\n static const uintptr_t kIndirectionTableBase = 0x80000000;\n // The code range is 512MB, but we know the total code games will have is\n // pretty small (dozens of mb at most) and our expansion is reasonablish\n // so 256MB should be more than enough.\n static const size_t kGeneratedCodeSize = 0x0FFFFFFF;\n static const uintptr_t kGeneratedCodeExecuteBase = 0xA0000000;\n // Used for writing when PageAccess::kExecuteReadWrite is not supported.\n static const uintptr_t kGeneratedCodeWriteBase =\n kGeneratedCodeExecuteBase + kGeneratedCodeSize + 1;\n\n // This is picked to be high enough to cover whatever we can reasonably\n // expect. If we hit issues with this it probably means some corner case\n // in analysis triggering.\n static const size_t kMaximumFunctionCount = 100000;\n\n struct UnwindReservation {\n size_t data_size = 0;\n size_t table_slot = 0;\n uint8_t* entry_address = 0;\n };\n\n X64CodeCache();\n\n virtual UnwindReservation RequestUnwindReservation(uint8_t* entry_address) {\n return UnwindReservation();\n }\n virtual void PlaceCode(uint32_t guest_address, void* machine_code,\n const EmitFunctionInfo& func_info,\n void* code_execute_address,\n UnwindReservation unwind_reservation) {}\n\n std::filesystem::path file_name_;\n xe::memory::FileMappingHandle mapping_ =\n xe::memory::kFileMappingHandleInvalid;\n\n // NOTE: the global critical region must be held when manipulating the offsets\n // or counts of anything, to keep the tables consistent and ordered.\n xe::global_critical_region global_critical_region_;\n\n // Value that the indirection table will be initialized with upon commit.\n uint32_t indirection_default_value_ = 0xFEEDF00D;\n\n // Fixed at kIndirectionTableBase in host space, holding 4 byte pointers into\n // the generated code table that correspond to the PPC functions in guest\n // space.\n uint8_t* indirection_table_base_ = nullptr;\n // Fixed at kGeneratedCodeExecuteBase and holding all generated code, growing\n // as needed.\n uint8_t* generated_code_execute_base_ = nullptr;\n // View of the memory that backs generated_code_execute_base_ when\n // PageAccess::kExecuteReadWrite is not supported, for writing the generated\n // code. Equals to generated_code_execute_base_ when it's supported.\n uint8_t* generated_code_write_base_ = nullptr;\n // Current offset to empty space in generated code.\n size_t generated_code_offset_ = 0;\n // Current high water mark of COMMITTED code.\n std::atomic generated_code_commit_mark_ = {0};\n // Sorted map by host PC base offsets to source function info.\n // This can be used to bsearch on host PC to find the guest function.\n // The key is [start address | end address].\n std::vector> generated_code_map_;\n};\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_CODE_CACHE_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_code_cache_posix.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2017 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_code_cache.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nclass PosixX64CodeCache : public X64CodeCache {\n public:\n PosixX64CodeCache();\n ~PosixX64CodeCache() override;\n\n bool Initialize() override;\n\n void* LookupUnwindInfo(uint64_t host_pc) override { return nullptr; }\n\n private:\n /*\n UnwindReservation RequestUnwindReservation(uint8_t* entry_address) override;\n void PlaceCode(uint32_t guest_address, void* machine_code, size_t code_size,\n size_t stack_size, void* code_execute_address,\n UnwindReservation unwind_reservation) override;\n\n void InitializeUnwindEntry(uint8_t* unwind_entry_address,\n size_t unwind_table_slot, void* code_address,\n size_t code_size, size_t stack_size);\n */\n};\n\nstd::unique_ptr X64CodeCache::Create() {\n return std::make_unique();\n}\n\nPosixX64CodeCache::PosixX64CodeCache() = default;\nPosixX64CodeCache::~PosixX64CodeCache() = default;\n\nbool PosixX64CodeCache::Initialize() { return X64CodeCache::Initialize(); }\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe" }, { "path": "src/xenia/cpu/backend/x64/x64_code_cache_win.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2019 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_code_cache.h\"\n\n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/clock.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/memory.h\"\n#include \"xenia/base/platform_win.h\"\n#include \"xenia/cpu/function.h\"\n\n// Function pointer definitions\nusing FnRtlAddGrowableFunctionTable = decltype(&RtlAddGrowableFunctionTable);\nusing FnRtlGrowFunctionTable = decltype(&RtlGrowFunctionTable);\nusing FnRtlDeleteGrowableFunctionTable =\n decltype(&RtlDeleteGrowableFunctionTable);\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\n// https://msdn.microsoft.com/en-us/library/ssa62fwe.aspx\ntypedef enum _UNWIND_OP_CODES {\n UWOP_PUSH_NONVOL = 0, /* info == register number */\n UWOP_ALLOC_LARGE, /* no info, alloc size in next 2 slots */\n UWOP_ALLOC_SMALL, /* info == size of allocation / 8 - 1 */\n UWOP_SET_FPREG, /* no info, FP = RSP + UNWIND_INFO.FPRegOffset*16 */\n UWOP_SAVE_NONVOL, /* info == register number, offset in next slot */\n UWOP_SAVE_NONVOL_FAR, /* info == register number, offset in next 2 slots */\n UWOP_SAVE_XMM128, /* info == XMM reg number, offset in next slot */\n UWOP_SAVE_XMM128_FAR, /* info == XMM reg number, offset in next 2 slots */\n UWOP_PUSH_MACHFRAME /* info == 0: no error-code, 1: error-code */\n} UNWIND_CODE_OPS;\nclass UNWIND_REGISTER {\n public:\n enum _ {\n RAX = 0,\n RCX = 1,\n RDX = 2,\n RBX = 3,\n RSP = 4,\n RBP = 5,\n RSI = 6,\n RDI = 7,\n R8 = 8,\n R9 = 9,\n R10 = 10,\n R11 = 11,\n R12 = 12,\n R13 = 13,\n R14 = 14,\n R15 = 15,\n };\n};\n\ntypedef union _UNWIND_CODE {\n struct {\n uint8_t CodeOffset;\n uint8_t UnwindOp : 4;\n uint8_t OpInfo : 4;\n };\n USHORT FrameOffset;\n} UNWIND_CODE, *PUNWIND_CODE;\n\ntypedef struct _UNWIND_INFO {\n uint8_t Version : 3;\n uint8_t Flags : 5;\n uint8_t SizeOfProlog;\n uint8_t CountOfCodes;\n uint8_t FrameRegister : 4;\n uint8_t FrameOffset : 4;\n UNWIND_CODE UnwindCode[1];\n /* UNWIND_CODE MoreUnwindCode[((CountOfCodes + 1) & ~1) - 1];\n * union {\n * OPTIONAL ULONG ExceptionHandler;\n * OPTIONAL ULONG FunctionEntry;\n * };\n * OPTIONAL ULONG ExceptionData[]; */\n} UNWIND_INFO, *PUNWIND_INFO;\n\n// Size of unwind info per function.\n// TODO(benvanik): move this to emitter.\nstatic const uint32_t kUnwindInfoSize =\n sizeof(UNWIND_INFO) + (sizeof(UNWIND_CODE) * (6 - 1));\n\nclass Win32X64CodeCache : public X64CodeCache {\n public:\n Win32X64CodeCache();\n ~Win32X64CodeCache() override;\n\n bool Initialize() override;\n\n void* LookupUnwindInfo(uint64_t host_pc) override;\n\n private:\n UnwindReservation RequestUnwindReservation(uint8_t* entry_address) override;\n void PlaceCode(uint32_t guest_address, void* machine_code,\n const EmitFunctionInfo& func_info, void* code_execute_address,\n UnwindReservation unwind_reservation) override;\n\n void InitializeUnwindEntry(uint8_t* unwind_entry_address,\n size_t unwind_table_slot,\n void* code_execute_address,\n const EmitFunctionInfo& func_info);\n\n // Growable function table system handle.\n void* unwind_table_handle_ = nullptr;\n // Actual unwind table entries.\n std::vector unwind_table_;\n // Current number of entries in the table.\n std::atomic unwind_table_count_ = {0};\n // Does this version of Windows support growable funciton tables?\n bool supports_growable_table_ = false;\n\n FnRtlAddGrowableFunctionTable add_growable_table_ = nullptr;\n FnRtlDeleteGrowableFunctionTable delete_growable_table_ = nullptr;\n FnRtlGrowFunctionTable grow_table_ = nullptr;\n};\n\nstd::unique_ptr X64CodeCache::Create() {\n return std::make_unique();\n}\n\nWin32X64CodeCache::Win32X64CodeCache() = default;\n\nWin32X64CodeCache::~Win32X64CodeCache() {\n if (supports_growable_table_) {\n if (unwind_table_handle_) {\n delete_growable_table_(unwind_table_handle_);\n }\n } else {\n if (generated_code_execute_base_) {\n RtlDeleteFunctionTable(reinterpret_cast(\n reinterpret_cast(generated_code_execute_base_) | 0x3));\n }\n }\n}\n\nbool Win32X64CodeCache::Initialize() {\n if (!X64CodeCache::Initialize()) {\n return false;\n }\n\n // Compute total number of unwind entries we should allocate.\n // We don't support reallocing right now, so this should be high.\n unwind_table_.resize(kMaximumFunctionCount);\n\n // Check if this version of Windows supports growable function tables.\n auto ntdll_handle = GetModuleHandleW(L\"ntdll.dll\");\n if (!ntdll_handle) {\n add_growable_table_ = nullptr;\n delete_growable_table_ = nullptr;\n grow_table_ = nullptr;\n } else {\n add_growable_table_ = (FnRtlAddGrowableFunctionTable)GetProcAddress(\n ntdll_handle, \"RtlAddGrowableFunctionTable\");\n delete_growable_table_ = (FnRtlDeleteGrowableFunctionTable)GetProcAddress(\n ntdll_handle, \"RtlDeleteGrowableFunctionTable\");\n grow_table_ = (FnRtlGrowFunctionTable)GetProcAddress(\n ntdll_handle, \"RtlGrowFunctionTable\");\n }\n supports_growable_table_ =\n add_growable_table_ && delete_growable_table_ && grow_table_;\n\n // Create table and register with the system. It's empty now, but we'll grow\n // it as functions are added.\n if (supports_growable_table_) {\n if (add_growable_table_(\n &unwind_table_handle_, unwind_table_.data(), unwind_table_count_,\n DWORD(unwind_table_.size()),\n reinterpret_cast(generated_code_execute_base_),\n reinterpret_cast(generated_code_execute_base_ +\n kGeneratedCodeSize))) {\n XELOGE(\"Unable to create unwind function table\");\n return false;\n }\n } else {\n // Install a callback that the debugger will use to lookup unwind info on\n // demand.\n if (!RtlInstallFunctionTableCallback(\n reinterpret_cast(generated_code_execute_base_) | 0x3,\n reinterpret_cast(generated_code_execute_base_),\n kGeneratedCodeSize,\n [](DWORD64 control_pc, PVOID context) {\n auto code_cache = reinterpret_cast(context);\n return reinterpret_cast(\n code_cache->LookupUnwindInfo(control_pc));\n },\n this, nullptr)) {\n XELOGE(\"Unable to install function table callback\");\n return false;\n }\n }\n\n return true;\n}\n\nWin32X64CodeCache::UnwindReservation\nWin32X64CodeCache::RequestUnwindReservation(uint8_t* entry_address) {\n assert_false(unwind_table_count_ >= kMaximumFunctionCount);\n UnwindReservation unwind_reservation;\n unwind_reservation.data_size = xe::round_up(kUnwindInfoSize, 16);\n unwind_reservation.table_slot = unwind_table_count_++;\n unwind_reservation.entry_address = entry_address;\n return unwind_reservation;\n}\n\nvoid Win32X64CodeCache::PlaceCode(uint32_t guest_address, void* machine_code,\n const EmitFunctionInfo& func_info,\n void* code_execute_address,\n UnwindReservation unwind_reservation) {\n // Add unwind info.\n InitializeUnwindEntry(unwind_reservation.entry_address,\n unwind_reservation.table_slot, code_execute_address,\n func_info);\n\n if (supports_growable_table_) {\n // Notify that the unwind table has grown.\n // We do this outside of the lock, but with the latest total count.\n grow_table_(unwind_table_handle_, unwind_table_count_);\n }\n\n // This isn't needed on x64 (probably), but is convention.\n // On UWP, FlushInstructionCache available starting from 10.0.16299.0.\n // https://docs.microsoft.com/en-us/uwp/win32-and-com/win32-apis\n FlushInstructionCache(GetCurrentProcess(), code_execute_address,\n func_info.code_size.total);\n}\n\nvoid Win32X64CodeCache::InitializeUnwindEntry(\n uint8_t* unwind_entry_address, size_t unwind_table_slot,\n void* code_execute_address, const EmitFunctionInfo& func_info) {\n auto unwind_info = reinterpret_cast(unwind_entry_address);\n UNWIND_CODE* unwind_code = nullptr;\n\n assert_true(func_info.code_size.prolog < 256); // needs to fit into a uint8_t\n auto prolog_size = static_cast(func_info.code_size.prolog);\n assert_true(func_info.prolog_stack_alloc_offset <\n 256); // needs to fit into a uint8_t\n auto prolog_stack_alloc_offset =\n static_cast(func_info.prolog_stack_alloc_offset);\n\n if (!func_info.stack_size) {\n // https://docs.microsoft.com/en-us/cpp/build/exception-handling-x64#struct-unwind_info\n unwind_info->Version = 1;\n unwind_info->Flags = 0;\n unwind_info->SizeOfProlog = prolog_size;\n unwind_info->CountOfCodes = 0;\n unwind_info->FrameRegister = 0;\n unwind_info->FrameOffset = 0;\n } else if (func_info.stack_size <= 128) {\n // https://docs.microsoft.com/en-us/cpp/build/exception-handling-x64#struct-unwind_info\n unwind_info->Version = 1;\n unwind_info->Flags = 0;\n unwind_info->SizeOfProlog = prolog_size;\n unwind_info->CountOfCodes = 0;\n unwind_info->FrameRegister = 0;\n unwind_info->FrameOffset = 0;\n\n // https://docs.microsoft.com/en-us/cpp/build/exception-handling-x64#struct-unwind_code\n unwind_code = &unwind_info->UnwindCode[unwind_info->CountOfCodes++];\n unwind_code->CodeOffset = prolog_stack_alloc_offset;\n unwind_code->UnwindOp = UWOP_ALLOC_SMALL;\n unwind_code->OpInfo = (func_info.stack_size / 8) - 1;\n } else {\n // TODO(benvanik): take as parameters?\n\n // https://docs.microsoft.com/en-us/cpp/build/exception-handling-x64#struct-unwind_info\n unwind_info->Version = 1;\n unwind_info->Flags = 0;\n unwind_info->SizeOfProlog = prolog_size;\n unwind_info->CountOfCodes = 0;\n unwind_info->FrameRegister = 0;\n unwind_info->FrameOffset = 0;\n\n // https://docs.microsoft.com/en-us/cpp/build/exception-handling-x64#struct-unwind_code\n unwind_code = &unwind_info->UnwindCode[unwind_info->CountOfCodes++];\n unwind_code->CodeOffset = prolog_stack_alloc_offset;\n unwind_code->UnwindOp = UWOP_ALLOC_LARGE;\n unwind_code->OpInfo = 0; // One slot for size\n\n assert_true((func_info.stack_size / 8) < 65536u);\n unwind_code = &unwind_info->UnwindCode[unwind_info->CountOfCodes++];\n unwind_code->FrameOffset = (USHORT)(func_info.stack_size) / 8;\n }\n\n if (unwind_info->CountOfCodes % 1) {\n // Count of unwind codes must always be even.\n std::memset(&unwind_info->UnwindCode[unwind_info->CountOfCodes + 1], 0,\n sizeof(UNWIND_CODE));\n }\n\n // Add entry.\n auto& fn_entry = unwind_table_[unwind_table_slot];\n fn_entry.BeginAddress =\n DWORD(reinterpret_cast(code_execute_address) -\n generated_code_execute_base_);\n fn_entry.EndAddress =\n DWORD(fn_entry.BeginAddress + func_info.code_size.total);\n fn_entry.UnwindData =\n DWORD(unwind_entry_address - generated_code_execute_base_);\n}\n\nvoid* Win32X64CodeCache::LookupUnwindInfo(uint64_t host_pc) {\n return std::bsearch(\n &host_pc, unwind_table_.data(), unwind_table_count_,\n sizeof(RUNTIME_FUNCTION),\n [](const void* key_ptr, const void* element_ptr) {\n auto key = *reinterpret_cast(key_ptr) -\n kGeneratedCodeExecuteBase;\n auto element = reinterpret_cast(element_ptr);\n if (key < element->BeginAddress) {\n return -1;\n } else if (key > element->EndAddress) {\n return 1;\n } else {\n return 0;\n }\n });\n}\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_emitter.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_emitter.h\"\n\n#include \n\n#include \n#include \n\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/atomic.h\"\n#include \"xenia/base/debugging.h\"\n#include \"xenia/base/literals.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/memory.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/base/vec128.h\"\n#include \"xenia/cpu/backend/x64/x64_backend.h\"\n#include \"xenia/cpu/backend/x64/x64_code_cache.h\"\n#include \"xenia/cpu/backend/x64/x64_function.h\"\n#include \"xenia/cpu/backend/x64/x64_sequences.h\"\n#include \"xenia/cpu/backend/x64/x64_stack_layout.h\"\n#include \"xenia/cpu/cpu_flags.h\"\n#include \"xenia/cpu/function.h\"\n#include \"xenia/cpu/function_debug_info.h\"\n#include \"xenia/cpu/processor.h\"\n#include \"xenia/cpu/symbol.h\"\n#include \"xenia/cpu/thread_state.h\"\n\nDEFINE_bool(debugprint_trap_log, false,\n \"Log debugprint traps to the active debugger\", \"CPU\");\nDEFINE_bool(ignore_undefined_externs, true,\n \"Don't exit when an undefined extern is called.\", \"CPU\");\nDEFINE_bool(emit_source_annotations, false,\n \"Add extra movs and nops to make disassembly easier to read.\",\n \"CPU\");\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::Instr;\nusing namespace xe::literals;\n\nstatic const size_t kMaxCodeSize = 1_MiB;\n\nstatic const size_t kStashOffset = 32;\n// static const size_t kStashOffsetHigh = 32 + 32;\n\nconst uint32_t X64Emitter::gpr_reg_map_[X64Emitter::GPR_COUNT] = {\n Xbyak::Operand::RBX, Xbyak::Operand::R10, Xbyak::Operand::R11,\n Xbyak::Operand::R12, Xbyak::Operand::R13, Xbyak::Operand::R14,\n Xbyak::Operand::R15,\n};\n\nconst uint32_t X64Emitter::xmm_reg_map_[X64Emitter::XMM_COUNT] = {\n 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,\n};\n\nX64Emitter::X64Emitter(X64Backend* backend, XbyakAllocator* allocator)\n : CodeGenerator(kMaxCodeSize, Xbyak::AutoGrow, allocator),\n processor_(backend->processor()),\n backend_(backend),\n code_cache_(backend->code_cache()),\n allocator_(allocator) {\n if (!cpu_.has(Xbyak::util::Cpu::tAVX)) {\n xe::FatalError(\n \"Your CPU does not support AVX, which is required by Xenia. See the \"\n \"FAQ for system requirements at https://xenia.jp\");\n return;\n }\n\n#define TEST_EMIT_FEATURE(emit, ext) \\\n if ((cvars::x64_extension_mask & emit) == emit) { \\\n feature_flags_ |= (cpu_.has(ext) ? emit : 0); \\\n }\n\n TEST_EMIT_FEATURE(kX64EmitAVX2, Xbyak::util::Cpu::tAVX2);\n TEST_EMIT_FEATURE(kX64EmitFMA, Xbyak::util::Cpu::tFMA);\n TEST_EMIT_FEATURE(kX64EmitLZCNT, Xbyak::util::Cpu::tLZCNT);\n TEST_EMIT_FEATURE(kX64EmitBMI1, Xbyak::util::Cpu::tBMI1);\n TEST_EMIT_FEATURE(kX64EmitBMI2, Xbyak::util::Cpu::tBMI2);\n TEST_EMIT_FEATURE(kX64EmitF16C, Xbyak::util::Cpu::tF16C);\n TEST_EMIT_FEATURE(kX64EmitMovbe, Xbyak::util::Cpu::tMOVBE);\n TEST_EMIT_FEATURE(kX64EmitGFNI, Xbyak::util::Cpu::tGFNI);\n TEST_EMIT_FEATURE(kX64EmitAVX512F, Xbyak::util::Cpu::tAVX512F);\n TEST_EMIT_FEATURE(kX64EmitAVX512VL, Xbyak::util::Cpu::tAVX512VL);\n TEST_EMIT_FEATURE(kX64EmitAVX512BW, Xbyak::util::Cpu::tAVX512BW);\n TEST_EMIT_FEATURE(kX64EmitAVX512DQ, Xbyak::util::Cpu::tAVX512DQ);\n TEST_EMIT_FEATURE(kX64EmitAVX512VBMI, Xbyak::util::Cpu::tAVX512_VBMI);\n\n#undef TEST_EMIT_FEATURE\n}\n\nX64Emitter::~X64Emitter() = default;\n\nbool X64Emitter::Emit(GuestFunction* function, HIRBuilder* builder,\n uint32_t debug_info_flags, FunctionDebugInfo* debug_info,\n void** out_code_address, size_t* out_code_size,\n std::vector* out_source_map) {\n SCOPE_profile_cpu_f(\"cpu\");\n\n // Reset.\n debug_info_ = debug_info;\n debug_info_flags_ = debug_info_flags;\n trace_data_ = &function->trace_data();\n source_map_arena_.Reset();\n\n // Fill the generator with code.\n EmitFunctionInfo func_info = {};\n if (!Emit(builder, func_info)) {\n return false;\n }\n\n // Copy the final code to the cache and relocate it.\n *out_code_size = getSize();\n *out_code_address = Emplace(func_info, function);\n\n // Stash source map.\n source_map_arena_.CloneContents(out_source_map);\n\n return true;\n}\n\nvoid* X64Emitter::Emplace(const EmitFunctionInfo& func_info,\n GuestFunction* function) {\n // To avoid changing xbyak, we do a switcharoo here.\n // top_ points to the Xbyak buffer, and since we are in AutoGrow mode\n // it has pending relocations. We copy the top_ to our buffer, swap the\n // pointer, relocate, then return the original scratch pointer for use.\n // top_ is used by Xbyak's ready() as both write base pointer and the absolute\n // address base, which would not work on platforms not supporting writable\n // executable memory, but Xenia doesn't use absolute label addresses in the\n // generated code.\n uint8_t* old_address = top_;\n void* new_execute_address;\n void* new_write_address;\n assert_true(func_info.code_size.total == size_);\n if (function) {\n code_cache_->PlaceGuestCode(function->address(), top_, func_info, function,\n new_execute_address, new_write_address);\n } else {\n code_cache_->PlaceHostCode(0, top_, func_info, new_execute_address,\n new_write_address);\n }\n top_ = reinterpret_cast(new_write_address);\n ready();\n top_ = old_address;\n reset();\n return new_execute_address;\n}\n\nbool X64Emitter::Emit(HIRBuilder* builder, EmitFunctionInfo& func_info) {\n Xbyak::Label epilog_label;\n epilog_label_ = &epilog_label;\n\n // Calculate stack size. We need to align things to their natural sizes.\n // This could be much better (sort by type/etc).\n auto locals = builder->locals();\n size_t stack_offset = StackLayout::GUEST_STACK_SIZE;\n for (auto it = locals.begin(); it != locals.end(); ++it) {\n auto slot = *it;\n size_t type_size = GetTypeSize(slot->type);\n\n // Align to natural size.\n stack_offset = xe::align(stack_offset, type_size);\n slot->set_constant((uint32_t)stack_offset);\n stack_offset += type_size;\n }\n\n // Ensure 16b alignment.\n stack_offset -= StackLayout::GUEST_STACK_SIZE;\n stack_offset = xe::align(stack_offset, static_cast(16));\n\n struct _code_offsets {\n size_t prolog;\n size_t prolog_stack_alloc;\n size_t body;\n size_t epilog;\n size_t tail;\n } code_offsets = {};\n\n code_offsets.prolog = getSize();\n\n // Function prolog.\n // Must be 16b aligned.\n // Windows is very strict about the form of this and the epilog:\n // https://docs.microsoft.com/en-us/cpp/build/prolog-and-epilog?view=vs-2017\n // IMPORTANT: any changes to the prolog must be kept in sync with\n // X64CodeCache, which dynamically generates exception information.\n // Adding or changing anything here must be matched!\n const size_t stack_size = StackLayout::GUEST_STACK_SIZE + stack_offset;\n assert_true((stack_size + 8) % 16 == 0);\n func_info.stack_size = stack_size;\n stack_size_ = stack_size;\n\n sub(rsp, (uint32_t)stack_size);\n\n code_offsets.prolog_stack_alloc = getSize();\n code_offsets.body = getSize();\n\n mov(qword[rsp + StackLayout::GUEST_CTX_HOME], GetContextReg());\n mov(qword[rsp + StackLayout::GUEST_RET_ADDR], rcx);\n mov(qword[rsp + StackLayout::GUEST_CALL_RET_ADDR], 0);\n\n // Safe now to do some tracing.\n if (debug_info_flags_ & DebugInfoFlags::kDebugInfoTraceFunctions) {\n // We require 32-bit addresses.\n assert_true(uint64_t(trace_data_->header()) < UINT_MAX);\n auto trace_header = trace_data_->header();\n\n // Call count.\n lock();\n inc(qword[low_address(&trace_header->function_call_count)]);\n\n // Get call history slot.\n static_assert(FunctionTraceData::kFunctionCallerHistoryCount == 4,\n \"bitmask depends on count\");\n mov(rax, qword[low_address(&trace_header->function_call_count)]);\n and_(rax, 0b00000011);\n\n // Record call history value into slot (guest addr in RDX).\n mov(dword[Xbyak::RegExp(uint32_t(uint64_t(\n low_address(&trace_header->function_caller_history)))) +\n rax * 4],\n edx);\n\n // Calling thread. Load ax with thread ID.\n EmitGetCurrentThreadId();\n lock();\n bts(qword[low_address(&trace_header->function_thread_use)], rax);\n }\n\n // Load membase.\n mov(GetMembaseReg(),\n qword[GetContextReg() + offsetof(ppc::PPCContext, virtual_membase)]);\n\n // Body.\n auto block = builder->first_block();\n while (block) {\n // Mark block labels.\n auto label = block->label_head;\n while (label) {\n L(label->name);\n label = label->next;\n }\n\n // Process instructions.\n const Instr* instr = block->instr_head;\n while (instr) {\n const Instr* new_tail = instr;\n if (!SelectSequence(this, instr, &new_tail)) {\n // No sequence found!\n // NOTE: If you encounter this after adding a new instruction, do a full\n // rebuild!\n assert_always();\n XELOGE(\"Unable to process HIR opcode {}\", instr->opcode->name);\n break;\n }\n instr = new_tail;\n }\n\n block = block->next;\n }\n\n // Function epilog.\n L(epilog_label);\n epilog_label_ = nullptr;\n EmitTraceUserCallReturn();\n mov(GetContextReg(), qword[rsp + StackLayout::GUEST_CTX_HOME]);\n\n code_offsets.epilog = getSize();\n\n add(rsp, (uint32_t)stack_size);\n ret();\n\n code_offsets.tail = getSize();\n\n if (cvars::emit_source_annotations) {\n nop();\n nop();\n nop();\n nop();\n nop();\n }\n\n assert_zero(code_offsets.prolog);\n func_info.code_size.total = getSize();\n func_info.code_size.prolog = code_offsets.body - code_offsets.prolog;\n func_info.code_size.body = code_offsets.epilog - code_offsets.body;\n func_info.code_size.epilog = code_offsets.tail - code_offsets.epilog;\n func_info.code_size.tail = getSize() - code_offsets.tail;\n func_info.prolog_stack_alloc_offset =\n code_offsets.prolog_stack_alloc - code_offsets.prolog;\n\n return true;\n}\n\nvoid X64Emitter::MarkSourceOffset(const Instr* i) {\n auto entry = source_map_arena_.Alloc();\n entry->guest_address = static_cast(i->src1.offset);\n entry->hir_offset = uint32_t(i->block->ordinal << 16) | i->ordinal;\n entry->code_offset = static_cast(getSize());\n\n if (cvars::emit_source_annotations) {\n nop();\n nop();\n mov(eax, entry->guest_address);\n nop();\n nop();\n }\n\n if (debug_info_flags_ & DebugInfoFlags::kDebugInfoTraceFunctionCoverage) {\n uint32_t instruction_index =\n (entry->guest_address - trace_data_->start_address()) / 4;\n lock();\n inc(qword[low_address(trace_data_->instruction_execute_counts() +\n instruction_index * 8)]);\n }\n}\n\nvoid X64Emitter::EmitGetCurrentThreadId() {\n // rsi must point to context. We could fetch from the stack if needed.\n mov(ax, word[GetContextReg() + offsetof(ppc::PPCContext, thread_id)]);\n}\n\nvoid X64Emitter::EmitTraceUserCallReturn() {}\n\nvoid X64Emitter::DebugBreak() {\n // TODO(benvanik): notify debugger.\n db(0xCC);\n}\n\nuint64_t TrapDebugPrint(void* raw_context, uint64_t address) {\n auto thread_state = *reinterpret_cast(raw_context);\n uint32_t str_ptr = uint32_t(thread_state->context()->r[3]);\n // uint16_t str_len = uint16_t(thread_state->context()->r[4]);\n auto str = thread_state->memory()->TranslateVirtual(str_ptr);\n // TODO(benvanik): truncate to length?\n XELOGD(\"(DebugPrint) {}\", str);\n\n if (cvars::debugprint_trap_log) {\n debugging::DebugPrint(\"(DebugPrint) {}\", str);\n }\n\n return 0;\n}\n\nuint64_t TrapDebugBreak(void* raw_context, uint64_t address) {\n auto thread_state = *reinterpret_cast(raw_context);\n XELOGE(\"tw/td forced trap hit! This should be a crash!\");\n if (cvars::break_on_debugbreak) {\n xe::debugging::Break();\n }\n return 0;\n}\n\nvoid X64Emitter::Trap(uint16_t trap_type) {\n switch (trap_type) {\n case 20:\n case 26:\n // 0x0FE00014 is a 'debug print' where r3 = buffer r4 = length\n CallNative(TrapDebugPrint, 0);\n break;\n case 0:\n case 22:\n // Always trap?\n // TODO(benvanik): post software interrupt to debugger.\n CallNative(TrapDebugBreak, 0);\n break;\n case 25:\n // ?\n break;\n default:\n XELOGW(\"Unknown trap type {}\", trap_type);\n db(0xCC);\n break;\n }\n}\n\nvoid X64Emitter::UnimplementedInstr(const hir::Instr* i) {\n // TODO(benvanik): notify debugger.\n db(0xCC);\n assert_always();\n}\n\n// This is used by the X64ThunkEmitter's ResolveFunctionThunk.\nuint64_t ResolveFunction(void* raw_context, uint64_t target_address) {\n auto thread_state = *reinterpret_cast(raw_context);\n\n // TODO(benvanik): required?\n assert_not_zero(target_address);\n\n auto fn = thread_state->processor()->ResolveFunction(\n static_cast(target_address));\n assert_not_null(fn);\n auto x64_fn = static_cast(fn);\n uint64_t addr = reinterpret_cast(x64_fn->machine_code());\n\n return addr;\n}\n\nvoid X64Emitter::Call(const hir::Instr* instr, GuestFunction* function) {\n assert_not_null(function);\n auto fn = static_cast(function);\n // Resolve address to the function to call and store in rax.\n if (fn->machine_code()) {\n // TODO(benvanik): is it worth it to do this? It removes the need for\n // a ResolveFunction call, but makes the table less useful.\n assert_zero(uint64_t(fn->machine_code()) & 0xFFFFFFFF00000000);\n mov(eax, uint32_t(uint64_t(fn->machine_code())));\n } else if (code_cache_->has_indirection_table()) {\n // Load the pointer to the indirection table maintained in X64CodeCache.\n // The target dword will either contain the address of the generated code\n // or a thunk to ResolveAddress.\n mov(ebx, function->address());\n mov(eax, dword[ebx]);\n } else {\n // Old-style resolve.\n // Not too important because indirection table is almost always available.\n // TODO: Overwrite the call-site with a straight call.\n CallNative(&ResolveFunction, function->address());\n }\n\n // Actually jump/call to rax.\n if (instr->flags & hir::CALL_TAIL) {\n // Since we skip the prolog we need to mark the return here.\n EmitTraceUserCallReturn();\n\n // Pass the callers return address over.\n mov(rcx, qword[rsp + StackLayout::GUEST_RET_ADDR]);\n\n add(rsp, static_cast(stack_size()));\n jmp(rax);\n } else {\n // Return address is from the previous SET_RETURN_ADDRESS.\n mov(rcx, qword[rsp + StackLayout::GUEST_CALL_RET_ADDR]);\n\n call(rax);\n }\n}\n\nvoid X64Emitter::CallIndirect(const hir::Instr* instr,\n const Xbyak::Reg64& reg) {\n // Check if return.\n if (instr->flags & hir::CALL_POSSIBLE_RETURN) {\n cmp(reg.cvt32(), dword[rsp + StackLayout::GUEST_RET_ADDR]);\n je(epilog_label(), CodeGenerator::T_NEAR);\n }\n\n // Load the pointer to the indirection table maintained in X64CodeCache.\n // The target dword will either contain the address of the generated code\n // or a thunk to ResolveAddress.\n if (code_cache_->has_indirection_table()) {\n if (reg.cvt32() != ebx) {\n mov(ebx, reg.cvt32());\n }\n mov(eax, dword[ebx]);\n } else {\n // Old-style resolve.\n // Not too important because indirection table is almost always available.\n mov(edx, reg.cvt32());\n mov(rax, reinterpret_cast(ResolveFunction));\n mov(rcx, GetContextReg());\n call(rax);\n }\n\n // Actually jump/call to rax.\n if (instr->flags & hir::CALL_TAIL) {\n // Since we skip the prolog we need to mark the return here.\n EmitTraceUserCallReturn();\n\n // Pass the callers return address over.\n mov(rcx, qword[rsp + StackLayout::GUEST_RET_ADDR]);\n\n add(rsp, static_cast(stack_size()));\n jmp(rax);\n } else {\n // Return address is from the previous SET_RETURN_ADDRESS.\n mov(rcx, qword[rsp + StackLayout::GUEST_CALL_RET_ADDR]);\n\n call(rax);\n }\n}\n\nuint64_t UndefinedCallExtern(void* raw_context, uint64_t function_ptr) {\n auto function = reinterpret_cast(function_ptr);\n if (!cvars::ignore_undefined_externs) {\n xe::FatalError(fmt::format(\"undefined extern call to {:08X} {}\",\n function->address(), function->name().c_str()));\n } else {\n XELOGE(\"undefined extern call to {:08X} {}\", function->address(),\n function->name());\n }\n return 0;\n}\nvoid X64Emitter::CallExtern(const hir::Instr* instr, const Function* function) {\n bool undefined = true;\n if (function->behavior() == Function::Behavior::kBuiltin) {\n auto builtin_function = static_cast(function);\n if (builtin_function->handler()) {\n undefined = false;\n // rcx = target function\n // rdx = arg0\n // r8 = arg1\n // r9 = arg2\n auto thunk = backend()->guest_to_host_thunk();\n mov(rax, reinterpret_cast(thunk));\n mov(rcx, reinterpret_cast(builtin_function->handler()));\n mov(rdx, reinterpret_cast(builtin_function->arg0()));\n mov(r8, reinterpret_cast(builtin_function->arg1()));\n call(rax);\n // rax = host return\n }\n } else if (function->behavior() == Function::Behavior::kExtern) {\n auto extern_function = static_cast(function);\n if (extern_function->extern_handler()) {\n undefined = false;\n // rcx = target function\n // rdx = arg0\n // r8 = arg1\n // r9 = arg2\n auto thunk = backend()->guest_to_host_thunk();\n mov(rax, reinterpret_cast(thunk));\n mov(rcx, reinterpret_cast(extern_function->extern_handler()));\n mov(rdx,\n qword[GetContextReg() + offsetof(ppc::PPCContext, kernel_state)]);\n call(rax);\n // rax = host return\n }\n }\n if (undefined) {\n CallNative(UndefinedCallExtern, reinterpret_cast(function));\n }\n}\n\nvoid X64Emitter::CallNative(void* fn) { CallNativeSafe(fn); }\n\nvoid X64Emitter::CallNative(uint64_t (*fn)(void* raw_context)) {\n CallNativeSafe(reinterpret_cast(fn));\n}\n\nvoid X64Emitter::CallNative(uint64_t (*fn)(void* raw_context, uint64_t arg0)) {\n CallNativeSafe(reinterpret_cast(fn));\n}\n\nvoid X64Emitter::CallNative(uint64_t (*fn)(void* raw_context, uint64_t arg0),\n uint64_t arg0) {\n mov(GetNativeParam(0), arg0);\n CallNativeSafe(reinterpret_cast(fn));\n}\n\nvoid X64Emitter::CallNativeSafe(void* fn) {\n // rcx = target function\n // rdx = arg0\n // r8 = arg1\n // r9 = arg2\n auto thunk = backend()->guest_to_host_thunk();\n mov(rax, reinterpret_cast(thunk));\n mov(rcx, reinterpret_cast(fn));\n call(rax);\n // rax = host return\n}\n\nvoid X64Emitter::SetReturnAddress(uint64_t value) {\n mov(rax, value);\n mov(qword[rsp + StackLayout::GUEST_CALL_RET_ADDR], rax);\n}\n\nXbyak::Reg64 X64Emitter::GetNativeParam(uint32_t param) {\n if (param == 0)\n return rdx;\n else if (param == 1)\n return r8;\n else if (param == 2)\n return r9;\n\n assert_always();\n return r9;\n}\n\n// Important: If you change these, you must update the thunks in x64_backend.cc!\nXbyak::Reg64 X64Emitter::GetContextReg() { return rsi; }\nXbyak::Reg64 X64Emitter::GetMembaseReg() { return rdi; }\n\nvoid X64Emitter::ReloadContext() {\n mov(GetContextReg(), qword[rsp + StackLayout::GUEST_CTX_HOME]);\n}\n\nvoid X64Emitter::ReloadMembase() {\n mov(GetMembaseReg(), qword[GetContextReg() + 8]); // membase\n}\n\n// Len Assembly Byte Sequence\n// ============================================================================\n// 1b NOP 90H\n// 2b 66 NOP 66 90H\n// 3b NOP DWORD ptr [EAX] 0F 1F 00H\n// 4b NOP DWORD ptr [EAX + 00H] 0F 1F 40 00H\n// 5b NOP DWORD ptr [EAX + EAX*1 + 00H] 0F 1F 44 00 00H\n// 6b 66 NOP DWORD ptr [EAX + EAX*1 + 00H] 66 0F 1F 44 00 00H\n// 7b NOP DWORD ptr [EAX + 00000000H] 0F 1F 80 00 00 00 00H\n// 8b NOP DWORD ptr [EAX + EAX*1 + 00000000H] 0F 1F 84 00 00 00 00 00H\n// 9b 66 NOP DWORD ptr [EAX + EAX*1 + 00000000H] 66 0F 1F 84 00 00 00 00 00H\nvoid X64Emitter::nop(size_t length) {\n // TODO(benvanik): fat nop\n for (size_t i = 0; i < length; ++i) {\n db(0x90);\n }\n}\n\nbool X64Emitter::ConstantFitsIn32Reg(uint64_t v) {\n if ((v & ~0x7FFFFFFF) == 0) {\n // Fits under 31 bits, so just load using normal mov.\n return true;\n } else if ((v & ~0x7FFFFFFF) == ~0x7FFFFFFF) {\n // Negative number that fits in 32bits.\n return true;\n }\n return false;\n}\n\nvoid X64Emitter::MovMem64(const Xbyak::RegExp& addr, uint64_t v) {\n if ((v & ~0x7FFFFFFF) == 0) {\n // Fits under 31 bits, so just load using normal mov.\n mov(qword[addr], v);\n } else if ((v & ~0x7FFFFFFF) == ~0x7FFFFFFF) {\n // Negative number that fits in 32bits.\n mov(qword[addr], v);\n } else if (!(v >> 32)) {\n // All high bits are zero. It'd be nice if we had a way to load a 32bit\n // immediate without sign extending!\n // TODO(benvanik): this is super common, find a better way.\n mov(dword[addr], static_cast(v));\n mov(dword[addr + 4], 0);\n } else {\n // 64bit number that needs double movs.\n mov(dword[addr], static_cast(v));\n mov(dword[addr + 4], static_cast(v >> 32));\n }\n}\n\nstatic const vec128_t xmm_consts[] = {\n /* XMMZero */ vec128f(0.0f),\n /* XMMOne */ vec128f(1.0f),\n /* XMMOnePD */ vec128d(1.0),\n /* XMMNegativeOne */ vec128f(-1.0f, -1.0f, -1.0f, -1.0f),\n /* XMMFFFF */\n vec128i(0xFFFFFFFFu, 0xFFFFFFFFu, 0xFFFFFFFFu, 0xFFFFFFFFu),\n /* XMMMaskX16Y16 */\n vec128i(0x0000FFFFu, 0xFFFF0000u, 0x00000000u, 0x00000000u),\n /* XMMFlipX16Y16 */\n vec128i(0x00008000u, 0x00000000u, 0x00000000u, 0x00000000u),\n /* XMMFixX16Y16 */ vec128f(-32768.0f, 0.0f, 0.0f, 0.0f),\n /* XMMNormalizeX16Y16 */\n vec128f(1.0f / 32767.0f, 1.0f / (32767.0f * 65536.0f), 0.0f, 0.0f),\n /* XMM0001 */ vec128f(0.0f, 0.0f, 0.0f, 1.0f),\n /* XMM3301 */ vec128f(3.0f, 3.0f, 0.0f, 1.0f),\n /* XMM3331 */ vec128f(3.0f, 3.0f, 3.0f, 1.0f),\n /* XMM3333 */ vec128f(3.0f, 3.0f, 3.0f, 3.0f),\n /* XMMSignMaskPS */\n vec128i(0x80000000u, 0x80000000u, 0x80000000u, 0x80000000u),\n /* XMMSignMaskPD */\n vec128i(0x00000000u, 0x80000000u, 0x00000000u, 0x80000000u),\n /* XMMAbsMaskPS */\n vec128i(0x7FFFFFFFu, 0x7FFFFFFFu, 0x7FFFFFFFu, 0x7FFFFFFFu),\n /* XMMAbsMaskPD */\n vec128i(0xFFFFFFFFu, 0x7FFFFFFFu, 0xFFFFFFFFu, 0x7FFFFFFFu),\n /* XMMByteSwapMask */\n vec128i(0x00010203u, 0x04050607u, 0x08090A0Bu, 0x0C0D0E0Fu),\n /* XMMByteOrderMask */\n vec128i(0x01000302u, 0x05040706u, 0x09080B0Au, 0x0D0C0F0Eu),\n /* XMMPermuteControl15 */ vec128b(15),\n /* XMMPermuteByteMask */ vec128b(0x1F),\n /* XMMPackD3DCOLORSat */ vec128i(0x404000FFu),\n /* XMMPackD3DCOLOR */\n vec128i(0xFFFFFFFFu, 0xFFFFFFFFu, 0xFFFFFFFFu, 0x0C000408u),\n /* XMMUnpackD3DCOLOR */\n vec128i(0xFFFFFF0Eu, 0xFFFFFF0Du, 0xFFFFFF0Cu, 0xFFFFFF0Fu),\n /* XMMPackFLOAT16_2 */\n vec128i(0xFFFFFFFFu, 0xFFFFFFFFu, 0xFFFFFFFFu, 0x01000302u),\n /* XMMUnpackFLOAT16_2 */\n vec128i(0x0D0C0F0Eu, 0xFFFFFFFFu, 0xFFFFFFFFu, 0xFFFFFFFFu),\n /* XMMPackFLOAT16_4 */\n vec128i(0xFFFFFFFFu, 0xFFFFFFFFu, 0x01000302u, 0x05040706u),\n /* XMMUnpackFLOAT16_4 */\n vec128i(0x09080B0Au, 0x0D0C0F0Eu, 0xFFFFFFFFu, 0xFFFFFFFFu),\n /* XMMPackSHORT_Min */ vec128i(0x403F8001u),\n /* XMMPackSHORT_Max */ vec128i(0x40407FFFu),\n /* XMMPackSHORT_2 */\n vec128i(0xFFFFFFFFu, 0xFFFFFFFFu, 0xFFFFFFFFu, 0x01000504u),\n /* XMMPackSHORT_4 */\n vec128i(0xFFFFFFFFu, 0xFFFFFFFFu, 0x01000504u, 0x09080D0Cu),\n /* XMMUnpackSHORT_2 */\n vec128i(0xFFFF0F0Eu, 0xFFFF0D0Cu, 0xFFFFFFFFu, 0xFFFFFFFFu),\n /* XMMUnpackSHORT_4 */\n vec128i(0xFFFF0B0Au, 0xFFFF0908u, 0xFFFF0F0Eu, 0xFFFF0D0Cu),\n /* XMMUnpackSHORT_Overflow */ vec128i(0x403F8000u),\n /* XMMPackUINT_2101010_MinUnpacked */\n vec128i(0x403FFE01u, 0x403FFE01u, 0x403FFE01u, 0x40400000u),\n /* XMMPackUINT_2101010_MaxUnpacked */\n vec128i(0x404001FFu, 0x404001FFu, 0x404001FFu, 0x40400003u),\n /* XMMPackUINT_2101010_MaskUnpacked */\n vec128i(0x3FFu, 0x3FFu, 0x3FFu, 0x3u),\n /* XMMPackUINT_2101010_MaskPacked */\n vec128i(0x3FFu, 0x3FFu << 10, 0x3FFu << 20, 0x3u << 30),\n /* XMMPackUINT_2101010_Shift */ vec128i(0, 10, 20, 30),\n /* XMMUnpackUINT_2101010_Overflow */ vec128i(0x403FFE00u),\n /* XMMPackULONG_4202020_MinUnpacked */\n vec128i(0x40380001u, 0x40380001u, 0x40380001u, 0x40400000u),\n /* XMMPackULONG_4202020_MaxUnpacked */\n vec128i(0x4047FFFFu, 0x4047FFFFu, 0x4047FFFFu, 0x4040000Fu),\n /* XMMPackULONG_4202020_MaskUnpacked */\n vec128i(0xFFFFFu, 0xFFFFFu, 0xFFFFFu, 0xFu),\n /* XMMPackULONG_4202020_PermuteXZ */\n vec128i(0xFFFFFFFFu, 0xFFFFFFFFu, 0x0A0908FFu, 0xFF020100u),\n /* XMMPackULONG_4202020_PermuteYW */\n vec128i(0xFFFFFFFFu, 0xFFFFFFFFu, 0x0CFFFF06u, 0x0504FFFFu),\n /* XMMUnpackULONG_4202020_Permute */\n vec128i(0xFF0E0D0Cu, 0xFF0B0A09u, 0xFF080F0Eu, 0xFFFFFF0Bu),\n /* XMMUnpackULONG_4202020_Overflow */ vec128i(0x40380000u),\n /* XMMOneOver255 */ vec128f(1.0f / 255.0f),\n /* XMMMaskEvenPI16 */\n vec128i(0x0000FFFFu, 0x0000FFFFu, 0x0000FFFFu, 0x0000FFFFu),\n /* XMMShiftMaskEvenPI16 */\n vec128i(0x0000000Fu, 0x0000000Fu, 0x0000000Fu, 0x0000000Fu),\n /* XMMShiftMaskPS */\n vec128i(0x0000001Fu, 0x0000001Fu, 0x0000001Fu, 0x0000001Fu),\n /* XMMShiftByteMask */\n vec128i(0x000000FFu, 0x000000FFu, 0x000000FFu, 0x000000FFu),\n /* XMMSwapWordMask */\n vec128i(0x03030303u, 0x03030303u, 0x03030303u, 0x03030303u),\n /* XMMUnsignedDwordMax */\n vec128i(0xFFFFFFFFu, 0x00000000u, 0xFFFFFFFFu, 0x00000000u),\n /* XMM255 */ vec128f(255.0f),\n /* XMMPI32 */ vec128i(32),\n /* XMMSignMaskI8 */\n vec128i(0x80808080u, 0x80808080u, 0x80808080u, 0x80808080u),\n /* XMMSignMaskI16 */\n vec128i(0x80008000u, 0x80008000u, 0x80008000u, 0x80008000u),\n /* XMMSignMaskI32 */\n vec128i(0x80000000u, 0x80000000u, 0x80000000u, 0x80000000u),\n /* XMMSignMaskF32 */\n vec128i(0x80000000u, 0x80000000u, 0x80000000u, 0x80000000u),\n /* XMMShortMinPS */ vec128f(SHRT_MIN),\n /* XMMShortMaxPS */ vec128f(SHRT_MAX),\n /* XMMIntMin */ vec128i(INT_MIN),\n /* XMMIntMax */ vec128i(INT_MAX),\n /* XMMIntMaxPD */ vec128d(INT_MAX),\n /* XMMPosIntMinPS */ vec128f((float)0x80000000u),\n /* XMMQNaN */ vec128i(0x7FC00000u),\n /* XMMInt127 */ vec128i(0x7Fu),\n /* XMM2To32 */ vec128f(0x1.0p32f),\n};\n\n// First location to try and place constants.\nstatic const uintptr_t kConstDataLocation = 0x20000000;\nstatic const uintptr_t kConstDataSize = sizeof(xmm_consts);\n\n// Increment the location by this amount for every allocation failure.\nstatic const uintptr_t kConstDataIncrement = 0x00001000;\n\n// This function places constant data that is used by the emitter later on.\n// Only called once and used by multiple instances of the emitter.\n//\n// TODO(DrChat): This should be placed in the code cache with the code, but\n// doing so requires RIP-relative addressing, which is difficult to support\n// given the current setup.\nuintptr_t X64Emitter::PlaceConstData() {\n uint8_t* ptr = reinterpret_cast(kConstDataLocation);\n void* mem = nullptr;\n while (!mem) {\n mem = memory::AllocFixed(\n ptr, xe::round_up(kConstDataSize, memory::page_size()),\n memory::AllocationType::kReserveCommit, memory::PageAccess::kReadWrite);\n\n ptr += kConstDataIncrement;\n }\n\n // The pointer must not be greater than 31 bits.\n assert_zero(reinterpret_cast(mem) & ~0x7FFFFFFF);\n std::memcpy(mem, xmm_consts, sizeof(xmm_consts));\n memory::Protect(mem, kConstDataSize, memory::PageAccess::kReadOnly, nullptr);\n\n return reinterpret_cast(mem);\n}\n\nvoid X64Emitter::FreeConstData(uintptr_t data) {\n memory::DeallocFixed(reinterpret_cast(data), 0,\n memory::DeallocationType::kRelease);\n}\n\nXbyak::Address X64Emitter::GetXmmConstPtr(XmmConst id) {\n // Load through fixed constant table setup by PlaceConstData.\n // It's important that the pointer is not signed, as it will be sign-extended.\n return ptr[reinterpret_cast(backend_->emitter_data() +\n sizeof(vec128_t) * id)];\n}\n\n// Implies possible StashXmm(0, ...)!\nvoid X64Emitter::LoadConstantXmm(Xbyak::Xmm dest, const vec128_t& v) {\n // https://www.agner.org/optimize/optimizing_assembly.pdf\n // 13.4 Generating constants\n if (!v.low && !v.high) {\n // 0000...\n vpxor(dest, dest);\n } else if (v.low == ~uint64_t(0) && v.high == ~uint64_t(0)) {\n // 1111...\n vpcmpeqb(dest, dest);\n } else {\n // TODO(benvanik): see what other common values are.\n // TODO(benvanik): build constant table - 99% are reused.\n MovMem64(rsp + kStashOffset, v.low);\n MovMem64(rsp + kStashOffset + 8, v.high);\n vmovdqa(dest, ptr[rsp + kStashOffset]);\n }\n}\n\nvoid X64Emitter::LoadConstantXmm(Xbyak::Xmm dest, float v) {\n union {\n float f;\n uint32_t i;\n } x = {v};\n if (!x.i) {\n // +0.0f (but not -0.0f because it may be used to flip the sign via xor).\n vpxor(dest, dest);\n } else if (x.i == ~uint32_t(0)) {\n // 1111...\n vpcmpeqb(dest, dest);\n } else {\n // TODO(benvanik): see what other common values are.\n // TODO(benvanik): build constant table - 99% are reused.\n mov(eax, x.i);\n vmovd(dest, eax);\n }\n}\n\nvoid X64Emitter::LoadConstantXmm(Xbyak::Xmm dest, double v) {\n union {\n double d;\n uint64_t i;\n } x = {v};\n if (!x.i) {\n // +0.0 (but not -0.0 because it may be used to flip the sign via xor).\n vpxor(dest, dest);\n } else if (x.i == ~uint64_t(0)) {\n // 1111...\n vpcmpeqb(dest, dest);\n } else {\n // TODO(benvanik): see what other common values are.\n // TODO(benvanik): build constant table - 99% are reused.\n mov(rax, x.i);\n vmovq(dest, rax);\n }\n}\n\nXbyak::Address X64Emitter::StashXmm(int index, const Xbyak::Xmm& r) {\n auto addr = ptr[rsp + kStashOffset + (index * 16)];\n vmovups(addr, r);\n return addr;\n}\n\nXbyak::Address X64Emitter::StashConstantXmm(int index, float v) {\n union {\n float f;\n uint32_t i;\n } x = {v};\n auto addr = rsp + kStashOffset + (index * 16);\n MovMem64(addr, x.i);\n MovMem64(addr + 8, 0);\n return ptr[addr];\n}\n\nXbyak::Address X64Emitter::StashConstantXmm(int index, double v) {\n union {\n double d;\n uint64_t i;\n } x = {v};\n auto addr = rsp + kStashOffset + (index * 16);\n MovMem64(addr, x.i);\n MovMem64(addr + 8, 0);\n return ptr[addr];\n}\n\nXbyak::Address X64Emitter::StashConstantXmm(int index, const vec128_t& v) {\n auto addr = rsp + kStashOffset + (index * 16);\n MovMem64(addr, v.low);\n MovMem64(addr + 8, v.high);\n return ptr[addr];\n}\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_emitter.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_EMITTER_H_\n#define XENIA_CPU_BACKEND_X64_X64_EMITTER_H_\n\n#include \n\n#include \"xenia/base/arena.h\"\n#include \"xenia/cpu/function.h\"\n#include \"xenia/cpu/function_trace_data.h\"\n#include \"xenia/cpu/hir/hir_builder.h\"\n#include \"xenia/cpu/hir/instr.h\"\n#include \"xenia/cpu/hir/value.h\"\n#include \"xenia/memory.h\"\n\n// NOTE: must be included last as it expects windows.h to already be included.\n#include \"third_party/xbyak/xbyak/xbyak.h\"\n#include \"third_party/xbyak/xbyak/xbyak_util.h\"\n\nnamespace xe {\nnamespace cpu {\nclass Processor;\n} // namespace cpu\n} // namespace xe\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nclass X64Backend;\nclass X64CodeCache;\n\nstruct EmitFunctionInfo;\n\nenum RegisterFlags {\n REG_DEST = (1 << 0),\n REG_ABCD = (1 << 1),\n};\n\nenum XmmConst {\n XMMZero = 0,\n XMMOne,\n XMMOnePD,\n XMMNegativeOne,\n XMMFFFF,\n XMMMaskX16Y16,\n XMMFlipX16Y16,\n XMMFixX16Y16,\n XMMNormalizeX16Y16,\n XMM0001,\n XMM3301,\n XMM3331,\n XMM3333,\n XMMSignMaskPS,\n XMMSignMaskPD,\n XMMAbsMaskPS,\n XMMAbsMaskPD,\n XMMByteSwapMask,\n XMMByteOrderMask,\n XMMPermuteControl15,\n XMMPermuteByteMask,\n XMMPackD3DCOLORSat,\n XMMPackD3DCOLOR,\n XMMUnpackD3DCOLOR,\n XMMPackFLOAT16_2,\n XMMUnpackFLOAT16_2,\n XMMPackFLOAT16_4,\n XMMUnpackFLOAT16_4,\n XMMPackSHORT_Min,\n XMMPackSHORT_Max,\n XMMPackSHORT_2,\n XMMPackSHORT_4,\n XMMUnpackSHORT_2,\n XMMUnpackSHORT_4,\n XMMUnpackSHORT_Overflow,\n XMMPackUINT_2101010_MinUnpacked,\n XMMPackUINT_2101010_MaxUnpacked,\n XMMPackUINT_2101010_MaskUnpacked,\n XMMPackUINT_2101010_MaskPacked,\n XMMPackUINT_2101010_Shift,\n XMMUnpackUINT_2101010_Overflow,\n XMMPackULONG_4202020_MinUnpacked,\n XMMPackULONG_4202020_MaxUnpacked,\n XMMPackULONG_4202020_MaskUnpacked,\n XMMPackULONG_4202020_PermuteXZ,\n XMMPackULONG_4202020_PermuteYW,\n XMMUnpackULONG_4202020_Permute,\n XMMUnpackULONG_4202020_Overflow,\n XMMOneOver255,\n XMMMaskEvenPI16,\n XMMShiftMaskEvenPI16,\n XMMShiftMaskPS,\n XMMShiftByteMask,\n XMMSwapWordMask,\n XMMUnsignedDwordMax,\n XMM255,\n XMMPI32,\n XMMSignMaskI8,\n XMMSignMaskI16,\n XMMSignMaskI32,\n XMMSignMaskF32,\n XMMShortMinPS,\n XMMShortMaxPS,\n XMMIntMin,\n XMMIntMax,\n XMMIntMaxPD,\n XMMPosIntMinPS,\n XMMQNaN,\n XMMInt127,\n XMM2To32,\n};\n\n// Unfortunately due to the design of xbyak we have to pass this to the ctor.\nclass XbyakAllocator : public Xbyak::Allocator {\n public:\n virtual bool useProtect() const { return false; }\n};\n\nenum X64EmitterFeatureFlags {\n kX64EmitAVX2 = 1 << 0,\n kX64EmitFMA = 1 << 1,\n kX64EmitLZCNT = 1 << 2,\n kX64EmitBMI1 = 1 << 3,\n kX64EmitBMI2 = 1 << 4,\n kX64EmitF16C = 1 << 5,\n kX64EmitMovbe = 1 << 6,\n kX64EmitGFNI = 1 << 7,\n\n kX64EmitAVX512F = 1 << 8,\n kX64EmitAVX512VL = 1 << 9,\n\n kX64EmitAVX512BW = 1 << 10,\n kX64EmitAVX512DQ = 1 << 11,\n\n kX64EmitAVX512VBMI = 1 << 12,\n\n kX64EmitAVX512Ortho = kX64EmitAVX512F | kX64EmitAVX512VL,\n kX64EmitAVX512Ortho64 = kX64EmitAVX512Ortho | kX64EmitAVX512DQ\n};\n\nclass X64Emitter : public Xbyak::CodeGenerator {\n public:\n X64Emitter(X64Backend* backend, XbyakAllocator* allocator);\n virtual ~X64Emitter();\n\n Processor* processor() const { return processor_; }\n X64Backend* backend() const { return backend_; }\n\n static uintptr_t PlaceConstData();\n static void FreeConstData(uintptr_t data);\n\n bool Emit(GuestFunction* function, hir::HIRBuilder* builder,\n uint32_t debug_info_flags, FunctionDebugInfo* debug_info,\n void** out_code_address, size_t* out_code_size,\n std::vector* out_source_map);\n\n public:\n // Reserved: rsp, rsi, rdi\n // Scratch: rax/rcx/rdx\n // xmm0-2\n // Available: rbx, r10-r15\n // xmm4-xmm15 (save to get xmm3)\n static const int GPR_COUNT = 7;\n static const int XMM_COUNT = 12;\n\n static void SetupReg(const hir::Value* v, Xbyak::Reg8& r) {\n auto idx = gpr_reg_map_[v->reg.index];\n r = Xbyak::Reg8(idx);\n }\n static void SetupReg(const hir::Value* v, Xbyak::Reg16& r) {\n auto idx = gpr_reg_map_[v->reg.index];\n r = Xbyak::Reg16(idx);\n }\n static void SetupReg(const hir::Value* v, Xbyak::Reg32& r) {\n auto idx = gpr_reg_map_[v->reg.index];\n r = Xbyak::Reg32(idx);\n }\n static void SetupReg(const hir::Value* v, Xbyak::Reg64& r) {\n auto idx = gpr_reg_map_[v->reg.index];\n r = Xbyak::Reg64(idx);\n }\n static void SetupReg(const hir::Value* v, Xbyak::Xmm& r) {\n auto idx = xmm_reg_map_[v->reg.index];\n r = Xbyak::Xmm(idx);\n }\n\n Xbyak::Label& epilog_label() { return *epilog_label_; }\n\n void MarkSourceOffset(const hir::Instr* i);\n\n void DebugBreak();\n void Trap(uint16_t trap_type = 0);\n void UnimplementedInstr(const hir::Instr* i);\n\n void Call(const hir::Instr* instr, GuestFunction* function);\n void CallIndirect(const hir::Instr* instr, const Xbyak::Reg64& reg);\n void CallExtern(const hir::Instr* instr, const Function* function);\n void CallNative(void* fn);\n void CallNative(uint64_t (*fn)(void* raw_context));\n void CallNative(uint64_t (*fn)(void* raw_context, uint64_t arg0));\n void CallNative(uint64_t (*fn)(void* raw_context, uint64_t arg0),\n uint64_t arg0);\n void CallNativeSafe(void* fn);\n void SetReturnAddress(uint64_t value);\n\n Xbyak::Reg64 GetNativeParam(uint32_t param);\n\n Xbyak::Reg64 GetContextReg();\n Xbyak::Reg64 GetMembaseReg();\n void ReloadContext();\n void ReloadMembase();\n\n void nop(size_t length = 1);\n\n // Moves a 64bit immediate into memory.\n bool ConstantFitsIn32Reg(uint64_t v);\n void MovMem64(const Xbyak::RegExp& addr, uint64_t v);\n\n Xbyak::Address GetXmmConstPtr(XmmConst id);\n void LoadConstantXmm(Xbyak::Xmm dest, float v);\n void LoadConstantXmm(Xbyak::Xmm dest, double v);\n void LoadConstantXmm(Xbyak::Xmm dest, const vec128_t& v);\n Xbyak::Address StashXmm(int index, const Xbyak::Xmm& r);\n Xbyak::Address StashConstantXmm(int index, float v);\n Xbyak::Address StashConstantXmm(int index, double v);\n Xbyak::Address StashConstantXmm(int index, const vec128_t& v);\n\n bool IsFeatureEnabled(uint32_t feature_flag) const {\n return (feature_flags_ & feature_flag) == feature_flag;\n }\n\n FunctionDebugInfo* debug_info() const { return debug_info_; }\n\n size_t stack_size() const { return stack_size_; }\n\n protected:\n void* Emplace(const EmitFunctionInfo& func_info,\n GuestFunction* function = nullptr);\n bool Emit(hir::HIRBuilder* builder, EmitFunctionInfo& func_info);\n void EmitGetCurrentThreadId();\n void EmitTraceUserCallReturn();\n\n protected:\n Processor* processor_ = nullptr;\n X64Backend* backend_ = nullptr;\n X64CodeCache* code_cache_ = nullptr;\n XbyakAllocator* allocator_ = nullptr;\n Xbyak::util::Cpu cpu_;\n uint32_t feature_flags_ = 0;\n\n Xbyak::Label* epilog_label_ = nullptr;\n\n hir::Instr* current_instr_ = nullptr;\n\n FunctionDebugInfo* debug_info_ = nullptr;\n uint32_t debug_info_flags_ = 0;\n FunctionTraceData* trace_data_ = nullptr;\n Arena source_map_arena_;\n\n size_t stack_size_ = 0;\n\n static const uint32_t gpr_reg_map_[GPR_COUNT];\n static const uint32_t xmm_reg_map_[XMM_COUNT];\n};\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_EMITTER_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_function.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_function.h\"\n\n#include \"xenia/cpu/backend/x64/x64_backend.h\"\n#include \"xenia/cpu/processor.h\"\n#include \"xenia/cpu/thread_state.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nX64Function::X64Function(Module* module, uint32_t address)\n : GuestFunction(module, address) {}\n\nX64Function::~X64Function() {\n // machine_code_ is freed by code cache.\n}\n\nvoid X64Function::Setup(uint8_t* machine_code, size_t machine_code_length) {\n machine_code_ = machine_code;\n machine_code_length_ = machine_code_length;\n}\n\nbool X64Function::CallImpl(ThreadState* thread_state, uint32_t return_address) {\n auto backend =\n reinterpret_cast(thread_state->processor()->backend());\n auto thunk = backend->host_to_guest_thunk();\n thunk(machine_code_, thread_state->context(),\n reinterpret_cast(uintptr_t(return_address)));\n return true;\n}\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_function.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_FUNCTION_H_\n#define XENIA_CPU_BACKEND_X64_X64_FUNCTION_H_\n\n#include \"xenia/cpu/function.h\"\n#include \"xenia/cpu/thread_state.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nclass X64Function : public GuestFunction {\n public:\n X64Function(Module* module, uint32_t address);\n ~X64Function() override;\n\n uint8_t* machine_code() const override { return machine_code_; }\n size_t machine_code_length() const override { return machine_code_length_; }\n\n void Setup(uint8_t* machine_code, size_t machine_code_length);\n\n protected:\n bool CallImpl(ThreadState* thread_state, uint32_t return_address) override;\n\n private:\n uint8_t* machine_code_ = nullptr;\n size_t machine_code_length_ = 0;\n};\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_FUNCTION_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_op.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2018 Xenia Developers. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n#ifndef XENIA_CPU_BACKEND_X64_X64_OP_H_\n#define XENIA_CPU_BACKEND_X64_X64_OP_H_\n\n#include \"xenia/cpu/backend/x64/x64_emitter.h\"\n\n#include \"xenia/cpu/hir/instr.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\n// TODO(benvanik): direct usings.\nusing namespace xe::cpu;\nusing namespace xe::cpu::hir;\nusing namespace Xbyak;\n\n// Selects the right byte/word/etc from a vector. We need to flip logical\n// indices (0,1,2,3,4,5,6,7,...) = (3,2,1,0,7,6,5,4,...)\n#define VEC128_B(n) ((n) ^ 0x3)\n#define VEC128_W(n) ((n) ^ 0x1)\n#define VEC128_D(n) (n)\n#define VEC128_F(n) (n)\n\nenum KeyType {\n KEY_TYPE_X = OPCODE_SIG_TYPE_X,\n KEY_TYPE_L = OPCODE_SIG_TYPE_L,\n KEY_TYPE_O = OPCODE_SIG_TYPE_O,\n KEY_TYPE_S = OPCODE_SIG_TYPE_S,\n KEY_TYPE_V_I8 = OPCODE_SIG_TYPE_V + INT8_TYPE,\n KEY_TYPE_V_I16 = OPCODE_SIG_TYPE_V + INT16_TYPE,\n KEY_TYPE_V_I32 = OPCODE_SIG_TYPE_V + INT32_TYPE,\n KEY_TYPE_V_I64 = OPCODE_SIG_TYPE_V + INT64_TYPE,\n KEY_TYPE_V_F32 = OPCODE_SIG_TYPE_V + FLOAT32_TYPE,\n KEY_TYPE_V_F64 = OPCODE_SIG_TYPE_V + FLOAT64_TYPE,\n KEY_TYPE_V_V128 = OPCODE_SIG_TYPE_V + VEC128_TYPE,\n};\n\n#pragma pack(push, 1)\nunion InstrKey {\n uint32_t value;\n struct {\n uint32_t opcode : 8;\n uint32_t dest : 5;\n uint32_t src1 : 5;\n uint32_t src2 : 5;\n uint32_t src3 : 5;\n uint32_t reserved : 4;\n };\n\n operator uint32_t() const { return value; }\n\n InstrKey() : value(0) { static_assert_size(*this, sizeof(value)); }\n InstrKey(uint32_t v) : value(v) {}\n InstrKey(const Instr* i) : value(0) {\n opcode = i->opcode->num;\n uint32_t sig = i->opcode->signature;\n dest =\n GET_OPCODE_SIG_TYPE_DEST(sig) ? OPCODE_SIG_TYPE_V + i->dest->type : 0;\n src1 = GET_OPCODE_SIG_TYPE_SRC1(sig);\n if (src1 == OPCODE_SIG_TYPE_V) {\n src1 += i->src1.value->type;\n }\n src2 = GET_OPCODE_SIG_TYPE_SRC2(sig);\n if (src2 == OPCODE_SIG_TYPE_V) {\n src2 += i->src2.value->type;\n }\n src3 = GET_OPCODE_SIG_TYPE_SRC3(sig);\n if (src3 == OPCODE_SIG_TYPE_V) {\n src3 += i->src3.value->type;\n }\n }\n\n template \n struct Construct {\n static const uint32_t value =\n (OPCODE) | (DEST << 8) | (SRC1 << 13) | (SRC2 << 18) | (SRC3 << 23);\n };\n};\n#pragma pack(pop)\nstatic_assert(sizeof(InstrKey) <= 4, \"Key must be 4 bytes\");\n\ntemplate \nstruct CombinedStruct;\ntemplate <>\nstruct CombinedStruct<> {};\ntemplate \nstruct CombinedStruct : T, CombinedStruct {};\n\nstruct OpBase {};\n\ntemplate \nstruct Op : OpBase {\n static const KeyType key_type = KEY_TYPE;\n};\n\nstruct VoidOp : Op {\n protected:\n friend struct Op;\n template \n friend struct I;\n void Load(const Instr::Op& op) {}\n};\n\nstruct OffsetOp : Op {\n uint64_t value;\n\n protected:\n friend struct Op;\n template \n friend struct I;\n void Load(const Instr::Op& op) { this->value = op.offset; }\n};\n\nstruct SymbolOp : Op {\n Function* value;\n\n protected:\n friend struct Op;\n template \n friend struct I;\n bool Load(const Instr::Op& op) {\n this->value = op.symbol;\n return true;\n }\n};\n\nstruct LabelOp : Op {\n hir::Label* value;\n\n protected:\n friend struct Op;\n template \n friend struct I;\n void Load(const Instr::Op& op) { this->value = op.label; }\n};\n\ntemplate \nstruct ValueOp : Op, KEY_TYPE> {\n typedef REG_TYPE reg_type;\n const Value* value;\n bool is_constant;\n virtual bool ConstantFitsIn32Reg() const { return true; }\n const REG_TYPE& reg() const {\n assert_true(!is_constant);\n return reg_;\n }\n operator const REG_TYPE&() const { return reg(); }\n bool IsEqual(const T& b) const {\n if (is_constant && b.is_constant) {\n return reinterpret_cast(this)->constant() == b.constant();\n } else if (!is_constant && !b.is_constant) {\n return reg_.getIdx() == b.reg_.getIdx();\n } else {\n return false;\n }\n }\n bool IsEqual(const Xbyak::Reg& b) const {\n if (is_constant) {\n return false;\n } else if (!is_constant) {\n return reg_.getIdx() == b.getIdx();\n } else {\n return false;\n }\n }\n bool operator==(const T& b) const { return IsEqual(b); }\n bool operator!=(const T& b) const { return !IsEqual(b); }\n bool operator==(const Xbyak::Reg& b) const { return IsEqual(b); }\n bool operator!=(const Xbyak::Reg& b) const { return !IsEqual(b); }\n void Load(const Instr::Op& op) {\n value = op.value;\n is_constant = value->IsConstant();\n if (!is_constant) {\n X64Emitter::SetupReg(value, reg_);\n }\n }\n\n protected:\n REG_TYPE reg_;\n};\n\nstruct I8Op : ValueOp {\n typedef ValueOp BASE;\n const int8_t constant() const {\n assert_true(BASE::is_constant);\n return BASE::value->constant.i8;\n }\n};\nstruct I16Op : ValueOp {\n typedef ValueOp BASE;\n const int16_t constant() const {\n assert_true(BASE::is_constant);\n return BASE::value->constant.i16;\n }\n};\nstruct I32Op : ValueOp {\n typedef ValueOp BASE;\n const int32_t constant() const {\n assert_true(BASE::is_constant);\n return BASE::value->constant.i32;\n }\n};\nstruct I64Op : ValueOp {\n typedef ValueOp BASE;\n const int64_t constant() const {\n assert_true(BASE::is_constant);\n return BASE::value->constant.i64;\n }\n bool ConstantFitsIn32Reg() const override {\n int64_t v = BASE::value->constant.i64;\n if ((v & ~0x7FFFFFFF) == 0) {\n // Fits under 31 bits, so just load using normal mov.\n return true;\n } else if ((v & ~0x7FFFFFFF) == ~0x7FFFFFFF) {\n // Negative number that fits in 32bits.\n return true;\n }\n return false;\n }\n};\nstruct F32Op : ValueOp {\n typedef ValueOp BASE;\n const float constant() const {\n assert_true(BASE::is_constant);\n return BASE::value->constant.f32;\n }\n};\nstruct F64Op : ValueOp {\n typedef ValueOp BASE;\n const double constant() const {\n assert_true(BASE::is_constant);\n return BASE::value->constant.f64;\n }\n};\nstruct V128Op : ValueOp {\n typedef ValueOp BASE;\n const vec128_t& constant() const {\n assert_true(BASE::is_constant);\n return BASE::value->constant.v128;\n }\n};\n\ntemplate \nstruct DestField;\ntemplate \nstruct DestField {\n DEST dest;\n\n protected:\n bool LoadDest(const Instr* i) {\n Instr::Op op;\n op.value = i->dest;\n dest.Load(op);\n return true;\n }\n};\ntemplate <>\nstruct DestField {\n protected:\n bool LoadDest(const Instr* i) { return true; }\n};\n\ntemplate \nstruct I;\ntemplate \nstruct I : DestField {\n typedef DestField BASE;\n static const hir::Opcode opcode = OPCODE;\n static const uint32_t key =\n InstrKey::Construct::value;\n static const KeyType dest_type = DEST::key_type;\n const Instr* instr;\n\n protected:\n template \n friend struct Sequence;\n bool Load(const Instr* i) {\n if (InstrKey(i).value == key && BASE::LoadDest(i)) {\n instr = i;\n return true;\n }\n return false;\n }\n};\ntemplate \nstruct I : DestField {\n typedef DestField BASE;\n static const hir::Opcode opcode = OPCODE;\n static const uint32_t key =\n InstrKey::Construct::value;\n static const KeyType dest_type = DEST::key_type;\n static const KeyType src1_type = SRC1::key_type;\n const Instr* instr;\n SRC1 src1;\n\n protected:\n template \n friend struct Sequence;\n bool Load(const Instr* i) {\n if (InstrKey(i).value == key && BASE::LoadDest(i)) {\n instr = i;\n src1.Load(i->src1);\n return true;\n }\n return false;\n }\n};\ntemplate \nstruct I : DestField {\n typedef DestField BASE;\n static const hir::Opcode opcode = OPCODE;\n static const uint32_t key =\n InstrKey::Construct::value;\n static const KeyType dest_type = DEST::key_type;\n static const KeyType src1_type = SRC1::key_type;\n static const KeyType src2_type = SRC2::key_type;\n const Instr* instr;\n SRC1 src1;\n SRC2 src2;\n\n protected:\n template \n friend struct Sequence;\n bool Load(const Instr* i) {\n if (InstrKey(i).value == key && BASE::LoadDest(i)) {\n instr = i;\n src1.Load(i->src1);\n src2.Load(i->src2);\n return true;\n }\n return false;\n }\n};\ntemplate \nstruct I : DestField {\n typedef DestField BASE;\n static const hir::Opcode opcode = OPCODE;\n static const uint32_t key =\n InstrKey::Construct::value;\n static const KeyType dest_type = DEST::key_type;\n static const KeyType src1_type = SRC1::key_type;\n static const KeyType src2_type = SRC2::key_type;\n static const KeyType src3_type = SRC3::key_type;\n const Instr* instr;\n SRC1 src1;\n SRC2 src2;\n SRC3 src3;\n\n protected:\n template \n friend struct Sequence;\n bool Load(const Instr* i) {\n if (InstrKey(i).value == key && BASE::LoadDest(i)) {\n instr = i;\n src1.Load(i->src1);\n src2.Load(i->src2);\n src3.Load(i->src3);\n return true;\n }\n return false;\n }\n};\n\ntemplate \nstatic const T GetTempReg(X64Emitter& e);\ntemplate <>\nconst Reg8 GetTempReg(X64Emitter& e) {\n return e.al;\n}\ntemplate <>\nconst Reg16 GetTempReg(X64Emitter& e) {\n return e.ax;\n}\ntemplate <>\nconst Reg32 GetTempReg(X64Emitter& e) {\n return e.eax;\n}\ntemplate <>\nconst Reg64 GetTempReg(X64Emitter& e) {\n return e.rax;\n}\n\ntemplate \nstruct Sequence {\n typedef T EmitArgType;\n\n static constexpr uint32_t head_key() { return T::key; }\n\n static bool Select(X64Emitter& e, const Instr* i) {\n T args;\n if (!args.Load(i)) {\n return false;\n }\n SEQ::Emit(e, args);\n return true;\n }\n\n template \n static void EmitUnaryOp(X64Emitter& e, const EmitArgType& i,\n const REG_FN& reg_fn) {\n if (i.src1.is_constant) {\n e.mov(i.dest, i.src1.constant());\n reg_fn(e, i.dest);\n } else {\n if (i.dest != i.src1) {\n e.mov(i.dest, i.src1);\n }\n reg_fn(e, i.dest);\n }\n }\n\n template \n static void EmitCommutativeBinaryOp(X64Emitter& e, const EmitArgType& i,\n const REG_REG_FN& reg_reg_fn,\n const REG_CONST_FN& reg_const_fn) {\n if (i.src1.is_constant) {\n if (i.src2.is_constant) {\n // Both constants.\n if (i.src1.ConstantFitsIn32Reg()) {\n e.mov(i.dest, i.src2.constant());\n reg_const_fn(e, i.dest, static_cast(i.src1.constant()));\n } else if (i.src2.ConstantFitsIn32Reg()) {\n e.mov(i.dest, i.src1.constant());\n reg_const_fn(e, i.dest, static_cast(i.src2.constant()));\n } else {\n e.mov(i.dest, i.src1.constant());\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2.constant());\n reg_reg_fn(e, i.dest, temp);\n }\n } else {\n // src1 constant.\n if (i.dest == i.src2) {\n if (i.src1.ConstantFitsIn32Reg()) {\n reg_const_fn(e, i.dest, static_cast(i.src1.constant()));\n } else {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src1.constant());\n reg_reg_fn(e, i.dest, temp);\n }\n } else {\n e.mov(i.dest, i.src1.constant());\n reg_reg_fn(e, i.dest, i.src2);\n }\n }\n } else if (i.src2.is_constant) {\n if (i.dest == i.src1) {\n if (i.src2.ConstantFitsIn32Reg()) {\n reg_const_fn(e, i.dest, static_cast(i.src2.constant()));\n } else {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2.constant());\n reg_reg_fn(e, i.dest, temp);\n }\n } else {\n e.mov(i.dest, i.src2.constant());\n reg_reg_fn(e, i.dest, i.src1);\n }\n } else {\n if (i.dest == i.src1) {\n reg_reg_fn(e, i.dest, i.src2);\n } else if (i.dest == i.src2) {\n reg_reg_fn(e, i.dest, i.src1);\n } else {\n e.mov(i.dest, i.src1);\n reg_reg_fn(e, i.dest, i.src2);\n }\n }\n }\n template \n static void EmitAssociativeBinaryOp(X64Emitter& e, const EmitArgType& i,\n const REG_REG_FN& reg_reg_fn,\n const REG_CONST_FN& reg_const_fn) {\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n if (i.dest == i.src2) {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2);\n e.mov(i.dest, i.src1.constant());\n reg_reg_fn(e, i.dest, temp);\n } else {\n e.mov(i.dest, i.src1.constant());\n reg_reg_fn(e, i.dest, i.src2);\n }\n } else if (i.src2.is_constant) {\n if (i.dest == i.src1) {\n if (i.src2.ConstantFitsIn32Reg()) {\n reg_const_fn(e, i.dest, static_cast(i.src2.constant()));\n } else {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2.constant());\n reg_reg_fn(e, i.dest, temp);\n }\n } else {\n e.mov(i.dest, i.src1);\n if (i.src2.ConstantFitsIn32Reg()) {\n reg_const_fn(e, i.dest, static_cast(i.src2.constant()));\n } else {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2.constant());\n reg_reg_fn(e, i.dest, temp);\n }\n }\n } else {\n if (i.dest == i.src1) {\n reg_reg_fn(e, i.dest, i.src2);\n } else if (i.dest == i.src2) {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2);\n e.mov(i.dest, i.src1);\n reg_reg_fn(e, i.dest, temp);\n } else {\n e.mov(i.dest, i.src1);\n reg_reg_fn(e, i.dest, i.src2);\n }\n }\n }\n\n template \n static void EmitCommutativeBinaryXmmOp(X64Emitter& e, const EmitArgType& i,\n const FN& fn) {\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n e.LoadConstantXmm(e.xmm0, i.src1.constant());\n fn(e, i.dest, e.xmm0, i.src2);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n e.LoadConstantXmm(e.xmm0, i.src2.constant());\n fn(e, i.dest, i.src1, e.xmm0);\n } else {\n fn(e, i.dest, i.src1, i.src2);\n }\n }\n\n template \n static void EmitAssociativeBinaryXmmOp(X64Emitter& e, const EmitArgType& i,\n const FN& fn) {\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n e.LoadConstantXmm(e.xmm0, i.src1.constant());\n fn(e, i.dest, e.xmm0, i.src2);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n e.LoadConstantXmm(e.xmm0, i.src2.constant());\n fn(e, i.dest, i.src1, e.xmm0);\n } else {\n fn(e, i.dest, i.src1, i.src2);\n }\n }\n\n template \n static void EmitCommutativeCompareOp(X64Emitter& e, const EmitArgType& i,\n const REG_REG_FN& reg_reg_fn,\n const REG_CONST_FN& reg_const_fn) {\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n if (i.src1.ConstantFitsIn32Reg()) {\n reg_const_fn(e, i.src2, static_cast(i.src1.constant()));\n } else {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src1.constant());\n reg_reg_fn(e, i.src2, temp);\n }\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n if (i.src2.ConstantFitsIn32Reg()) {\n reg_const_fn(e, i.src1, static_cast(i.src2.constant()));\n } else {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2.constant());\n reg_reg_fn(e, i.src1, temp);\n }\n } else {\n reg_reg_fn(e, i.src1, i.src2);\n }\n }\n template \n static void EmitAssociativeCompareOp(X64Emitter& e, const EmitArgType& i,\n const REG_REG_FN& reg_reg_fn,\n const REG_CONST_FN& reg_const_fn) {\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n if (i.src1.ConstantFitsIn32Reg()) {\n reg_const_fn(e, i.dest, i.src2, static_cast(i.src1.constant()),\n true);\n } else {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src1.constant());\n reg_reg_fn(e, i.dest, i.src2, temp, true);\n }\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n if (i.src2.ConstantFitsIn32Reg()) {\n reg_const_fn(e, i.dest, i.src1, static_cast(i.src2.constant()),\n false);\n } else {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2.constant());\n reg_reg_fn(e, i.dest, i.src1, temp, false);\n }\n } else {\n reg_reg_fn(e, i.dest, i.src1, i.src2, false);\n }\n }\n};\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_OP_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_seq_control.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2018 Xenia Developers. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_sequences.h\"\n\n#include \n#include \n\n#include \"xenia/cpu/backend/x64/x64_op.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nvolatile int anchor_control = 0;\n\n// ============================================================================\n// OPCODE_DEBUG_BREAK\n// ============================================================================\nstruct DEBUG_BREAK : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) { e.DebugBreak(); }\n};\nEMITTER_OPCODE_TABLE(OPCODE_DEBUG_BREAK, DEBUG_BREAK);\n\n// ============================================================================\n// OPCODE_DEBUG_BREAK_TRUE\n// ============================================================================\nstruct DEBUG_BREAK_TRUE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.DebugBreak();\n e.L(skip);\n }\n};\nstruct DEBUG_BREAK_TRUE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.DebugBreak();\n e.L(skip);\n }\n};\nstruct DEBUG_BREAK_TRUE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.DebugBreak();\n e.L(skip);\n }\n};\nstruct DEBUG_BREAK_TRUE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.DebugBreak();\n e.L(skip);\n }\n};\nstruct DEBUG_BREAK_TRUE_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.DebugBreak();\n e.L(skip);\n }\n};\nstruct DEBUG_BREAK_TRUE_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.DebugBreak();\n e.L(skip);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_DEBUG_BREAK_TRUE, DEBUG_BREAK_TRUE_I8,\n DEBUG_BREAK_TRUE_I16, DEBUG_BREAK_TRUE_I32,\n DEBUG_BREAK_TRUE_I64, DEBUG_BREAK_TRUE_F32,\n DEBUG_BREAK_TRUE_F64);\n\n// ============================================================================\n// OPCODE_TRAP\n// ============================================================================\nstruct TRAP : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.Trap(i.instr->flags);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_TRAP, TRAP);\n\n// ============================================================================\n// OPCODE_TRAP_TRUE\n// ============================================================================\nstruct TRAP_TRUE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Trap(i.instr->flags);\n e.L(skip);\n }\n};\nstruct TRAP_TRUE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Trap(i.instr->flags);\n e.L(skip);\n }\n};\nstruct TRAP_TRUE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Trap(i.instr->flags);\n e.L(skip);\n }\n};\nstruct TRAP_TRUE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Trap(i.instr->flags);\n e.L(skip);\n }\n};\nstruct TRAP_TRUE_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Trap(i.instr->flags);\n e.L(skip);\n }\n};\nstruct TRAP_TRUE_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Trap(i.instr->flags);\n e.L(skip);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_TRAP_TRUE, TRAP_TRUE_I8, TRAP_TRUE_I16,\n TRAP_TRUE_I32, TRAP_TRUE_I64, TRAP_TRUE_F32,\n TRAP_TRUE_F64);\n\n// ============================================================================\n// OPCODE_CALL\n// ============================================================================\nstruct CALL : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src1.value->is_guest());\n e.Call(i.instr, static_cast(i.src1.value));\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CALL, CALL);\n\n// ============================================================================\n// OPCODE_CALL_TRUE\n// ============================================================================\nstruct CALL_TRUE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->is_guest());\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Call(i.instr, static_cast(i.src2.value));\n e.L(skip);\n }\n};\nstruct CALL_TRUE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->is_guest());\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Call(i.instr, static_cast(i.src2.value));\n e.L(skip);\n }\n};\nstruct CALL_TRUE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->is_guest());\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Call(i.instr, static_cast(i.src2.value));\n e.L(skip);\n }\n};\nstruct CALL_TRUE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->is_guest());\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Call(i.instr, static_cast(i.src2.value));\n e.L(skip);\n }\n};\nstruct CALL_TRUE_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->is_guest());\n e.vptest(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Call(i.instr, static_cast(i.src2.value));\n e.L(skip);\n }\n};\nstruct CALL_TRUE_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->is_guest());\n e.vptest(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip);\n e.Call(i.instr, static_cast(i.src2.value));\n e.L(skip);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CALL_TRUE, CALL_TRUE_I8, CALL_TRUE_I16,\n CALL_TRUE_I32, CALL_TRUE_I64, CALL_TRUE_F32,\n CALL_TRUE_F64);\n\n// ============================================================================\n// OPCODE_CALL_INDIRECT\n// ============================================================================\nstruct CALL_INDIRECT\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.CallIndirect(i.instr, i.src1);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CALL_INDIRECT, CALL_INDIRECT);\n\n// ============================================================================\n// OPCODE_CALL_INDIRECT_TRUE\n// ============================================================================\nstruct CALL_INDIRECT_TRUE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip, CodeGenerator::T_NEAR);\n e.CallIndirect(i.instr, i.src2);\n e.L(skip);\n }\n};\nstruct CALL_INDIRECT_TRUE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip, CodeGenerator::T_NEAR);\n e.CallIndirect(i.instr, i.src2);\n e.L(skip);\n }\n};\nstruct CALL_INDIRECT_TRUE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip, CodeGenerator::T_NEAR);\n e.CallIndirect(i.instr, i.src2);\n e.L(skip);\n }\n};\nstruct CALL_INDIRECT_TRUE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip, CodeGenerator::T_NEAR);\n e.CallIndirect(i.instr, i.src2);\n e.L(skip);\n }\n};\nstruct CALL_INDIRECT_TRUE_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip, CodeGenerator::T_NEAR);\n e.CallIndirect(i.instr, i.src2);\n e.L(skip);\n }\n};\nstruct CALL_INDIRECT_TRUE_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n Xbyak::Label skip;\n e.jz(skip, CodeGenerator::T_NEAR);\n e.CallIndirect(i.instr, i.src2);\n e.L(skip);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CALL_INDIRECT_TRUE, CALL_INDIRECT_TRUE_I8,\n CALL_INDIRECT_TRUE_I16, CALL_INDIRECT_TRUE_I32,\n CALL_INDIRECT_TRUE_I64, CALL_INDIRECT_TRUE_F32,\n CALL_INDIRECT_TRUE_F64);\n\n// ============================================================================\n// OPCODE_CALL_EXTERN\n// ============================================================================\nstruct CALL_EXTERN\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.CallExtern(i.instr, i.src1.value);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CALL_EXTERN, CALL_EXTERN);\n\n// ============================================================================\n// OPCODE_RETURN\n// ============================================================================\nstruct RETURN : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // If this is the last instruction in the last block, just let us\n // fall through.\n if (i.instr->next || i.instr->block->next) {\n e.jmp(e.epilog_label(), CodeGenerator::T_NEAR);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_RETURN, RETURN);\n\n// ============================================================================\n// OPCODE_RETURN_TRUE\n// ============================================================================\nstruct RETURN_TRUE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jnz(e.epilog_label(), CodeGenerator::T_NEAR);\n }\n};\nstruct RETURN_TRUE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jnz(e.epilog_label(), CodeGenerator::T_NEAR);\n }\n};\nstruct RETURN_TRUE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jnz(e.epilog_label(), CodeGenerator::T_NEAR);\n }\n};\nstruct RETURN_TRUE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jnz(e.epilog_label(), CodeGenerator::T_NEAR);\n }\n};\nstruct RETURN_TRUE_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.jnz(e.epilog_label(), CodeGenerator::T_NEAR);\n }\n};\nstruct RETURN_TRUE_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.jnz(e.epilog_label(), CodeGenerator::T_NEAR);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_RETURN_TRUE, RETURN_TRUE_I8, RETURN_TRUE_I16,\n RETURN_TRUE_I32, RETURN_TRUE_I64, RETURN_TRUE_F32,\n RETURN_TRUE_F64);\n\n// ============================================================================\n// OPCODE_SET_RETURN_ADDRESS\n// ============================================================================\nstruct SET_RETURN_ADDRESS\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.SetReturnAddress(i.src1.constant());\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SET_RETURN_ADDRESS, SET_RETURN_ADDRESS);\n\n// ============================================================================\n// OPCODE_BRANCH\n// ============================================================================\nstruct BRANCH : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.jmp(i.src1.value->name, e.T_NEAR);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_BRANCH, BRANCH);\n\n// ============================================================================\n// OPCODE_BRANCH_TRUE\n// ============================================================================\nstruct BRANCH_TRUE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jnz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_TRUE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jnz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_TRUE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jnz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_TRUE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jnz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_TRUE_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.jnz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_TRUE_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.jnz(i.src2.value->name, e.T_NEAR);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_BRANCH_TRUE, BRANCH_TRUE_I8, BRANCH_TRUE_I16,\n BRANCH_TRUE_I32, BRANCH_TRUE_I64, BRANCH_TRUE_F32,\n BRANCH_TRUE_F64);\n\n// ============================================================================\n// OPCODE_BRANCH_FALSE\n// ============================================================================\nstruct BRANCH_FALSE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_FALSE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_FALSE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_FALSE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.jz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_FALSE_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.jz(i.src2.value->name, e.T_NEAR);\n }\n};\nstruct BRANCH_FALSE_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.jz(i.src2.value->name, e.T_NEAR);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_BRANCH_FALSE, BRANCH_FALSE_I8, BRANCH_FALSE_I16,\n BRANCH_FALSE_I32, BRANCH_FALSE_I64, BRANCH_FALSE_F32,\n BRANCH_FALSE_F64);\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe" }, { "path": "src/xenia/cpu/backend/x64/x64_seq_memory.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2018 Xenia Developers. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_sequences.h\"\n\n#include \n#include \n\n#include \"xenia/base/memory.h\"\n#include \"xenia/cpu/backend/x64/x64_op.h\"\n#include \"xenia/cpu/backend/x64/x64_tracers.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nvolatile int anchor_memory = 0;\n\n// Note: all types are always aligned in the context.\nRegExp ComputeContextAddress(X64Emitter& e, const OffsetOp& offset) {\n return e.GetContextReg() + offset.value;\n}\n\ntemplate \nRegExp ComputeMemoryAddressOffset(X64Emitter& e, const T& guest,\n const T& offset) {\n assert_true(offset.is_constant);\n int32_t offset_const = static_cast(offset.constant());\n\n if (guest.is_constant) {\n uint32_t address = static_cast(guest.constant());\n address += offset_const;\n if (address < 0x80000000) {\n return e.GetMembaseReg() + address;\n } else {\n if (address >= 0xE0000000 &&\n xe::memory::allocation_granularity() > 0x1000) {\n e.mov(e.eax, address + 0x1000);\n } else {\n e.mov(e.eax, address);\n }\n return e.GetMembaseReg() + e.rax;\n }\n } else {\n if (xe::memory::allocation_granularity() > 0x1000) {\n // Emulate the 4 KB physical address offset in 0xE0000000+ when can't do\n // it via memory mapping.\n e.cmp(guest.reg().cvt32(), 0xE0000000 - offset_const);\n e.setae(e.al);\n e.movzx(e.eax, e.al);\n e.shl(e.eax, 12);\n e.add(e.eax, guest.reg().cvt32());\n } else {\n // Clear the top 32 bits, as they are likely garbage.\n // TODO(benvanik): find a way to avoid doing this.\n e.mov(e.eax, guest.reg().cvt32());\n }\n return e.GetMembaseReg() + e.rax + offset_const;\n }\n}\n\n// Note: most *should* be aligned, but needs to be checked!\ntemplate \nRegExp ComputeMemoryAddress(X64Emitter& e, const T& guest) {\n if (guest.is_constant) {\n // TODO(benvanik): figure out how to do this without a temp.\n // Since the constant is often 0x8... if we tried to use that as a\n // displacement it would be sign extended and mess things up.\n uint32_t address = static_cast(guest.constant());\n if (address < 0x80000000) {\n return e.GetMembaseReg() + address;\n } else {\n if (address >= 0xE0000000 &&\n xe::memory::allocation_granularity() > 0x1000) {\n e.mov(e.eax, address + 0x1000);\n } else {\n e.mov(e.eax, address);\n }\n return e.GetMembaseReg() + e.rax;\n }\n } else {\n if (xe::memory::allocation_granularity() > 0x1000) {\n // Emulate the 4 KB physical address offset in 0xE0000000+ when can't do\n // it via memory mapping.\n e.cmp(guest.reg().cvt32(), 0xE0000000);\n e.setae(e.al);\n e.movzx(e.eax, e.al);\n e.shl(e.eax, 12);\n e.add(e.eax, guest.reg().cvt32());\n } else {\n // Clear the top 32 bits, as they are likely garbage.\n // TODO(benvanik): find a way to avoid doing this.\n e.mov(e.eax, guest.reg().cvt32());\n }\n return e.GetMembaseReg() + e.rax;\n }\n}\n\n// ============================================================================\n// OPCODE_ATOMIC_EXCHANGE\n// ============================================================================\n// Note that the address we use here is a real, host address!\n// This is weird, and should be fixed.\ntemplate \nvoid EmitAtomicExchangeXX(X64Emitter& e, const ARGS& i) {\n if (i.dest == i.src1) {\n e.mov(e.rax, i.src1);\n if (i.dest != i.src2) {\n if (i.src2.is_constant) {\n e.mov(i.dest, i.src2.constant());\n } else {\n e.mov(i.dest, i.src2);\n }\n }\n e.lock();\n e.xchg(e.dword[e.rax], i.dest);\n } else {\n if (i.dest != i.src2) {\n if (i.src2.is_constant) {\n e.mov(i.dest, i.src2.constant());\n } else {\n e.mov(i.dest, i.src2);\n }\n }\n e.lock();\n e.xchg(e.dword[i.src1.reg()], i.dest);\n }\n}\nstruct ATOMIC_EXCHANGE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAtomicExchangeXX(e, i);\n }\n};\nstruct ATOMIC_EXCHANGE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAtomicExchangeXX(e, i);\n }\n};\nstruct ATOMIC_EXCHANGE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAtomicExchangeXX(e, i);\n }\n};\nstruct ATOMIC_EXCHANGE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAtomicExchangeXX(e, i);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ATOMIC_EXCHANGE, ATOMIC_EXCHANGE_I8,\n ATOMIC_EXCHANGE_I16, ATOMIC_EXCHANGE_I32,\n ATOMIC_EXCHANGE_I64);\n\n// ============================================================================\n// OPCODE_ATOMIC_COMPARE_EXCHANGE\n// ============================================================================\nstruct ATOMIC_COMPARE_EXCHANGE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(e.eax, i.src2);\n if (xe::memory::allocation_granularity() > 0x1000) {\n // Emulate the 4 KB physical address offset in 0xE0000000+ when can't do\n // it via memory mapping.\n e.cmp(i.src1.reg().cvt32(), 0xE0000000);\n e.setae(e.cl);\n e.movzx(e.ecx, e.cl);\n e.shl(e.ecx, 12);\n e.add(e.ecx, i.src1.reg().cvt32());\n } else {\n e.mov(e.ecx, i.src1.reg().cvt32());\n }\n e.lock();\n e.cmpxchg(e.dword[e.GetMembaseReg() + e.rcx], i.src3);\n e.sete(i.dest);\n }\n};\nstruct ATOMIC_COMPARE_EXCHANGE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(e.rax, i.src2);\n if (xe::memory::allocation_granularity() > 0x1000) {\n // Emulate the 4 KB physical address offset in 0xE0000000+ when can't do\n // it via memory mapping.\n e.cmp(i.src1.reg().cvt32(), 0xE0000000);\n e.setae(e.cl);\n e.movzx(e.ecx, e.cl);\n e.shl(e.ecx, 12);\n e.add(e.ecx, i.src1.reg().cvt32());\n } else {\n e.mov(e.ecx, i.src1.reg().cvt32());\n }\n e.lock();\n e.cmpxchg(e.qword[e.GetMembaseReg() + e.rcx], i.src3);\n e.sete(i.dest);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ATOMIC_COMPARE_EXCHANGE,\n ATOMIC_COMPARE_EXCHANGE_I32, ATOMIC_COMPARE_EXCHANGE_I64);\n\n// ============================================================================\n// OPCODE_LOAD_LOCAL\n// ============================================================================\n// Note: all types are always aligned on the stack.\nstruct LOAD_LOCAL_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, e.byte[e.rsp + i.src1.constant()]);\n // e.TraceLoadI8(DATA_LOCAL, i.src1.constant, i.dest);\n }\n};\nstruct LOAD_LOCAL_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, e.word[e.rsp + i.src1.constant()]);\n // e.TraceLoadI16(DATA_LOCAL, i.src1.constant, i.dest);\n }\n};\nstruct LOAD_LOCAL_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, e.dword[e.rsp + i.src1.constant()]);\n // e.TraceLoadI32(DATA_LOCAL, i.src1.constant, i.dest);\n }\n};\nstruct LOAD_LOCAL_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, e.qword[e.rsp + i.src1.constant()]);\n // e.TraceLoadI64(DATA_LOCAL, i.src1.constant, i.dest);\n }\n};\nstruct LOAD_LOCAL_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovss(i.dest, e.dword[e.rsp + i.src1.constant()]);\n // e.TraceLoadF32(DATA_LOCAL, i.src1.constant, i.dest);\n }\n};\nstruct LOAD_LOCAL_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovsd(i.dest, e.qword[e.rsp + i.src1.constant()]);\n // e.TraceLoadF64(DATA_LOCAL, i.src1.constant, i.dest);\n }\n};\nstruct LOAD_LOCAL_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovaps(i.dest, e.ptr[e.rsp + i.src1.constant()]);\n // e.TraceLoadV128(DATA_LOCAL, i.src1.constant, i.dest);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOAD_LOCAL, LOAD_LOCAL_I8, LOAD_LOCAL_I16,\n LOAD_LOCAL_I32, LOAD_LOCAL_I64, LOAD_LOCAL_F32,\n LOAD_LOCAL_F64, LOAD_LOCAL_V128);\n\n// ============================================================================\n// OPCODE_STORE_LOCAL\n// ============================================================================\n// Note: all types are always aligned on the stack.\nstruct STORE_LOCAL_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // e.TraceStoreI8(DATA_LOCAL, i.src1.constant, i.src2);\n e.mov(e.byte[e.rsp + i.src1.constant()], i.src2);\n }\n};\nstruct STORE_LOCAL_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // e.TraceStoreI16(DATA_LOCAL, i.src1.constant, i.src2);\n e.mov(e.word[e.rsp + i.src1.constant()], i.src2);\n }\n};\nstruct STORE_LOCAL_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // e.TraceStoreI32(DATA_LOCAL, i.src1.constant, i.src2);\n e.mov(e.dword[e.rsp + i.src1.constant()], i.src2);\n }\n};\nstruct STORE_LOCAL_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // e.TraceStoreI64(DATA_LOCAL, i.src1.constant, i.src2);\n e.mov(e.qword[e.rsp + i.src1.constant()], i.src2);\n }\n};\nstruct STORE_LOCAL_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // e.TraceStoreF32(DATA_LOCAL, i.src1.constant, i.src2);\n e.vmovss(e.dword[e.rsp + i.src1.constant()], i.src2);\n }\n};\nstruct STORE_LOCAL_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // e.TraceStoreF64(DATA_LOCAL, i.src1.constant, i.src2);\n e.vmovsd(e.qword[e.rsp + i.src1.constant()], i.src2);\n }\n};\nstruct STORE_LOCAL_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // e.TraceStoreV128(DATA_LOCAL, i.src1.constant, i.src2);\n e.vmovaps(e.ptr[e.rsp + i.src1.constant()], i.src2);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_STORE_LOCAL, STORE_LOCAL_I8, STORE_LOCAL_I16,\n STORE_LOCAL_I32, STORE_LOCAL_I64, STORE_LOCAL_F32,\n STORE_LOCAL_F64, STORE_LOCAL_V128);\n\n// ============================================================================\n// OPCODE_LOAD_CONTEXT\n// ============================================================================\nstruct LOAD_CONTEXT_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n e.mov(i.dest, e.byte[addr]);\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.mov(e.GetNativeParam(1), e.byte[addr]);\n e.CallNative(reinterpret_cast(TraceContextLoadI8));\n }\n }\n};\nstruct LOAD_CONTEXT_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n e.mov(i.dest, e.word[addr]);\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1), e.word[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextLoadI16));\n }\n }\n};\nstruct LOAD_CONTEXT_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n e.mov(i.dest, e.dword[addr]);\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1), e.dword[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextLoadI32));\n }\n }\n};\nstruct LOAD_CONTEXT_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n e.mov(i.dest, e.qword[addr]);\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1), e.qword[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextLoadI64));\n }\n }\n};\nstruct LOAD_CONTEXT_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n e.vmovss(i.dest, e.dword[addr]);\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.dword[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextLoadF32));\n }\n }\n};\nstruct LOAD_CONTEXT_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n e.vmovsd(i.dest, e.qword[addr]);\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.qword[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextLoadF64));\n }\n }\n};\nstruct LOAD_CONTEXT_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n e.vmovaps(i.dest, e.ptr[addr]);\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.ptr[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextLoadV128));\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOAD_CONTEXT, LOAD_CONTEXT_I8, LOAD_CONTEXT_I16,\n LOAD_CONTEXT_I32, LOAD_CONTEXT_I64, LOAD_CONTEXT_F32,\n LOAD_CONTEXT_F64, LOAD_CONTEXT_V128);\n\n// ============================================================================\n// OPCODE_STORE_CONTEXT\n// ============================================================================\n// Note: all types are always aligned on the stack.\nstruct STORE_CONTEXT_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n if (i.src2.is_constant) {\n e.mov(e.byte[addr], i.src2.constant());\n } else {\n e.mov(e.byte[addr], i.src2);\n }\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1), e.byte[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextStoreI8));\n }\n }\n};\nstruct STORE_CONTEXT_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n if (i.src2.is_constant) {\n e.mov(e.word[addr], i.src2.constant());\n } else {\n e.mov(e.word[addr], i.src2);\n }\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1), e.word[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextStoreI16));\n }\n }\n};\nstruct STORE_CONTEXT_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n if (i.src2.is_constant) {\n e.mov(e.dword[addr], i.src2.constant());\n } else {\n e.mov(e.dword[addr], i.src2);\n }\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1), e.dword[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextStoreI32));\n }\n }\n};\nstruct STORE_CONTEXT_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n if (i.src2.is_constant) {\n e.MovMem64(addr, i.src2.constant());\n } else {\n e.mov(e.qword[addr], i.src2);\n }\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1), e.qword[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextStoreI64));\n }\n }\n};\nstruct STORE_CONTEXT_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n if (i.src2.is_constant) {\n e.mov(e.dword[addr], i.src2.value->constant.i32);\n } else {\n e.vmovss(e.dword[addr], i.src2);\n }\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.dword[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextStoreF32));\n }\n }\n};\nstruct STORE_CONTEXT_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n if (i.src2.is_constant) {\n e.MovMem64(addr, i.src2.value->constant.i64);\n } else {\n e.vmovsd(e.qword[addr], i.src2);\n }\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.qword[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextStoreF64));\n }\n }\n};\nstruct STORE_CONTEXT_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeContextAddress(e, i.src1);\n if (i.src2.is_constant) {\n e.LoadConstantXmm(e.xmm0, i.src2.constant());\n e.vmovaps(e.ptr[addr], e.xmm0);\n } else {\n e.vmovaps(e.ptr[addr], i.src2);\n }\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.ptr[addr]);\n e.mov(e.GetNativeParam(0), i.src1.value);\n e.CallNative(reinterpret_cast(TraceContextStoreV128));\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_STORE_CONTEXT, STORE_CONTEXT_I8, STORE_CONTEXT_I16,\n STORE_CONTEXT_I32, STORE_CONTEXT_I64, STORE_CONTEXT_F32,\n STORE_CONTEXT_F64, STORE_CONTEXT_V128);\n\n// ============================================================================\n// OPCODE_LOAD_MMIO\n// ============================================================================\n// Note: all types are always aligned in the context.\nstruct LOAD_MMIO_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // uint64_t (context, addr)\n auto mmio_range = reinterpret_cast(i.src1.value);\n auto read_address = uint32_t(i.src2.value);\n e.mov(e.GetNativeParam(0), uint64_t(mmio_range->callback_context));\n e.mov(e.GetNativeParam(1).cvt32(), read_address);\n e.CallNativeSafe(reinterpret_cast(mmio_range->read));\n e.bswap(e.eax);\n e.mov(i.dest, e.eax);\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(0), i.dest);\n e.mov(e.edx, read_address);\n e.CallNative(reinterpret_cast(TraceContextLoadI32));\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOAD_MMIO, LOAD_MMIO_I32);\n\n// ============================================================================\n// OPCODE_STORE_MMIO\n// ============================================================================\n// Note: all types are always aligned on the stack.\nstruct STORE_MMIO_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // void (context, addr, value)\n auto mmio_range = reinterpret_cast(i.src1.value);\n auto write_address = uint32_t(i.src2.value);\n e.mov(e.GetNativeParam(0), uint64_t(mmio_range->callback_context));\n e.mov(e.GetNativeParam(1).cvt32(), write_address);\n if (i.src3.is_constant) {\n e.mov(e.GetNativeParam(2).cvt32(), xe::byte_swap(i.src3.constant()));\n } else {\n e.mov(e.GetNativeParam(2).cvt32(), i.src3);\n e.bswap(e.GetNativeParam(2).cvt32());\n }\n e.CallNativeSafe(reinterpret_cast(mmio_range->write));\n if (IsTracingData()) {\n if (i.src3.is_constant) {\n e.mov(e.GetNativeParam(0).cvt32(), i.src3.constant());\n } else {\n e.mov(e.GetNativeParam(0).cvt32(), i.src3);\n }\n e.mov(e.edx, write_address);\n e.CallNative(reinterpret_cast(TraceContextStoreI32));\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_STORE_MMIO, STORE_MMIO_I32);\n\n// ============================================================================\n// OPCODE_LOAD_OFFSET\n// ============================================================================\nstruct LOAD_OFFSET_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddressOffset(e, i.src1, i.src2);\n e.mov(i.dest, e.byte[addr]);\n }\n};\n\nstruct LOAD_OFFSET_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddressOffset(e, i.src1, i.src2);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(i.dest, e.word[addr]);\n } else {\n e.mov(i.dest, e.word[addr]);\n e.ror(i.dest, 8);\n }\n } else {\n e.mov(i.dest, e.word[addr]);\n }\n }\n};\n\nstruct LOAD_OFFSET_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddressOffset(e, i.src1, i.src2);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(i.dest, e.dword[addr]);\n } else {\n e.mov(i.dest, e.dword[addr]);\n e.bswap(i.dest);\n }\n } else {\n e.mov(i.dest, e.dword[addr]);\n }\n }\n};\n\nstruct LOAD_OFFSET_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddressOffset(e, i.src1, i.src2);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(i.dest, e.qword[addr]);\n } else {\n e.mov(i.dest, e.qword[addr]);\n e.bswap(i.dest);\n }\n } else {\n e.mov(i.dest, e.qword[addr]);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOAD_OFFSET, LOAD_OFFSET_I8, LOAD_OFFSET_I16,\n LOAD_OFFSET_I32, LOAD_OFFSET_I64);\n\n// ============================================================================\n// OPCODE_STORE_OFFSET\n// ============================================================================\nstruct STORE_OFFSET_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddressOffset(e, i.src1, i.src2);\n if (i.src3.is_constant) {\n e.mov(e.byte[addr], i.src3.constant());\n } else {\n e.mov(e.byte[addr], i.src3);\n }\n }\n};\n\nstruct STORE_OFFSET_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddressOffset(e, i.src1, i.src2);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src3.is_constant);\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(e.word[addr], i.src3);\n } else {\n assert_always(\"not implemented\");\n }\n } else {\n if (i.src3.is_constant) {\n e.mov(e.word[addr], i.src3.constant());\n } else {\n e.mov(e.word[addr], i.src3);\n }\n }\n }\n};\n\nstruct STORE_OFFSET_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddressOffset(e, i.src1, i.src2);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src3.is_constant);\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(e.dword[addr], i.src3);\n } else {\n assert_always(\"not implemented\");\n }\n } else {\n if (i.src3.is_constant) {\n e.mov(e.dword[addr], i.src3.constant());\n } else {\n e.mov(e.dword[addr], i.src3);\n }\n }\n }\n};\n\nstruct STORE_OFFSET_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddressOffset(e, i.src1, i.src2);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src3.is_constant);\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(e.qword[addr], i.src3);\n } else {\n assert_always(\"not implemented\");\n }\n } else {\n if (i.src3.is_constant) {\n e.MovMem64(addr, i.src3.constant());\n } else {\n e.mov(e.qword[addr], i.src3);\n }\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_STORE_OFFSET, STORE_OFFSET_I8, STORE_OFFSET_I16,\n STORE_OFFSET_I32, STORE_OFFSET_I64);\n\n// ============================================================================\n// OPCODE_LOAD\n// ============================================================================\nstruct LOAD_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n e.mov(i.dest, e.byte[addr]);\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1).cvt8(), i.dest);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryLoadI8));\n }\n }\n};\nstruct LOAD_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(i.dest, e.word[addr]);\n } else {\n e.mov(i.dest, e.word[addr]);\n e.ror(i.dest, 8);\n }\n } else {\n e.mov(i.dest, e.word[addr]);\n }\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1).cvt16(), i.dest);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryLoadI16));\n }\n }\n};\nstruct LOAD_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(i.dest, e.dword[addr]);\n } else {\n e.mov(i.dest, e.dword[addr]);\n e.bswap(i.dest);\n }\n } else {\n e.mov(i.dest, e.dword[addr]);\n }\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1).cvt32(), i.dest);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryLoadI32));\n }\n }\n};\nstruct LOAD_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(i.dest, e.qword[addr]);\n } else {\n e.mov(i.dest, e.qword[addr]);\n e.bswap(i.dest);\n }\n } else {\n e.mov(i.dest, e.qword[addr]);\n }\n if (IsTracingData()) {\n e.mov(e.GetNativeParam(1), i.dest);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryLoadI64));\n }\n }\n};\nstruct LOAD_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n e.vmovss(i.dest, e.dword[addr]);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_always(\"not implemented yet\");\n }\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.dword[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryLoadF32));\n }\n }\n};\nstruct LOAD_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n e.vmovsd(i.dest, e.qword[addr]);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_always(\"not implemented yet\");\n }\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.qword[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryLoadF64));\n }\n }\n};\nstruct LOAD_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n // TODO(benvanik): we should try to stick to movaps if possible.\n e.vmovups(i.dest, e.ptr[addr]);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n // TODO(benvanik): find a way to do this without the memory load.\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMByteSwapMask));\n }\n if (IsTracingData()) {\n e.lea(e.GetNativeParam(1), e.ptr[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryLoadV128));\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOAD, LOAD_I8, LOAD_I16, LOAD_I32, LOAD_I64,\n LOAD_F32, LOAD_F64, LOAD_V128);\n\n// ============================================================================\n// OPCODE_STORE\n// ============================================================================\n// Note: most *should* be aligned, but needs to be checked!\nstruct STORE_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.src2.is_constant) {\n e.mov(e.byte[addr], i.src2.constant());\n } else {\n e.mov(e.byte[addr], i.src2);\n }\n if (IsTracingData()) {\n addr = ComputeMemoryAddress(e, i.src1);\n e.mov(e.GetNativeParam(1).cvt8(), e.byte[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryStoreI8));\n }\n }\n};\nstruct STORE_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src2.is_constant);\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(e.word[addr], i.src2);\n } else {\n assert_always(\"not implemented\");\n }\n } else {\n if (i.src2.is_constant) {\n e.mov(e.word[addr], i.src2.constant());\n } else {\n e.mov(e.word[addr], i.src2);\n }\n }\n if (IsTracingData()) {\n addr = ComputeMemoryAddress(e, i.src1);\n e.mov(e.GetNativeParam(1).cvt16(), e.word[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryStoreI16));\n }\n }\n};\nstruct STORE_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src2.is_constant);\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(e.dword[addr], i.src2);\n } else {\n assert_always(\"not implemented\");\n }\n } else {\n if (i.src2.is_constant) {\n e.mov(e.dword[addr], i.src2.constant());\n } else {\n e.mov(e.dword[addr], i.src2);\n }\n }\n if (IsTracingData()) {\n addr = ComputeMemoryAddress(e, i.src1);\n e.mov(e.GetNativeParam(1).cvt32(), e.dword[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryStoreI32));\n }\n }\n};\nstruct STORE_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src2.is_constant);\n if (e.IsFeatureEnabled(kX64EmitMovbe)) {\n e.movbe(e.qword[addr], i.src2);\n } else {\n assert_always(\"not implemented\");\n }\n } else {\n if (i.src2.is_constant) {\n e.MovMem64(addr, i.src2.constant());\n } else {\n e.mov(e.qword[addr], i.src2);\n }\n }\n if (IsTracingData()) {\n addr = ComputeMemoryAddress(e, i.src1);\n e.mov(e.GetNativeParam(1), e.qword[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryStoreI64));\n }\n }\n};\nstruct STORE_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src2.is_constant);\n assert_always(\"not yet implemented\");\n } else {\n if (i.src2.is_constant) {\n e.mov(e.dword[addr], i.src2.value->constant.i32);\n } else {\n e.vmovss(e.dword[addr], i.src2);\n }\n }\n if (IsTracingData()) {\n addr = ComputeMemoryAddress(e, i.src1);\n e.lea(e.GetNativeParam(1), e.ptr[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryStoreF32));\n }\n }\n};\nstruct STORE_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src2.is_constant);\n assert_always(\"not yet implemented\");\n } else {\n if (i.src2.is_constant) {\n e.MovMem64(addr, i.src2.value->constant.i64);\n } else {\n e.vmovsd(e.qword[addr], i.src2);\n }\n }\n if (IsTracingData()) {\n addr = ComputeMemoryAddress(e, i.src1);\n e.lea(e.GetNativeParam(1), e.ptr[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryStoreF64));\n }\n }\n};\nstruct STORE_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto addr = ComputeMemoryAddress(e, i.src1);\n if (i.instr->flags & LoadStoreFlags::LOAD_STORE_BYTE_SWAP) {\n assert_false(i.src2.is_constant);\n e.vpshufb(e.xmm0, i.src2, e.GetXmmConstPtr(XMMByteSwapMask));\n e.vmovaps(e.ptr[addr], e.xmm0);\n } else {\n if (i.src2.is_constant) {\n e.LoadConstantXmm(e.xmm0, i.src2.constant());\n e.vmovaps(e.ptr[addr], e.xmm0);\n } else {\n e.vmovaps(e.ptr[addr], i.src2);\n }\n }\n if (IsTracingData()) {\n addr = ComputeMemoryAddress(e, i.src1);\n e.lea(e.GetNativeParam(1), e.ptr[addr]);\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemoryStoreV128));\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_STORE, STORE_I8, STORE_I16, STORE_I32, STORE_I64,\n STORE_F32, STORE_F64, STORE_V128);\n\n// ============================================================================\n// OPCODE_CACHE_CONTROL\n// ============================================================================\nstruct CACHE_CONTROL\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n bool is_clflush = false, is_prefetch = false;\n switch (CacheControlType(i.instr->flags)) {\n case CacheControlType::CACHE_CONTROL_TYPE_DATA_TOUCH:\n case CacheControlType::CACHE_CONTROL_TYPE_DATA_TOUCH_FOR_STORE:\n is_prefetch = true;\n break;\n case CacheControlType::CACHE_CONTROL_TYPE_DATA_STORE:\n case CacheControlType::CACHE_CONTROL_TYPE_DATA_STORE_AND_FLUSH:\n is_clflush = true;\n break;\n default:\n assert_unhandled_case(CacheControlType(i.instr->flags));\n return;\n }\n size_t cache_line_size = i.src2.value;\n\n RegExp addr;\n uint32_t address_constant;\n if (i.src1.is_constant) {\n // TODO(benvanik): figure out how to do this without a temp.\n // Since the constant is often 0x8... if we tried to use that as a\n // displacement it would be sign extended and mess things up.\n address_constant = static_cast(i.src1.constant());\n if (address_constant < 0x80000000) {\n addr = e.GetMembaseReg() + address_constant;\n } else {\n if (address_constant >= 0xE0000000 &&\n xe::memory::allocation_granularity() > 0x1000) {\n e.mov(e.eax, address_constant + 0x1000);\n } else {\n e.mov(e.eax, address_constant);\n }\n addr = e.GetMembaseReg() + e.rax;\n }\n } else {\n if (xe::memory::allocation_granularity() > 0x1000) {\n // Emulate the 4 KB physical address offset in 0xE0000000+ when can't do\n // it via memory mapping.\n e.cmp(i.src1.reg().cvt32(), 0xE0000000);\n e.setae(e.al);\n e.movzx(e.eax, e.al);\n e.shl(e.eax, 12);\n e.add(e.eax, i.src1.reg().cvt32());\n } else {\n // Clear the top 32 bits, as they are likely garbage.\n // TODO(benvanik): find a way to avoid doing this.\n e.mov(e.eax, i.src1.reg().cvt32());\n }\n addr = e.GetMembaseReg() + e.rax;\n }\n if (is_clflush) {\n e.clflush(e.ptr[addr]);\n }\n if (is_prefetch) {\n e.prefetcht0(e.ptr[addr]);\n }\n\n if (cache_line_size >= 128) {\n // Prefetch the other 64 bytes of the 128-byte cache line.\n if (i.src1.is_constant && address_constant < 0x80000000) {\n addr = e.GetMembaseReg() + (address_constant ^ 64);\n } else {\n e.xor_(e.eax, 64);\n }\n if (is_clflush) {\n e.clflush(e.ptr[addr]);\n }\n if (is_prefetch) {\n e.prefetcht0(e.ptr[addr]);\n }\n assert_true(cache_line_size == 128);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CACHE_CONTROL, CACHE_CONTROL);\n\n// ============================================================================\n// OPCODE_MEMORY_BARRIER\n// ============================================================================\nstruct MEMORY_BARRIER\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) { e.mfence(); }\n};\nEMITTER_OPCODE_TABLE(OPCODE_MEMORY_BARRIER, MEMORY_BARRIER);\n\n// ============================================================================\n// OPCODE_MEMSET\n// ============================================================================\nstruct MEMSET_I64_I8_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.is_constant);\n assert_true(i.src3.is_constant);\n assert_true(i.src2.constant() == 0);\n e.vpxor(e.xmm0, e.xmm0);\n auto addr = ComputeMemoryAddress(e, i.src1);\n switch (i.src3.constant()) {\n case 32:\n e.vmovaps(e.ptr[addr + 0 * 16], e.xmm0);\n e.vmovaps(e.ptr[addr + 1 * 16], e.xmm0);\n break;\n case 128:\n e.vmovaps(e.ptr[addr + 0 * 16], e.xmm0);\n e.vmovaps(e.ptr[addr + 1 * 16], e.xmm0);\n e.vmovaps(e.ptr[addr + 2 * 16], e.xmm0);\n e.vmovaps(e.ptr[addr + 3 * 16], e.xmm0);\n e.vmovaps(e.ptr[addr + 4 * 16], e.xmm0);\n e.vmovaps(e.ptr[addr + 5 * 16], e.xmm0);\n e.vmovaps(e.ptr[addr + 6 * 16], e.xmm0);\n e.vmovaps(e.ptr[addr + 7 * 16], e.xmm0);\n break;\n default:\n assert_unhandled_case(i.src3.constant());\n break;\n }\n if (IsTracingData()) {\n addr = ComputeMemoryAddress(e, i.src1);\n e.mov(e.GetNativeParam(2), i.src3.constant());\n e.mov(e.GetNativeParam(1), i.src2.constant());\n e.lea(e.GetNativeParam(0), e.ptr[addr]);\n e.CallNative(reinterpret_cast(TraceMemset));\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_MEMSET, MEMSET_I64_I8_I64);\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_seq_vector.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Xenia Developers. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_sequences.h\"\n\n#include \n#include \n\n#include \"xenia/cpu/backend/x64/x64_op.h\"\n\n// For OPCODE_PACK/OPCODE_UNPACK\n#include \"third_party/half/include/half.hpp\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nvolatile int anchor_vector = 0;\n\n// ============================================================================\n// OPCODE_VECTOR_CONVERT_I2F\n// ============================================================================\nstruct VECTOR_CONVERT_I2F\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // flags = ARITHMETIC_UNSIGNED\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n // Round manually to (1.stored mantissa bits * 2^31) or to 2^32 to the\n // nearest even (the only rounding mode used on AltiVec) if the number is\n // 0x80000000 or greater, instead of converting src & 0x7FFFFFFF and then\n // adding 2147483648.0f, which results in double rounding that can give a\n // result larger than needed - see OPCODE_VECTOR_CONVERT_I2F notes.\n\n // [0x80000000, 0xFFFFFFFF] case:\n\n // Round to the nearest even, from (0x80000000 | 31 stored mantissa bits)\n // to ((-1 << 23) | 23 stored mantissa bits), or to 0 if the result should\n // be 4294967296.0f.\n // xmm0 = src + 0b01111111 + ((src >> 8) & 1)\n // (xmm1 also used to launch reg + mem early and to require it late)\n e.vpaddd(e.xmm1, i.src1, e.GetXmmConstPtr(XMMInt127));\n e.vpslld(e.xmm0, i.src1, 31 - 8);\n e.vpsrld(e.xmm0, e.xmm0, 31);\n e.vpaddd(e.xmm0, e.xmm0, e.xmm1);\n // xmm0 = (0xFF800000 | 23 explicit mantissa bits), or 0 if overflowed\n e.vpsrad(e.xmm0, e.xmm0, 8);\n // Calculate the result for the [0x80000000, 0xFFFFFFFF] case - take the\n // rounded mantissa, and add -1 or 0 to the exponent of 32, depending on\n // whether the number should be (1.stored mantissa bits * 2^31) or 2^32.\n // xmm0 = [0x80000000, 0xFFFFFFFF] case result\n e.vpaddd(e.xmm0, e.xmm0, e.GetXmmConstPtr(XMM2To32));\n\n // [0x00000000, 0x7FFFFFFF] case\n // (during vblendvps reg -> vpaddd reg -> vpaddd mem dependency):\n\n // Convert from signed integer to float.\n // xmm1 = [0x00000000, 0x7FFFFFFF] case result\n e.vcvtdq2ps(e.xmm1, i.src1);\n\n // Merge the two ways depending on whether the number is >= 0x80000000\n // (has high bit set).\n e.vblendvps(i.dest, e.xmm1, e.xmm0, i.src1);\n } else {\n e.vcvtdq2ps(i.dest, i.src1);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_CONVERT_I2F, VECTOR_CONVERT_I2F);\n\n// ============================================================================\n// OPCODE_VECTOR_CONVERT_F2I\n// ============================================================================\nstruct VECTOR_CONVERT_F2I\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n Opmask mask = e.k1;\n // Mask positive values and unordered values\n // _CMP_NLT_UQ\n e.vcmpps(mask, i.src1, e.GetXmmConstPtr(XMMZero), 0x15);\n\n // vcvttps2udq will saturate overflowing positive values and unordered\n // values to UINT_MAX. Mask registers will write zero everywhere\n // else (negative values)\n e.vcvttps2udq(i.dest.reg() | mask | e.T_z, i.src1);\n return;\n }\n\n // clamp to min 0\n e.vmaxps(e.xmm0, i.src1, e.GetXmmConstPtr(XMMZero));\n\n // xmm1 = mask of values >= (unsigned)INT_MIN\n e.vcmpgeps(e.xmm1, e.xmm0, e.GetXmmConstPtr(XMMPosIntMinPS));\n\n // scale any values >= (unsigned)INT_MIN back to [0, ...]\n e.vsubps(e.xmm2, e.xmm0, e.GetXmmConstPtr(XMMPosIntMinPS));\n e.vblendvps(e.xmm0, e.xmm0, e.xmm2, e.xmm1);\n\n // xmm0 = [0, INT_MAX]\n // this may still contain values > INT_MAX (if src has vals > UINT_MAX)\n e.vcvttps2dq(i.dest, e.xmm0);\n\n // xmm0 = mask of values that need saturation\n e.vpcmpeqd(e.xmm0, i.dest, e.GetXmmConstPtr(XMMIntMin));\n\n // scale values back above [INT_MIN, UINT_MAX]\n e.vpand(e.xmm1, e.xmm1, e.GetXmmConstPtr(XMMIntMin));\n e.vpaddd(i.dest, i.dest, e.xmm1);\n\n // saturate values > UINT_MAX\n e.vpor(i.dest, i.dest, e.xmm0);\n } else {\n // xmm2 = NaN mask\n e.vcmpunordps(e.xmm2, i.src1, i.src1);\n\n // convert packed floats to packed dwords\n e.vcvttps2dq(e.xmm0, i.src1);\n\n // (high bit) xmm1 = dest is indeterminate and i.src1 >= 0\n e.vpcmpeqd(e.xmm1, e.xmm0, e.GetXmmConstPtr(XMMIntMin));\n e.vpandn(e.xmm1, i.src1, e.xmm1);\n\n // saturate positive values\n e.vblendvps(i.dest, e.xmm0, e.GetXmmConstPtr(XMMIntMax), e.xmm1);\n\n // mask NaNs\n e.vpandn(i.dest, e.xmm2, i.dest);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_CONVERT_F2I, VECTOR_CONVERT_F2I);\n\n// ============================================================================\n// OPCODE_LOAD_VECTOR_SHL\n// ============================================================================\nstatic const vec128_t lvsl_table[16] = {\n vec128b(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15),\n vec128b(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16),\n vec128b(2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17),\n vec128b(3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18),\n vec128b(4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19),\n vec128b(5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20),\n vec128b(6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21),\n vec128b(7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22),\n vec128b(8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23),\n vec128b(9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24),\n vec128b(10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25),\n vec128b(11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26),\n vec128b(12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27),\n vec128b(13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28),\n vec128b(14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29),\n vec128b(15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30),\n};\nstruct LOAD_VECTOR_SHL_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.src1.is_constant) {\n auto sh = i.src1.constant();\n assert_true(sh < xe::countof(lvsl_table));\n e.mov(e.rax, (uintptr_t)&lvsl_table[sh]);\n e.vmovaps(i.dest, e.ptr[e.rax]);\n } else {\n // TODO(benvanik): find a cheaper way of doing this.\n e.movzx(e.rdx, i.src1);\n e.and_(e.dx, 0xF);\n e.shl(e.dx, 4);\n e.mov(e.rax, (uintptr_t)lvsl_table);\n e.vmovaps(i.dest, e.ptr[e.rax + e.rdx]);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOAD_VECTOR_SHL, LOAD_VECTOR_SHL_I8);\n\n// ============================================================================\n// OPCODE_LOAD_VECTOR_SHR\n// ============================================================================\nstatic const vec128_t lvsr_table[16] = {\n vec128b(16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31),\n vec128b(15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30),\n vec128b(14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29),\n vec128b(13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28),\n vec128b(12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27),\n vec128b(11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26),\n vec128b(10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25),\n vec128b(9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24),\n vec128b(8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23),\n vec128b(7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22),\n vec128b(6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21),\n vec128b(5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20),\n vec128b(4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19),\n vec128b(3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18),\n vec128b(2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17),\n vec128b(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16),\n};\nstruct LOAD_VECTOR_SHR_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.src1.is_constant) {\n auto sh = i.src1.constant();\n assert_true(sh < xe::countof(lvsr_table));\n e.mov(e.rax, (uintptr_t)&lvsr_table[sh]);\n e.vmovaps(i.dest, e.ptr[e.rax]);\n } else {\n // TODO(benvanik): find a cheaper way of doing this.\n e.movzx(e.rdx, i.src1);\n e.and_(e.dx, 0xF);\n e.shl(e.dx, 4);\n e.mov(e.rax, (uintptr_t)lvsr_table);\n e.vmovaps(i.dest, e.ptr[e.rax + e.rdx]);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOAD_VECTOR_SHR, LOAD_VECTOR_SHR_I8);\n\n// ============================================================================\n// OPCODE_VECTOR_MAX\n// ============================================================================\nstruct VECTOR_MAX\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n uint32_t part_type = i.instr->flags >> 8;\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n switch (part_type) {\n case INT8_TYPE:\n e.vpmaxub(dest, src1, src2);\n break;\n case INT16_TYPE:\n e.vpmaxuw(dest, src1, src2);\n break;\n case INT32_TYPE:\n e.vpmaxud(dest, src1, src2);\n break;\n default:\n assert_unhandled_case(part_type);\n break;\n }\n } else {\n switch (part_type) {\n case INT8_TYPE:\n e.vpmaxsb(dest, src1, src2);\n break;\n case INT16_TYPE:\n e.vpmaxsw(dest, src1, src2);\n break;\n case INT32_TYPE:\n e.vpmaxsd(dest, src1, src2);\n break;\n default:\n assert_unhandled_case(part_type);\n break;\n }\n }\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_MAX, VECTOR_MAX);\n\n// ============================================================================\n// OPCODE_VECTOR_MIN\n// ============================================================================\nstruct VECTOR_MIN\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n uint32_t part_type = i.instr->flags >> 8;\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n switch (part_type) {\n case INT8_TYPE:\n e.vpminub(dest, src1, src2);\n break;\n case INT16_TYPE:\n e.vpminuw(dest, src1, src2);\n break;\n case INT32_TYPE:\n e.vpminud(dest, src1, src2);\n break;\n default:\n assert_unhandled_case(part_type);\n break;\n }\n } else {\n switch (part_type) {\n case INT8_TYPE:\n e.vpminsb(dest, src1, src2);\n break;\n case INT16_TYPE:\n e.vpminsw(dest, src1, src2);\n break;\n case INT32_TYPE:\n e.vpminsd(dest, src1, src2);\n break;\n default:\n assert_unhandled_case(part_type);\n break;\n }\n }\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_MIN, VECTOR_MIN);\n\n// ============================================================================\n// OPCODE_VECTOR_COMPARE_EQ\n// ============================================================================\nstruct VECTOR_COMPARE_EQ_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n switch (i.instr->flags) {\n case INT8_TYPE:\n e.vpcmpeqb(dest, src1, src2);\n break;\n case INT16_TYPE:\n e.vpcmpeqw(dest, src1, src2);\n break;\n case INT32_TYPE:\n e.vpcmpeqd(dest, src1, src2);\n break;\n case FLOAT32_TYPE:\n e.vcmpeqps(dest, src1, src2);\n break;\n }\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_COMPARE_EQ, VECTOR_COMPARE_EQ_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_COMPARE_SGT\n// ============================================================================\nstruct VECTOR_COMPARE_SGT_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAssociativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n switch (i.instr->flags) {\n case INT8_TYPE:\n e.vpcmpgtb(dest, src1, src2);\n break;\n case INT16_TYPE:\n e.vpcmpgtw(dest, src1, src2);\n break;\n case INT32_TYPE:\n e.vpcmpgtd(dest, src1, src2);\n break;\n case FLOAT32_TYPE:\n e.vcmpgtps(dest, src1, src2);\n break;\n }\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_COMPARE_SGT, VECTOR_COMPARE_SGT_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_COMPARE_SGE\n// ============================================================================\nstruct VECTOR_COMPARE_SGE_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAssociativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n switch (i.instr->flags) {\n case INT8_TYPE:\n e.vpcmpeqb(e.xmm0, src1, src2);\n e.vpcmpgtb(dest, src1, src2);\n e.vpor(dest, e.xmm0);\n break;\n case INT16_TYPE:\n e.vpcmpeqw(e.xmm0, src1, src2);\n e.vpcmpgtw(dest, src1, src2);\n e.vpor(dest, e.xmm0);\n break;\n case INT32_TYPE:\n e.vpcmpeqd(e.xmm0, src1, src2);\n e.vpcmpgtd(dest, src1, src2);\n e.vpor(dest, e.xmm0);\n break;\n case FLOAT32_TYPE:\n e.vcmpgeps(dest, src1, src2);\n break;\n }\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_COMPARE_SGE, VECTOR_COMPARE_SGE_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_COMPARE_UGT\n// ============================================================================\nstruct VECTOR_COMPARE_UGT_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho | kX64EmitAVX512BW |\n kX64EmitAVX512DQ) &&\n (i.instr->flags != FLOAT32_TYPE)) {\n Xmm src1 = e.xmm0;\n if (i.src1.is_constant) {\n e.LoadConstantXmm(src1, i.src1.constant());\n } else {\n src1 = i.src1;\n }\n\n Xmm src2 = e.xmm1;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(src2, i.src2.constant());\n } else {\n src2 = i.src2;\n }\n\n switch (i.instr->flags) {\n case INT8_TYPE:\n e.vpcmpub(e.k1, src1, src2, 0x6);\n e.vpmovm2b(i.dest, e.k1);\n break;\n case INT16_TYPE:\n e.vpcmpuw(e.k1, src1, src2, 0x6);\n e.vpmovm2w(i.dest, e.k1);\n break;\n case INT32_TYPE:\n e.vpcmpud(e.k1, src1, src2, 0x6);\n e.vpmovm2d(i.dest, e.k1);\n break;\n default:\n assert_always();\n break;\n }\n return;\n }\n\n Xbyak::Address sign_addr = e.ptr[e.rax]; // dummy\n switch (i.instr->flags) {\n case INT8_TYPE:\n sign_addr = e.GetXmmConstPtr(XMMSignMaskI8);\n break;\n case INT16_TYPE:\n sign_addr = e.GetXmmConstPtr(XMMSignMaskI16);\n break;\n case INT32_TYPE:\n sign_addr = e.GetXmmConstPtr(XMMSignMaskI32);\n break;\n case FLOAT32_TYPE:\n sign_addr = e.GetXmmConstPtr(XMMSignMaskF32);\n break;\n default:\n assert_always();\n break;\n }\n if (i.src1.is_constant) {\n // TODO(benvanik): make this constant.\n e.LoadConstantXmm(e.xmm0, i.src1.constant());\n e.vpxor(e.xmm0, sign_addr);\n } else {\n e.vpxor(e.xmm0, i.src1, sign_addr);\n }\n if (i.src2.is_constant) {\n // TODO(benvanik): make this constant.\n e.LoadConstantXmm(e.xmm1, i.src2.constant());\n e.vpxor(e.xmm1, sign_addr);\n } else {\n e.vpxor(e.xmm1, i.src2, sign_addr);\n }\n switch (i.instr->flags) {\n case INT8_TYPE:\n e.vpcmpgtb(i.dest, e.xmm0, e.xmm1);\n break;\n case INT16_TYPE:\n e.vpcmpgtw(i.dest, e.xmm0, e.xmm1);\n break;\n case INT32_TYPE:\n e.vpcmpgtd(i.dest, e.xmm0, e.xmm1);\n break;\n case FLOAT32_TYPE:\n e.vcmpgtps(i.dest, e.xmm0, e.xmm1);\n break;\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_COMPARE_UGT, VECTOR_COMPARE_UGT_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_COMPARE_UGE\n// ============================================================================\nstruct VECTOR_COMPARE_UGE_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Xbyak::Address sign_addr = e.ptr[e.rax]; // dummy\n switch (i.instr->flags) {\n case INT8_TYPE:\n sign_addr = e.GetXmmConstPtr(XMMSignMaskI8);\n break;\n case INT16_TYPE:\n sign_addr = e.GetXmmConstPtr(XMMSignMaskI16);\n break;\n case INT32_TYPE:\n sign_addr = e.GetXmmConstPtr(XMMSignMaskI32);\n break;\n case FLOAT32_TYPE:\n sign_addr = e.GetXmmConstPtr(XMMSignMaskF32);\n break;\n }\n if (i.src1.is_constant) {\n // TODO(benvanik): make this constant.\n e.LoadConstantXmm(e.xmm0, i.src1.constant());\n e.vpxor(e.xmm0, sign_addr);\n } else {\n e.vpxor(e.xmm0, i.src1, sign_addr);\n }\n if (i.src2.is_constant) {\n // TODO(benvanik): make this constant.\n e.LoadConstantXmm(e.xmm1, i.src2.constant());\n e.vpxor(e.xmm1, sign_addr);\n } else {\n e.vpxor(e.xmm1, i.src2, sign_addr);\n }\n switch (i.instr->flags) {\n case INT8_TYPE:\n e.vpcmpeqb(e.xmm2, e.xmm0, e.xmm1);\n e.vpcmpgtb(i.dest, e.xmm0, e.xmm1);\n e.vpor(i.dest, e.xmm2);\n break;\n case INT16_TYPE:\n e.vpcmpeqw(e.xmm2, e.xmm0, e.xmm1);\n e.vpcmpgtw(i.dest, e.xmm0, e.xmm1);\n e.vpor(i.dest, e.xmm2);\n break;\n case INT32_TYPE:\n e.vpcmpeqd(e.xmm2, e.xmm0, e.xmm1);\n e.vpcmpgtd(i.dest, e.xmm0, e.xmm1);\n e.vpor(i.dest, e.xmm2);\n break;\n case FLOAT32_TYPE:\n e.vcmpgeps(i.dest, e.xmm0, e.xmm1);\n break;\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_COMPARE_UGE, VECTOR_COMPARE_UGE_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_ADD\n// ============================================================================\nstruct VECTOR_ADD\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, const Xmm& dest, Xmm src1, Xmm src2) {\n const TypeName part_type =\n static_cast(i.instr->flags & 0xFF);\n const uint32_t arithmetic_flags = i.instr->flags >> 8;\n bool is_unsigned = !!(arithmetic_flags & ARITHMETIC_UNSIGNED);\n bool saturate = !!(arithmetic_flags & ARITHMETIC_SATURATE);\n switch (part_type) {\n case INT8_TYPE:\n if (saturate) {\n // TODO(benvanik): trace DID_SATURATE\n if (is_unsigned) {\n e.vpaddusb(dest, src1, src2);\n } else {\n e.vpaddsb(dest, src1, src2);\n }\n } else {\n e.vpaddb(dest, src1, src2);\n }\n break;\n case INT16_TYPE:\n if (saturate) {\n // TODO(benvanik): trace DID_SATURATE\n if (is_unsigned) {\n e.vpaddusw(dest, src1, src2);\n } else {\n e.vpaddsw(dest, src1, src2);\n }\n } else {\n e.vpaddw(dest, src1, src2);\n }\n break;\n case INT32_TYPE:\n if (saturate) {\n if (is_unsigned) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vpaddd(dest, src1, src2);\n Opmask saturate = e.k1;\n // _mm_cmplt_epu32_mask\n e.vpcmpud(saturate, dest, src1, 0x1);\n e.vpternlogd(dest | saturate, dest, dest, 0xFF);\n return;\n }\n\n // xmm0 is the only temp register that can be used by\n // src1/src2.\n e.vpaddd(e.xmm1, src1, src2);\n\n // If result is smaller than either of the inputs, we've\n // overflowed (only need to check one input)\n // if (src1 > res) then overflowed\n // http://locklessinc.com/articles/sat_arithmetic/\n e.vpxor(e.xmm2, src1, e.GetXmmConstPtr(XMMSignMaskI32));\n e.vpxor(e.xmm0, e.xmm1, e.GetXmmConstPtr(XMMSignMaskI32));\n e.vpcmpgtd(e.xmm0, e.xmm2, e.xmm0);\n e.vpor(dest, e.xmm1, e.xmm0);\n } else {\n e.vpaddd(e.xmm1, src1, src2);\n\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho |\n kX64EmitAVX512DQ)) {\n e.vmovdqa32(e.xmm3, src1);\n e.vpternlogd(e.xmm3, e.xmm1, src2, 0b00100100);\n\n const Opmask saturate = e.k1;\n e.vpmovd2m(saturate, e.xmm3);\n\n e.vpsrad(e.xmm2, e.xmm1, 31);\n e.vpxord(e.xmm2, e.xmm2, e.GetXmmConstPtr(XMMSignMaskI32));\n e.vpblendmd(dest | saturate, e.xmm1, e.xmm2);\n return;\n }\n\n // Overflow results if two inputs are the same sign and the\n // result isn't the same sign. if ((s32b)(~(src1 ^ src2) &\n // (src1 ^ res)) < 0) then overflowed\n // http://locklessinc.com/articles/sat_arithmetic/\n e.vpxor(e.xmm2, src1, src2);\n e.vpxor(e.xmm3, src1, e.xmm1);\n e.vpandn(e.xmm2, e.xmm2, e.xmm3);\n\n // Set any negative overflowed elements of src1 to INT_MIN\n e.vpand(e.xmm3, src1, e.xmm2);\n e.vblendvps(e.xmm1, e.xmm1, e.GetXmmConstPtr(XMMSignMaskI32),\n e.xmm3);\n\n // Set any positive overflowed elements of src1 to INT_MAX\n e.vpandn(e.xmm3, src1, e.xmm2);\n e.vblendvps(dest, e.xmm1, e.GetXmmConstPtr(XMMAbsMaskPS),\n e.xmm3);\n }\n } else {\n e.vpaddd(dest, src1, src2);\n }\n break;\n case FLOAT32_TYPE:\n assert_false(is_unsigned);\n assert_false(saturate);\n e.vaddps(dest, src1, src2);\n break;\n default:\n assert_unhandled_case(part_type);\n break;\n }\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_ADD, VECTOR_ADD);\n\n// ============================================================================\n// OPCODE_VECTOR_SUB\n// ============================================================================\nstruct VECTOR_SUB\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, const Xmm& dest, Xmm src1, Xmm src2) {\n const TypeName part_type =\n static_cast(i.instr->flags & 0xFF);\n const uint32_t arithmetic_flags = i.instr->flags >> 8;\n bool is_unsigned = !!(arithmetic_flags & ARITHMETIC_UNSIGNED);\n bool saturate = !!(arithmetic_flags & ARITHMETIC_SATURATE);\n switch (part_type) {\n case INT8_TYPE:\n if (saturate) {\n // TODO(benvanik): trace DID_SATURATE\n if (is_unsigned) {\n e.vpsubusb(dest, src1, src2);\n } else {\n e.vpsubsb(dest, src1, src2);\n }\n } else {\n e.vpsubb(dest, src1, src2);\n }\n break;\n case INT16_TYPE:\n if (saturate) {\n // TODO(benvanik): trace DID_SATURATE\n if (is_unsigned) {\n e.vpsubusw(dest, src1, src2);\n } else {\n e.vpsubsw(dest, src1, src2);\n }\n } else {\n e.vpsubw(dest, src1, src2);\n }\n break;\n case INT32_TYPE:\n if (saturate) {\n if (is_unsigned) {\n // xmm0 is the only temp register that can be used by\n // src1/src2.\n e.vpsubd(e.xmm1, src1, src2);\n\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n // If the result is less or equal to the first operand then\n // we did not underflow\n Opmask not_underflow = e.k1;\n // _mm_cmple_epu32_mask\n e.vpcmpud(not_underflow, e.xmm1, src1, 0x2);\n\n // Copy over values that did not underflow, write zero\n // everywhere else\n e.vmovdqa32(dest | not_underflow | e.T_z, e.xmm1);\n return;\n }\n\n // If result is greater than either of the inputs, we've\n // underflowed (only need to check one input)\n // if (res > src1) then underflowed\n // http://locklessinc.com/articles/sat_arithmetic/\n e.vpxor(e.xmm2, src1, e.GetXmmConstPtr(XMMSignMaskI32));\n e.vpxor(e.xmm0, e.xmm1, e.GetXmmConstPtr(XMMSignMaskI32));\n e.vpcmpgtd(e.xmm0, e.xmm0, e.xmm2);\n e.vpandn(dest, e.xmm0, e.xmm1);\n } else {\n e.vpsubd(e.xmm1, src1, src2);\n\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho |\n kX64EmitAVX512DQ)) {\n e.vmovdqa32(e.xmm3, src1);\n e.vpternlogd(e.xmm3, e.xmm1, src2, 0b00011000);\n\n const Opmask saturate = e.k1;\n e.vpmovd2m(saturate, e.xmm3);\n\n e.vpsrad(e.xmm2, e.xmm1, 31);\n e.vpxord(e.xmm2, e.xmm2, e.GetXmmConstPtr(XMMSignMaskI32));\n\n e.vpblendmd(dest | saturate, e.xmm1, e.xmm2);\n return;\n }\n\n // We can only overflow if the signs of the operands are\n // opposite. If signs are opposite and result sign isn't the\n // same as src1's sign, we've overflowed. if ((s32b)((src1 ^\n // src2) & (src1 ^ res)) < 0) then overflowed\n // http://locklessinc.com/articles/sat_arithmetic/\n e.vpxor(e.xmm2, src1, src2);\n e.vpxor(e.xmm3, src1, e.xmm1);\n e.vpand(e.xmm2, e.xmm2, e.xmm3);\n\n // Set any negative overflowed elements of src1 to INT_MIN\n e.vpand(e.xmm3, src1, e.xmm2);\n e.vblendvps(e.xmm1, e.xmm1, e.GetXmmConstPtr(XMMSignMaskI32),\n e.xmm3);\n\n // Set any positive overflowed elements of src1 to INT_MAX\n e.vpandn(e.xmm3, src1, e.xmm2);\n e.vblendvps(dest, e.xmm1, e.GetXmmConstPtr(XMMAbsMaskPS),\n e.xmm3);\n }\n } else {\n e.vpsubd(dest, src1, src2);\n }\n break;\n case FLOAT32_TYPE:\n e.vsubps(dest, src1, src2);\n break;\n default:\n assert_unhandled_case(part_type);\n break;\n }\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_SUB, VECTOR_SUB);\n\n// ============================================================================\n// OPCODE_VECTOR_SHL\n// ============================================================================\ntemplate ::value, int> = 0>\nstatic __m128i EmulateVectorShl(void*, __m128i src1, __m128i src2) {\n alignas(16) T value[16 / sizeof(T)];\n alignas(16) T shamt[16 / sizeof(T)];\n\n // Load SSE registers into a C array.\n _mm_store_si128(reinterpret_cast<__m128i*>(value), src1);\n _mm_store_si128(reinterpret_cast<__m128i*>(shamt), src2);\n\n for (size_t i = 0; i < (16 / sizeof(T)); ++i) {\n value[i] = value[i] << (shamt[i] & ((sizeof(T) * 8) - 1));\n }\n\n // Store result and return it.\n return _mm_load_si128(reinterpret_cast<__m128i*>(value));\n}\n\nstruct VECTOR_SHL_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags) {\n case INT8_TYPE:\n EmitInt8(e, i);\n break;\n case INT16_TYPE:\n EmitInt16(e, i);\n break;\n case INT32_TYPE:\n EmitInt32(e, i);\n break;\n default:\n assert_always();\n break;\n }\n }\n\n static void EmitInt8(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): native version (with shift magic).\n if (i.src2.is_constant) {\n if (e.IsFeatureEnabled(kX64EmitGFNI)) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 16 - n; ++n) {\n if (shamt.u8[n] != shamt.u8[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use gf2p8affineqb.\n const uint8_t shift_amount = shamt.u8[0] & 0b111;\n const uint64_t shift_matrix =\n UINT64_C(0x0102040810204080) >> (shift_amount * 8);\n e.vgf2p8affineqb(i.dest, i.src1,\n e.StashConstantXmm(0, vec128q(shift_matrix)), 0);\n return;\n }\n }\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShl));\n e.vmovaps(i.dest, e.xmm0);\n }\n\n static void EmitInt16(X64Emitter& e, const EmitArgType& i) {\n Xmm src1;\n if (i.src1.is_constant) {\n src1 = e.xmm2;\n e.LoadConstantXmm(src1, i.src1.constant());\n } else {\n src1 = i.src1;\n }\n\n if (i.src2.is_constant) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 8 - n; ++n) {\n if (shamt.u16[n] != shamt.u16[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use vpsllw.\n e.vpsllw(i.dest, src1, shamt.u16[0] & 0xF);\n return;\n }\n }\n\n // Shift 8 words in src1 by amount specified in src2.\n Xbyak::Label emu, end;\n\n // Only bother with this check if shift amt isn't constant.\n if (!i.src2.is_constant) {\n // See if the shift is equal first for a shortcut.\n e.vpshuflw(e.xmm0, i.src2, 0b00000000);\n e.vpshufd(e.xmm0, e.xmm0, 0b00000000);\n e.vpxor(e.xmm1, e.xmm0, i.src2);\n e.vptest(e.xmm1, e.xmm1);\n e.jnz(emu);\n\n // Equal. Shift using vpsllw.\n e.mov(e.rax, 0xF);\n e.vmovq(e.xmm1, e.rax);\n e.vpand(e.xmm0, e.xmm0, e.xmm1);\n e.vpsllw(i.dest, src1, e.xmm0);\n e.jmp(end);\n }\n\n // TODO(benvanik): native version (with shift magic).\n e.L(emu);\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShl));\n e.vmovaps(i.dest, e.xmm0);\n\n e.L(end);\n }\n\n static void EmitInt32(X64Emitter& e, const EmitArgType& i) {\n Xmm src1;\n if (i.src1.is_constant) {\n src1 = e.xmm2;\n e.LoadConstantXmm(src1, i.src1.constant());\n } else {\n src1 = i.src1;\n }\n\n if (i.src2.is_constant) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 4 - n; ++n) {\n if (shamt.u32[n] != shamt.u32[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use vpslld.\n e.vpslld(i.dest, src1, shamt.u8[0] & 0x1F);\n return;\n }\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n if (i.src2.is_constant) {\n const auto& shamt = i.src2.constant();\n // Counts differ, so pre-mask and load constant.\n vec128_t masked = i.src2.constant();\n for (size_t n = 0; n < 4; ++n) {\n masked.u32[n] &= 0x1F;\n }\n e.LoadConstantXmm(e.xmm0, masked);\n e.vpsllvd(i.dest, src1, e.xmm0);\n } else {\n // Fully variable shift.\n // src shift mask may have values >31, and x86 sets to zero when\n // that happens so we mask.\n e.vandps(e.xmm0, i.src2, e.GetXmmConstPtr(XMMShiftMaskPS));\n e.vpsllvd(i.dest, src1, e.xmm0);\n }\n } else {\n // Shift 4 words in src1 by amount specified in src2.\n Xbyak::Label emu, end;\n\n // See if the shift is equal first for a shortcut.\n // Only bother with this check if shift amt isn't constant.\n if (!i.src2.is_constant) {\n e.vpshufd(e.xmm0, i.src2, 0b00000000);\n e.vpxor(e.xmm1, e.xmm0, i.src2);\n e.vptest(e.xmm1, e.xmm1);\n e.jnz(emu);\n\n // Equal. Shift using vpsrad.\n e.mov(e.rax, 0x1F);\n e.vmovq(e.xmm1, e.rax);\n e.vpand(e.xmm0, e.xmm0, e.xmm1);\n e.vpslld(i.dest, src1, e.xmm0);\n e.jmp(end);\n }\n\n // TODO(benvanik): native version (with shift magic).\n e.L(emu);\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShl));\n e.vmovaps(i.dest, e.xmm0);\n\n e.L(end);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_SHL, VECTOR_SHL_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_SHR\n// ============================================================================\ntemplate ::value, int> = 0>\nstatic __m128i EmulateVectorShr(void*, __m128i src1, __m128i src2) {\n alignas(16) T value[16 / sizeof(T)];\n alignas(16) T shamt[16 / sizeof(T)];\n\n // Load SSE registers into a C array.\n _mm_store_si128(reinterpret_cast<__m128i*>(value), src1);\n _mm_store_si128(reinterpret_cast<__m128i*>(shamt), src2);\n\n for (size_t i = 0; i < (16 / sizeof(T)); ++i) {\n value[i] = value[i] >> (shamt[i] & ((sizeof(T) * 8) - 1));\n }\n\n // Store result and return it.\n return _mm_load_si128(reinterpret_cast<__m128i*>(value));\n}\n\nstruct VECTOR_SHR_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags) {\n case INT8_TYPE:\n EmitInt8(e, i);\n break;\n case INT16_TYPE:\n EmitInt16(e, i);\n break;\n case INT32_TYPE:\n EmitInt32(e, i);\n break;\n default:\n assert_always();\n break;\n }\n }\n\n static void EmitInt8(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): native version (with shift magic).\n if (i.src2.is_constant) {\n if (e.IsFeatureEnabled(kX64EmitGFNI)) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 16 - n; ++n) {\n if (shamt.u8[n] != shamt.u8[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use gf2p8affineqb.\n const uint8_t shift_amount = shamt.u8[0] & 0b111;\n const uint64_t shift_matrix = UINT64_C(0x0102040810204080)\n << (shift_amount * 8);\n e.vgf2p8affineqb(i.dest, i.src1,\n e.StashConstantXmm(0, vec128q(shift_matrix)), 0);\n return;\n }\n }\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShr));\n e.vmovaps(i.dest, e.xmm0);\n }\n\n static void EmitInt16(X64Emitter& e, const EmitArgType& i) {\n if (i.src2.is_constant) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 8 - n; ++n) {\n if (shamt.u16[n] != shamt.u16[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use vpsllw.\n e.vpsrlw(i.dest, i.src1, shamt.u16[0] & 0xF);\n return;\n }\n }\n\n // Shift 8 words in src1 by amount specified in src2.\n Xbyak::Label emu, end;\n\n // See if the shift is equal first for a shortcut.\n // Only bother with this check if shift amt isn't constant.\n if (!i.src2.is_constant) {\n e.vpshuflw(e.xmm0, i.src2, 0b00000000);\n e.vpshufd(e.xmm0, e.xmm0, 0b00000000);\n e.vpxor(e.xmm1, e.xmm0, i.src2);\n e.vptest(e.xmm1, e.xmm1);\n e.jnz(emu);\n\n // Equal. Shift using vpsrlw.\n e.mov(e.rax, 0xF);\n e.vmovq(e.xmm1, e.rax);\n e.vpand(e.xmm0, e.xmm0, e.xmm1);\n e.vpsrlw(i.dest, i.src1, e.xmm0);\n e.jmp(end);\n }\n\n // TODO(benvanik): native version (with shift magic).\n e.L(emu);\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShr));\n e.vmovaps(i.dest, e.xmm0);\n\n e.L(end);\n }\n\n static void EmitInt32(X64Emitter& e, const EmitArgType& i) {\n Xmm src1;\n if (i.src1.is_constant) {\n src1 = e.xmm2;\n e.LoadConstantXmm(src1, i.src1.constant());\n } else {\n src1 = i.src1;\n }\n\n if (i.src2.is_constant) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 4 - n; ++n) {\n if (shamt.u32[n] != shamt.u32[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use vpsrld.\n e.vpsrld(i.dest, src1, shamt.u8[0] & 0x1F);\n return;\n } else {\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n // Counts differ, so pre-mask and load constant.\n vec128_t masked = i.src2.constant();\n for (size_t n = 0; n < 4; ++n) {\n masked.u32[n] &= 0x1F;\n }\n e.LoadConstantXmm(e.xmm0, masked);\n e.vpsrlvd(i.dest, src1, e.xmm0);\n return;\n }\n }\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n // Fully variable shift.\n // src shift mask may have values >31, and x86 sets to zero when\n // that happens so we mask.\n e.vandps(e.xmm0, i.src2, e.GetXmmConstPtr(XMMShiftMaskPS));\n e.vpsrlvd(i.dest, src1, e.xmm0);\n } else {\n // Shift 4 words in src1 by amount specified in src2.\n Xbyak::Label emu, end;\n\n // See if the shift is equal first for a shortcut.\n // Only bother with this check if shift amt isn't constant.\n if (!i.src2.is_constant) {\n e.vpshufd(e.xmm0, i.src2, 0b00000000);\n e.vpxor(e.xmm1, e.xmm0, i.src2);\n e.vptest(e.xmm1, e.xmm1);\n e.jnz(emu);\n\n // Equal. Shift using vpsrld.\n e.mov(e.rax, 0x1F);\n e.vmovq(e.xmm1, e.rax);\n e.vpand(e.xmm0, e.xmm0, e.xmm1);\n e.vpsrld(i.dest, src1, e.xmm0);\n e.jmp(end);\n }\n\n // TODO(benvanik): native version.\n e.L(emu);\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShr));\n e.vmovaps(i.dest, e.xmm0);\n\n e.L(end);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_SHR, VECTOR_SHR_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_SHA\n// ============================================================================\nstruct VECTOR_SHA_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags) {\n case INT8_TYPE:\n EmitInt8(e, i);\n break;\n case INT16_TYPE:\n EmitInt16(e, i);\n break;\n case INT32_TYPE:\n EmitInt32(e, i);\n break;\n default:\n assert_always();\n break;\n }\n }\n\n static void EmitInt8(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): native version (with shift magic).\n if (i.src2.is_constant) {\n if (e.IsFeatureEnabled(kX64EmitGFNI)) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 16 - n; ++n) {\n if (shamt.u8[n] != shamt.u8[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use gf2p8affineqb.\n const uint8_t shift_amount = shamt.u8[0] & 0b111;\n const uint64_t shift_matrix =\n (UINT64_C(0x0102040810204080) << (shift_amount * 8)) |\n (UINT64_C(0x8080808080808080) >> (64 - shift_amount * 8));\n ;\n e.vgf2p8affineqb(i.dest, i.src1,\n e.StashConstantXmm(0, vec128q(shift_matrix)), 0);\n return;\n }\n }\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShr));\n e.vmovaps(i.dest, e.xmm0);\n }\n\n static void EmitInt16(X64Emitter& e, const EmitArgType& i) {\n if (i.src2.is_constant) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 8 - n; ++n) {\n if (shamt.u16[n] != shamt.u16[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use vpsraw.\n e.vpsraw(i.dest, i.src1, shamt.u16[0] & 0xF);\n return;\n }\n }\n\n // Shift 8 words in src1 by amount specified in src2.\n Xbyak::Label emu, end;\n\n // See if the shift is equal first for a shortcut.\n // Only bother with this check if shift amt isn't constant.\n if (!i.src2.is_constant) {\n e.vpshuflw(e.xmm0, i.src2, 0b00000000);\n e.vpshufd(e.xmm0, e.xmm0, 0b00000000);\n e.vpxor(e.xmm1, e.xmm0, i.src2);\n e.vptest(e.xmm1, e.xmm1);\n e.jnz(emu);\n\n // Equal. Shift using vpsraw.\n e.mov(e.rax, 0xF);\n e.vmovq(e.xmm1, e.rax);\n e.vpand(e.xmm0, e.xmm0, e.xmm1);\n e.vpsraw(i.dest, i.src1, e.xmm0);\n e.jmp(end);\n }\n\n // TODO(benvanik): native version (with shift magic).\n e.L(emu);\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShr));\n e.vmovaps(i.dest, e.xmm0);\n\n e.L(end);\n }\n\n static void EmitInt32(X64Emitter& e, const EmitArgType& i) {\n if (i.src2.is_constant) {\n const auto& shamt = i.src2.constant();\n bool all_same = true;\n for (size_t n = 0; n < 4 - n; ++n) {\n if (shamt.u32[n] != shamt.u32[n + 1]) {\n all_same = false;\n break;\n }\n }\n if (all_same) {\n // Every count is the same, so we can use vpsrad.\n e.vpsrad(i.dest, i.src1, shamt.u32[0] & 0x1F);\n return;\n }\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n // src shift mask may have values >31, and x86 sets to zero when\n // that happens so we mask.\n if (i.src2.is_constant) {\n e.LoadConstantXmm(e.xmm0, i.src2.constant());\n e.vandps(e.xmm0, e.GetXmmConstPtr(XMMShiftMaskPS));\n } else {\n e.vandps(e.xmm0, i.src2, e.GetXmmConstPtr(XMMShiftMaskPS));\n }\n e.vpsravd(i.dest, i.src1, e.xmm0);\n } else {\n // Shift 4 words in src1 by amount specified in src2.\n Xbyak::Label emu, end;\n\n // See if the shift is equal first for a shortcut.\n // Only bother with this check if shift amt isn't constant.\n if (!i.src2.is_constant) {\n e.vpshufd(e.xmm0, i.src2, 0b00000000);\n e.vpxor(e.xmm1, e.xmm0, i.src2);\n e.vptest(e.xmm1, e.xmm1);\n e.jnz(emu);\n\n // Equal. Shift using vpsrad.\n e.mov(e.rax, 0x1F);\n e.vmovq(e.xmm1, e.rax);\n e.vpand(e.xmm0, e.xmm0, e.xmm1);\n e.vpsrad(i.dest, i.src1, e.xmm0);\n e.jmp(end);\n }\n\n // TODO(benvanik): native version.\n e.L(emu);\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulateVectorShr));\n e.vmovaps(i.dest, e.xmm0);\n\n e.L(end);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_SHA, VECTOR_SHA_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_ROTATE_LEFT\n// ============================================================================\ntemplate ::value, int> = 0>\nstatic __m128i EmulateVectorRotateLeft(void*, __m128i src1, __m128i src2) {\n alignas(16) T value[16 / sizeof(T)];\n alignas(16) T shamt[16 / sizeof(T)];\n\n // Load SSE registers into a C array.\n _mm_store_si128(reinterpret_cast<__m128i*>(value), src1);\n _mm_store_si128(reinterpret_cast<__m128i*>(shamt), src2);\n\n for (size_t i = 0; i < (16 / sizeof(T)); ++i) {\n value[i] = xe::rotate_left(value[i], shamt[i] & ((sizeof(T) * 8) - 1));\n }\n\n // Store result and return it.\n return _mm_load_si128(reinterpret_cast<__m128i*>(value));\n}\n\nstruct VECTOR_ROTATE_LEFT_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags) {\n case INT8_TYPE:\n // TODO(benvanik): native version (with shift magic).\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(\n reinterpret_cast(EmulateVectorRotateLeft));\n e.vmovaps(i.dest, e.xmm0);\n break;\n case INT16_TYPE:\n // TODO(benvanik): native version (with shift magic).\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1), e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(\n reinterpret_cast(EmulateVectorRotateLeft));\n e.vmovaps(i.dest, e.xmm0);\n break;\n case INT32_TYPE: {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vprolvd(i.dest, i.src1, i.src2);\n } else if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n Xmm temp = i.dest;\n if (i.dest == i.src1 || i.dest == i.src2) {\n temp = e.xmm2;\n }\n // Shift left (to get high bits):\n if (i.src2.is_constant) {\n e.LoadConstantXmm(temp, i.src2.constant());\n e.vpand(e.xmm0, temp, e.GetXmmConstPtr(XMMShiftMaskPS));\n } else {\n e.vpand(e.xmm0, i.src2, e.GetXmmConstPtr(XMMShiftMaskPS));\n }\n e.vpsllvd(e.xmm1, i.src1, e.xmm0);\n // Shift right (to get low bits):\n e.vmovaps(temp, e.GetXmmConstPtr(XMMPI32));\n e.vpsubd(temp, e.xmm0);\n e.vpsrlvd(i.dest, i.src1, temp);\n // Merge:\n e.vpor(i.dest, e.xmm1);\n } else {\n // TODO(benvanik): non-AVX2 native version.\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1),\n e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(\n reinterpret_cast(EmulateVectorRotateLeft));\n e.vmovaps(i.dest, e.xmm0);\n }\n break;\n }\n default:\n assert_always();\n break;\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_ROTATE_LEFT, VECTOR_ROTATE_LEFT_V128);\n\n// ============================================================================\n// OPCODE_VECTOR_AVERAGE\n// ============================================================================\ntemplate ::value, int> = 0>\nstatic __m128i EmulateVectorAverage(void*, __m128i src1, __m128i src2) {\n alignas(16) T src1v[16 / sizeof(T)];\n alignas(16) T src2v[16 / sizeof(T)];\n alignas(16) T value[16 / sizeof(T)];\n\n // Load SSE registers into a C array.\n _mm_store_si128(reinterpret_cast<__m128i*>(src1v), src1);\n _mm_store_si128(reinterpret_cast<__m128i*>(src2v), src2);\n\n for (size_t i = 0; i < (16 / sizeof(T)); ++i) {\n auto t = (uint64_t(src1v[i]) + uint64_t(src2v[i]) + 1) / 2;\n value[i] = T(t);\n }\n\n // Store result and return it.\n return _mm_load_si128(reinterpret_cast<__m128i*>(value));\n}\n\nstruct VECTOR_AVERAGE\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(\n e, i,\n [&i](X64Emitter& e, const Xmm& dest, const Xmm& src1, const Xmm& src2) {\n const TypeName part_type =\n static_cast(i.instr->flags & 0xFF);\n const uint32_t arithmetic_flags = i.instr->flags >> 8;\n bool is_unsigned = !!(arithmetic_flags & ARITHMETIC_UNSIGNED);\n switch (part_type) {\n case INT8_TYPE:\n if (is_unsigned) {\n e.vpavgb(dest, src1, src2);\n } else {\n assert_always();\n }\n break;\n case INT16_TYPE:\n if (is_unsigned) {\n e.vpavgw(dest, src1, src2);\n } else {\n assert_always();\n }\n break;\n case INT32_TYPE:\n // No 32bit averages in AVX.\n if (is_unsigned) {\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1),\n e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(\n reinterpret_cast(EmulateVectorAverage));\n e.vmovaps(i.dest, e.xmm0);\n } else {\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1),\n e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(\n reinterpret_cast(EmulateVectorAverage));\n e.vmovaps(i.dest, e.xmm0);\n }\n break;\n default:\n assert_unhandled_case(part_type);\n break;\n }\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_VECTOR_AVERAGE, VECTOR_AVERAGE);\n\n// ============================================================================\n// OPCODE_INSERT\n// ============================================================================\nstruct INSERT_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.is_constant);\n e.vpinsrb(i.dest, i.src3.reg().cvt32(), i.src2.constant() ^ 0x3);\n }\n};\nstruct INSERT_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.is_constant);\n e.vpinsrw(i.dest, i.src3.reg().cvt32(), i.src2.constant() ^ 0x1);\n }\n};\nstruct INSERT_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.is_constant);\n e.vpinsrd(i.dest, i.src3, i.src2.constant());\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_INSERT, INSERT_I8, INSERT_I16, INSERT_I32);\n\n// ============================================================================\n// OPCODE_EXTRACT\n// ============================================================================\n// TODO(benvanik): sequence extract/splat:\n// v0.i32 = extract v0.v128, 0\n// v0.v128 = splat v0.i32\n// This can be a single broadcast.\nstruct EXTRACT_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.src2.is_constant) {\n e.vpextrb(i.dest.reg().cvt32(), i.src1, VEC128_B(i.src2.constant()));\n } else {\n e.mov(e.eax, 0x00000003);\n e.xor_(e.al, i.src2);\n e.and_(e.al, 0x1F);\n e.vmovd(e.xmm0, e.eax);\n e.vpshufb(e.xmm0, i.src1, e.xmm0);\n e.vmovd(i.dest.reg().cvt32(), e.xmm0);\n e.and_(i.dest, uint8_t(0xFF));\n }\n }\n};\nstruct EXTRACT_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.src2.is_constant) {\n e.vpextrw(i.dest.reg().cvt32(), i.src1, VEC128_W(i.src2.constant()));\n } else {\n e.mov(e.al, i.src2);\n e.xor_(e.al, 0x01);\n e.shl(e.al, 1);\n e.mov(e.ah, e.al);\n e.add(e.ah, 1);\n e.vmovd(e.xmm0, e.eax);\n e.vpshufb(e.xmm0, i.src1, e.xmm0);\n e.vmovd(i.dest.reg().cvt32(), e.xmm0);\n e.and_(i.dest.reg().cvt32(), 0xFFFFu);\n }\n }\n};\nstruct EXTRACT_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n static const vec128_t extract_table_32[4] = {\n vec128b(3, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0),\n vec128b(7, 6, 5, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0),\n vec128b(11, 10, 9, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0),\n vec128b(15, 14, 13, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0),\n };\n if (i.src2.is_constant) {\n // TODO(gibbed): add support to constant propagation pass for\n // OPCODE_EXTRACT.\n Xmm src1;\n if (i.src1.is_constant) {\n src1 = e.xmm0;\n e.LoadConstantXmm(src1, i.src1.constant());\n } else {\n src1 = i.src1;\n }\n if (i.src2.constant() == 0) {\n e.vmovd(i.dest, src1);\n } else {\n e.vpextrd(i.dest, src1, VEC128_D(i.src2.constant()));\n }\n } else {\n // TODO(benvanik): try out hlide's version:\n // e.mov(e.eax, 3);\n // e.and_(e.al, i.src2); // eax = [(i&3), 0, 0, 0]\n // e.imul(e.eax, 0x04040404); // [(i&3)*4, (i&3)*4, (i&3)*4, (i&3)*4]\n // e.add(e.eax, 0x00010203); // [((i&3)*4)+3, ((i&3)*4)+2, ((i&3)*4)+1,\n // ((i&3)*4)+0]\n // e.vmovd(e.xmm0, e.eax);\n // e.vpshufb(e.xmm0, i.src1, e.xmm0);\n // e.vmovd(i.dest.reg().cvt32(), e.xmm0);\n // Get the desired word in xmm0, then extract that.\n Xmm src1;\n if (i.src1.is_constant) {\n src1 = e.xmm1;\n e.LoadConstantXmm(src1, i.src1.constant());\n } else {\n src1 = i.src1.reg();\n }\n\n e.xor_(e.rax, e.rax);\n e.mov(e.al, i.src2);\n e.and_(e.al, 0x03);\n e.shl(e.al, 4);\n e.mov(e.rdx, reinterpret_cast(extract_table_32));\n e.vmovaps(e.xmm0, e.ptr[e.rdx + e.rax]);\n e.vpshufb(e.xmm0, src1, e.xmm0);\n e.vpextrd(i.dest, e.xmm0, 0);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_EXTRACT, EXTRACT_I8, EXTRACT_I16, EXTRACT_I32);\n\n// ============================================================================\n// OPCODE_SPLAT\n// ============================================================================\n// Copy a value into all elements of a vector\nstruct SPLAT_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.src1.is_constant) {\n if (e.IsFeatureEnabled(kX64EmitGFNI)) {\n e.pxor(e.xmm0, e.xmm0);\n e.gf2p8affineqb(i.dest, e.xmm0, i.src1.constant());\n return;\n }\n e.mov(e.eax, i.src1.constant());\n e.vmovd(e.xmm0, e.eax);\n } else {\n e.vmovd(e.xmm0, i.src1.reg().cvt32());\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n e.vpbroadcastb(i.dest, e.xmm0);\n } else {\n e.vpunpcklbw(e.xmm0, e.xmm0);\n e.vpunpcklwd(e.xmm0, e.xmm0);\n e.vpshufd(i.dest, e.xmm0, 0);\n }\n }\n};\nstruct SPLAT_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.src1.is_constant) {\n // TODO(benvanik): faster constant splats.\n e.mov(e.eax, i.src1.constant());\n e.vmovd(e.xmm0, e.eax);\n } else {\n e.vmovd(e.xmm0, i.src1.reg().cvt32());\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n e.vpbroadcastw(i.dest, e.xmm0);\n } else {\n e.vpunpcklwd(e.xmm0, e.xmm0); // unpack low word data\n e.vpshufd(i.dest, e.xmm0, 0);\n }\n }\n};\nstruct SPLAT_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.src1.is_constant) {\n // TODO(benvanik): faster constant splats.\n e.mov(e.eax, i.src1.constant());\n e.vmovd(e.xmm0, e.eax);\n } else {\n e.vmovd(e.xmm0, i.src1);\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n e.vpbroadcastd(i.dest, e.xmm0);\n } else {\n e.vpshufd(i.dest, e.xmm0, 0);\n }\n }\n};\nstruct SPLAT_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n if (i.src1.is_constant) {\n // TODO(benvanik): faster constant splats.\n e.mov(e.eax, i.src1.value->constant.i32);\n e.vmovd(e.xmm0, e.eax);\n e.vbroadcastss(i.dest, e.xmm0);\n } else {\n e.vbroadcastss(i.dest, i.src1);\n }\n } else {\n if (i.src1.is_constant) {\n e.mov(e.eax, i.src1.value->constant.i32);\n e.vmovd(i.dest, e.eax);\n e.vshufps(i.dest, i.dest, i.dest, 0);\n } else {\n e.vshufps(i.dest, i.src1, i.src1, 0);\n }\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SPLAT, SPLAT_I8, SPLAT_I16, SPLAT_I32, SPLAT_F32);\n\n// ============================================================================\n// OPCODE_PERMUTE\n// ============================================================================\nstruct PERMUTE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.instr->flags == INT32_TYPE);\n // Permute words between src2 and src3.\n // TODO(benvanik): check src3 for zero. if 0, we can use pshufb.\n if (i.src1.is_constant) {\n uint32_t control = i.src1.constant();\n // Shuffle things into the right places in dest & xmm0,\n // then we blend them together.\n uint32_t src_control =\n (((control >> 24) & 0x3) << 6) | (((control >> 16) & 0x3) << 4) |\n (((control >> 8) & 0x3) << 2) | (((control >> 0) & 0x3) << 0);\n\n uint32_t blend_control = 0;\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n // Blender for vpblendd\n blend_control =\n (((control >> 26) & 0x1) << 3) | (((control >> 18) & 0x1) << 2) |\n (((control >> 10) & 0x1) << 1) | (((control >> 2) & 0x1) << 0);\n } else {\n // Blender for vpblendw\n blend_control =\n (((control >> 26) & 0x1) << 6) | (((control >> 18) & 0x1) << 4) |\n (((control >> 10) & 0x1) << 2) | (((control >> 2) & 0x1) << 0);\n blend_control |= blend_control << 1;\n }\n\n // TODO(benvanik): if src2/src3 are constants, shuffle now!\n Xmm src2;\n if (i.src2.is_constant) {\n src2 = e.xmm1;\n e.LoadConstantXmm(src2, i.src2.constant());\n } else {\n src2 = i.src2;\n }\n Xmm src3;\n if (i.src3.is_constant) {\n src3 = e.xmm2;\n e.LoadConstantXmm(src3, i.src3.constant());\n } else {\n src3 = i.src3;\n }\n if (i.dest != src3) {\n e.vpshufd(i.dest, src2, src_control);\n e.vpshufd(e.xmm0, src3, src_control);\n } else {\n e.vmovaps(e.xmm0, src3);\n e.vpshufd(i.dest, src2, src_control);\n e.vpshufd(e.xmm0, e.xmm0, src_control);\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n e.vpblendd(i.dest, e.xmm0, blend_control); // $0 = $1 $2\n } else {\n e.vpblendw(i.dest, e.xmm0, blend_control); // $0 = $1 $2\n }\n } else {\n // Permute by non-constant.\n assert_always();\n }\n }\n};\nstruct PERMUTE_V128\n : Sequence> {\n static void EmitByInt8(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): find out how to do this with only one temp register!\n // Permute bytes between src2 and src3.\n // src1 is an array of indices corresponding to positions within src2 and\n // src3.\n if (i.src3.value->IsConstantZero()) {\n // Permuting with src2/zero, so just shuffle/mask.\n if (i.src2.value->IsConstantZero()) {\n // src2 & src3 are zero, so result will always be zero.\n e.vpxor(i.dest, i.dest);\n } else {\n // Control mask needs to be shuffled.\n if (i.src1.is_constant) {\n e.LoadConstantXmm(e.xmm0, i.src1.constant());\n e.vxorps(e.xmm0, e.xmm0, e.GetXmmConstPtr(XMMSwapWordMask));\n } else {\n e.vxorps(e.xmm0, i.src1, e.GetXmmConstPtr(XMMSwapWordMask));\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho | kX64EmitAVX512VBMI)) {\n Xmm table_lo = e.xmm1;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(table_lo, i.src2.constant());\n } else {\n table_lo = i.src2;\n }\n Opmask zeroes = e.k1;\n // _mm_cmple_epu8_mask\n e.vpcmpub(zeroes, e.xmm0, e.GetXmmConstPtr(XMMPermuteControl15), 2);\n e.vpermb(i.dest.reg() | zeroes | e.T_z, e.xmm0, table_lo);\n return;\n }\n\n e.vpand(e.xmm0, e.GetXmmConstPtr(XMMPermuteByteMask));\n\n if (i.src2.is_constant) {\n e.LoadConstantXmm(i.dest, i.src2.constant());\n e.vpshufb(i.dest, i.dest, e.xmm0);\n } else {\n e.vpshufb(i.dest, i.src2, e.xmm0);\n }\n // Build a mask with values in src2 having 0 and values in src3 having\n // 1.\n e.vpcmpgtb(e.xmm0, e.xmm0, e.GetXmmConstPtr(XMMPermuteControl15));\n e.vpandn(i.dest, e.xmm0, i.dest);\n }\n } else {\n // General permute.\n // Control mask needs to be shuffled.\n // TODO(benvanik): do constants here instead of in generated code.\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho | kX64EmitAVX512BW |\n kX64EmitAVX512VBMI)) {\n Xmm table_idx = e.xmm0;\n if (i.src1.is_constant) {\n e.LoadConstantXmm(table_idx, i.src1.constant());\n e.vxorps(table_idx, table_idx, e.GetXmmConstPtr(XMMSwapWordMask));\n } else {\n e.vxorps(table_idx, i.src1, e.GetXmmConstPtr(XMMSwapWordMask));\n }\n\n Xmm table_lo = e.xmm1;\n if (i.src2.value->IsConstantZero()) {\n e.vpxor(table_lo, table_lo);\n } else if (i.src2.is_constant) {\n e.LoadConstantXmm(table_lo, i.src2.constant());\n } else {\n table_lo = i.src2;\n }\n\n Xmm table_hi = e.xmm2;\n if (i.src3.value->IsConstantZero()) {\n e.vpxor(table_hi, table_hi);\n } else if (i.src3.is_constant) {\n e.LoadConstantXmm(table_hi, i.src3.constant());\n } else {\n table_hi = i.src3;\n }\n\n e.vpermi2b(table_idx, table_lo, table_hi);\n e.vmovdqu8(i.dest, table_idx);\n return;\n }\n\n if (i.src1.is_constant) {\n e.LoadConstantXmm(e.xmm2, i.src1.constant());\n e.vxorps(e.xmm2, e.xmm2, e.GetXmmConstPtr(XMMSwapWordMask));\n } else {\n e.vxorps(e.xmm2, i.src1, e.GetXmmConstPtr(XMMSwapWordMask));\n }\n e.vpand(e.xmm2, e.GetXmmConstPtr(XMMPermuteByteMask));\n\n Xmm src2_shuf = e.xmm0;\n if (i.src2.value->IsConstantZero()) {\n e.vpxor(src2_shuf, src2_shuf);\n } else if (i.src2.is_constant) {\n e.LoadConstantXmm(src2_shuf, i.src2.constant());\n e.vpshufb(src2_shuf, src2_shuf, e.xmm2);\n } else {\n e.vpshufb(src2_shuf, i.src2, e.xmm2);\n }\n Xmm src3_shuf = e.xmm1;\n if (i.src3.value->IsConstantZero()) {\n e.vpxor(src3_shuf, src3_shuf);\n } else if (i.src3.is_constant) {\n e.LoadConstantXmm(src3_shuf, i.src3.constant());\n e.vpshufb(src3_shuf, src3_shuf, e.xmm2);\n } else {\n e.vpshufb(src3_shuf, i.src3, e.xmm2);\n }\n // Build a mask with values in src2 having 0 and values in src3 having 1.\n e.vpcmpgtb(i.dest, e.xmm2, e.GetXmmConstPtr(XMMPermuteControl15));\n e.vpblendvb(i.dest, src2_shuf, src3_shuf, i.dest);\n }\n }\n\n static void EmitByInt16(X64Emitter& e, const EmitArgType& i) {\n // src1 is an array of indices corresponding to positions within src2 and\n // src3\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho | kX64EmitAVX512BW)) {\n e.LoadConstantXmm(e.xmm1, vec128s(0x1));\n\n Xmm table_idx = e.xmm0;\n if (i.src1.is_constant) {\n e.LoadConstantXmm(table_idx, i.src1.constant());\n e.vpxord(table_idx, table_idx, e.xmm1);\n } else {\n e.vpxord(table_idx, i.src1, e.xmm1);\n }\n\n Xmm table_lo = e.xmm1;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(table_lo, i.src2.constant());\n } else {\n table_lo = i.src2;\n }\n\n Xmm table_hi = e.xmm2;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(table_hi, i.src3.constant());\n } else {\n table_hi = i.src3;\n }\n\n e.vpermi2w(table_idx, table_lo, table_hi);\n e.vmovdqu8(i.dest, table_idx);\n return;\n }\n\n assert_true(i.src1.is_constant);\n\n vec128_t perm = (i.src1.constant() & vec128s(0xF)) ^ vec128s(0x1);\n vec128_t perm_ctrl = vec128b(0);\n for (int i = 0; i < 8; i++) {\n perm_ctrl.i16[i] = perm.i16[i] > 7 ? -1 : 0;\n\n auto v = uint8_t(perm.u16[i]);\n perm.u8[i * 2] = v * 2;\n perm.u8[i * 2 + 1] = v * 2 + 1;\n }\n e.LoadConstantXmm(e.xmm0, perm);\n\n if (i.src2.is_constant) {\n e.LoadConstantXmm(e.xmm1, i.src2.constant());\n } else {\n e.vmovdqa(e.xmm1, i.src2);\n }\n if (i.src3.is_constant) {\n e.LoadConstantXmm(e.xmm2, i.src3.constant());\n } else {\n e.vmovdqa(e.xmm2, i.src3);\n }\n\n e.vpshufb(e.xmm1, e.xmm1, e.xmm0);\n e.vpshufb(e.xmm2, e.xmm2, e.xmm0);\n\n uint8_t mask = 0;\n for (int i = 0; i < 8; i++) {\n if (perm_ctrl.i16[i] == 0) {\n mask |= 1 << (7 - i);\n }\n }\n e.vpblendw(i.dest, e.xmm1, e.xmm2, mask);\n }\n\n static void EmitByInt32(X64Emitter& e, const EmitArgType& i) {\n assert_always();\n }\n\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags) {\n case INT8_TYPE:\n EmitByInt8(e, i);\n break;\n case INT16_TYPE:\n EmitByInt16(e, i);\n break;\n case INT32_TYPE:\n EmitByInt32(e, i);\n break;\n default:\n assert_unhandled_case(i.instr->flags);\n return;\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_PERMUTE, PERMUTE_I32, PERMUTE_V128);\n\n// ============================================================================\n// OPCODE_SWIZZLE\n// ============================================================================\nstruct SWIZZLE\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n auto element_type = i.instr->flags;\n if (element_type == INT8_TYPE) {\n assert_always();\n } else if (element_type == INT16_TYPE) {\n assert_always();\n } else if (element_type == INT32_TYPE || element_type == FLOAT32_TYPE) {\n uint8_t swizzle_mask = static_cast(i.src2.value);\n Xmm src1;\n if (i.src1.is_constant) {\n src1 = e.xmm0;\n e.LoadConstantXmm(src1, i.src1.constant());\n } else {\n src1 = i.src1;\n }\n e.vpshufd(i.dest, src1, swizzle_mask);\n } else if (element_type == INT64_TYPE || element_type == FLOAT64_TYPE) {\n assert_always();\n } else {\n assert_always();\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SWIZZLE, SWIZZLE);\n\n// ============================================================================\n// OPCODE_PACK\n// ============================================================================\nstruct PACK : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags & PACK_TYPE_MODE) {\n case PACK_TYPE_D3DCOLOR:\n EmitD3DCOLOR(e, i);\n break;\n case PACK_TYPE_FLOAT16_2:\n EmitFLOAT16_2(e, i);\n break;\n case PACK_TYPE_FLOAT16_4:\n EmitFLOAT16_4(e, i);\n break;\n case PACK_TYPE_SHORT_2:\n EmitSHORT_2(e, i);\n break;\n case PACK_TYPE_SHORT_4:\n EmitSHORT_4(e, i);\n break;\n case PACK_TYPE_UINT_2101010:\n EmitUINT_2101010(e, i);\n break;\n case PACK_TYPE_ULONG_4202020:\n EmitULONG_4202020(e, i);\n break;\n case PACK_TYPE_8_IN_16:\n Emit8_IN_16(e, i, i.instr->flags);\n break;\n case PACK_TYPE_16_IN_32:\n Emit16_IN_32(e, i, i.instr->flags);\n break;\n default:\n assert_unhandled_case(i.instr->flags);\n break;\n }\n }\n static void EmitD3DCOLOR(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->IsConstantZero());\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Saturate to [3,3....] so that only values between 3...[00] and 3...[FF]\n // are valid - max before min to pack NaN as zero (5454082B is heavily\n // affected by the order - packs 0xFFFFFFFF in matrix code to get a 0\n // constant).\n e.vmaxps(i.dest, src, e.GetXmmConstPtr(XMM3333));\n e.vminps(i.dest, i.dest, e.GetXmmConstPtr(XMMPackD3DCOLORSat));\n // Extract bytes.\n // RGBA (XYZW) -> ARGB (WXYZ)\n // w = ((src1.uw & 0xFF) << 24) | ((src1.ux & 0xFF) << 16) |\n // ((src1.uy & 0xFF) << 8) | (src1.uz & 0xFF)\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMPackD3DCOLOR));\n }\n static __m128i EmulateFLOAT16_2(void*, __m128 src1) {\n alignas(16) float a[4];\n alignas(16) uint16_t b[8];\n _mm_store_ps(a, src1);\n std::memset(b, 0, sizeof(b));\n\n for (int i = 0; i < 2; i++) {\n b[7 - i] = half_float::detail::float2half(a[i]);\n }\n\n return _mm_load_si128(reinterpret_cast<__m128i*>(b));\n }\n static void EmitFLOAT16_2(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->IsConstantZero());\n // http://blogs.msdn.com/b/chuckw/archive/2012/09/11/directxmath-f16c-and-fma.aspx\n // dest = [(src1.x | src1.y), 0, 0, 0]\n\n if (e.IsFeatureEnabled(kX64EmitF16C)) {\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // 0|0|0|0|W|Z|Y|X\n e.vcvtps2ph(i.dest, src, 0b00000011);\n // Shuffle to X|Y|0|0|0|0|0|0\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMPackFLOAT16_2));\n } else {\n if (i.src1.is_constant) {\n e.lea(e.GetNativeParam(0), e.StashConstantXmm(0, i.src1.constant()));\n } else {\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n }\n e.CallNativeSafe(reinterpret_cast(EmulateFLOAT16_2));\n e.vmovaps(i.dest, e.xmm0);\n }\n }\n static __m128i EmulateFLOAT16_4(void*, __m128 src1) {\n alignas(16) float a[4];\n alignas(16) uint16_t b[8];\n _mm_store_ps(a, src1);\n std::memset(b, 0, sizeof(b));\n\n for (int i = 0; i < 4; i++) {\n b[7 - (i ^ 2)] =\n half_float::detail::float2half(a[i]);\n }\n\n return _mm_load_si128(reinterpret_cast<__m128i*>(b));\n }\n static void EmitFLOAT16_4(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->IsConstantZero());\n // dest = [(src1.z | src1.w), (src1.x | src1.y), 0, 0]\n\n if (e.IsFeatureEnabled(kX64EmitF16C)) {\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // 0|0|0|0|W|Z|Y|X\n e.vcvtps2ph(i.dest, src, 0b00000011);\n // Shuffle to Z|W|X|Y|0|0|0|0\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMPackFLOAT16_4));\n } else {\n if (i.src1.is_constant) {\n e.lea(e.GetNativeParam(0), e.StashConstantXmm(0, i.src1.constant()));\n } else {\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n }\n e.CallNativeSafe(reinterpret_cast(EmulateFLOAT16_4));\n e.vmovaps(i.dest, e.xmm0);\n }\n }\n static void EmitSHORT_2(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->IsConstantZero());\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Saturate.\n e.vmaxps(i.dest, src, e.GetXmmConstPtr(XMMPackSHORT_Min));\n e.vminps(i.dest, i.dest, e.GetXmmConstPtr(XMMPackSHORT_Max));\n // Pack.\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMPackSHORT_2));\n }\n static void EmitSHORT_4(X64Emitter& e, const EmitArgType& i) {\n assert_true(i.src2.value->IsConstantZero());\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Saturate.\n e.vmaxps(i.dest, src, e.GetXmmConstPtr(XMMPackSHORT_Min));\n e.vminps(i.dest, i.dest, e.GetXmmConstPtr(XMMPackSHORT_Max));\n // Pack.\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMPackSHORT_4));\n }\n static void EmitUINT_2101010(X64Emitter& e, const EmitArgType& i) {\n // https://www.opengl.org/registry/specs/ARB/vertex_type_2_10_10_10_rev.txt\n // XYZ are 10 bits, signed and saturated.\n // W is 2 bits, unsigned and saturated.\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Saturate.\n e.vmaxps(i.dest, src, e.GetXmmConstPtr(XMMPackUINT_2101010_MinUnpacked));\n e.vminps(i.dest, i.dest, e.GetXmmConstPtr(XMMPackUINT_2101010_MaxUnpacked));\n // Remove the unneeded bits of the floats.\n e.vpand(i.dest, e.GetXmmConstPtr(XMMPackUINT_2101010_MaskUnpacked));\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n // Shift the components up.\n e.vpsllvd(i.dest, i.dest, e.GetXmmConstPtr(XMMPackUINT_2101010_Shift));\n } else {\n // Duplicate all the components into bits 10-19.\n e.vpslld(e.xmm0, i.dest, 10);\n e.vpor(i.dest, e.xmm0);\n // Duplicate all the components into bits 20-39\n // (so alpha will be in 30-31).\n e.vpslld(e.xmm0, i.dest, 20);\n e.vpor(i.dest, e.xmm0);\n // Leave only the needed components.\n e.vpand(i.dest, e.GetXmmConstPtr(XMMPackUINT_2101010_MaskPacked));\n }\n // Combine the components.\n e.vshufps(e.xmm0, i.dest, i.dest, _MM_SHUFFLE(2, 3, 0, 1));\n e.vorps(i.dest, e.xmm0);\n e.vshufps(e.xmm0, i.dest, i.dest, _MM_SHUFFLE(1, 0, 3, 2));\n e.vorps(i.dest, e.xmm0);\n }\n static void EmitULONG_4202020(X64Emitter& e, const EmitArgType& i) {\n // XYZ are 20 bits, signed and saturated.\n // W is 4 bits, unsigned and saturated.\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Saturate.\n e.vmaxps(i.dest, src, e.GetXmmConstPtr(XMMPackULONG_4202020_MinUnpacked));\n e.vminps(i.dest, i.dest,\n e.GetXmmConstPtr(XMMPackULONG_4202020_MaxUnpacked));\n // Remove the unneeded bits of the floats (so excess nibbles will also be\n // cleared).\n e.vpand(i.dest, e.GetXmmConstPtr(XMMPackULONG_4202020_MaskUnpacked));\n // Store Y and W shifted left by 4 so vpshufb can be used with them.\n e.vpslld(e.xmm0, i.dest, 4);\n // Place XZ where they're supposed to be.\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMPackULONG_4202020_PermuteXZ));\n // Place YW.\n e.vpshufb(e.xmm0, e.xmm0, e.GetXmmConstPtr(XMMPackULONG_4202020_PermuteYW));\n // Merge XZ and YW.\n e.vorps(i.dest, e.xmm0);\n }\n static __m128i EmulatePack8_IN_16_UN_UN_SAT(void*, __m128i src1,\n __m128i src2) {\n alignas(16) uint16_t a[8];\n alignas(16) uint16_t b[8];\n alignas(16) uint8_t c[16];\n _mm_store_si128(reinterpret_cast<__m128i*>(a), src1);\n _mm_store_si128(reinterpret_cast<__m128i*>(b), src2);\n for (int i = 0; i < 8; ++i) {\n c[i] = uint8_t(std::max(uint16_t(0), std::min(uint16_t(255), a[i])));\n c[i + 8] = uint8_t(std::max(uint16_t(0), std::min(uint16_t(255), b[i])));\n }\n return _mm_load_si128(reinterpret_cast<__m128i*>(c));\n }\n static __m128i EmulatePack8_IN_16_UN_UN(void*, __m128i src1, __m128i src2) {\n alignas(16) uint8_t a[16];\n alignas(16) uint8_t b[16];\n alignas(16) uint8_t c[16];\n _mm_store_si128(reinterpret_cast<__m128i*>(a), src1);\n _mm_store_si128(reinterpret_cast<__m128i*>(b), src2);\n for (int i = 0; i < 8; ++i) {\n c[i] = a[i * 2];\n c[i + 8] = b[i * 2];\n }\n return _mm_load_si128(reinterpret_cast<__m128i*>(c));\n }\n static void Emit8_IN_16(X64Emitter& e, const EmitArgType& i, uint32_t flags) {\n // TODO(benvanik): handle src2 (or src1) being constant zero\n if (IsPackInUnsigned(flags)) {\n if (IsPackOutUnsigned(flags)) {\n if (IsPackOutSaturate(flags)) {\n // unsigned -> unsigned + saturate\n if (i.src2.is_constant) {\n e.lea(e.GetNativeParam(1),\n e.StashConstantXmm(1, i.src2.constant()));\n } else {\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(\n reinterpret_cast(EmulatePack8_IN_16_UN_UN_SAT));\n e.vmovaps(i.dest, e.xmm0);\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMByteOrderMask));\n } else {\n // unsigned -> unsigned\n e.lea(e.GetNativeParam(1), e.StashXmm(1, i.src2));\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulatePack8_IN_16_UN_UN));\n e.vmovaps(i.dest, e.xmm0);\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMByteOrderMask));\n }\n } else {\n if (IsPackOutSaturate(flags)) {\n // unsigned -> signed + saturate\n assert_always();\n } else {\n // unsigned -> signed\n assert_always();\n }\n }\n } else {\n if (IsPackOutUnsigned(flags)) {\n if (IsPackOutSaturate(flags)) {\n // signed -> unsigned + saturate\n // PACKUSWB / SaturateSignedWordToUnsignedByte\n Xbyak::Xmm src2 = i.src2.is_constant ? e.xmm0 : i.src2;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(src2, i.src2.constant());\n }\n\n e.vpackuswb(i.dest, i.src1, src2);\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMByteOrderMask));\n } else {\n // signed -> unsigned\n assert_always();\n }\n } else {\n if (IsPackOutSaturate(flags)) {\n // signed -> signed + saturate\n // PACKSSWB / SaturateSignedWordToSignedByte\n e.vpacksswb(i.dest, i.src1, i.src2);\n e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMByteOrderMask));\n } else {\n // signed -> signed\n assert_always();\n }\n }\n }\n }\n // Pack 2 32-bit vectors into a 16-bit vector.\n static void Emit16_IN_32(X64Emitter& e, const EmitArgType& i,\n uint32_t flags) {\n // TODO(benvanik): handle src2 (or src1) being constant zero\n if (IsPackInUnsigned(flags)) {\n if (IsPackOutUnsigned(flags)) {\n if (IsPackOutSaturate(flags)) {\n // unsigned -> unsigned + saturate\n // Construct a saturation max value\n e.mov(e.eax, 0xFFFFu);\n e.vmovd(e.xmm0, e.eax);\n e.vpshufd(e.xmm0, e.xmm0, 0b00000000);\n\n if (!i.src1.is_constant) {\n e.vpminud(e.xmm1, i.src1, e.xmm0); // Saturate src1\n e.vpshuflw(e.xmm1, e.xmm1, 0b00100010);\n e.vpshufhw(e.xmm1, e.xmm1, 0b00100010);\n e.vpshufd(e.xmm1, e.xmm1, 0b00001000);\n } else {\n // TODO(DrChat): Non-zero constants\n assert_true(i.src1.constant().u64[0] == 0 &&\n i.src1.constant().u64[1] == 0);\n e.vpxor(e.xmm1, e.xmm1);\n }\n\n if (!i.src2.is_constant) {\n e.vpminud(i.dest, i.src2, e.xmm0); // Saturate src2\n e.vpshuflw(i.dest, i.dest, 0b00100010);\n e.vpshufhw(i.dest, i.dest, 0b00100010);\n e.vpshufd(i.dest, i.dest, 0b10000000);\n } else {\n // TODO(DrChat): Non-zero constants\n assert_true(i.src2.constant().u64[0] == 0 &&\n i.src2.constant().u64[1] == 0);\n e.vpxor(i.dest, i.dest);\n }\n\n e.vpblendw(i.dest, i.dest, e.xmm1, 0b00001111);\n } else {\n // unsigned -> unsigned\n e.vmovaps(e.xmm0, i.src1);\n e.vpshuflw(e.xmm0, e.xmm0, 0b00100010);\n e.vpshufhw(e.xmm0, e.xmm0, 0b00100010);\n e.vpshufd(e.xmm0, e.xmm0, 0b00001000);\n\n e.vmovaps(i.dest, i.src2);\n e.vpshuflw(i.dest, i.dest, 0b00100010);\n e.vpshufhw(i.dest, i.dest, 0b00100010);\n e.vpshufd(i.dest, i.dest, 0b10000000);\n\n e.vpblendw(i.dest, i.dest, e.xmm0, 0b00001111);\n }\n } else {\n if (IsPackOutSaturate(flags)) {\n // unsigned -> signed + saturate\n assert_always();\n } else {\n // unsigned -> signed\n assert_always();\n }\n }\n } else {\n if (IsPackOutUnsigned(flags)) {\n if (IsPackOutSaturate(flags)) {\n // signed -> unsigned + saturate\n // PACKUSDW\n // TMP[15:0] <- (DEST[31:0] < 0) ? 0 : DEST[15:0];\n // DEST[15:0] <- (DEST[31:0] > FFFFH) ? FFFFH : TMP[15:0];\n e.vpackusdw(i.dest, i.src1, i.src2);\n e.vpshuflw(i.dest, i.dest, 0b10110001);\n e.vpshufhw(i.dest, i.dest, 0b10110001);\n } else {\n // signed -> unsigned\n assert_always();\n }\n } else {\n if (IsPackOutSaturate(flags)) {\n // signed -> signed + saturate\n // PACKSSDW / SaturateSignedDwordToSignedWord\n Xmm src2;\n if (!i.src2.is_constant) {\n src2 = i.src2;\n } else {\n assert_false(i.src1 == e.xmm0);\n src2 = e.xmm0;\n e.LoadConstantXmm(src2, i.src2.constant());\n }\n e.vpackssdw(i.dest, i.src1, src2);\n e.vpshuflw(i.dest, i.dest, 0b10110001);\n e.vpshufhw(i.dest, i.dest, 0b10110001);\n } else {\n // signed -> signed\n assert_always();\n }\n }\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_PACK, PACK);\n\n// ============================================================================\n// OPCODE_UNPACK\n// ============================================================================\nstruct UNPACK : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags & PACK_TYPE_MODE) {\n case PACK_TYPE_D3DCOLOR:\n EmitD3DCOLOR(e, i);\n break;\n case PACK_TYPE_FLOAT16_2:\n EmitFLOAT16_2(e, i);\n break;\n case PACK_TYPE_FLOAT16_4:\n EmitFLOAT16_4(e, i);\n break;\n case PACK_TYPE_SHORT_2:\n EmitSHORT_2(e, i);\n break;\n case PACK_TYPE_SHORT_4:\n EmitSHORT_4(e, i);\n break;\n case PACK_TYPE_UINT_2101010:\n EmitUINT_2101010(e, i);\n break;\n case PACK_TYPE_ULONG_4202020:\n EmitULONG_4202020(e, i);\n break;\n case PACK_TYPE_8_IN_16:\n Emit8_IN_16(e, i, i.instr->flags);\n break;\n case PACK_TYPE_16_IN_32:\n Emit16_IN_32(e, i, i.instr->flags);\n break;\n default:\n assert_unhandled_case(i.instr->flags);\n break;\n }\n }\n static void EmitD3DCOLOR(X64Emitter& e, const EmitArgType& i) {\n // ARGB (WXYZ) -> RGBA (XYZW)\n Xmm src;\n if (i.src1.is_constant) {\n if (i.src1.value->IsConstantZero()) {\n e.vmovaps(i.dest, e.GetXmmConstPtr(XMMOne));\n return;\n }\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // src = ZZYYXXWW\n // Unpack to 000000ZZ,000000YY,000000XX,000000WW\n e.vpshufb(i.dest, src, e.GetXmmConstPtr(XMMUnpackD3DCOLOR));\n // Add 1.0f to each.\n e.vpor(i.dest, e.GetXmmConstPtr(XMMOne));\n // To convert to 0 to 1, games multiply by 0x47008081 and add 0xC7008081.\n }\n static __m128 EmulateFLOAT16_2(void*, __m128i src1) {\n alignas(16) uint16_t a[8];\n alignas(16) float b[4];\n _mm_store_si128(reinterpret_cast<__m128i*>(a), src1);\n\n for (int i = 0; i < 2; i++) {\n b[i] = half_float::detail::half2float(a[VEC128_W(6 + i)]);\n }\n\n // Constants, or something\n b[2] = 0.f;\n b[3] = 1.f;\n\n return _mm_load_ps(b);\n }\n static void EmitFLOAT16_2(X64Emitter& e, const EmitArgType& i) {\n // 1 bit sign, 5 bit exponent, 10 bit mantissa\n // D3D10 half float format\n // TODO(benvanik):\n // http://blogs.msdn.com/b/chuckw/archive/2012/09/11/directxmath-f16c-and-fma.aspx\n // Use _mm_cvtph_ps -- requires very modern processors (SSE5+)\n // Unpacking half floats:\n // http://fgiesen.wordpress.com/2012/03/28/half-to-float-done-quic/\n // Packing half floats: https://gist.github.com/rygorous/2156668\n // Load source, move from tight pack of X16Y16.... to X16...Y16...\n // Also zero out the high end.\n // TODO(benvanik): special case constant unpacks that just get 0/1/etc.\n\n if (e.IsFeatureEnabled(kX64EmitF16C)) {\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // sx = src.iw >> 16;\n // sy = src.iw & 0xFFFF;\n // dest = { XMConvertHalfToFloat(sx),\n // XMConvertHalfToFloat(sy),\n // 0.0,\n // 1.0 };\n // Shuffle to 0|0|0|0|0|0|Y|X\n e.vpshufb(i.dest, src, e.GetXmmConstPtr(XMMUnpackFLOAT16_2));\n e.vcvtph2ps(i.dest, i.dest);\n e.vpshufd(i.dest, i.dest, 0b10100100);\n e.vpor(i.dest, e.GetXmmConstPtr(XMM0001));\n } else {\n if (i.src1.is_constant) {\n e.lea(e.GetNativeParam(0), e.StashConstantXmm(0, i.src1.constant()));\n } else {\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n }\n e.CallNativeSafe(reinterpret_cast(EmulateFLOAT16_2));\n e.vmovaps(i.dest, e.xmm0);\n }\n }\n static __m128 EmulateFLOAT16_4(void*, __m128i src1) {\n alignas(16) uint16_t a[8];\n alignas(16) float b[4];\n _mm_store_si128(reinterpret_cast<__m128i*>(a), src1);\n\n for (int i = 0; i < 4; i++) {\n b[i] = half_float::detail::half2float(a[VEC128_W(4 + i)]);\n }\n\n return _mm_load_ps(b);\n }\n static void EmitFLOAT16_4(X64Emitter& e, const EmitArgType& i) {\n // src = [(dest.x | dest.y), (dest.z | dest.w), 0, 0]\n if (e.IsFeatureEnabled(kX64EmitF16C)) {\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Shuffle to 0|0|0|0|W|Z|Y|X\n e.vpshufb(i.dest, src, e.GetXmmConstPtr(XMMUnpackFLOAT16_4));\n e.vcvtph2ps(i.dest, i.dest);\n } else {\n if (i.src1.is_constant) {\n e.lea(e.GetNativeParam(0), e.StashConstantXmm(0, i.src1.constant()));\n } else {\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n }\n e.CallNativeSafe(reinterpret_cast(EmulateFLOAT16_4));\n e.vmovaps(i.dest, e.xmm0);\n }\n }\n static void EmitSHORT_2(X64Emitter& e, const EmitArgType& i) {\n // (VD.x) = 3.0 + (VB.x>>16)*2^-22\n // (VD.y) = 3.0 + (VB.x)*2^-22\n // (VD.z) = 0.0\n // (VD.w) = 1.0 (games splat W after unpacking to get vectors of 1.0f)\n // src is (xx,xx,xx,VALUE)\n Xmm src;\n if (i.src1.is_constant) {\n if (i.src1.value->IsConstantZero()) {\n e.vmovdqa(i.dest, e.GetXmmConstPtr(XMM3301));\n return;\n }\n // TODO(benvanik): check other common constants/perform shuffle/or here.\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Shuffle bytes.\n e.vpshufb(i.dest, src, e.GetXmmConstPtr(XMMUnpackSHORT_2));\n // If negative, make smaller than 3 - sign extend before adding.\n e.vpslld(i.dest, 16);\n e.vpsrad(i.dest, 16);\n // Add 3,3,0,1.\n e.vpaddd(i.dest, e.GetXmmConstPtr(XMM3301));\n // Return quiet NaNs in case of negative overflow.\n e.vcmpeqps(e.xmm0, i.dest, e.GetXmmConstPtr(XMMUnpackSHORT_Overflow));\n e.vblendvps(i.dest, i.dest, e.GetXmmConstPtr(XMMQNaN), e.xmm0);\n }\n static void EmitSHORT_4(X64Emitter& e, const EmitArgType& i) {\n // (VD.x) = 3.0 + (VB.x>>16)*2^-22\n // (VD.y) = 3.0 + (VB.x)*2^-22\n // (VD.z) = 3.0 + (VB.y>>16)*2^-22\n // (VD.w) = 3.0 + (VB.y)*2^-22\n // src is (xx,xx,VALUE,VALUE)\n Xmm src;\n if (i.src1.is_constant) {\n if (i.src1.value->IsConstantZero()) {\n e.vmovdqa(i.dest, e.GetXmmConstPtr(XMM3333));\n return;\n }\n // TODO(benvanik): check other common constants/perform shuffle/or here.\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Shuffle bytes.\n e.vpshufb(i.dest, src, e.GetXmmConstPtr(XMMUnpackSHORT_4));\n // If negative, make smaller than 3 - sign extend before adding.\n e.vpslld(i.dest, 16);\n e.vpsrad(i.dest, 16);\n // Add 3,3,3,3.\n e.vpaddd(i.dest, e.GetXmmConstPtr(XMM3333));\n // Return quiet NaNs in case of negative overflow.\n e.vcmpeqps(e.xmm0, i.dest, e.GetXmmConstPtr(XMMUnpackSHORT_Overflow));\n e.vblendvps(i.dest, i.dest, e.GetXmmConstPtr(XMMQNaN), e.xmm0);\n }\n static void EmitUINT_2101010(X64Emitter& e, const EmitArgType& i) {\n Xmm src;\n if (i.src1.is_constant) {\n if (i.src1.value->IsConstantZero()) {\n e.vmovdqa(i.dest, e.GetXmmConstPtr(XMM3331));\n return;\n }\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Splat W.\n e.vshufps(i.dest, src, src, _MM_SHUFFLE(3, 3, 3, 3));\n // Keep only the needed components.\n // Red in 0-9 now, green in 10-19, blue in 20-29, alpha in 30-31.\n e.vpand(i.dest, e.GetXmmConstPtr(XMMPackUINT_2101010_MaskPacked));\n if (e.IsFeatureEnabled(kX64EmitAVX2)) {\n // Shift the components down.\n e.vpsrlvd(i.dest, i.dest, e.GetXmmConstPtr(XMMPackUINT_2101010_Shift));\n } else {\n // Duplicate green in 0-9 and alpha in 20-21.\n e.vpsrld(e.xmm0, i.dest, 10);\n e.vpor(i.dest, e.xmm0);\n // Duplicate blue in 0-9 and alpha in 0-1.\n e.vpsrld(e.xmm0, i.dest, 20);\n e.vpor(i.dest, e.xmm0);\n // Remove higher duplicate components.\n e.vpand(i.dest, e.GetXmmConstPtr(XMMPackUINT_2101010_MaskUnpacked));\n }\n // If XYZ are negative, make smaller than 3 - sign extend XYZ before adding.\n // W is unsigned.\n e.vpslld(i.dest, 22);\n e.vpsrad(i.dest, 22);\n // Add 3,3,3,1.\n e.vpaddd(i.dest, e.GetXmmConstPtr(XMM3331));\n // Return quiet NaNs in case of negative overflow.\n e.vcmpeqps(e.xmm0, i.dest,\n e.GetXmmConstPtr(XMMUnpackUINT_2101010_Overflow));\n e.vblendvps(i.dest, i.dest, e.GetXmmConstPtr(XMMQNaN), e.xmm0);\n // To convert XYZ to -1 to 1, games multiply by 0x46004020 & sub 0x46C06030.\n // For W to 0 to 1, they multiply by and subtract 0x4A2AAAAB.\n }\n static void EmitULONG_4202020(X64Emitter& e, const EmitArgType& i) {\n Xmm src;\n if (i.src1.is_constant) {\n if (i.src1.value->IsConstantZero()) {\n e.vmovdqa(i.dest, e.GetXmmConstPtr(XMM3331));\n return;\n }\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n // Extract pairs of nibbles to XZYW. XZ will have excess 4 upper bits, YW\n // will have excess 4 lower bits.\n e.vpshufb(i.dest, src, e.GetXmmConstPtr(XMMUnpackULONG_4202020_Permute));\n // Drop the excess nibble of YW.\n e.vpsrld(e.xmm0, i.dest, 4);\n // Merge XZ and YW now both starting at offset 0.\n e.vshufps(i.dest, i.dest, e.xmm0, _MM_SHUFFLE(3, 2, 1, 0));\n // Reorder as XYZW.\n e.vshufps(i.dest, i.dest, _MM_SHUFFLE(3, 1, 2, 0));\n // Drop the excess upper nibble in XZ and sign-extend XYZ.\n e.vpslld(i.dest, 12);\n e.vpsrad(i.dest, 12);\n // Add 3,3,3,1.\n e.vpaddd(i.dest, e.GetXmmConstPtr(XMM3331));\n // Return quiet NaNs in case of negative overflow.\n e.vcmpeqps(e.xmm0, i.dest,\n e.GetXmmConstPtr(XMMUnpackULONG_4202020_Overflow));\n e.vblendvps(i.dest, i.dest, e.GetXmmConstPtr(XMMQNaN), e.xmm0);\n }\n static void Emit8_IN_16(X64Emitter& e, const EmitArgType& i, uint32_t flags) {\n assert_false(IsPackOutSaturate(flags));\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n if (IsPackToLo(flags)) {\n // Unpack to LO.\n if (IsPackInUnsigned(flags)) {\n if (IsPackOutUnsigned(flags)) {\n // unsigned -> unsigned\n assert_always();\n } else {\n // unsigned -> signed\n assert_always();\n }\n } else {\n if (IsPackOutUnsigned(flags)) {\n // signed -> unsigned\n assert_always();\n } else {\n // signed -> signed\n e.vpshufb(i.dest, src, e.GetXmmConstPtr(XMMByteOrderMask));\n e.vpunpckhbw(i.dest, i.dest, i.dest);\n e.vpsraw(i.dest, 8);\n }\n }\n } else {\n // Unpack to HI.\n if (IsPackInUnsigned(flags)) {\n if (IsPackOutUnsigned(flags)) {\n // unsigned -> unsigned\n assert_always();\n } else {\n // unsigned -> signed\n assert_always();\n }\n } else {\n if (IsPackOutUnsigned(flags)) {\n // signed -> unsigned\n assert_always();\n } else {\n // signed -> signed\n e.vpshufb(i.dest, src, e.GetXmmConstPtr(XMMByteOrderMask));\n e.vpunpcklbw(i.dest, i.dest, i.dest);\n e.vpsraw(i.dest, 8);\n }\n }\n }\n }\n static void Emit16_IN_32(X64Emitter& e, const EmitArgType& i,\n uint32_t flags) {\n assert_false(IsPackOutSaturate(flags));\n Xmm src;\n if (i.src1.is_constant) {\n src = i.dest;\n e.LoadConstantXmm(src, i.src1.constant());\n } else {\n src = i.src1;\n }\n if (IsPackToLo(flags)) {\n // Unpack to LO.\n if (IsPackInUnsigned(flags)) {\n if (IsPackOutUnsigned(flags)) {\n // unsigned -> unsigned\n assert_always();\n } else {\n // unsigned -> signed\n assert_always();\n }\n } else {\n if (IsPackOutUnsigned(flags)) {\n // signed -> unsigned\n assert_always();\n } else {\n // signed -> signed\n e.vpunpckhwd(i.dest, src, src);\n e.vpsrad(i.dest, 16);\n }\n }\n } else {\n // Unpack to HI.\n if (IsPackInUnsigned(flags)) {\n if (IsPackOutUnsigned(flags)) {\n // unsigned -> unsigned\n assert_always();\n } else {\n // unsigned -> signed\n assert_always();\n }\n } else {\n if (IsPackOutUnsigned(flags)) {\n // signed -> unsigned\n assert_always();\n } else {\n // signed -> signed\n e.vpunpcklwd(i.dest, src, src);\n e.vpsrad(i.dest, 16);\n }\n }\n }\n e.vpshufd(i.dest, i.dest, 0xB1);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_UNPACK, UNPACK);\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_sequences.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n// A note about vectors:\n// Xenia represents vectors as xyzw pairs, with indices 0123.\n// XMM registers are xyzw pairs with indices 3210, making them more like wzyx.\n// This makes things somewhat confusing. It'd be nice to just shuffle the\n// registers around on load/store, however certain operations require that\n// data be in the right offset.\n// Basically, this identity must hold:\n// shuffle(vec, b00011011) -> {x,y,z,w} => {x,y,z,w}\n// All indices and operations must respect that.\n//\n// Memory (big endian):\n// [00 01 02 03] [04 05 06 07] [08 09 0A 0B] [0C 0D 0E 0F] (x, y, z, w)\n// load into xmm register:\n// [0F 0E 0D 0C] [0B 0A 09 08] [07 06 05 04] [03 02 01 00] (w, z, y, x)\n\n#include \"xenia/cpu/backend/x64/x64_sequences.h\"\n\n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/clock.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/string.h\"\n#include \"xenia/base/threading.h\"\n#include \"xenia/cpu/backend/x64/x64_emitter.h\"\n#include \"xenia/cpu/backend/x64/x64_op.h\"\n#include \"xenia/cpu/backend/x64/x64_tracers.h\"\n#include \"xenia/cpu/backend/x64/x64_util.h\"\n#include \"xenia/cpu/hir/hir_builder.h\"\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nusing namespace Xbyak;\n\n// TODO(benvanik): direct usings.\nusing namespace xe::cpu;\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::Instr;\n\ntypedef bool (*SequenceSelectFn)(X64Emitter&, const Instr*);\nstd::unordered_map sequence_table;\n\n// ============================================================================\n// OPCODE_COMMENT\n// ============================================================================\nstruct COMMENT : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (IsTracingInstr()) {\n auto str = reinterpret_cast(i.src1.value);\n // TODO(benvanik): pass through.\n // TODO(benvanik): don't just leak this memory.\n auto str_copy = xe_strdup(str);\n e.mov(e.rdx, reinterpret_cast(str_copy));\n e.CallNative(reinterpret_cast(TraceString));\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_COMMENT, COMMENT);\n\n// ============================================================================\n// OPCODE_NOP\n// ============================================================================\nstruct NOP : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) { e.nop(); }\n};\nEMITTER_OPCODE_TABLE(OPCODE_NOP, NOP);\n\n// ============================================================================\n// OPCODE_SOURCE_OFFSET\n// ============================================================================\nstruct SOURCE_OFFSET\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.MarkSourceOffset(i.instr);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SOURCE_OFFSET, SOURCE_OFFSET);\n\n// ============================================================================\n// OPCODE_ASSIGN\n// ============================================================================\nstruct ASSIGN_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, i.src1);\n }\n};\nstruct ASSIGN_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, i.src1);\n }\n};\nstruct ASSIGN_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, i.src1);\n }\n};\nstruct ASSIGN_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, i.src1);\n }\n};\nstruct ASSIGN_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovaps(i.dest, i.src1);\n }\n};\nstruct ASSIGN_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovaps(i.dest, i.src1);\n }\n};\nstruct ASSIGN_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovaps(i.dest, i.src1);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ASSIGN, ASSIGN_I8, ASSIGN_I16, ASSIGN_I32,\n ASSIGN_I64, ASSIGN_F32, ASSIGN_F64, ASSIGN_V128);\n\n// ============================================================================\n// OPCODE_CAST\n// ============================================================================\nstruct CAST_I32_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovd(i.dest, i.src1);\n }\n};\nstruct CAST_I64_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovq(i.dest, i.src1);\n }\n};\nstruct CAST_F32_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovd(i.dest, i.src1);\n }\n};\nstruct CAST_F64_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vmovq(i.dest, i.src1);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CAST, CAST_I32_F32, CAST_I64_F64, CAST_F32_I32,\n CAST_F64_I64);\n\n// ============================================================================\n// OPCODE_ZERO_EXTEND\n// ============================================================================\nstruct ZERO_EXTEND_I16_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest, i.src1);\n }\n};\nstruct ZERO_EXTEND_I32_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest, i.src1);\n }\n};\nstruct ZERO_EXTEND_I64_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest, i.src1);\n }\n};\nstruct ZERO_EXTEND_I32_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest, i.src1);\n }\n};\nstruct ZERO_EXTEND_I64_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest, i.src1);\n }\n};\nstruct ZERO_EXTEND_I64_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest.reg().cvt32(), i.src1);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ZERO_EXTEND, ZERO_EXTEND_I16_I8, ZERO_EXTEND_I32_I8,\n ZERO_EXTEND_I64_I8, ZERO_EXTEND_I32_I16,\n ZERO_EXTEND_I64_I16, ZERO_EXTEND_I64_I32);\n\n// ============================================================================\n// OPCODE_SIGN_EXTEND\n// ============================================================================\nstruct SIGN_EXTEND_I16_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movsx(i.dest, i.src1);\n }\n};\nstruct SIGN_EXTEND_I32_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movsx(i.dest, i.src1);\n }\n};\nstruct SIGN_EXTEND_I64_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movsx(i.dest, i.src1);\n }\n};\nstruct SIGN_EXTEND_I32_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movsx(i.dest, i.src1);\n }\n};\nstruct SIGN_EXTEND_I64_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movsx(i.dest, i.src1);\n }\n};\nstruct SIGN_EXTEND_I64_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movsxd(i.dest, i.src1);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SIGN_EXTEND, SIGN_EXTEND_I16_I8, SIGN_EXTEND_I32_I8,\n SIGN_EXTEND_I64_I8, SIGN_EXTEND_I32_I16,\n SIGN_EXTEND_I64_I16, SIGN_EXTEND_I64_I32);\n\n// ============================================================================\n// OPCODE_TRUNCATE\n// ============================================================================\nstruct TRUNCATE_I8_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest.reg().cvt32(), i.src1.reg().cvt8());\n }\n};\nstruct TRUNCATE_I8_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest.reg().cvt32(), i.src1.reg().cvt8());\n }\n};\nstruct TRUNCATE_I8_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest.reg().cvt32(), i.src1.reg().cvt8());\n }\n};\nstruct TRUNCATE_I16_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest.reg().cvt32(), i.src1.reg().cvt16());\n }\n};\nstruct TRUNCATE_I16_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.movzx(i.dest.reg().cvt32(), i.src1.reg().cvt16());\n }\n};\nstruct TRUNCATE_I32_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(i.dest, i.src1.reg().cvt32());\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_TRUNCATE, TRUNCATE_I8_I16, TRUNCATE_I8_I32,\n TRUNCATE_I8_I64, TRUNCATE_I16_I32, TRUNCATE_I16_I64,\n TRUNCATE_I32_I64);\n\n// ============================================================================\n// OPCODE_CONVERT\n// ============================================================================\nstruct CONVERT_I32_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): saturation check? cvtt* (trunc?)\n if (i.instr->flags == ROUND_TO_ZERO) {\n e.vcvttss2si(i.dest, i.src1);\n } else {\n e.vcvtss2si(i.dest, i.src1);\n }\n }\n};\nstruct CONVERT_I32_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // Intel returns 0x80000000 if the double value does not fit within an int32\n // PPC saturates the value instead.\n // So, we can clamp the double value to (double)0x7FFFFFFF.\n e.vminsd(e.xmm0, i.src1, e.GetXmmConstPtr(XMMIntMaxPD));\n if (i.instr->flags == ROUND_TO_ZERO) {\n e.vcvttsd2si(i.dest, e.xmm0);\n } else {\n e.vcvtsd2si(i.dest, e.xmm0);\n }\n }\n};\nstruct CONVERT_I64_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // Copy src1.\n e.movq(e.rcx, i.src1);\n\n // TODO(benvanik): saturation check? cvtt* (trunc?)\n if (i.instr->flags == ROUND_TO_ZERO) {\n e.vcvttsd2si(i.dest, i.src1);\n } else {\n e.vcvtsd2si(i.dest, i.src1);\n }\n\n // 0x8000000000000000\n e.mov(e.rax, 0x1);\n e.shl(e.rax, 63);\n\n // Saturate positive overflow\n // TODO(DrChat): Find a shorter equivalent sequence.\n // if (result ind. && src1 >= 0)\n // result = 0x7FFFFFFFFFFFFFFF;\n e.cmp(e.rax, i.dest);\n e.sete(e.al);\n e.movzx(e.rax, e.al);\n e.shr(e.rcx, 63);\n e.xor_(e.rcx, 0x01);\n e.and_(e.rax, e.rcx);\n\n e.sub(i.dest, e.rax);\n }\n};\nstruct CONVERT_F32_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): saturation check? cvtt* (trunc?)\n e.vcvtsi2ss(i.dest, i.src1);\n }\n};\nstruct CONVERT_F32_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): saturation check? cvtt* (trunc?)\n e.vcvtsd2ss(i.dest, i.src1);\n }\n};\nstruct CONVERT_F64_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): saturation check? cvtt* (trunc?)\n e.vcvtsi2sd(i.dest, i.src1);\n }\n};\nstruct CONVERT_F64_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vcvtss2sd(i.dest, i.src1);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CONVERT, CONVERT_I32_F32, CONVERT_I32_F64,\n CONVERT_I64_F64, CONVERT_F32_I32, CONVERT_F32_F64,\n CONVERT_F64_I64, CONVERT_F64_F32);\n\n// ============================================================================\n// OPCODE_ROUND\n// ============================================================================\nstruct ROUND_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags) {\n case ROUND_TO_ZERO:\n e.vroundss(i.dest, i.src1, 0b00000011);\n break;\n case ROUND_TO_NEAREST:\n e.vroundss(i.dest, i.src1, 0b00000000);\n break;\n case ROUND_TO_MINUS_INFINITY:\n e.vroundss(i.dest, i.src1, 0b00000001);\n break;\n case ROUND_TO_POSITIVE_INFINITY:\n e.vroundss(i.dest, i.src1, 0b00000010);\n break;\n }\n }\n};\nstruct ROUND_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags) {\n case ROUND_TO_ZERO:\n e.vroundsd(i.dest, i.src1, 0b00000011);\n break;\n case ROUND_TO_NEAREST:\n e.vroundsd(i.dest, i.src1, 0b00000000);\n break;\n case ROUND_TO_MINUS_INFINITY:\n e.vroundsd(i.dest, i.src1, 0b00000001);\n break;\n case ROUND_TO_POSITIVE_INFINITY:\n e.vroundsd(i.dest, i.src1, 0b00000010);\n break;\n }\n }\n};\nstruct ROUND_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n switch (i.instr->flags) {\n case ROUND_TO_ZERO:\n e.vroundps(i.dest, i.src1, 0b00000011);\n break;\n case ROUND_TO_NEAREST:\n e.vroundps(i.dest, i.src1, 0b00000000);\n break;\n case ROUND_TO_MINUS_INFINITY:\n e.vroundps(i.dest, i.src1, 0b00000001);\n break;\n case ROUND_TO_POSITIVE_INFINITY:\n e.vroundps(i.dest, i.src1, 0b00000010);\n break;\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ROUND, ROUND_F32, ROUND_F64, ROUND_V128);\n\n// ============================================================================\n// OPCODE_LOAD_CLOCK\n// ============================================================================\nstruct LOAD_CLOCK : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // When scaling is disabled and the raw clock source is selected, the code\n // in the Clock class is actually just forwarding tick counts after one\n // simple multiply and division. In that case we rather bake the scaling in\n // here to cut extra function calls with CPU cache misses and stack frame\n // overhead.\n if (cvars::clock_no_scaling && cvars::clock_source_raw) {\n auto ratio = Clock::guest_tick_ratio();\n // The 360 CPU is an in-order CPU, AMD64 usually isn't. Without\n // mfence/lfence magic the rdtsc instruction can be executed sooner or\n // later in the cache window. Since it's resolution however is much higher\n // than the 360's mftb instruction this can safely be ignored.\n\n // Read time stamp in edx (high part) and eax (low part).\n e.rdtsc();\n // Make it a 64 bit number in rax.\n e.shl(e.rdx, 32);\n e.or_(e.rax, e.rdx);\n // Apply tick frequency scaling.\n e.mov(e.rcx, ratio.first);\n e.mul(e.rcx);\n // We actually now have a 128 bit number in rdx:rax.\n e.mov(e.rcx, ratio.second);\n e.div(e.rcx);\n e.mov(i.dest, e.rax);\n } else {\n e.CallNative(LoadClock);\n e.mov(i.dest, e.rax);\n }\n }\n static uint64_t LoadClock(void* raw_context) {\n return Clock::QueryGuestTickCount();\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOAD_CLOCK, LOAD_CLOCK);\n\n// ============================================================================\n// OPCODE_CONTEXT_BARRIER\n// ============================================================================\nstruct CONTEXT_BARRIER\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {}\n};\nEMITTER_OPCODE_TABLE(OPCODE_CONTEXT_BARRIER, CONTEXT_BARRIER);\n\n// ============================================================================\n// OPCODE_MAX\n// ============================================================================\nstruct MAX_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmaxss(dest, src1, src2);\n });\n }\n};\nstruct MAX_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmaxsd(dest, src1, src2);\n });\n }\n};\nstruct MAX_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmaxps(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_MAX, MAX_F32, MAX_F64, MAX_V128);\n\n// ============================================================================\n// OPCODE_MIN\n// ============================================================================\nstruct MIN_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const Reg8& dest_src, const Reg8& src) {\n e.cmp(dest_src, src);\n e.cmovg(dest_src.cvt32(), src.cvt32());\n },\n [](X64Emitter& e, const Reg8& dest_src, int32_t constant) {\n e.mov(e.al, constant);\n e.cmp(dest_src, e.al);\n e.cmovg(dest_src.cvt32(), e.eax);\n });\n }\n};\nstruct MIN_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const Reg16& dest_src, const Reg16& src) {\n e.cmp(dest_src, src);\n e.cmovg(dest_src.cvt32(), src.cvt32());\n },\n [](X64Emitter& e, const Reg16& dest_src, int32_t constant) {\n e.mov(e.ax, constant);\n e.cmp(dest_src, e.ax);\n e.cmovg(dest_src.cvt32(), e.eax);\n });\n }\n};\nstruct MIN_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const Reg32& dest_src, const Reg32& src) {\n e.cmp(dest_src, src);\n e.cmovg(dest_src, src);\n },\n [](X64Emitter& e, const Reg32& dest_src, int32_t constant) {\n e.mov(e.eax, constant);\n e.cmp(dest_src, e.eax);\n e.cmovg(dest_src, e.eax);\n });\n }\n};\nstruct MIN_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const Reg64& dest_src, const Reg64& src) {\n e.cmp(dest_src, src);\n e.cmovg(dest_src, src);\n },\n [](X64Emitter& e, const Reg64& dest_src, int64_t constant) {\n e.mov(e.rax, constant);\n e.cmp(dest_src, e.rax);\n e.cmovg(dest_src, e.rax);\n });\n }\n};\nstruct MIN_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vminss(dest, src1, src2);\n });\n }\n};\nstruct MIN_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vminsd(dest, src1, src2);\n });\n }\n};\nstruct MIN_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vminps(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_MIN, MIN_I8, MIN_I16, MIN_I32, MIN_I64, MIN_F32,\n MIN_F64, MIN_V128);\n\n// ============================================================================\n// OPCODE_SELECT\n// ============================================================================\n// dest = src1 ? src2 : src3\n// TODO(benvanik): match compare + select sequences, as often it's something\n// like SELECT(VECTOR_COMPARE_SGE(a, b), a, b)\nstruct SELECT_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Reg8 src2;\n if (i.src2.is_constant) {\n src2 = e.al;\n e.mov(src2, i.src2.constant());\n } else {\n src2 = i.src2;\n }\n e.test(i.src1, i.src1);\n e.cmovnz(i.dest.reg().cvt32(), src2.cvt32());\n e.cmovz(i.dest.reg().cvt32(), i.src3.reg().cvt32());\n }\n};\nstruct SELECT_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Reg16 src2;\n if (i.src2.is_constant) {\n src2 = e.ax;\n e.mov(src2, i.src2.constant());\n } else {\n src2 = i.src2;\n }\n e.test(i.src1, i.src1);\n e.cmovnz(i.dest.reg().cvt32(), src2.cvt32());\n e.cmovz(i.dest.reg().cvt32(), i.src3.reg().cvt32());\n }\n};\nstruct SELECT_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Reg32 src2;\n if (i.src2.is_constant) {\n src2 = e.eax;\n e.mov(src2, i.src2.constant());\n } else {\n src2 = i.src2;\n }\n e.test(i.src1, i.src1);\n e.cmovnz(i.dest, src2);\n e.cmovz(i.dest, i.src3);\n }\n};\nstruct SELECT_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Reg64 src2;\n if (i.src2.is_constant) {\n src2 = e.rax;\n e.mov(src2, i.src2.constant());\n } else {\n src2 = i.src2;\n }\n e.test(i.src1, i.src1);\n e.cmovnz(i.dest, src2);\n e.cmovz(i.dest, i.src3);\n }\n};\nstruct SELECT_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): find a shorter sequence.\n // dest = src1 != 0 ? src2 : src3\n e.movzx(e.eax, i.src1);\n e.vmovd(e.xmm1, e.eax);\n e.vxorps(e.xmm0, e.xmm0);\n e.vpcmpeqd(e.xmm0, e.xmm1);\n\n Xmm src2 = i.src2.is_constant ? e.xmm2 : i.src2;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(src2, i.src2.constant());\n }\n e.vpandn(e.xmm1, e.xmm0, src2);\n\n Xmm src3 = i.src3.is_constant ? e.xmm2 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n e.vpand(i.dest, e.xmm0, src3);\n e.vpor(i.dest, e.xmm1);\n }\n};\nstruct SELECT_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // dest = src1 != 0 ? src2 : src3\n\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.movzx(e.rax, i.src1);\n e.vmovq(e.xmm0, e.rax);\n e.vptestmq(e.k1, e.xmm0, e.xmm0);\n\n const Xmm src2 = i.src2.is_constant ? e.xmm1 : i.src2;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(src2, i.src2.constant());\n }\n\n const Xmm src3 = i.src3.is_constant ? e.xmm2 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n\n e.vpblendmq(i.dest.reg() | e.k1, src3, src2);\n return;\n }\n\n e.movzx(e.eax, i.src1);\n e.vmovd(e.xmm1, e.eax);\n e.vpxor(e.xmm0, e.xmm0);\n e.vpcmpeqq(e.xmm0, e.xmm1);\n\n Xmm src2 = i.src2.is_constant ? e.xmm2 : i.src2;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(src2, i.src2.constant());\n }\n e.vpandn(e.xmm1, e.xmm0, src2);\n\n Xmm src3 = i.src3.is_constant ? e.xmm2 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n e.vpand(i.dest, e.xmm0, src3);\n e.vpor(i.dest, e.xmm1);\n }\n};\nstruct SELECT_V128_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): find a shorter sequence.\n // dest = src1 != 0 ? src2 : src3\n e.movzx(e.eax, i.src1);\n e.vmovd(e.xmm1, e.eax);\n e.vpbroadcastd(e.xmm1, e.xmm1);\n e.vxorps(e.xmm0, e.xmm0);\n e.vpcmpeqd(e.xmm0, e.xmm1);\n\n Xmm src2 = i.src2.is_constant ? e.xmm2 : i.src2;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(src2, i.src2.constant());\n }\n e.vpandn(e.xmm1, e.xmm0, src2);\n\n Xmm src3 = i.src3.is_constant ? e.xmm2 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n e.vpand(i.dest, e.xmm0, src3);\n e.vpor(i.dest, e.xmm1);\n }\n};\nstruct SELECT_V128_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n const Xmm src1 = i.src1.is_constant ? e.xmm0 : i.src1;\n if (i.src1.is_constant) {\n e.LoadConstantXmm(src1, i.src1.constant());\n }\n\n const Xmm src2 = i.src2.is_constant ? e.xmm1 : i.src2;\n if (i.src2.is_constant) {\n e.LoadConstantXmm(src2, i.src2.constant());\n }\n\n const Xmm src3 = i.src3.is_constant ? e.xmm2 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vmovdqa(e.xmm3, src1);\n e.vpternlogd(e.xmm3, src2, src3,\n (~TernaryOperand::a & TernaryOperand::b) |\n (TernaryOperand::c & TernaryOperand::a));\n e.vmovdqa(i.dest, e.xmm3);\n return;\n }\n\n // src1 ? src2 : src3;\n e.vpandn(e.xmm3, src1, src2);\n e.vpand(i.dest, src1, src3);\n e.vpor(i.dest, i.dest, e.xmm3);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SELECT, SELECT_I8, SELECT_I16, SELECT_I32,\n SELECT_I64, SELECT_F32, SELECT_F64, SELECT_V128_I8,\n SELECT_V128_V128);\n\n// ============================================================================\n// OPCODE_IS_TRUE\n// ============================================================================\nstruct IS_TRUE_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.setnz(i.dest);\n }\n};\nstruct IS_TRUE_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.setnz(i.dest);\n }\n};\nstruct IS_TRUE_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.setnz(i.dest);\n }\n};\nstruct IS_TRUE_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.setnz(i.dest);\n }\n};\nstruct IS_TRUE_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.setnz(i.dest);\n }\n};\nstruct IS_TRUE_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.setnz(i.dest);\n }\n};\nstruct IS_TRUE_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.setnz(i.dest);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_IS_TRUE, IS_TRUE_I8, IS_TRUE_I16, IS_TRUE_I32,\n IS_TRUE_I64, IS_TRUE_F32, IS_TRUE_F64, IS_TRUE_V128);\n\n// ============================================================================\n// OPCODE_IS_FALSE\n// ============================================================================\nstruct IS_FALSE_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.setz(i.dest);\n }\n};\nstruct IS_FALSE_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.setz(i.dest);\n }\n};\nstruct IS_FALSE_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.setz(i.dest);\n }\n};\nstruct IS_FALSE_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.test(i.src1, i.src1);\n e.setz(i.dest);\n }\n};\nstruct IS_FALSE_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.setz(i.dest);\n }\n};\nstruct IS_FALSE_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.setz(i.dest);\n }\n};\nstruct IS_FALSE_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vptest(i.src1, i.src1);\n e.setz(i.dest);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_IS_FALSE, IS_FALSE_I8, IS_FALSE_I16, IS_FALSE_I32,\n IS_FALSE_I64, IS_FALSE_F32, IS_FALSE_F64, IS_FALSE_V128);\n\n// ============================================================================\n// OPCODE_IS_NAN\n// ============================================================================\nstruct IS_NAN_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vucomiss(i.src1, i.src1);\n e.setp(i.dest);\n }\n};\n\nstruct IS_NAN_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vucomisd(i.src1, i.src1);\n e.setp(i.dest);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_IS_NAN, IS_NAN_F32, IS_NAN_F64);\n\n// ============================================================================\n// OPCODE_COMPARE_EQ\n// ============================================================================\nstruct COMPARE_EQ_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeCompareOp(\n e, i,\n [](X64Emitter& e, const Reg8& src1, const Reg8& src2) {\n e.cmp(src1, src2);\n },\n [](X64Emitter& e, const Reg8& src1, int32_t constant) {\n e.cmp(src1, constant);\n });\n e.sete(i.dest);\n }\n};\nstruct COMPARE_EQ_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeCompareOp(\n e, i,\n [](X64Emitter& e, const Reg16& src1, const Reg16& src2) {\n e.cmp(src1, src2);\n },\n [](X64Emitter& e, const Reg16& src1, int32_t constant) {\n e.cmp(src1, constant);\n });\n e.sete(i.dest);\n }\n};\nstruct COMPARE_EQ_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeCompareOp(\n e, i,\n [](X64Emitter& e, const Reg32& src1, const Reg32& src2) {\n e.cmp(src1, src2);\n },\n [](X64Emitter& e, const Reg32& src1, int32_t constant) {\n e.cmp(src1, constant);\n });\n e.sete(i.dest);\n }\n};\nstruct COMPARE_EQ_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeCompareOp(\n e, i,\n [](X64Emitter& e, const Reg64& src1, const Reg64& src2) {\n e.cmp(src1, src2);\n },\n [](X64Emitter& e, const Reg64& src1, int32_t constant) {\n e.cmp(src1, constant);\n });\n e.sete(i.dest);\n }\n};\nstruct COMPARE_EQ_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, I8Op dest, const Xmm& src1, const Xmm& src2) {\n e.vcomiss(src1, src2);\n });\n e.sete(i.dest);\n }\n};\nstruct COMPARE_EQ_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, I8Op dest, const Xmm& src1, const Xmm& src2) {\n e.vcomisd(src1, src2);\n });\n e.sete(i.dest);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_COMPARE_EQ, COMPARE_EQ_I8, COMPARE_EQ_I16,\n COMPARE_EQ_I32, COMPARE_EQ_I64, COMPARE_EQ_F32,\n COMPARE_EQ_F64);\n\n// ============================================================================\n// OPCODE_COMPARE_NE\n// ============================================================================\nstruct COMPARE_NE_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeCompareOp(\n e, i,\n [](X64Emitter& e, const Reg8& src1, const Reg8& src2) {\n e.cmp(src1, src2);\n },\n [](X64Emitter& e, const Reg8& src1, int32_t constant) {\n e.cmp(src1, constant);\n });\n e.setne(i.dest);\n }\n};\nstruct COMPARE_NE_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeCompareOp(\n e, i,\n [](X64Emitter& e, const Reg16& src1, const Reg16& src2) {\n e.cmp(src1, src2);\n },\n [](X64Emitter& e, const Reg16& src1, int32_t constant) {\n e.cmp(src1, constant);\n });\n e.setne(i.dest);\n }\n};\nstruct COMPARE_NE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeCompareOp(\n e, i,\n [](X64Emitter& e, const Reg32& src1, const Reg32& src2) {\n e.cmp(src1, src2);\n },\n [](X64Emitter& e, const Reg32& src1, int32_t constant) {\n e.cmp(src1, constant);\n });\n e.setne(i.dest);\n }\n};\nstruct COMPARE_NE_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeCompareOp(\n e, i,\n [](X64Emitter& e, const Reg64& src1, const Reg64& src2) {\n e.cmp(src1, src2);\n },\n [](X64Emitter& e, const Reg64& src1, int32_t constant) {\n e.cmp(src1, constant);\n });\n e.setne(i.dest);\n }\n};\nstruct COMPARE_NE_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vcomiss(i.src1, i.src2);\n e.setne(i.dest);\n }\n};\nstruct COMPARE_NE_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vcomisd(i.src1, i.src2);\n e.setne(i.dest);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_COMPARE_NE, COMPARE_NE_I8, COMPARE_NE_I16,\n COMPARE_NE_I32, COMPARE_NE_I64, COMPARE_NE_F32,\n COMPARE_NE_F64);\n\n// ============================================================================\n// OPCODE_COMPARE_*\n// ============================================================================\n#define EMITTER_ASSOCIATIVE_COMPARE_INT(op, instr, inverse_instr, type, \\\n reg_type) \\\n struct COMPARE_##op##_##type \\\n : Sequence> { \\\n static void Emit(X64Emitter& e, const EmitArgType& i) { \\\n EmitAssociativeCompareOp( \\\n e, i, \\\n [](X64Emitter& e, const Reg8& dest, const reg_type& src1, \\\n const reg_type& src2, bool inverse) { \\\n e.cmp(src1, src2); \\\n if (!inverse) { \\\n e.instr(dest); \\\n } else { \\\n e.inverse_instr(dest); \\\n } \\\n }, \\\n [](X64Emitter& e, const Reg8& dest, const reg_type& src1, \\\n int32_t constant, bool inverse) { \\\n e.cmp(src1, constant); \\\n if (!inverse) { \\\n e.instr(dest); \\\n } else { \\\n e.inverse_instr(dest); \\\n } \\\n }); \\\n } \\\n };\n#define EMITTER_ASSOCIATIVE_COMPARE_XX(op, instr, inverse_instr) \\\n EMITTER_ASSOCIATIVE_COMPARE_INT(op, instr, inverse_instr, I8Op, Reg8); \\\n EMITTER_ASSOCIATIVE_COMPARE_INT(op, instr, inverse_instr, I16Op, Reg16); \\\n EMITTER_ASSOCIATIVE_COMPARE_INT(op, instr, inverse_instr, I32Op, Reg32); \\\n EMITTER_ASSOCIATIVE_COMPARE_INT(op, instr, inverse_instr, I64Op, Reg64); \\\n EMITTER_OPCODE_TABLE(OPCODE_COMPARE_##op, COMPARE_##op##_I8Op, \\\n COMPARE_##op##_I16Op, COMPARE_##op##_I32Op, \\\n COMPARE_##op##_I64Op);\nEMITTER_ASSOCIATIVE_COMPARE_XX(SLT, setl, setg);\nEMITTER_ASSOCIATIVE_COMPARE_XX(SLE, setle, setge);\nEMITTER_ASSOCIATIVE_COMPARE_XX(SGT, setg, setl);\nEMITTER_ASSOCIATIVE_COMPARE_XX(SGE, setge, setle);\nEMITTER_ASSOCIATIVE_COMPARE_XX(ULT, setb, seta);\nEMITTER_ASSOCIATIVE_COMPARE_XX(ULE, setbe, setae);\nEMITTER_ASSOCIATIVE_COMPARE_XX(UGT, seta, setb);\nEMITTER_ASSOCIATIVE_COMPARE_XX(UGE, setae, setbe);\n\n// https://web.archive.org/web/20171129015931/https://x86.renejeschke.de/html/file_module_x86_id_288.html\n// Original link: https://x86.renejeschke.de/html/file_module_x86_id_288.html\n#define EMITTER_ASSOCIATIVE_COMPARE_FLT_XX(op, instr) \\\n struct COMPARE_##op##_F32 \\\n : Sequence> { \\\n static void Emit(X64Emitter& e, const EmitArgType& i) { \\\n e.vcomiss(i.src1, i.src2); \\\n e.instr(i.dest); \\\n } \\\n }; \\\n struct COMPARE_##op##_F64 \\\n : Sequence> { \\\n static void Emit(X64Emitter& e, const EmitArgType& i) { \\\n if (i.src1.is_constant) { \\\n e.LoadConstantXmm(e.xmm0, i.src1.constant()); \\\n e.vcomisd(e.xmm0, i.src2); \\\n } else if (i.src2.is_constant) { \\\n e.LoadConstantXmm(e.xmm0, i.src2.constant()); \\\n e.vcomisd(i.src1, e.xmm0); \\\n } else { \\\n e.vcomisd(i.src1, i.src2); \\\n } \\\n e.instr(i.dest); \\\n } \\\n }; \\\n EMITTER_OPCODE_TABLE(OPCODE_COMPARE_##op##_FLT, COMPARE_##op##_F32, \\\n COMPARE_##op##_F64);\nEMITTER_ASSOCIATIVE_COMPARE_FLT_XX(SLT, setb);\nEMITTER_ASSOCIATIVE_COMPARE_FLT_XX(SLE, setbe);\nEMITTER_ASSOCIATIVE_COMPARE_FLT_XX(SGT, seta);\nEMITTER_ASSOCIATIVE_COMPARE_FLT_XX(SGE, setae);\nEMITTER_ASSOCIATIVE_COMPARE_FLT_XX(ULT, setb);\nEMITTER_ASSOCIATIVE_COMPARE_FLT_XX(ULE, setbe);\nEMITTER_ASSOCIATIVE_COMPARE_FLT_XX(UGT, seta);\nEMITTER_ASSOCIATIVE_COMPARE_FLT_XX(UGE, setae);\n\n// ============================================================================\n// OPCODE_DID_SATURATE\n// ============================================================================\nstruct DID_SATURATE\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): implement saturation check (VECTOR_ADD, etc).\n e.xor_(i.dest, i.dest);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_DID_SATURATE, DID_SATURATE);\n\n// ============================================================================\n// OPCODE_ADD\n// ============================================================================\n// TODO(benvanik): put dest/src1|2 together.\ntemplate \nvoid EmitAddXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const REG& src) {\n e.add(dest_src, src);\n },\n [](X64Emitter& e, const REG& dest_src, int32_t constant) {\n e.add(dest_src, constant);\n });\n}\nstruct ADD_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAddXX(e, i);\n }\n};\nstruct ADD_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAddXX(e, i);\n }\n};\nstruct ADD_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAddXX(e, i);\n }\n};\nstruct ADD_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAddXX(e, i);\n }\n};\nstruct ADD_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vaddss(dest, src1, src2);\n });\n }\n};\nstruct ADD_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vaddsd(dest, src1, src2);\n });\n }\n};\nstruct ADD_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vaddps(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ADD, ADD_I8, ADD_I16, ADD_I32, ADD_I64, ADD_F32,\n ADD_F64, ADD_V128);\n\n// ============================================================================\n// OPCODE_ADD_CARRY\n// ============================================================================\n// TODO(benvanik): put dest/src1|2 together.\ntemplate \nvoid EmitAddCarryXX(X64Emitter& e, const ARGS& i) {\n // TODO(benvanik): faster setting? we could probably do some fun math tricks\n // here to get the carry flag set.\n if (i.src3.is_constant) {\n if (i.src3.constant()) {\n e.stc();\n } else {\n e.clc();\n }\n } else {\n if (i.src3.reg().getIdx() <= 4) {\n // Can move from A/B/C/DX to AH.\n e.mov(e.ah, i.src3.reg().cvt8());\n } else {\n e.mov(e.al, i.src3);\n e.mov(e.ah, e.al);\n }\n e.sahf();\n }\n SEQ::EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const REG& src) {\n e.adc(dest_src, src);\n },\n [](X64Emitter& e, const REG& dest_src, int32_t constant) {\n e.adc(dest_src, constant);\n });\n}\nstruct ADD_CARRY_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAddCarryXX(e, i);\n }\n};\nstruct ADD_CARRY_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAddCarryXX(e, i);\n }\n};\nstruct ADD_CARRY_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAddCarryXX(e, i);\n }\n};\nstruct ADD_CARRY_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAddCarryXX(e, i);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ADD_CARRY, ADD_CARRY_I8, ADD_CARRY_I16,\n ADD_CARRY_I32, ADD_CARRY_I64);\n\n// ============================================================================\n// OPCODE_SUB\n// ============================================================================\n// TODO(benvanik): put dest/src1|2 together.\ntemplate \nvoid EmitSubXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitAssociativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const REG& src) {\n e.sub(dest_src, src);\n },\n [](X64Emitter& e, const REG& dest_src, int32_t constant) {\n e.sub(dest_src, constant);\n });\n}\nstruct SUB_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitSubXX(e, i);\n }\n};\nstruct SUB_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitSubXX(e, i);\n }\n};\nstruct SUB_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitSubXX(e, i);\n }\n};\nstruct SUB_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitSubXX(e, i);\n }\n};\nstruct SUB_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitAssociativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vsubss(dest, src1, src2);\n });\n }\n};\nstruct SUB_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitAssociativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vsubsd(dest, src1, src2);\n });\n }\n};\nstruct SUB_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitAssociativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vsubps(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SUB, SUB_I8, SUB_I16, SUB_I32, SUB_I64, SUB_F32,\n SUB_F64, SUB_V128);\n\n// ============================================================================\n// OPCODE_MUL\n// ============================================================================\n// Sign doesn't matter here, as we don't use the high bits.\n// We exploit mulx here to avoid creating too much register pressure.\nstruct MUL_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n // mulx: $1:$2 = EDX * $3\n\n // TODO(benvanik): place src2 in edx?\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n e.movzx(e.edx, i.src2);\n e.mov(e.eax, static_cast(i.src1.constant()));\n e.mulx(e.edx, i.dest.reg().cvt32(), e.eax);\n } else if (i.src2.is_constant) {\n e.movzx(e.edx, i.src1);\n e.mov(e.eax, static_cast(i.src2.constant()));\n e.mulx(e.edx, i.dest.reg().cvt32(), e.eax);\n } else {\n e.movzx(e.edx, i.src2);\n e.mulx(e.edx, i.dest.reg().cvt32(), i.src1.reg().cvt32());\n }\n } else {\n // x86 mul instruction\n // AH:AL = AL * $1;\n\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n e.mov(e.al, i.src1.constant());\n e.mul(i.src2);\n e.mov(i.dest, e.al);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n e.mov(e.al, i.src2.constant());\n e.mul(i.src1);\n e.mov(i.dest, e.al);\n } else {\n e.movzx(e.al, i.src1);\n e.mul(i.src2);\n e.mov(i.dest, e.al);\n }\n }\n }\n};\nstruct MUL_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n // mulx: $1:$2 = EDX * $3\n\n // TODO(benvanik): place src2 in edx?\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n e.movzx(e.edx, i.src2);\n e.mov(e.ax, static_cast(i.src1.constant()));\n e.mulx(e.edx, i.dest.reg().cvt32(), e.eax);\n } else if (i.src2.is_constant) {\n e.movzx(e.edx, i.src1);\n e.mov(e.ax, static_cast(i.src2.constant()));\n e.mulx(e.edx, i.dest.reg().cvt32(), e.eax);\n } else {\n e.movzx(e.edx, i.src2);\n e.mulx(e.edx, i.dest.reg().cvt32(), i.src1.reg().cvt32());\n }\n } else {\n // x86 mul instruction\n // DX:AX = AX * $1;\n\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n e.mov(e.ax, i.src1.constant());\n e.mul(i.src2);\n e.movzx(i.dest, e.ax);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n e.mov(e.ax, i.src2.constant());\n e.mul(i.src1);\n e.movzx(i.dest, e.ax);\n } else {\n e.movzx(e.ax, i.src1);\n e.mul(i.src2);\n e.movzx(i.dest, e.ax);\n }\n }\n }\n};\nstruct MUL_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n // mulx: $1:$2 = EDX * $3\n\n // TODO(benvanik): place src2 in edx?\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n e.mov(e.edx, i.src2);\n e.mov(e.eax, i.src1.constant());\n e.mulx(e.edx, i.dest, e.eax);\n } else if (i.src2.is_constant) {\n e.mov(e.edx, i.src1);\n e.mov(e.eax, i.src2.constant());\n e.mulx(e.edx, i.dest, e.eax);\n } else {\n e.mov(e.edx, i.src2);\n e.mulx(e.edx, i.dest, i.src1);\n }\n } else {\n // x86 mul instruction\n // EDX:EAX = EAX * $1;\n\n // is_constant AKA not a register\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant); // can't multiply 2 constants\n e.mov(e.eax, i.src1.constant());\n e.mul(i.src2);\n e.mov(i.dest, e.eax);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant); // can't multiply 2 constants\n e.mov(e.eax, i.src2.constant());\n e.mul(i.src1);\n e.mov(i.dest, e.eax);\n } else {\n e.mov(e.eax, i.src1);\n e.mul(i.src2);\n e.mov(i.dest, e.eax);\n }\n }\n }\n};\nstruct MUL_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n // mulx: $1:$2 = RDX * $3\n\n // TODO(benvanik): place src2 in edx?\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant);\n e.mov(e.rdx, i.src2);\n e.mov(e.rax, i.src1.constant());\n e.mulx(e.rdx, i.dest, e.rax);\n } else if (i.src2.is_constant) {\n e.mov(e.rdx, i.src1);\n e.mov(e.rax, i.src2.constant());\n e.mulx(e.rdx, i.dest, e.rax);\n } else {\n e.mov(e.rdx, i.src2);\n e.mulx(e.rdx, i.dest, i.src1);\n }\n } else {\n // x86 mul instruction\n // RDX:RAX = RAX * $1;\n\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant); // can't multiply 2 constants\n e.mov(e.rax, i.src1.constant());\n e.mul(i.src2);\n e.mov(i.dest, e.rax);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant); // can't multiply 2 constants\n e.mov(e.rax, i.src2.constant());\n e.mul(i.src1);\n e.mov(i.dest, e.rax);\n } else {\n e.mov(e.rax, i.src1);\n e.mul(i.src2);\n e.mov(i.dest, e.rax);\n }\n }\n }\n};\nstruct MUL_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulss(dest, src1, src2);\n });\n }\n};\nstruct MUL_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulsd(dest, src1, src2);\n });\n }\n};\nstruct MUL_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulps(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_MUL, MUL_I8, MUL_I16, MUL_I32, MUL_I64, MUL_F32,\n MUL_F64, MUL_V128);\n\n// ============================================================================\n// OPCODE_MUL_HI\n// ============================================================================\nstruct MUL_HI_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n // mulx: $1:$2 = EDX * $3\n\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n // TODO(benvanik): place src1 in eax? still need to sign extend\n e.movzx(e.edx, i.src1);\n e.mulx(i.dest.reg().cvt32(), e.eax, i.src2.reg().cvt32());\n } else {\n // x86 mul instruction\n // AH:AL = AL * $1;\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant); // can't multiply 2 constants\n e.mov(e.al, i.src1.constant());\n e.mul(i.src2);\n e.mov(i.dest, e.ah);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant); // can't multiply 2 constants\n e.mov(e.al, i.src2.constant());\n e.mul(i.src1);\n e.mov(i.dest, e.ah);\n } else {\n e.mov(e.al, i.src1);\n e.mul(i.src2);\n e.mov(i.dest, e.ah);\n }\n }\n } else {\n if (i.src1.is_constant) {\n e.mov(e.al, i.src1.constant());\n } else {\n e.mov(e.al, i.src1);\n }\n if (i.src2.is_constant) {\n e.mov(e.al, i.src2.constant());\n e.imul(e.al);\n } else {\n e.imul(i.src2);\n }\n e.mov(i.dest, e.ah);\n }\n }\n};\nstruct MUL_HI_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n // TODO(benvanik): place src1 in eax? still need to sign extend\n e.movzx(e.edx, i.src1);\n e.mulx(i.dest.reg().cvt32(), e.eax, i.src2.reg().cvt32());\n } else {\n // x86 mul instruction\n // DX:AX = AX * $1;\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant); // can't multiply 2 constants\n e.mov(e.ax, i.src1.constant());\n e.mul(i.src2);\n e.mov(i.dest, e.dx);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant); // can't multiply 2 constants\n e.mov(e.ax, i.src2.constant());\n e.mul(i.src1);\n e.mov(i.dest, e.dx);\n } else {\n e.mov(e.ax, i.src1);\n e.mul(i.src2);\n e.mov(i.dest, e.dx);\n }\n }\n } else {\n if (i.src1.is_constant) {\n e.mov(e.ax, i.src1.constant());\n } else {\n e.mov(e.ax, i.src1);\n }\n if (i.src2.is_constant) {\n e.mov(e.dx, i.src2.constant());\n e.imul(e.dx);\n } else {\n e.imul(i.src2);\n }\n e.mov(i.dest, e.dx);\n }\n }\n};\nstruct MUL_HI_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n // TODO(benvanik): place src1 in eax? still need to sign extend\n e.mov(e.edx, i.src1);\n if (i.src2.is_constant) {\n e.mov(e.eax, i.src2.constant());\n e.mulx(i.dest, e.edx, e.eax);\n } else {\n e.mulx(i.dest, e.edx, i.src2);\n }\n } else {\n // x86 mul instruction\n // EDX:EAX = EAX * $1;\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant); // can't multiply 2 constants\n e.mov(e.eax, i.src1.constant());\n e.mul(i.src2);\n e.mov(i.dest, e.edx);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant); // can't multiply 2 constants\n e.mov(e.eax, i.src2.constant());\n e.mul(i.src1);\n e.mov(i.dest, e.edx);\n } else {\n e.mov(e.eax, i.src1);\n e.mul(i.src2);\n e.mov(i.dest, e.edx);\n }\n }\n } else {\n if (i.src1.is_constant) {\n e.mov(e.eax, i.src1.constant());\n } else {\n e.mov(e.eax, i.src1);\n }\n if (i.src2.is_constant) {\n e.mov(e.edx, i.src2.constant());\n e.imul(e.edx);\n } else {\n e.imul(i.src2);\n }\n e.mov(i.dest, e.edx);\n }\n }\n};\nstruct MUL_HI_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n // TODO(benvanik): place src1 in eax? still need to sign extend\n e.mov(e.rdx, i.src1);\n if (i.src2.is_constant) {\n e.mov(e.rax, i.src2.constant());\n e.mulx(i.dest, e.rdx, e.rax);\n } else {\n e.mulx(i.dest, e.rax, i.src2);\n }\n } else {\n // x86 mul instruction\n // RDX:RAX < RAX * REG(op1);\n if (i.src1.is_constant) {\n assert_true(!i.src2.is_constant); // can't multiply 2 constants\n e.mov(e.rax, i.src1.constant());\n e.mul(i.src2);\n e.mov(i.dest, e.rdx);\n } else if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant); // can't multiply 2 constants\n e.mov(e.rax, i.src2.constant());\n e.mul(i.src1);\n e.mov(i.dest, e.rdx);\n } else {\n e.mov(e.rax, i.src1);\n e.mul(i.src2);\n e.mov(i.dest, e.rdx);\n }\n }\n } else {\n if (i.src1.is_constant) {\n e.mov(e.rax, i.src1.constant());\n } else {\n e.mov(e.rax, i.src1);\n }\n if (i.src2.is_constant) {\n e.mov(e.rdx, i.src2.constant());\n e.imul(e.rdx);\n } else {\n e.imul(i.src2);\n }\n e.mov(i.dest, e.rdx);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_MUL_HI, MUL_HI_I8, MUL_HI_I16, MUL_HI_I32,\n MUL_HI_I64);\n\n// ============================================================================\n// OPCODE_DIV\n// ============================================================================\n// TODO(benvanik): optimize common constant cases.\n// TODO(benvanik): simplify code!\nstruct DIV_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Xbyak::Label skip;\n e.inLocalLabel();\n\n if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n e.mov(e.cl, i.src2.constant());\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n e.movzx(e.ax, i.src1);\n e.div(e.cl);\n } else {\n e.movsx(e.ax, i.src1);\n e.idiv(e.cl);\n }\n } else {\n // Skip if src2 is zero.\n e.test(i.src2, i.src2);\n e.jz(skip, CodeGenerator::T_SHORT);\n\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n if (i.src1.is_constant) {\n e.mov(e.ax, static_cast(i.src1.constant()));\n } else {\n e.movzx(e.ax, i.src1);\n }\n e.div(i.src2);\n } else {\n if (i.src1.is_constant) {\n e.mov(e.ax, static_cast(i.src1.constant()));\n } else {\n e.movsx(e.ax, i.src1);\n }\n e.idiv(i.src2);\n }\n }\n\n e.L(skip);\n e.outLocalLabel();\n e.mov(i.dest, e.al);\n }\n};\nstruct DIV_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Xbyak::Label skip;\n e.inLocalLabel();\n\n if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n e.mov(e.cx, i.src2.constant());\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n e.mov(e.ax, i.src1);\n // Zero upper bits.\n e.xor_(e.dx, e.dx);\n e.div(e.cx);\n } else {\n e.mov(e.ax, i.src1);\n e.cwd(); // dx:ax = sign-extend ax\n e.idiv(e.cx);\n }\n } else {\n // Skip if src2 is zero.\n e.test(i.src2, i.src2);\n e.jz(skip, CodeGenerator::T_SHORT);\n\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n if (i.src1.is_constant) {\n e.mov(e.ax, i.src1.constant());\n } else {\n e.mov(e.ax, i.src1);\n }\n // Zero upper bits.\n e.xor_(e.dx, e.dx);\n e.div(i.src2);\n } else {\n if (i.src1.is_constant) {\n e.mov(e.ax, i.src1.constant());\n } else {\n e.mov(e.ax, i.src1);\n }\n e.cwd(); // dx:ax = sign-extend ax\n e.idiv(i.src2);\n }\n }\n\n e.L(skip);\n e.outLocalLabel();\n e.mov(i.dest, e.ax);\n }\n};\nstruct DIV_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Xbyak::Label skip;\n e.inLocalLabel();\n\n if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n e.mov(e.ecx, i.src2.constant());\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n e.mov(e.eax, i.src1);\n // Zero upper bits.\n e.xor_(e.edx, e.edx);\n e.div(e.ecx);\n } else {\n e.mov(e.eax, i.src1);\n e.cdq(); // edx:eax = sign-extend eax\n e.idiv(e.ecx);\n }\n } else {\n // Skip if src2 is zero.\n e.test(i.src2, i.src2);\n e.jz(skip, CodeGenerator::T_SHORT);\n\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n if (i.src1.is_constant) {\n e.mov(e.eax, i.src1.constant());\n } else {\n e.mov(e.eax, i.src1);\n }\n // Zero upper bits.\n e.xor_(e.edx, e.edx);\n e.div(i.src2);\n } else {\n if (i.src1.is_constant) {\n e.mov(e.eax, i.src1.constant());\n } else {\n e.mov(e.eax, i.src1);\n }\n e.cdq(); // edx:eax = sign-extend eax\n e.idiv(i.src2);\n }\n }\n\n e.L(skip);\n e.outLocalLabel();\n e.mov(i.dest, e.eax);\n }\n};\nstruct DIV_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n Xbyak::Label skip;\n e.inLocalLabel();\n\n if (i.src2.is_constant) {\n assert_true(!i.src1.is_constant);\n e.mov(e.rcx, i.src2.constant());\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n e.mov(e.rax, i.src1);\n // Zero upper bits.\n e.xor_(e.rdx, e.rdx);\n e.div(e.rcx);\n } else {\n e.mov(e.rax, i.src1);\n e.cqo(); // rdx:rax = sign-extend rax\n e.idiv(e.rcx);\n }\n } else {\n // Skip if src2 is zero.\n e.test(i.src2, i.src2);\n e.jz(skip, CodeGenerator::T_SHORT);\n\n if (i.instr->flags & ARITHMETIC_UNSIGNED) {\n if (i.src1.is_constant) {\n e.mov(e.rax, i.src1.constant());\n } else {\n e.mov(e.rax, i.src1);\n }\n // Zero upper bits.\n e.xor_(e.rdx, e.rdx);\n e.div(i.src2);\n } else {\n if (i.src1.is_constant) {\n e.mov(e.rax, i.src1.constant());\n } else {\n e.mov(e.rax, i.src1);\n }\n e.cqo(); // rdx:rax = sign-extend rax\n e.idiv(i.src2);\n }\n }\n\n e.L(skip);\n e.outLocalLabel();\n e.mov(i.dest, e.rax);\n }\n};\nstruct DIV_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitAssociativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vdivss(dest, src1, src2);\n });\n }\n};\nstruct DIV_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitAssociativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vdivsd(dest, src1, src2);\n });\n }\n};\nstruct DIV_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n EmitAssociativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vdivps(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_DIV, DIV_I8, DIV_I16, DIV_I32, DIV_I64, DIV_F32,\n DIV_F64, DIV_V128);\n\n// ============================================================================\n// OPCODE_MUL_ADD\n// ============================================================================\n// d = 1 * 2 + 3\n// $0 = $1x$0 + $2\n// Forms of vfmadd/vfmsub:\n// - 132 -> $1 = $1 * $3 + $2\n// - 213 -> $1 = $2 * $1 + $3\n// - 231 -> $1 = $2 * $3 + $1\nstruct MUL_ADD_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // FMA extension\n if (e.IsFeatureEnabled(kX64EmitFMA)) {\n EmitCommutativeBinaryXmmOp(e, i,\n [&i](X64Emitter& e, const Xmm& dest,\n const Xmm& src1, const Xmm& src2) {\n Xmm src3 =\n i.src3.is_constant ? e.xmm1 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n if (i.dest == src1) {\n e.vfmadd213ss(i.dest, src2, src3);\n } else if (i.dest == src2) {\n e.vfmadd213ss(i.dest, src1, src3);\n } else if (i.dest == i.src3) {\n e.vfmadd231ss(i.dest, src1, src2);\n } else {\n // Dest not equal to anything\n e.vmovss(i.dest, src1);\n e.vfmadd213ss(i.dest, src2, src3);\n }\n });\n } else {\n Xmm src3;\n if (i.src3.is_constant) {\n src3 = e.xmm1;\n e.LoadConstantXmm(src3, i.src3.constant());\n } else {\n // If i.dest == i.src3, back up i.src3 so we don't overwrite it.\n src3 = i.src3;\n if (i.dest == i.src3) {\n e.vmovss(e.xmm1, i.src3);\n src3 = e.xmm1;\n }\n }\n\n // Multiply operation is commutative.\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulss(dest, src1, src2); // $0 = $1 * $2\n });\n\n e.vaddss(i.dest, i.dest, src3); // $0 = $1 + $2\n }\n }\n};\nstruct MUL_ADD_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // FMA extension\n if (e.IsFeatureEnabled(kX64EmitFMA)) {\n EmitCommutativeBinaryXmmOp(e, i,\n [&i](X64Emitter& e, const Xmm& dest,\n const Xmm& src1, const Xmm& src2) {\n Xmm src3 =\n i.src3.is_constant ? e.xmm1 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n if (i.dest == src1) {\n e.vfmadd213sd(i.dest, src2, src3);\n } else if (i.dest == src2) {\n e.vfmadd213sd(i.dest, src1, src3);\n } else if (i.dest == i.src3) {\n e.vfmadd231sd(i.dest, src1, src2);\n } else {\n // Dest not equal to anything\n e.vmovsd(i.dest, src1);\n e.vfmadd213sd(i.dest, src2, src3);\n }\n });\n } else {\n Xmm src3;\n if (i.src3.is_constant) {\n src3 = e.xmm1;\n e.LoadConstantXmm(src3, i.src3.constant());\n } else {\n // If i.dest == i.src3, back up i.src3 so we don't overwrite it.\n src3 = i.src3;\n if (i.dest == i.src3) {\n e.vmovsd(e.xmm1, i.src3);\n src3 = e.xmm1;\n }\n }\n\n // Multiply operation is commutative.\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulsd(dest, src1, src2); // $0 = $1 * $2\n });\n\n e.vaddsd(i.dest, i.dest, src3); // $0 = $1 + $2\n }\n }\n};\nstruct MUL_ADD_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): the vfmadd sequence produces slightly different results\n // than vmul+vadd and it'd be nice to know why. Until we know, it's\n // disabled so tests pass.\n if (false && e.IsFeatureEnabled(kX64EmitFMA)) {\n EmitCommutativeBinaryXmmOp(e, i,\n [&i](X64Emitter& e, const Xmm& dest,\n const Xmm& src1, const Xmm& src2) {\n Xmm src3 =\n i.src3.is_constant ? e.xmm1 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n if (i.dest == src1) {\n e.vfmadd213ps(i.dest, src2, src3);\n } else if (i.dest == src2) {\n e.vfmadd213ps(i.dest, src1, src3);\n } else if (i.dest == i.src3) {\n e.vfmadd231ps(i.dest, src1, src2);\n } else {\n // Dest not equal to anything\n e.vmovdqa(i.dest, src1);\n e.vfmadd213ps(i.dest, src2, src3);\n }\n });\n } else {\n Xmm src3;\n if (i.src3.is_constant) {\n src3 = e.xmm1;\n e.LoadConstantXmm(src3, i.src3.constant());\n } else {\n // If i.dest == i.src3, back up i.src3 so we don't overwrite it.\n src3 = i.src3;\n if (i.dest == i.src3) {\n e.vmovdqa(e.xmm1, i.src3);\n src3 = e.xmm1;\n }\n }\n\n // Multiply operation is commutative.\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulps(dest, src1, src2); // $0 = $1 * $2\n });\n\n e.vaddps(i.dest, i.dest, src3); // $0 = $1 + $2\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_MUL_ADD, MUL_ADD_F32, MUL_ADD_F64, MUL_ADD_V128);\n\n// ============================================================================\n// OPCODE_MUL_SUB\n// ============================================================================\n// d = 1 * 2 - 3\n// $0 = $2x$0 - $3\n// TODO(benvanik): use other forms (132/213/etc) to avoid register shuffling.\n// dest could be src2 or src3 - need to ensure it's not before overwriting dest\n// perhaps use other 132/213/etc\n// Forms:\n// - 132 -> $1 = $1 * $3 - $2\n// - 213 -> $1 = $2 * $1 - $3\n// - 231 -> $1 = $2 * $3 - $1\nstruct MUL_SUB_F32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // FMA extension\n if (e.IsFeatureEnabled(kX64EmitFMA)) {\n EmitCommutativeBinaryXmmOp(e, i,\n [&i](X64Emitter& e, const Xmm& dest,\n const Xmm& src1, const Xmm& src2) {\n Xmm src3 =\n i.src3.is_constant ? e.xmm1 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n if (i.dest == src1) {\n e.vfmsub213ss(i.dest, src2, src3);\n } else if (i.dest == src2) {\n e.vfmsub213ss(i.dest, src1, src3);\n } else if (i.dest == i.src3) {\n e.vfmsub231ss(i.dest, src1, src2);\n } else {\n // Dest not equal to anything\n e.vmovss(i.dest, src1);\n e.vfmsub213ss(i.dest, src2, src3);\n }\n });\n } else {\n Xmm src3;\n if (i.src3.is_constant) {\n src3 = e.xmm1;\n e.LoadConstantXmm(src3, i.src3.constant());\n } else {\n // If i.dest == i.src3, back up i.src3 so we don't overwrite it.\n src3 = i.src3;\n if (i.dest == i.src3) {\n e.vmovss(e.xmm1, i.src3);\n src3 = e.xmm1;\n }\n }\n\n // Multiply operation is commutative.\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulss(dest, src1, src2); // $0 = $1 * $2\n });\n\n e.vsubss(i.dest, i.dest, src3); // $0 = $1 - $2\n }\n }\n};\nstruct MUL_SUB_F64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // FMA extension\n if (e.IsFeatureEnabled(kX64EmitFMA)) {\n EmitCommutativeBinaryXmmOp(e, i,\n [&i](X64Emitter& e, const Xmm& dest,\n const Xmm& src1, const Xmm& src2) {\n Xmm src3 =\n i.src3.is_constant ? e.xmm1 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n if (i.dest == src1) {\n e.vfmsub213sd(i.dest, src2, src3);\n } else if (i.dest == src2) {\n e.vfmsub213sd(i.dest, src1, src3);\n } else if (i.dest == i.src3) {\n e.vfmsub231sd(i.dest, src1, src2);\n } else {\n // Dest not equal to anything\n e.vmovsd(i.dest, src1);\n e.vfmsub213sd(i.dest, src2, src3);\n }\n });\n } else {\n Xmm src3;\n if (i.src3.is_constant) {\n src3 = e.xmm1;\n e.LoadConstantXmm(src3, i.src3.constant());\n } else {\n // If i.dest == i.src3, back up i.src3 so we don't overwrite it.\n src3 = i.src3;\n if (i.dest == i.src3) {\n e.vmovsd(e.xmm1, i.src3);\n src3 = e.xmm1;\n }\n }\n\n // Multiply operation is commutative.\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulsd(dest, src1, src2); // $0 = $1 * $2\n });\n\n e.vsubsd(i.dest, i.dest, src3); // $0 = $1 - $2\n }\n }\n};\nstruct MUL_SUB_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // FMA extension\n if (e.IsFeatureEnabled(kX64EmitFMA)) {\n EmitCommutativeBinaryXmmOp(e, i,\n [&i](X64Emitter& e, const Xmm& dest,\n const Xmm& src1, const Xmm& src2) {\n Xmm src3 =\n i.src3.is_constant ? e.xmm1 : i.src3;\n if (i.src3.is_constant) {\n e.LoadConstantXmm(src3, i.src3.constant());\n }\n if (i.dest == src1) {\n e.vfmsub213ps(i.dest, src2, src3);\n } else if (i.dest == src2) {\n e.vfmsub213ps(i.dest, src1, src3);\n } else if (i.dest == i.src3) {\n e.vfmsub231ps(i.dest, src1, src2);\n } else {\n // Dest not equal to anything\n e.vmovdqa(i.dest, src1);\n e.vfmsub213ps(i.dest, src2, src3);\n }\n });\n } else {\n Xmm src3;\n if (i.src3.is_constant) {\n src3 = e.xmm1;\n e.LoadConstantXmm(src3, i.src3.constant());\n } else {\n // If i.dest == i.src3, back up i.src3 so we don't overwrite it.\n src3 = i.src3;\n if (i.dest == i.src3) {\n e.vmovdqa(e.xmm1, i.src3);\n src3 = e.xmm1;\n }\n }\n\n // Multiply operation is commutative.\n EmitCommutativeBinaryXmmOp(\n e, i, [&i](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vmulps(dest, src1, src2); // $0 = $1 * $2\n });\n\n e.vsubps(i.dest, i.dest, src3); // $0 = $1 - $2\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_MUL_SUB, MUL_SUB_F32, MUL_SUB_F64, MUL_SUB_V128);\n\n// ============================================================================\n// OPCODE_NEG\n// ============================================================================\n// TODO(benvanik): put dest/src1 together.\ntemplate \nvoid EmitNegXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitUnaryOp(e, i,\n [](X64Emitter& e, const REG& dest_src) { e.neg(dest_src); });\n}\nstruct NEG_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitNegXX(e, i);\n }\n};\nstruct NEG_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitNegXX(e, i);\n }\n};\nstruct NEG_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitNegXX(e, i);\n }\n};\nstruct NEG_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitNegXX(e, i);\n }\n};\nstruct NEG_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vxorps(i.dest, i.src1, e.GetXmmConstPtr(XMMSignMaskPS));\n }\n};\nstruct NEG_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vxorpd(i.dest, i.src1, e.GetXmmConstPtr(XMMSignMaskPD));\n }\n};\nstruct NEG_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_true(!i.instr->flags);\n e.vxorps(i.dest, i.src1, e.GetXmmConstPtr(XMMSignMaskPS));\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_NEG, NEG_I8, NEG_I16, NEG_I32, NEG_I64, NEG_F32,\n NEG_F64, NEG_V128);\n\n// ============================================================================\n// OPCODE_ABS\n// ============================================================================\nstruct ABS_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vpand(i.dest, i.src1, e.GetXmmConstPtr(XMMAbsMaskPS));\n }\n};\nstruct ABS_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vpand(i.dest, i.src1, e.GetXmmConstPtr(XMMAbsMaskPD));\n }\n};\nstruct ABS_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vpand(i.dest, i.src1, e.GetXmmConstPtr(XMMAbsMaskPS));\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ABS, ABS_F32, ABS_F64, ABS_V128);\n\n// ============================================================================\n// OPCODE_SQRT\n// ============================================================================\nstruct SQRT_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vsqrtss(i.dest, i.src1);\n }\n};\nstruct SQRT_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vsqrtsd(i.dest, i.src1);\n }\n};\nstruct SQRT_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.vsqrtps(i.dest, i.src1);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SQRT, SQRT_F32, SQRT_F64, SQRT_V128);\n\n// ============================================================================\n// OPCODE_RSQRT\n// ============================================================================\n// Altivec guarantees an error of < 1/4096 for vrsqrtefp while AVX only gives\n// < 1.5*2^-12 ≈ 1/2730 for vrsqrtps.\nstruct RSQRT_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vrsqrt14ss(i.dest, i.src1, i.src1);\n } else {\n e.vmovaps(e.xmm0, e.GetXmmConstPtr(XMMOne));\n e.vsqrtss(e.xmm1, i.src1, i.src1);\n e.vdivss(i.dest, e.xmm0, e.xmm1);\n }\n }\n};\nstruct RSQRT_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vrsqrt14sd(i.dest, i.src1, i.src1);\n } else {\n e.vmovapd(e.xmm0, e.GetXmmConstPtr(XMMOnePD));\n e.vsqrtsd(e.xmm1, i.src1, i.src1);\n e.vdivsd(i.dest, e.xmm0, e.xmm1);\n }\n }\n};\nstruct RSQRT_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vrsqrt14ps(i.dest, i.src1);\n } else {\n e.vmovaps(e.xmm0, e.GetXmmConstPtr(XMMOne));\n e.vsqrtps(e.xmm1, i.src1);\n e.vdivps(i.dest, e.xmm0, e.xmm1);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_RSQRT, RSQRT_F32, RSQRT_F64, RSQRT_V128);\n\n// ============================================================================\n// OPCODE_RECIP\n// ============================================================================\n// Altivec guarantees an error of < 1/4096 for vrefp while AVX only gives\n// < 1.5*2^-12 ≈ 1/2730 for rcpps. This breaks camp, horse and random event\n// spawning, breaks cactus collision as well as flickering grass in 5454082B\nstruct RECIP_F32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vrcp14ss(i.dest, i.src1, i.src1);\n } else {\n e.vmovaps(e.xmm0, e.GetXmmConstPtr(XMMOne));\n e.vdivss(i.dest, e.xmm0, i.src1);\n }\n }\n};\nstruct RECIP_F64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vrcp14sd(i.dest, i.src1, i.src1);\n } else {\n e.vmovapd(e.xmm0, e.GetXmmConstPtr(XMMOnePD));\n e.vdivsd(i.dest, e.xmm0, i.src1);\n }\n }\n};\nstruct RECIP_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vrcp14ps(i.dest, i.src1);\n } else {\n e.vmovaps(e.xmm0, e.GetXmmConstPtr(XMMOne));\n e.vdivps(i.dest, e.xmm0, i.src1);\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_RECIP, RECIP_F32, RECIP_F64, RECIP_V128);\n\n// ============================================================================\n// OPCODE_POW2\n// ============================================================================\n// TODO(benvanik): use approx here:\n// https://jrfonseca.blogspot.com/2008/09/fast-sse2-pow-tables-or-polynomials.html\nstruct POW2_F32 : Sequence> {\n static __m128 EmulatePow2(void*, __m128 src) {\n float src_value;\n _mm_store_ss(&src_value, src);\n float result = std::exp2(src_value);\n return _mm_load_ss(&result);\n }\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_always();\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulatePow2));\n e.vmovaps(i.dest, e.xmm0);\n }\n};\nstruct POW2_F64 : Sequence> {\n static __m128d EmulatePow2(void*, __m128d src) {\n double src_value;\n _mm_store_sd(&src_value, src);\n double result = std::exp2(src_value);\n return _mm_load_sd(&result);\n }\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_always();\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulatePow2));\n e.vmovaps(i.dest, e.xmm0);\n }\n};\nstruct POW2_V128 : Sequence> {\n static __m128 EmulatePow2(void*, __m128 src) {\n alignas(16) float values[4];\n _mm_store_ps(values, src);\n for (size_t i = 0; i < 4; ++i) {\n values[i] = std::exp2(values[i]);\n }\n return _mm_load_ps(values);\n }\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulatePow2));\n e.vmovaps(i.dest, e.xmm0);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_POW2, POW2_F32, POW2_F64, POW2_V128);\n\n// ============================================================================\n// OPCODE_LOG2\n// ============================================================================\n// TODO(benvanik): use approx here:\n// https://jrfonseca.blogspot.com/2008/09/fast-sse2-pow-tables-or-polynomials.html\n// TODO(benvanik): this emulated fn destroys all xmm registers! don't do it!\nstruct LOG2_F32 : Sequence> {\n static __m128 EmulateLog2(void*, __m128 src) {\n float src_value;\n _mm_store_ss(&src_value, src);\n float result = std::log2(src_value);\n return _mm_load_ss(&result);\n }\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_always();\n if (i.src1.is_constant) {\n e.lea(e.GetNativeParam(0), e.StashConstantXmm(0, i.src1.constant()));\n } else {\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n }\n e.CallNativeSafe(reinterpret_cast(EmulateLog2));\n e.vmovaps(i.dest, e.xmm0);\n }\n};\nstruct LOG2_F64 : Sequence> {\n static __m128d EmulateLog2(void*, __m128d src) {\n double src_value;\n _mm_store_sd(&src_value, src);\n double result = std::log2(src_value);\n return _mm_load_sd(&result);\n }\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n assert_always();\n if (i.src1.is_constant) {\n e.lea(e.GetNativeParam(0), e.StashConstantXmm(0, i.src1.constant()));\n } else {\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n }\n e.CallNativeSafe(reinterpret_cast(EmulateLog2));\n e.vmovaps(i.dest, e.xmm0);\n }\n};\nstruct LOG2_V128 : Sequence> {\n static __m128 EmulateLog2(void*, __m128 src) {\n alignas(16) float values[4];\n _mm_store_ps(values, src);\n for (size_t i = 0; i < 4; ++i) {\n values[i] = std::log2(values[i]);\n }\n return _mm_load_ps(values);\n }\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (i.src1.is_constant) {\n e.lea(e.GetNativeParam(0), e.StashConstantXmm(0, i.src1.constant()));\n } else {\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n }\n e.CallNativeSafe(reinterpret_cast(EmulateLog2));\n e.vmovaps(i.dest, e.xmm0);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_LOG2, LOG2_F32, LOG2_F64, LOG2_V128);\n\nstruct DOT_PRODUCT_V128 {\n static void Emit(X64Emitter& e, Xmm dest, Xmm src1, Xmm src2, uint8_t imm) {\n // TODO(benvanik): apparently this is very slow\n // - find alternative?\n Xbyak::Label end;\n e.inLocalLabel();\n\n // Grab space to put MXCSR.\n // TODO(gibbed): stick this in TLS or\n // something?\n e.sub(e.rsp, 8);\n\n // Grab MXCSR and mask off the overflow flag,\n // because it's sticky.\n e.vstmxcsr(e.dword[e.rsp]);\n e.mov(e.eax, e.dword[e.rsp]);\n e.and_(e.eax, uint32_t(~8));\n e.mov(e.dword[e.rsp], e.eax);\n e.vldmxcsr(e.dword[e.rsp]);\n\n // Hey we can do the dot product now.\n e.vdpps(dest, src1, src2, imm);\n\n // Load MXCSR...\n e.vstmxcsr(e.dword[e.rsp]);\n\n // ..free our temporary space and get MXCSR at\n // the same time\n e.pop(e.rax);\n\n // Did we overflow?\n e.test(e.al, 8);\n e.jz(end);\n\n // Infinity? HA! Give NAN.\n e.vmovdqa(dest, e.GetXmmConstPtr(XMMQNaN));\n\n e.L(end);\n e.outLocalLabel();\n }\n};\n\n// ============================================================================\n// OPCODE_DOT_PRODUCT_3\n// ============================================================================\nstruct DOT_PRODUCT_3_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // https://msdn.microsoft.com/en-us/library/bb514054(v=vs.90).aspx\n EmitCommutativeBinaryXmmOp(\n e, i, [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n DOT_PRODUCT_V128::Emit(e, dest, src1, src2, 0b01110001);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_DOT_PRODUCT_3, DOT_PRODUCT_3_V128);\n\n// ============================================================================\n// OPCODE_DOT_PRODUCT_4\n// ============================================================================\nstruct DOT_PRODUCT_4_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // https://msdn.microsoft.com/en-us/library/bb514054(v=vs.90).aspx\n EmitCommutativeBinaryXmmOp(\n e, i, [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n DOT_PRODUCT_V128::Emit(e, dest, src1, src2, 0b11110001);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_DOT_PRODUCT_4, DOT_PRODUCT_4_V128);\n\n// ============================================================================\n// OPCODE_AND\n// ============================================================================\n// TODO(benvanik): put dest/src1|2 together.\ntemplate \nvoid EmitAndXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const REG& src) {\n e.and_(dest_src, src);\n },\n [](X64Emitter& e, const REG& dest_src, int32_t constant) {\n e.and_(dest_src, constant);\n });\n}\nstruct AND_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAndXX(e, i);\n }\n};\nstruct AND_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAndXX(e, i);\n }\n};\nstruct AND_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAndXX(e, i);\n }\n};\nstruct AND_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAndXX(e, i);\n }\n};\nstruct AND_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vpand(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_AND, AND_I8, AND_I16, AND_I32, AND_I64, AND_V128);\n\n// ============================================================================\n// OPCODE_AND_NOT\n// ============================================================================\ntemplate \nvoid EmitAndNotXX(X64Emitter& e, const ARGS& i) {\n if (i.src1.is_constant) {\n // src1 constant.\n // `and` instruction only supports up to 32-bit immediate constants\n // 64-bit constants will need a temp register\n if (i.dest.reg().getBit() == 64) {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src1.constant());\n\n if (e.IsFeatureEnabled(kX64EmitBMI1)) {\n if (i.dest.reg().getBit() == 64) {\n e.andn(i.dest.reg().cvt64(), i.src2.reg().cvt64(), temp.cvt64());\n } else {\n e.andn(i.dest.reg().cvt32(), i.src2.reg().cvt32(), temp.cvt32());\n }\n } else {\n e.mov(i.dest, i.src2);\n e.not_(i.dest);\n e.and_(i.dest, temp);\n }\n } else {\n e.mov(i.dest, i.src2);\n e.not_(i.dest);\n e.and_(i.dest, uint32_t(i.src1.constant()));\n }\n } else if (i.src2.is_constant) {\n // src2 constant.\n if (i.dest == i.src1) {\n auto temp = GetTempReg(e);\n e.mov(temp, ~i.src2.constant());\n e.and_(i.dest, temp);\n } else {\n e.mov(i.dest, i.src1);\n auto temp = GetTempReg(e);\n e.mov(temp, ~i.src2.constant());\n e.and_(i.dest, temp);\n }\n } else {\n // neither are constant\n if (e.IsFeatureEnabled(kX64EmitBMI1)) {\n if (i.dest.reg().getBit() == 64) {\n e.andn(i.dest.reg().cvt64(), i.src2.reg().cvt64(),\n i.src1.reg().cvt64());\n } else {\n e.andn(i.dest.reg().cvt32(), i.src2.reg().cvt32(),\n i.src1.reg().cvt32());\n }\n } else {\n if (i.dest == i.src2) {\n e.not_(i.dest);\n e.and_(i.dest, i.src1);\n } else if (i.dest == i.src1) {\n auto temp = GetTempReg(e);\n e.mov(temp, i.src2);\n e.not_(temp);\n e.and_(i.dest, temp);\n } else {\n e.mov(i.dest, i.src2);\n e.not_(i.dest);\n e.and_(i.dest, i.src1);\n }\n }\n }\n}\nstruct AND_NOT_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAndNotXX(e, i);\n }\n};\nstruct AND_NOT_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAndNotXX(e, i);\n }\n};\nstruct AND_NOT_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAndNotXX(e, i);\n }\n};\nstruct AND_NOT_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitAndNotXX(e, i);\n }\n};\nstruct AND_NOT_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vpandn(dest, src2, src1);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_AND_NOT, AND_NOT_I8, AND_NOT_I16, AND_NOT_I32,\n AND_NOT_I64, AND_NOT_V128);\n\n// ============================================================================\n// OPCODE_OR\n// ============================================================================\n// TODO(benvanik): put dest/src1|2 together.\ntemplate \nvoid EmitOrXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const REG& src) {\n e.or_(dest_src, src);\n },\n [](X64Emitter& e, const REG& dest_src, int32_t constant) {\n e.or_(dest_src, constant);\n });\n}\nstruct OR_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitOrXX(e, i);\n }\n};\nstruct OR_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitOrXX(e, i);\n }\n};\nstruct OR_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitOrXX(e, i);\n }\n};\nstruct OR_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitOrXX(e, i);\n }\n};\nstruct OR_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vpor(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_OR, OR_I8, OR_I16, OR_I32, OR_I64, OR_V128);\n\n// ============================================================================\n// OPCODE_XOR\n// ============================================================================\n// TODO(benvanik): put dest/src1|2 together.\ntemplate \nvoid EmitXorXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitCommutativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const REG& src) {\n e.xor_(dest_src, src);\n },\n [](X64Emitter& e, const REG& dest_src, int32_t constant) {\n e.xor_(dest_src, constant);\n });\n}\nstruct XOR_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitXorXX(e, i);\n }\n};\nstruct XOR_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitXorXX(e, i);\n }\n};\nstruct XOR_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitXorXX(e, i);\n }\n};\nstruct XOR_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitXorXX(e, i);\n }\n};\nstruct XOR_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitCommutativeBinaryXmmOp(e, i,\n [](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {\n e.vpxor(dest, src1, src2);\n });\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_XOR, XOR_I8, XOR_I16, XOR_I32, XOR_I64, XOR_V128);\n\n// ============================================================================\n// OPCODE_NOT\n// ============================================================================\n// TODO(benvanik): put dest/src1 together.\ntemplate \nvoid EmitNotXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitUnaryOp(\n e, i, [](X64Emitter& e, const REG& dest_src) { e.not_(dest_src); });\n}\nstruct NOT_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitNotXX(e, i);\n }\n};\nstruct NOT_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitNotXX(e, i);\n }\n};\nstruct NOT_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitNotXX(e, i);\n }\n};\nstruct NOT_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitNotXX(e, i);\n }\n};\nstruct NOT_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitAVX512Ortho)) {\n e.vpternlogd(i.dest, i.src1, i.src1, 0b01010101);\n return;\n }\n // dest = src ^ 0xFFFF...\n e.vpxor(i.dest, i.src1, e.GetXmmConstPtr(XMMFFFF /* FF... */));\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_NOT, NOT_I8, NOT_I16, NOT_I32, NOT_I64, NOT_V128);\n\n// ============================================================================\n// OPCODE_SHL\n// ============================================================================\n// TODO(benvanik): optimize common shifts.\ntemplate \nvoid EmitShlXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitAssociativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const Reg8& src) {\n // shlx: $1 = $2 << $3\n // shl: $1 = $1 << $2\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n if (dest_src.getBit() == 64) {\n e.shlx(dest_src.cvt64(), dest_src.cvt64(), src.cvt64());\n } else {\n e.shlx(dest_src.cvt32(), dest_src.cvt32(), src.cvt32());\n }\n } else {\n e.mov(e.cl, src);\n e.shl(dest_src, e.cl);\n }\n },\n [](X64Emitter& e, const REG& dest_src, int8_t constant) {\n e.shl(dest_src, constant);\n });\n}\nstruct SHL_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitShlXX(e, i);\n }\n};\nstruct SHL_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitShlXX(e, i);\n }\n};\nstruct SHL_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitShlXX(e, i);\n }\n};\nstruct SHL_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitShlXX(e, i);\n }\n};\nstruct SHL_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): native version (with shift magic).\n if (i.src2.is_constant) {\n e.mov(e.GetNativeParam(1), i.src2.constant());\n } else {\n e.mov(e.GetNativeParam(1), i.src2);\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulateShlV128));\n e.vmovaps(i.dest, e.xmm0);\n }\n static __m128i EmulateShlV128(void*, __m128i src1, uint8_t src2) {\n // Almost all instances are shamt = 1, but non-constant.\n // shamt is [0,7]\n uint8_t shamt = src2 & 0x7;\n alignas(16) vec128_t value;\n _mm_store_si128(reinterpret_cast<__m128i*>(&value), src1);\n for (int i = 0; i < 15; ++i) {\n value.u8[i ^ 0x3] = (value.u8[i ^ 0x3] << shamt) |\n (value.u8[(i + 1) ^ 0x3] >> (8 - shamt));\n }\n value.u8[15 ^ 0x3] = value.u8[15 ^ 0x3] << shamt;\n return _mm_load_si128(reinterpret_cast<__m128i*>(&value));\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SHL, SHL_I8, SHL_I16, SHL_I32, SHL_I64, SHL_V128);\n\n// ============================================================================\n// OPCODE_SHR\n// ============================================================================\n// TODO(benvanik): optimize common shifts.\ntemplate \nvoid EmitShrXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitAssociativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const Reg8& src) {\n // shrx: op1 dest, op2 src, op3 count\n // shr: op1 src/dest, op2 count\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n if (dest_src.getBit() == 64) {\n e.shrx(dest_src.cvt64(), dest_src.cvt64(), src.cvt64());\n } else if (dest_src.getBit() == 32) {\n e.shrx(dest_src.cvt32(), dest_src.cvt32(), src.cvt32());\n } else {\n e.movzx(dest_src.cvt32(), dest_src);\n e.shrx(dest_src.cvt32(), dest_src.cvt32(), src.cvt32());\n }\n } else {\n e.mov(e.cl, src);\n e.shr(dest_src, e.cl);\n }\n },\n [](X64Emitter& e, const REG& dest_src, int8_t constant) {\n e.shr(dest_src, constant);\n });\n}\nstruct SHR_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitShrXX(e, i);\n }\n};\nstruct SHR_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitShrXX(e, i);\n }\n};\nstruct SHR_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitShrXX(e, i);\n }\n};\nstruct SHR_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitShrXX(e, i);\n }\n};\nstruct SHR_V128 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): native version (with shift magic).\n if (i.src2.is_constant) {\n e.mov(e.GetNativeParam(1), i.src2.constant());\n } else {\n e.mov(e.GetNativeParam(1), i.src2);\n }\n e.lea(e.GetNativeParam(0), e.StashXmm(0, i.src1));\n e.CallNativeSafe(reinterpret_cast(EmulateShrV128));\n e.vmovaps(i.dest, e.xmm0);\n }\n static __m128i EmulateShrV128(void*, __m128i src1, uint8_t src2) {\n // Almost all instances are shamt = 1, but non-constant.\n // shamt is [0,7]\n uint8_t shamt = src2 & 0x7;\n alignas(16) vec128_t value;\n _mm_store_si128(reinterpret_cast<__m128i*>(&value), src1);\n for (int i = 15; i > 0; --i) {\n value.u8[i ^ 0x3] = (value.u8[i ^ 0x3] >> shamt) |\n (value.u8[(i - 1) ^ 0x3] << (8 - shamt));\n }\n value.u8[0 ^ 0x3] = value.u8[0 ^ 0x3] >> shamt;\n return _mm_load_si128(reinterpret_cast<__m128i*>(&value));\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SHR, SHR_I8, SHR_I16, SHR_I32, SHR_I64, SHR_V128);\n\n// ============================================================================\n// OPCODE_SHA\n// ============================================================================\n// TODO(benvanik): optimize common shifts.\ntemplate \nvoid EmitSarXX(X64Emitter& e, const ARGS& i) {\n SEQ::EmitAssociativeBinaryOp(\n e, i,\n [](X64Emitter& e, const REG& dest_src, const Reg8& src) {\n if (e.IsFeatureEnabled(kX64EmitBMI2)) {\n if (dest_src.getBit() == 64) {\n e.sarx(dest_src.cvt64(), dest_src.cvt64(), src.cvt64());\n } else if (dest_src.getBit() == 32) {\n e.sarx(dest_src.cvt32(), dest_src.cvt32(), src.cvt32());\n } else {\n e.movsx(dest_src.cvt32(), dest_src);\n e.sarx(dest_src.cvt32(), dest_src.cvt32(), src.cvt32());\n }\n } else {\n e.mov(e.cl, src);\n e.sar(dest_src, e.cl);\n }\n },\n [](X64Emitter& e, const REG& dest_src, int8_t constant) {\n e.sar(dest_src, constant);\n });\n}\nstruct SHA_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitSarXX(e, i);\n }\n};\nstruct SHA_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitSarXX(e, i);\n }\n};\nstruct SHA_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitSarXX(e, i);\n }\n};\nstruct SHA_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitSarXX(e, i);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SHA, SHA_I8, SHA_I16, SHA_I32, SHA_I64);\n\n// ============================================================================\n// OPCODE_ROTATE_LEFT\n// ============================================================================\n// TODO(benvanik): put dest/src1 together, src2 in cl.\ntemplate \nvoid EmitRotateLeftXX(X64Emitter& e, const ARGS& i) {\n if (i.src2.is_constant) {\n // Constant rotate.\n if (i.dest != i.src1) {\n if (i.src1.is_constant) {\n e.mov(i.dest, i.src1.constant());\n } else {\n e.mov(i.dest, i.src1);\n }\n }\n e.rol(i.dest, i.src2.constant());\n } else {\n // Variable rotate.\n if (i.src2.reg().getIdx() != e.cl.getIdx()) {\n e.mov(e.cl, i.src2);\n }\n if (i.dest != i.src1) {\n if (i.src1.is_constant) {\n e.mov(i.dest, i.src1.constant());\n } else {\n e.mov(i.dest, i.src1);\n }\n }\n e.rol(i.dest, e.cl);\n }\n}\nstruct ROTATE_LEFT_I8\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitRotateLeftXX(e, i);\n }\n};\nstruct ROTATE_LEFT_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitRotateLeftXX(e, i);\n }\n};\nstruct ROTATE_LEFT_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitRotateLeftXX(e, i);\n }\n};\nstruct ROTATE_LEFT_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitRotateLeftXX(e, i);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_ROTATE_LEFT, ROTATE_LEFT_I8, ROTATE_LEFT_I16,\n ROTATE_LEFT_I32, ROTATE_LEFT_I64);\n\n// ============================================================================\n// OPCODE_BYTE_SWAP\n// ============================================================================\n// TODO(benvanik): put dest/src1 together.\nstruct BYTE_SWAP_I16\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitUnaryOp(\n e, i, [](X64Emitter& e, const Reg16& dest_src) { e.ror(dest_src, 8); });\n }\n};\nstruct BYTE_SWAP_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitUnaryOp(\n e, i, [](X64Emitter& e, const Reg32& dest_src) { e.bswap(dest_src); });\n }\n};\nstruct BYTE_SWAP_I64\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n EmitUnaryOp(\n e, i, [](X64Emitter& e, const Reg64& dest_src) { e.bswap(dest_src); });\n }\n};\nstruct BYTE_SWAP_V128\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n // TODO(benvanik): find a way to do this without the memory load.\n e.vpshufb(i.dest, i.src1, e.GetXmmConstPtr(XMMByteSwapMask));\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_BYTE_SWAP, BYTE_SWAP_I16, BYTE_SWAP_I32,\n BYTE_SWAP_I64, BYTE_SWAP_V128);\n\n// ============================================================================\n// OPCODE_CNTLZ\n// Count leading zeroes\n// ============================================================================\nstruct CNTLZ_I8 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitLZCNT)) {\n // No 8bit lzcnt, so do 16 and sub 8.\n e.movzx(i.dest.reg().cvt16(), i.src1);\n e.lzcnt(i.dest.reg().cvt16(), i.dest.reg().cvt16());\n e.sub(i.dest, 8);\n } else {\n Xbyak::Label end;\n e.inLocalLabel();\n\n e.bsr(e.rax, i.src1); // ZF set if i.src1 is 0\n e.mov(i.dest, 0x8);\n e.jz(end);\n\n e.xor_(e.rax, 0x7);\n e.mov(i.dest, e.rax);\n\n e.L(end);\n e.outLocalLabel();\n }\n }\n};\nstruct CNTLZ_I16 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitLZCNT)) {\n // LZCNT: searches $2 until MSB 1 found, stores idx (from last bit) in $1\n e.lzcnt(i.dest.reg().cvt32(), i.src1);\n } else {\n Xbyak::Label end;\n e.inLocalLabel();\n\n e.bsr(e.rax, i.src1); // ZF set if i.src1 is 0\n e.mov(i.dest, 0x10);\n e.jz(end);\n\n e.xor_(e.rax, 0x0F);\n e.mov(i.dest, e.rax);\n\n e.L(end);\n e.outLocalLabel();\n }\n }\n};\nstruct CNTLZ_I32 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitLZCNT)) {\n e.lzcnt(i.dest.reg().cvt32(), i.src1);\n } else {\n Xbyak::Label end;\n e.inLocalLabel();\n\n e.bsr(e.rax, i.src1); // ZF set if i.src1 is 0\n e.mov(i.dest, 0x20);\n e.jz(end);\n\n e.xor_(e.rax, 0x1F);\n e.mov(i.dest, e.rax);\n\n e.L(end);\n e.outLocalLabel();\n }\n }\n};\nstruct CNTLZ_I64 : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n if (e.IsFeatureEnabled(kX64EmitLZCNT)) {\n e.lzcnt(i.dest.reg().cvt64(), i.src1);\n } else {\n Xbyak::Label end;\n e.inLocalLabel();\n\n e.bsr(e.rax, i.src1); // ZF set if i.src1 is 0\n e.mov(i.dest, 0x40);\n e.jz(end);\n\n e.xor_(e.rax, 0x3F);\n e.mov(i.dest, e.rax);\n\n e.L(end);\n e.outLocalLabel();\n }\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_CNTLZ, CNTLZ_I8, CNTLZ_I16, CNTLZ_I32, CNTLZ_I64);\n\n// ============================================================================\n// OPCODE_SET_ROUNDING_MODE\n// ============================================================================\n// Input: FPSCR (PPC format)\nstatic const uint32_t mxcsr_table[] = {\n 0x1F80, 0x7F80, 0x5F80, 0x3F80, 0x9F80, 0xFF80, 0xDF80, 0xBF80,\n};\nstruct SET_ROUNDING_MODE_I32\n : Sequence> {\n static void Emit(X64Emitter& e, const EmitArgType& i) {\n e.mov(e.rcx, i.src1);\n e.and_(e.rcx, 0x7);\n e.mov(e.rax, uintptr_t(mxcsr_table));\n e.vldmxcsr(e.ptr[e.rax + e.rcx * 4]);\n }\n};\nEMITTER_OPCODE_TABLE(OPCODE_SET_ROUNDING_MODE, SET_ROUNDING_MODE_I32);\n\n// Include anchors to other sequence sources so they get included in the build.\nextern volatile int anchor_control;\nstatic int anchor_control_dest = anchor_control;\n\nextern volatile int anchor_memory;\nstatic int anchor_memory_dest = anchor_memory;\n\nextern volatile int anchor_vector;\nstatic int anchor_vector_dest = anchor_vector;\n\nbool SelectSequence(X64Emitter* e, const Instr* i, const Instr** new_tail) {\n const InstrKey key(i);\n auto it = sequence_table.find(key);\n if (it != sequence_table.end()) {\n if (it->second(*e, i)) {\n *new_tail = i->next;\n return true;\n }\n }\n XELOGE(\"No sequence match for variant {}\", i->opcode->name);\n return false;\n}\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_sequences.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_SEQUENCES_H_\n#define XENIA_CPU_BACKEND_X64_X64_SEQUENCES_H_\n\n#include \"xenia/cpu/hir/instr.h\"\n\n#include \n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nclass X64Emitter;\n\ntypedef bool (*SequenceSelectFn)(X64Emitter&, const hir::Instr*);\nextern std::unordered_map sequence_table;\n\ntemplate \nbool Register() {\n sequence_table.insert({T::head_key(), T::Select});\n return true;\n}\n\ntemplate \nstatic bool Register() {\n bool b = true;\n b = b && Register(); // Call the above function\n b = b && Register(); // Call ourself again (recursively)\n return b;\n}\n#define EMITTER_OPCODE_TABLE(name, ...) \\\n const auto X64_INSTR_##name = Register<__VA_ARGS__>();\n\nbool SelectSequence(X64Emitter* e, const hir::Instr* i,\n const hir::Instr** new_tail);\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_SEQUENCES_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_stack_layout.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_STACK_LAYOUT_H_\n#define XENIA_CPU_BACKEND_X64_X64_STACK_LAYOUT_H_\n\n#include \"xenia/base/vec128.h\"\n#include \"xenia/cpu/backend/x64/x64_backend.h\"\n#include \"xenia/cpu/backend/x64/x64_emitter.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\nclass StackLayout {\n public:\n /**\n * Stack Layout\n * ----------------------------\n * NOTE: stack must always be 16b aligned.\n *\n * Thunk stack:\n * Non-Volatile Volatile\n * +------------------+------------------+\n * | arg temp, 3 * 8 | arg temp, 3 * 8 | rsp + 0x000\n * | | |\n * | | |\n * +------------------+------------------+\n * | rbx | (unused) | rsp + 0x018\n * +------------------+------------------+\n * | rbp | rcx | rsp + 0x020\n * +------------------+------------------+\n * | rcx (Win32) | rdx | rsp + 0x028\n * +------------------+------------------+\n * | rsi (Win32) | rsi (Linux) | rsp + 0x030\n * +------------------+------------------+\n * | rdi (Win32) | rdi (Linux) | rsp + 0x038\n * +------------------+------------------+\n * | r12 | r8 | rsp + 0x040\n * +------------------+------------------+\n * | r13 | r9 | rsp + 0x048\n * +------------------+------------------+\n * | r14 | r10 | rsp + 0x050\n * +------------------+------------------+\n * | r15 | r11 | rsp + 0x058\n * +------------------+------------------+\n * | xmm6 (Win32) | (unused) | rsp + 0x060\n * | | |\n * +------------------+------------------+\n * | xmm7 (Win32) | xmm1 | rsp + 0x070\n * | | |\n * +------------------+------------------+\n * | xmm8 (Win32) | xmm2 | rsp + 0x080\n * | | |\n * +------------------+------------------+\n * | xmm9 (Win32) | xmm3 | rsp + 0x090\n * | | |\n * +------------------+------------------+\n * | xmm10 (Win32) | xmm4 | rsp + 0x0A0\n * | | |\n * +------------------+------------------+\n * | xmm11 (Win32) | xmm5 | rsp + 0x0B0\n * | | |\n * +------------------+------------------+\n * | xmm12 (Win32) | (unused) | rsp + 0x0C0\n * | | |\n * +------------------+------------------+\n * | xmm13 (Win32) | (unused) | rsp + 0x0D0\n * | | |\n * +------------------+------------------+\n * | xmm14 (Win32) | (unused) | rsp + 0x0E0\n * | | |\n * +------------------+------------------+\n * | xmm15 (Win32) | (unused) | rsp + 0x0F0\n * | | |\n * +------------------+------------------+\n * | (return address) | (return address) | rsp + 0x100\n * +------------------+------------------+\n * | (rcx home) | (rcx home) | rsp + 0x108\n * +------------------+------------------+\n * | (rdx home) | (rdx home) | rsp + 0x110\n * +------------------+------------------+\n */\n XEPACKEDSTRUCT(Thunk, {\n uint64_t arg_temp[3];\n uint64_t r[9];\n vec128_t xmm[10];\n });\n static_assert(sizeof(Thunk) % 16 == 0,\n \"sizeof(Thunk) must be a multiple of 16!\");\n static const size_t THUNK_STACK_SIZE = sizeof(Thunk) + 8;\n\n /**\n *\n *\n * Guest stack:\n * +------------------+\n * | arg temp, 3 * 8 | rsp + 0\n * | |\n * | |\n * +------------------+\n * | scratch, 48b | rsp + 32\n * | |\n * +------------------+\n * | rcx / context | rsp + 80\n * +------------------+\n * | guest ret addr | rsp + 88\n * +------------------+\n * | call ret addr | rsp + 96\n * +------------------+\n * ... locals ...\n * +------------------+\n * | (return address) |\n * +------------------+\n *\n */\n static const size_t GUEST_STACK_SIZE = 104;\n static const size_t GUEST_CTX_HOME = 80;\n static const size_t GUEST_RET_ADDR = 88;\n static const size_t GUEST_CALL_RET_ADDR = 96;\n};\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_STACK_LAYOUT_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_tracers.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/backend/x64/x64_tracers.h\"\n\n#include \n\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/vec128.h\"\n#include \"xenia/cpu/backend/x64/x64_emitter.h\"\n#include \"xenia/cpu/processor.h\"\n#include \"xenia/cpu/thread_state.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\n#define ITRACE 0\n#define DTRACE 0\n\n#define TARGET_THREAD 0\n\nbool trace_enabled = true;\n\n#define THREAD_MATCH \\\n (!TARGET_THREAD || thread_state->thread_id() == TARGET_THREAD)\n#define IFLUSH()\n#define IPRINT(s) \\\n if (trace_enabled && THREAD_MATCH) \\\n xe::logging::AppendLogLine(xe::LogLevel::Debug, 't', s)\n#define DFLUSH()\n#define DPRINT(...) \\\n if (trace_enabled && THREAD_MATCH) \\\n xe::logging::AppendLogLineFormat(xe::LogLevel::Debug, 't', __VA_ARGS__)\n\nuint32_t GetTracingMode() {\n uint32_t mode = 0;\n#if ITRACE\n mode |= TRACING_INSTR;\n#endif // ITRACE\n#if DTRACE\n mode |= TRACING_DATA;\n#endif // DTRACE\n return mode;\n}\n\nvoid TraceString(void* raw_context, const char* str) {\n auto thread_state = *reinterpret_cast(raw_context);\n IPRINT(str);\n IFLUSH();\n}\n\nvoid TraceContextLoadI8(void* raw_context, uint64_t offset, uint8_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = ctx i8 +{}\\n\", (int8_t)value, value, offset);\n}\nvoid TraceContextLoadI16(void* raw_context, uint64_t offset, uint16_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = ctx i16 +{}\\n\", (int16_t)value, value, offset);\n}\nvoid TraceContextLoadI32(void* raw_context, uint64_t offset, uint32_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = ctx i32 +{}\\n\", (int32_t)value, value, offset);\n}\nvoid TraceContextLoadI64(void* raw_context, uint64_t offset, uint64_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = ctx i64 +{}\\n\", (int64_t)value, value, offset);\n}\nvoid TraceContextLoadF32(void* raw_context, uint64_t offset, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = ctx f32 +{}\\n\", xe::m128_f32<0>(value),\n xe::m128_i32<0>(value), offset);\n}\nvoid TraceContextLoadF64(void* raw_context, uint64_t offset,\n const double* value) {\n auto thread_state = *reinterpret_cast(raw_context);\n auto v = _mm_loadu_pd(value);\n DPRINT(\"{} ({:X}) = ctx f64 +{}\\n\", xe::m128_f64<0>(v), xe::m128_i64<0>(v),\n offset);\n}\nvoid TraceContextLoadV128(void* raw_context, uint64_t offset, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"[{}, {}, {}, {}] [{:08X}, {:08X}, {:08X}, {:08X}] = ctx v128 +{}\\n\",\n xe::m128_f32<0>(value), xe::m128_f32<1>(value), xe::m128_f32<2>(value),\n xe::m128_f32<3>(value), xe::m128_i32<0>(value), xe::m128_i32<1>(value),\n xe::m128_i32<2>(value), xe::m128_i32<3>(value), offset);\n}\n\nvoid TraceContextStoreI8(void* raw_context, uint64_t offset, uint8_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"ctx i8 +{} = {} ({:X})\\n\", offset, (int8_t)value, value);\n}\nvoid TraceContextStoreI16(void* raw_context, uint64_t offset, uint16_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"ctx i16 +{} = {} ({:X})\\n\", offset, (int16_t)value, value);\n}\nvoid TraceContextStoreI32(void* raw_context, uint64_t offset, uint32_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"ctx i32 +{} = {} ({:X})\\n\", offset, (int32_t)value, value);\n}\nvoid TraceContextStoreI64(void* raw_context, uint64_t offset, uint64_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"ctx i64 +{} = {} ({:X})\\n\", offset, (int64_t)value, value);\n}\nvoid TraceContextStoreF32(void* raw_context, uint64_t offset, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"ctx f32 +{} = {} ({:X})\\n\", offset, xe::m128_f32<0>(value),\n xe::m128_i32<0>(value));\n}\nvoid TraceContextStoreF64(void* raw_context, uint64_t offset,\n const double* value) {\n auto thread_state = *reinterpret_cast(raw_context);\n auto v = _mm_loadu_pd(value);\n DPRINT(\"ctx f64 +{} = {} ({:X})\\n\", offset, xe::m128_f64<0>(v),\n xe::m128_i64<0>(v));\n}\nvoid TraceContextStoreV128(void* raw_context, uint64_t offset, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"ctx v128 +{} = [{}, {}, {}, {}] [{:08X}, {:08X}, {:08X}, {:08X}]\\n\",\n offset, xe::m128_f32<0>(value), xe::m128_f32<1>(value),\n xe::m128_f32<2>(value), xe::m128_f32<3>(value), xe::m128_i32<0>(value),\n xe::m128_i32<1>(value), xe::m128_i32<2>(value),\n xe::m128_i32<3>(value));\n}\n\nvoid TraceMemoryLoadI8(void* raw_context, uint32_t address, uint8_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = load.i8 {:08X}\\n\", (int8_t)value, value, address);\n}\nvoid TraceMemoryLoadI16(void* raw_context, uint32_t address, uint16_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = load.i16 {:08X}\\n\", (int16_t)value, value, address);\n}\nvoid TraceMemoryLoadI32(void* raw_context, uint32_t address, uint32_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = load.i32 {:08X}\\n\", (int32_t)value, value, address);\n}\nvoid TraceMemoryLoadI64(void* raw_context, uint32_t address, uint64_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = load.i64 {:08X}\\n\", (int64_t)value, value, address);\n}\nvoid TraceMemoryLoadF32(void* raw_context, uint32_t address, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = load.f32 {:08X}\\n\", xe::m128_f32<0>(value),\n xe::m128_i32<0>(value), address);\n}\nvoid TraceMemoryLoadF64(void* raw_context, uint32_t address, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"{} ({:X}) = load.f64 {:08X}\\n\", xe::m128_f64<0>(value),\n xe::m128_i64<0>(value), address);\n}\nvoid TraceMemoryLoadV128(void* raw_context, uint32_t address, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\n \"[{}, {}, {}, {}] [{:08X}, {:08X}, {:08X}, {:08X}] = load.v128 {:08X}\\n\",\n xe::m128_f32<0>(value), xe::m128_f32<1>(value), xe::m128_f32<2>(value),\n xe::m128_f32<3>(value), xe::m128_i32<0>(value), xe::m128_i32<1>(value),\n xe::m128_i32<2>(value), xe::m128_i32<3>(value), address);\n}\n\nvoid TraceMemoryStoreI8(void* raw_context, uint32_t address, uint8_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"store.i8 {:08X} = {} ({:X})\\n\", address, (int8_t)value, value);\n}\nvoid TraceMemoryStoreI16(void* raw_context, uint32_t address, uint16_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"store.i16 {:08X} = {} ({:X})\\n\", address, (int16_t)value, value);\n}\nvoid TraceMemoryStoreI32(void* raw_context, uint32_t address, uint32_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"store.i32 {:08X} = {} ({:X})\\n\", address, (int32_t)value, value);\n}\nvoid TraceMemoryStoreI64(void* raw_context, uint32_t address, uint64_t value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"store.i64 {:08X} = {} ({:X})\\n\", address, (int64_t)value, value);\n}\nvoid TraceMemoryStoreF32(void* raw_context, uint32_t address, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"store.f32 {:08X} = {} ({:X})\\n\", address, xe::m128_f32<0>(value),\n xe::m128_i32<0>(value));\n}\nvoid TraceMemoryStoreF64(void* raw_context, uint32_t address, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"store.f64 {:08X} = {} ({:X})\\n\", address, xe::m128_f64<0>(value),\n xe::m128_i64<0>(value));\n}\nvoid TraceMemoryStoreV128(void* raw_context, uint32_t address, __m128 value) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\n \"store.v128 {:08X} = [{}, {}, {}, {}] [{:08X}, {:08X}, {:08X}, {:08X}]\\n\",\n address, xe::m128_f32<0>(value), xe::m128_f32<1>(value),\n xe::m128_f32<2>(value), xe::m128_f32<3>(value), xe::m128_i32<0>(value),\n xe::m128_i32<1>(value), xe::m128_i32<2>(value), xe::m128_i32<3>(value));\n}\n\nvoid TraceMemset(void* raw_context, uint32_t address, uint8_t value,\n uint32_t length) {\n auto thread_state = *reinterpret_cast(raw_context);\n DPRINT(\"memset {:08X}-{:08X} ({}) = {:02X}\", address, address + length,\n length, value);\n}\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/backend/x64/x64_tracers.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_TRACERS_H_\n#define XENIA_CPU_BACKEND_X64_X64_TRACERS_H_\n\n#include \n#include \n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\nclass X64Emitter;\n\nenum TracingMode {\n TRACING_INSTR = (1 << 1),\n TRACING_DATA = (1 << 2),\n};\n\nuint32_t GetTracingMode();\ninline bool IsTracingInstr() { return (GetTracingMode() & TRACING_INSTR) != 0; }\ninline bool IsTracingData() { return (GetTracingMode() & TRACING_DATA) != 0; }\n\nvoid TraceString(void* raw_context, const char* str);\n\nvoid TraceContextLoadI8(void* raw_context, uint64_t offset, uint8_t value);\nvoid TraceContextLoadI16(void* raw_context, uint64_t offset, uint16_t value);\nvoid TraceContextLoadI32(void* raw_context, uint64_t offset, uint32_t value);\nvoid TraceContextLoadI64(void* raw_context, uint64_t offset, uint64_t value);\nvoid TraceContextLoadF32(void* raw_context, uint64_t offset, __m128 value);\nvoid TraceContextLoadF64(void* raw_context, uint64_t offset,\n const double* value);\nvoid TraceContextLoadV128(void* raw_context, uint64_t offset, __m128 value);\n\nvoid TraceContextStoreI8(void* raw_context, uint64_t offset, uint8_t value);\nvoid TraceContextStoreI16(void* raw_context, uint64_t offset, uint16_t value);\nvoid TraceContextStoreI32(void* raw_context, uint64_t offset, uint32_t value);\nvoid TraceContextStoreI64(void* raw_context, uint64_t offset, uint64_t value);\nvoid TraceContextStoreF32(void* raw_context, uint64_t offset, __m128 value);\nvoid TraceContextStoreF64(void* raw_context, uint64_t offset,\n const double* value);\nvoid TraceContextStoreV128(void* raw_context, uint64_t offset, __m128 value);\n\nvoid TraceMemoryLoadI8(void* raw_context, uint32_t address, uint8_t value);\nvoid TraceMemoryLoadI16(void* raw_context, uint32_t address, uint16_t value);\nvoid TraceMemoryLoadI32(void* raw_context, uint32_t address, uint32_t value);\nvoid TraceMemoryLoadI64(void* raw_context, uint32_t address, uint64_t value);\nvoid TraceMemoryLoadF32(void* raw_context, uint32_t address, __m128 value);\nvoid TraceMemoryLoadF64(void* raw_context, uint32_t address, __m128 value);\nvoid TraceMemoryLoadV128(void* raw_context, uint32_t address, __m128 value);\n\nvoid TraceMemoryStoreI8(void* raw_context, uint32_t address, uint8_t value);\nvoid TraceMemoryStoreI16(void* raw_context, uint32_t address, uint16_t value);\nvoid TraceMemoryStoreI32(void* raw_context, uint32_t address, uint32_t value);\nvoid TraceMemoryStoreI64(void* raw_context, uint32_t address, uint64_t value);\nvoid TraceMemoryStoreF32(void* raw_context, uint32_t address, __m128 value);\nvoid TraceMemoryStoreF64(void* raw_context, uint32_t address, __m128 value);\nvoid TraceMemoryStoreV128(void* raw_context, uint32_t address, __m128 value);\n\nvoid TraceMemset(void* raw_context, uint32_t address, uint8_t value,\n uint32_t length);\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_TRACERS_H_\n" }, { "path": "src/xenia/cpu/backend/x64/x64_util.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2022 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BACKEND_X64_X64_UTIL_H_\n#define XENIA_CPU_BACKEND_X64_X64_UTIL_H_\n\n#include \"xenia/base/vec128.h\"\n#include \"xenia/cpu/backend/x64/x64_backend.h\"\n#include \"xenia/cpu/backend/x64/x64_emitter.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace backend {\nnamespace x64 {\n\n// Used to generate ternary logic truth-tables for vpternlog\n// Use these to directly refer to terms and perform binary operations upon them\n// and the resulting value will be the ternary lookup table\n// ex:\n// (TernaryOperand::a | ~TernaryOperand::b) & TernaryOperand::c\n// = 0b10100010\n// = 0xa2\n// vpternlog a, b, c, 0xa2\n//\n// ~(TernaryOperand::a ^ TernaryOperand::b) & TernaryOperand::c\n// = 0b10000010\n// = 0x82\n// vpternlog a, b, c, 0x82\nnamespace TernaryOperand {\nconstexpr uint8_t a = 0b11110000;\nconstexpr uint8_t b = 0b11001100;\nconstexpr uint8_t c = 0b10101010;\n} // namespace TernaryOperand\n\n} // namespace x64\n} // namespace backend\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BACKEND_X64_X64_UTIL_H_\n" }, { "path": "src/xenia/cpu/breakpoint.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/breakpoint.h\"\n\n#include \"xenia/base/string_util.h\"\n#include \"xenia/cpu/backend/backend.h\"\n#include \"xenia/cpu/backend/code_cache.h\"\n\nnamespace xe {\nnamespace cpu {\n\nBreakpoint::Breakpoint(Processor* processor, AddressType address_type,\n uint64_t address, HitCallback hit_callback)\n : processor_(processor),\n address_type_(address_type),\n address_(address),\n hit_callback_(hit_callback) {}\n\nBreakpoint::~Breakpoint() { assert_false(installed_); }\n\nvoid Breakpoint::Install() {\n assert_false(installed_);\n processor_->backend()->InstallBreakpoint(this);\n installed_ = true;\n}\n\nvoid Breakpoint::Uninstall() {\n assert_true(installed_);\n processor_->backend()->UninstallBreakpoint(this);\n installed_ = false;\n}\n\nstd::string Breakpoint::to_string() const {\n if (address_type_ == AddressType::kGuest) {\n auto str =\n std::string(\"PPC \") + xe::string_util::to_hex_string(guest_address());\n auto functions = processor_->FindFunctionsWithAddress(guest_address());\n if (functions.empty()) {\n return str;\n }\n str += \" \" + functions[0]->name();\n return str;\n } else {\n return std::string(\"x64 \") + xe::string_util::to_hex_string(host_address());\n }\n}\n\nGuestFunction* Breakpoint::guest_function() const {\n if (address_type_ == AddressType::kGuest) {\n auto functions = processor_->FindFunctionsWithAddress(guest_address());\n if (functions.empty()) {\n return nullptr;\n } else if (functions[0]->is_guest()) {\n return static_cast(functions[0]);\n }\n return nullptr;\n } else {\n return processor_->backend()->code_cache()->LookupFunction(host_address());\n }\n}\n\nbool Breakpoint::ContainsHostAddress(uintptr_t search_address) const {\n bool contains = false;\n ForEachHostAddress([&contains, search_address](uintptr_t host_address) {\n if (host_address == search_address) {\n contains = true;\n }\n });\n return contains;\n}\n\nvoid Breakpoint::ForEachHostAddress(\n std::function callback) const {\n if (address_type_ == AddressType::kGuest) {\n auto guest_address = this->guest_address();\n\n // Lookup all functions that contain this guest address and patch them.\n auto functions = processor_->FindFunctionsWithAddress(guest_address);\n\n if (functions.empty()) {\n // If function does not exist demand it, as we need someplace to put our\n // breakpoint. Note that this follows the same resolution rules as the\n // JIT, so what's returned is the function the JIT would have jumped to.\n auto fn = processor_->ResolveFunction(guest_address);\n if (!fn) {\n // TODO(benvanik): error out better with 'invalid breakpoint'?\n assert_not_null(fn);\n return;\n }\n functions.push_back(fn);\n }\n assert_false(functions.empty());\n\n for (auto function : functions) {\n // TODO(benvanik): other function types.\n assert_true(function->is_guest());\n auto guest_function = reinterpret_cast(function);\n uintptr_t host_address =\n guest_function->MapGuestAddressToMachineCode(guest_address);\n assert_not_zero(host_address);\n callback(host_address);\n }\n } else {\n // Direct host address patching.\n callback(host_address());\n }\n}\n\n} // namespace cpu\n} // namespace xe" }, { "path": "src/xenia/cpu/breakpoint.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_BREAKPOINT_H_\n#define XENIA_CPU_BREAKPOINT_H_\n\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\n\nclass Breakpoint {\n public:\n enum class AddressType {\n kGuest,\n kHost,\n };\n typedef std::function\n HitCallback;\n\n Breakpoint(Processor* processor, AddressType address_type, uint64_t address,\n HitCallback hit_callback);\n ~Breakpoint();\n\n AddressType address_type() const { return address_type_; }\n uint64_t address() const { return address_; }\n uint32_t guest_address() const {\n assert_true(address_type_ == AddressType::kGuest);\n return static_cast(address_);\n }\n uintptr_t host_address() const {\n assert_true(address_type_ == AddressType::kHost);\n return static_cast(address_);\n }\n\n // Whether the breakpoint has been enabled by the user.\n bool is_enabled() const { return enabled_; }\n\n // Toggles the breakpoint state.\n // Assumes the caller holds the global lock.\n void set_enabled(bool is_enabled) {\n enabled_ = is_enabled;\n if (!enabled_ && installed_) {\n Uninstall();\n } else if (enabled_ && !installed_ && !suspend_count_) {\n Install();\n }\n }\n\n // Whether the breakpoint is currently installed and active.\n bool is_installed() const { return installed_; }\n\n std::string to_string() const;\n\n // Returns a guest function that contains the guest address, if any.\n // If there are multiple functions that contain the address a random one will\n // be returned. If this is a host-address code breakpoint this will attempt to\n // find a function with that address and otherwise return nullptr.\n GuestFunction* guest_function() const;\n\n // Returns true if this breakpoint, when installed, contains the given host\n // address.\n bool ContainsHostAddress(uintptr_t search_address) const;\n\n // Enumerates all host addresses that correspond to this breakpoint.\n // If this is a host type it will return the single host address, if it is a\n // guest type it may return multiple host addresses.\n void ForEachHostAddress(std::function callback) const;\n\n // CPU backend data. Implementation specific - DO NOT TOUCH THIS!\n std::vector> backend_data() const {\n return backend_data_;\n }\n std::vector>& backend_data() {\n return backend_data_;\n }\n\n private:\n friend class Processor;\n\n void OnHit(ThreadDebugInfo* thread_info, uint64_t host_pc) {\n hit_callback_(this, thread_info, host_pc);\n }\n\n // Suspends the breakpoint until it is resumed with Resume.\n // This preserves the user enabled state.\n // Assumes the caller holds the global lock.\n void Suspend() {\n ++suspend_count_;\n if (installed_) {\n Uninstall();\n }\n }\n\n // Resumes a previously-suspended breakpoint, reinstalling it if required.\n // Assumes the caller holds the global lock.\n void Resume() {\n --suspend_count_;\n if (!suspend_count_ && enabled_) {\n Install();\n }\n }\n\n private:\n Processor* processor_ = nullptr;\n\n void Install();\n void Uninstall();\n\n // True if patched into code.\n bool installed_ = false;\n // True if user enabled.\n bool enabled_ = true;\n // Depth of suspends (must be 0 to be installed). Defaults to suspended so\n // that it's never installed unless the debugger knows about it.\n int suspend_count_ = 1;\n\n AddressType address_type_;\n uint64_t address_ = 0;\n\n HitCallback hit_callback_;\n\n // Opaque backend data. Don't touch this.\n std::vector> backend_data_;\n};\n\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_BREAKPOINT_H_" }, { "path": "src/xenia/cpu/compiler/compiler.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/compiler.h\"\n\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\n\nCompiler::Compiler(Processor* processor) : processor_(processor) {}\n\nCompiler::~Compiler() { Reset(); }\n\nvoid Compiler::AddPass(std::unique_ptr pass) {\n pass->Initialize(this);\n passes_.push_back(std::move(pass));\n}\n\nvoid Compiler::Reset() {}\n\nbool Compiler::Compile(xe::cpu::hir::HIRBuilder* builder) {\n // TODO(benvanik): sophisticated stuff. Run passes in parallel, run until they\n // stop changing things, etc.\n for (size_t i = 0; i < passes_.size(); ++i) {\n auto& pass = passes_[i];\n scratch_arena_.Reset();\n if (!pass->Run(builder)) {\n return false;\n }\n }\n\n return true;\n}\n\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/compiler.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_COMPILER_H_\n#define XENIA_CPU_COMPILER_COMPILER_H_\n\n#include \n#include \n\n#include \"xenia/base/arena.h\"\n#include \"xenia/cpu/hir/hir_builder.h\"\n\nnamespace xe {\nnamespace cpu {\nclass Processor;\n} // namespace cpu\n} // namespace xe\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\n\nclass CompilerPass;\n\nclass Compiler {\n public:\n explicit Compiler(Processor* processor);\n ~Compiler();\n\n Processor* processor() const { return processor_; }\n Arena* scratch_arena() { return &scratch_arena_; }\n\n void AddPass(std::unique_ptr pass);\n\n void Reset();\n\n bool Compile(hir::HIRBuilder* builder);\n\n private:\n Processor* processor_;\n Arena scratch_arena_;\n\n std::vector> passes_;\n};\n\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_COMPILER_H_\n" }, { "path": "src/xenia/cpu/compiler/compiler_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\n#include \"xenia/cpu/compiler/compiler.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\n\nCompilerPass::CompilerPass() : processor_(nullptr), compiler_(nullptr) {}\n\nCompilerPass::~CompilerPass() = default;\n\nbool CompilerPass::Initialize(Compiler* compiler) {\n processor_ = compiler->processor();\n compiler_ = compiler;\n return true;\n}\n\nArena* CompilerPass::scratch_arena() const {\n return compiler_->scratch_arena();\n}\n\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/compiler_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_COMPILER_PASS_H_\n#define XENIA_CPU_COMPILER_COMPILER_PASS_H_\n\n#include \"xenia/base/arena.h\"\n#include \"xenia/cpu/hir/hir_builder.h\"\n\nnamespace xe {\nnamespace cpu {\nclass Processor;\n} // namespace cpu\n} // namespace xe\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\n\nclass Compiler;\n\nclass CompilerPass {\n public:\n CompilerPass();\n virtual ~CompilerPass();\n\n virtual bool Initialize(Compiler* compiler);\n\n virtual bool Run(hir::HIRBuilder* builder) = 0;\n\n protected:\n Arena* scratch_arena() const;\n\n protected:\n Processor* processor_;\n Compiler* compiler_;\n};\n\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_COMPILER_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/compiler_passes.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_COMPILER_PASSES_H_\n#define XENIA_CPU_COMPILER_COMPILER_PASSES_H_\n\n#include \"xenia/cpu/compiler/passes/conditional_group_pass.h\"\n#include \"xenia/cpu/compiler/passes/conditional_group_subpass.h\"\n#include \"xenia/cpu/compiler/passes/constant_propagation_pass.h\"\n#include \"xenia/cpu/compiler/passes/context_promotion_pass.h\"\n#include \"xenia/cpu/compiler/passes/control_flow_analysis_pass.h\"\n#include \"xenia/cpu/compiler/passes/control_flow_simplification_pass.h\"\n#include \"xenia/cpu/compiler/passes/data_flow_analysis_pass.h\"\n#include \"xenia/cpu/compiler/passes/dead_code_elimination_pass.h\"\n#include \"xenia/cpu/compiler/passes/finalization_pass.h\"\n#include \"xenia/cpu/compiler/passes/memory_sequence_combination_pass.h\"\n#include \"xenia/cpu/compiler/passes/register_allocation_pass.h\"\n#include \"xenia/cpu/compiler/passes/simplification_pass.h\"\n#include \"xenia/cpu/compiler/passes/validation_pass.h\"\n#include \"xenia/cpu/compiler/passes/value_reduction_pass.h\"\n\n#endif // XENIA_CPU_COMPILER_COMPILER_PASSES_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/conditional_group_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/conditional_group_pass.h\"\n\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/compiler/compiler.h\"\n#include \"xenia/cpu/ppc/ppc_context.h\"\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::Block;\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::Instr;\nusing xe::cpu::hir::Value;\n\nConditionalGroupPass::ConditionalGroupPass() : CompilerPass() {}\n\nConditionalGroupPass::~ConditionalGroupPass() {}\n\nbool ConditionalGroupPass::Initialize(Compiler* compiler) {\n if (!CompilerPass::Initialize(compiler)) {\n return false;\n }\n\n for (size_t i = 0; i < passes_.size(); ++i) {\n auto& pass = passes_[i];\n if (!pass->Initialize(compiler)) {\n return false;\n }\n }\n\n return true;\n}\n\nbool ConditionalGroupPass::Run(HIRBuilder* builder) {\n bool dirty;\n int loops = 0;\n do {\n assert_true(loops < 20); // arbitrary number\n dirty = false;\n for (size_t i = 0; i < passes_.size(); ++i) {\n scratch_arena()->Reset();\n auto& pass = passes_[i];\n auto subpass = dynamic_cast(pass.get());\n if (!subpass) {\n if (!pass->Run(builder)) {\n return false;\n }\n } else {\n bool result = false;\n if (!subpass->Run(builder, result)) {\n return false;\n }\n dirty |= result;\n }\n }\n loops++;\n } while (dirty);\n return true;\n}\n\nvoid ConditionalGroupPass::AddPass(std::unique_ptr pass) {\n passes_.push_back(std::move(pass));\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/conditional_group_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_CONDITIONAL_GROUP_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_CONDITIONAL_GROUP_PASS_H_\n\n#include \n#include \n\n#include \"xenia/base/platform.h\"\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n#include \"xenia/cpu/compiler/passes/conditional_group_subpass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass ConditionalGroupPass : public CompilerPass {\n public:\n ConditionalGroupPass();\n virtual ~ConditionalGroupPass() override;\n\n bool Initialize(Compiler* compiler) override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n void AddPass(std::unique_ptr pass);\n\n private:\n std::vector> passes_;\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_CONDITIONAL_GROUP_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/conditional_group_subpass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/conditional_group_subpass.h\"\n\n#include \"xenia/cpu/compiler/compiler.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nConditionalGroupSubpass::ConditionalGroupSubpass() : CompilerPass() {}\n\nConditionalGroupSubpass::~ConditionalGroupSubpass() = default;\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/conditional_group_subpass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_CONDITIONAL_GROUP_SUBPASS_H_\n#define XENIA_CPU_COMPILER_PASSES_CONDITIONAL_GROUP_SUBPASS_H_\n\n#include \"xenia/base/arena.h\"\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n#include \"xenia/cpu/hir/hir_builder.h\"\n\nnamespace xe {\nnamespace cpu {\nclass Processor;\n} // namespace cpu\n} // namespace xe\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nclass Compiler;\nnamespace passes {\n\nclass ConditionalGroupSubpass : public CompilerPass {\n public:\n ConditionalGroupSubpass();\n virtual ~ConditionalGroupSubpass();\n\n bool Run(hir::HIRBuilder* builder) override {\n bool dummy;\n return Run(builder, dummy);\n }\n\n virtual bool Run(hir::HIRBuilder* builder, bool& result) = 0;\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_CONDITIONAL_GROUP_SUBPASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/constant_propagation_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/constant_propagation_pass.h\"\n\n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/function.h\"\n#include \"xenia/cpu/processor.h\"\n\nDEFINE_bool(inline_mmio_access, true, \"Inline constant MMIO loads and stores.\",\n \"CPU\");\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::TypeName;\nusing xe::cpu::hir::Value;\n\nConstantPropagationPass::ConstantPropagationPass()\n : ConditionalGroupSubpass() {}\n\nConstantPropagationPass::~ConstantPropagationPass() {}\n\nbool ConstantPropagationPass::Run(HIRBuilder* builder, bool& result) {\n // Once ContextPromotion has run there will likely be a whole slew of\n // constants that can be pushed through the function.\n // Example:\n // store_context +100, 1000\n // v0 = load_context +100\n // v1 = add v0, v0\n // store_context +200, v1\n // after PromoteContext:\n // store_context +100, 1000\n // v0 = 1000\n // v1 = add v0, v0\n // store_context +200, v1\n // after PropagateConstants:\n // store_context +100, 1000\n // v0 = 1000\n // v1 = add 1000, 1000\n // store_context +200, 2000\n // A DCE run after this should clean up any of the values no longer needed.\n //\n // Special care needs to be taken with paired instructions. For example,\n // DID_CARRY needs to be set as a constant:\n // v1 = sub.2 20, 1\n // v2 = did_carry v1\n // should become:\n // v1 = 19\n // v2 = 0\n\n result = false;\n auto block = builder->first_block();\n while (block) {\n auto i = block->instr_head;\n while (i) {\n auto v = i->dest;\n switch (i->opcode->num) {\n case OPCODE_DEBUG_BREAK_TRUE:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->IsConstantTrue()) {\n i->Replace(&OPCODE_DEBUG_BREAK_info, i->flags);\n } else {\n i->Remove();\n }\n result = true;\n }\n break;\n\n case OPCODE_TRAP_TRUE:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->IsConstantTrue()) {\n i->Replace(&OPCODE_TRAP_info, i->flags);\n } else {\n i->Remove();\n }\n result = true;\n }\n break;\n\n case OPCODE_CALL_TRUE:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->IsConstantTrue()) {\n auto symbol = i->src2.symbol;\n i->Replace(&OPCODE_CALL_info, i->flags);\n i->src1.symbol = symbol;\n } else {\n i->Remove();\n }\n result = true;\n }\n break;\n case OPCODE_CALL_INDIRECT:\n if (i->src1.value->IsConstant()) {\n auto function = processor_->LookupFunction(\n uint32_t(i->src1.value->constant.i32));\n if (!function) {\n break;\n }\n i->Replace(&OPCODE_CALL_info, i->flags);\n i->src1.symbol = function;\n result = true;\n }\n break;\n case OPCODE_CALL_INDIRECT_TRUE:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->IsConstantTrue()) {\n auto value = i->src2.value;\n i->Replace(&OPCODE_CALL_INDIRECT_info, i->flags);\n i->set_src1(value);\n } else {\n i->Remove();\n }\n result = true;\n }\n break;\n\n case OPCODE_BRANCH_TRUE:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->IsConstantTrue()) {\n auto label = i->src2.label;\n i->Replace(&OPCODE_BRANCH_info, i->flags);\n i->src1.label = label;\n } else {\n i->Remove();\n }\n result = true;\n }\n break;\n case OPCODE_BRANCH_FALSE:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->IsConstantFalse()) {\n auto label = i->src2.label;\n i->Replace(&OPCODE_BRANCH_info, i->flags);\n i->src1.label = label;\n } else {\n i->Remove();\n }\n result = true;\n }\n break;\n\n case OPCODE_CAST:\n if (i->src1.value->IsConstant()) {\n TypeName target_type = v->type;\n v->set_from(i->src1.value);\n v->Cast(target_type);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_CONVERT:\n if (i->src1.value->IsConstant()) {\n TypeName target_type = v->type;\n v->set_from(i->src1.value);\n v->Convert(target_type, RoundMode(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_ROUND:\n if (i->src1.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Round(RoundMode(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_ZERO_EXTEND:\n if (i->src1.value->IsConstant()) {\n TypeName target_type = v->type;\n v->set_from(i->src1.value);\n v->ZeroExtend(target_type);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_SIGN_EXTEND:\n if (i->src1.value->IsConstant()) {\n TypeName target_type = v->type;\n v->set_from(i->src1.value);\n v->SignExtend(target_type);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_TRUNCATE:\n if (i->src1.value->IsConstant()) {\n TypeName target_type = v->type;\n v->set_from(i->src1.value);\n v->Truncate(target_type);\n i->Remove();\n result = true;\n }\n break;\n\n case OPCODE_LOAD:\n case OPCODE_LOAD_OFFSET:\n if (i->src1.value->IsConstant()) {\n assert_false(i->flags & LOAD_STORE_BYTE_SWAP);\n auto memory = processor_->memory();\n auto address = i->src1.value->constant.i32;\n if (i->opcode->num == OPCODE_LOAD_OFFSET) {\n address += i->src2.value->constant.i32;\n }\n\n auto mmio_range =\n processor_->memory()->LookupVirtualMappedRange(address);\n if (cvars::inline_mmio_access && mmio_range) {\n i->Replace(&OPCODE_LOAD_MMIO_info, 0);\n i->src1.offset = reinterpret_cast(mmio_range);\n i->src2.offset = address;\n result = true;\n } else {\n auto heap = memory->LookupHeap(address);\n uint32_t protect;\n if (heap && heap->QueryProtect(address, &protect) &&\n !(protect & kMemoryProtectWrite) &&\n (protect & kMemoryProtectRead)) {\n // Memory is readonly - can just return the value.\n auto host_addr = memory->TranslateVirtual(address);\n switch (v->type) {\n case INT8_TYPE:\n v->set_constant(xe::load(host_addr));\n i->Remove();\n result = true;\n break;\n case INT16_TYPE:\n v->set_constant(xe::load(host_addr));\n i->Remove();\n result = true;\n break;\n case INT32_TYPE:\n v->set_constant(xe::load(host_addr));\n i->Remove();\n result = true;\n break;\n case INT64_TYPE:\n v->set_constant(xe::load(host_addr));\n i->Remove();\n result = true;\n break;\n case VEC128_TYPE:\n vec128_t val;\n val.low = xe::load(host_addr);\n val.high = xe::load(host_addr + 8);\n v->set_constant(val);\n i->Remove();\n result = true;\n break;\n default:\n assert_unhandled_case(v->type);\n break;\n }\n }\n }\n }\n break;\n case OPCODE_STORE:\n case OPCODE_STORE_OFFSET:\n if (cvars::inline_mmio_access && i->src1.value->IsConstant()) {\n auto address = i->src1.value->constant.i32;\n if (i->opcode->num == OPCODE_STORE_OFFSET) {\n address += i->src2.value->constant.i32;\n }\n\n auto mmio_range =\n processor_->memory()->LookupVirtualMappedRange(address);\n if (mmio_range) {\n auto value = i->src2.value;\n if (i->opcode->num == OPCODE_STORE_OFFSET) {\n value = i->src3.value;\n }\n\n i->Replace(&OPCODE_STORE_MMIO_info, 0);\n i->src1.offset = reinterpret_cast(mmio_range);\n i->src2.offset = address;\n i->set_src3(value);\n result = true;\n }\n }\n break;\n\n case OPCODE_SELECT:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->type != VEC128_TYPE) {\n if (i->src1.value->IsConstantTrue()) {\n auto src2 = i->src2.value;\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(src2);\n result = true;\n } else if (i->src1.value->IsConstantFalse()) {\n auto src3 = i->src3.value;\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(src3);\n result = true;\n } else if (i->src2.value->IsConstant() &&\n i->src3.value->IsConstant()) {\n // TODO: Select\n // v->set_from(i->src2.value);\n // v->Select(i->src3.value, i->src1.value);\n // i->Remove();\n }\n } else {\n // TODO: vec128 select\n }\n }\n break;\n case OPCODE_IS_TRUE:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->IsConstantTrue()) {\n v->set_constant(uint8_t(1));\n } else {\n v->set_constant(uint8_t(0));\n }\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_IS_FALSE:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->IsConstantFalse()) {\n v->set_constant(uint8_t(1));\n } else {\n v->set_constant(uint8_t(0));\n }\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_IS_NAN:\n if (i->src1.value->IsConstant()) {\n if (i->src1.value->type == FLOAT32_TYPE &&\n std::isnan(i->src1.value->constant.f32)) {\n v->set_constant(uint8_t(1));\n } else if (i->src1.value->type == FLOAT64_TYPE &&\n std::isnan(i->src1.value->constant.f64)) {\n v->set_constant(uint8_t(1));\n } else {\n v->set_constant(uint8_t(0));\n }\n i->Remove();\n result = true;\n }\n break;\n\n // TODO(benvanik): compares\n case OPCODE_COMPARE_EQ:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantEQ(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_NE:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantNE(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_SLT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantSLT(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_SLE:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantSLE(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_SGT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantSGT(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_SGE:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantSGE(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_ULT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantULT(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_ULE:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantULE(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_UGT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantUGT(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_COMPARE_UGE:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n bool value = i->src1.value->IsConstantUGE(i->src2.value);\n i->dest->set_constant(uint8_t(value));\n i->Remove();\n result = true;\n }\n break;\n\n case OPCODE_DID_SATURATE:\n // assert_true(!i->src1.value->IsConstant());\n break;\n\n case OPCODE_ADD:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Add(i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_ADD_CARRY:\n if (i->src1.value->IsConstantZero() &&\n i->src2.value->IsConstantZero()) {\n Value* ca = i->src3.value;\n if (ca->IsConstant()) {\n TypeName target_type = v->type;\n v->set_from(ca);\n v->ZeroExtend(target_type);\n i->Remove();\n } else {\n if (i->dest->type == ca->type) {\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(ca);\n } else {\n i->Replace(&OPCODE_ZERO_EXTEND_info, 0);\n i->set_src1(ca);\n }\n }\n result = true;\n }\n break;\n case OPCODE_SUB:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Sub(i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_MUL:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Mul(i->src2.value);\n i->Remove();\n result = true;\n } else if (i->src1.value->IsConstant() ||\n i->src2.value->IsConstant()) {\n // Reorder the sources to make things simpler.\n // s1 = non-const, s2 = const\n auto s1 =\n i->src1.value->IsConstant() ? i->src2.value : i->src1.value;\n auto s2 =\n i->src1.value->IsConstant() ? i->src1.value : i->src2.value;\n\n // Multiplication by one = no-op\n if (s2->type != VEC128_TYPE && s2->IsConstantOne()) {\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(s1);\n result = true;\n } else if (s2->type == VEC128_TYPE) {\n auto& c = s2->constant;\n if (c.v128.f32[0] == 1.f && c.v128.f32[1] == 1.f &&\n c.v128.f32[2] == 1.f && c.v128.f32[3] == 1.f) {\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(s1);\n result = true;\n }\n }\n }\n break;\n case OPCODE_MUL_HI:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->MulHi(i->src2.value, (i->flags & ARITHMETIC_UNSIGNED) != 0);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_DIV:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Div(i->src2.value, (i->flags & ARITHMETIC_UNSIGNED) != 0);\n i->Remove();\n result = true;\n } else if (i->src2.value->IsConstant()) {\n // Division by one = no-op.\n Value* src1 = i->src1.value;\n if (i->src2.value->type != VEC128_TYPE &&\n i->src2.value->IsConstantOne()) {\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(src1);\n result = true;\n } else if (i->src2.value->type == VEC128_TYPE) {\n auto& c = i->src2.value->constant;\n if (c.v128.f32[0] == 1.f && c.v128.f32[1] == 1.f &&\n c.v128.f32[2] == 1.f && c.v128.f32[3] == 1.f) {\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(src1);\n result = true;\n }\n }\n }\n break;\n case OPCODE_MUL_ADD:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n if (i->src3.value->IsConstant()) {\n v->set_from(i->src1.value);\n Value::MulAdd(v, i->src1.value, i->src2.value, i->src3.value);\n i->Remove();\n result = true;\n } else {\n // Multiply part is constant.\n Value* mul = builder->AllocValue();\n mul->set_from(i->src1.value);\n mul->Mul(i->src2.value);\n\n Value* add = i->src3.value;\n i->Replace(&OPCODE_ADD_info, 0);\n i->set_src1(mul);\n i->set_src2(add);\n\n result = true;\n }\n }\n break;\n case OPCODE_MUL_SUB:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n // Multiply part is constant.\n if (i->src3.value->IsConstant()) {\n v->set_from(i->src1.value);\n Value::MulSub(v, i->src1.value, i->src2.value, i->src3.value);\n i->Remove();\n result = true;\n } else {\n // Multiply part is constant.\n Value* mul = builder->AllocValue();\n mul->set_from(i->src1.value);\n mul->Mul(i->src2.value);\n\n Value* add = i->src3.value;\n i->Replace(&OPCODE_SUB_info, 0);\n i->set_src1(mul);\n i->set_src2(add);\n\n result = true;\n }\n }\n break;\n case OPCODE_MAX:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Max(i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_NEG:\n if (i->src1.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Neg();\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_ABS:\n if (i->src1.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Abs();\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_SQRT:\n if (i->src1.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Sqrt();\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_RSQRT:\n if (i->src1.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->RSqrt();\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_RECIP:\n if (i->src1.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Recip();\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_AND:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->And(i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_AND_NOT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src2.value);\n v->Not();\n v->And(i->src1.value);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_OR:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Or(i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_XOR:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Xor(i->src2.value);\n i->Remove();\n result = true;\n } else if (!i->src1.value->IsConstant() &&\n !i->src2.value->IsConstant() &&\n i->src1.value == i->src2.value) {\n v->set_zero(v->type);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_NOT:\n if (i->src1.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Not();\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_SHL:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Shl(i->src2.value);\n i->Remove();\n result = true;\n } else if (i->src2.value->IsConstantZero()) {\n auto src1 = i->src1.value;\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(src1);\n result = true;\n }\n break;\n case OPCODE_SHR:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Shr(i->src2.value);\n i->Remove();\n result = true;\n } else if (i->src2.value->IsConstantZero()) {\n auto src1 = i->src1.value;\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(src1);\n result = true;\n }\n break;\n case OPCODE_SHA:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->Sha(i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n // TODO(benvanik): ROTATE_LEFT\n case OPCODE_BYTE_SWAP:\n if (i->src1.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->ByteSwap();\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_CNTLZ:\n if (i->src1.value->IsConstant()) {\n v->set_zero(v->type);\n v->CountLeadingZeros(i->src1.value);\n i->Remove();\n result = true;\n }\n break;\n // TODO(benvanik): INSERT/EXTRACT\n // TODO(benvanik): PERMUTE/SWIZZLE\n case OPCODE_EXTRACT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_zero(v->type);\n v->Extract(i->src1.value, i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_SPLAT:\n if (i->src1.value->IsConstant()) {\n v->set_zero(v->type);\n v->Splat(i->src1.value);\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_COMPARE_EQ:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->VectorCompareEQ(i->src2.value, hir::TypeName(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_COMPARE_SGT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->VectorCompareSGT(i->src2.value, hir::TypeName(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_COMPARE_SGE:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->VectorCompareSGE(i->src2.value, hir::TypeName(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_COMPARE_UGT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->VectorCompareUGT(i->src2.value, hir::TypeName(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_COMPARE_UGE:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->VectorCompareUGE(i->src2.value, hir::TypeName(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_CONVERT_F2I:\n if (i->src1.value->IsConstant()) {\n v->set_zero(VEC128_TYPE);\n v->VectorConvertF2I(i->src1.value,\n !!(i->flags & ARITHMETIC_UNSIGNED));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_CONVERT_I2F:\n if (i->src1.value->IsConstant()) {\n v->set_zero(VEC128_TYPE);\n v->VectorConvertI2F(i->src1.value,\n !!(i->flags & ARITHMETIC_UNSIGNED));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_SHL:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->VectorShl(i->src2.value, hir::TypeName(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_SHR:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->VectorShr(i->src2.value, hir::TypeName(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_ROTATE_LEFT:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->VectorRol(i->src2.value, hir::TypeName(i->flags));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_ADD:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n uint32_t arith_flags = i->flags >> 8;\n v->VectorAdd(i->src2.value, hir::TypeName(i->flags & 0xFF),\n !!(arith_flags & ARITHMETIC_UNSIGNED),\n !!(arith_flags & ARITHMETIC_SATURATE));\n i->Remove();\n result = true;\n }\n break;\n case OPCODE_VECTOR_SUB:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n uint32_t arith_flags = i->flags >> 8;\n v->VectorSub(i->src2.value, hir::TypeName(i->flags & 0xFF),\n !!(arith_flags & ARITHMETIC_UNSIGNED),\n !!(arith_flags & ARITHMETIC_SATURATE));\n i->Remove();\n result = true;\n }\n break;\n\n case OPCODE_DOT_PRODUCT_3:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->DotProduct3(i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n\n case OPCODE_DOT_PRODUCT_4:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n v->DotProduct4(i->src2.value);\n i->Remove();\n result = true;\n }\n break;\n\n case OPCODE_VECTOR_AVERAGE:\n if (i->src1.value->IsConstant() && i->src2.value->IsConstant()) {\n v->set_from(i->src1.value);\n uint32_t arith_flags = i->flags >> 8;\n v->VectorAverage(i->src2.value, hir::TypeName(i->flags & 0xFF),\n !!(arith_flags & ARITHMETIC_UNSIGNED),\n !!(arith_flags & ARITHMETIC_SATURATE));\n i->Remove();\n result = true;\n }\n break;\n\n default:\n // Ignored.\n break;\n }\n i = i->next;\n }\n\n block = block->next;\n }\n\n return true;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/constant_propagation_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_CONSTANT_PROPAGATION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_CONSTANT_PROPAGATION_PASS_H_\n\n#include \"xenia/cpu/compiler/passes/conditional_group_subpass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass ConstantPropagationPass : public ConditionalGroupSubpass {\n public:\n ConstantPropagationPass();\n ~ConstantPropagationPass() override;\n\n bool Run(hir::HIRBuilder* builder, bool& result) override;\n\n private:\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_CONSTANT_PROPAGATION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/context_promotion_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/context_promotion_pass.h\"\n\n#include \"xenia/apu/apu_flags.h\"\n#include \"xenia/base/cvar.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/compiler/compiler.h\"\n#include \"xenia/cpu/ppc/ppc_context.h\"\n#include \"xenia/cpu/processor.h\"\n\nDECLARE_bool(debug);\n\nDEFINE_bool(store_all_context_values, false,\n \"Don't strip dead context stores to aid in debugging.\", \"CPU\");\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::Block;\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::Instr;\nusing xe::cpu::hir::Value;\n\nContextPromotionPass::ContextPromotionPass() : CompilerPass() {}\n\nContextPromotionPass::~ContextPromotionPass() {}\n\nbool ContextPromotionPass::Initialize(Compiler* compiler) {\n if (!CompilerPass::Initialize(compiler)) {\n return false;\n }\n\n // This is a terrible implementation.\n context_values_.resize(sizeof(ppc::PPCContext));\n context_validity_.resize(static_cast(sizeof(ppc::PPCContext)));\n\n return true;\n}\n\nbool ContextPromotionPass::Run(HIRBuilder* builder) {\n // Like mem2reg, but because context memory is unaliasable it's easier to\n // check and convert LoadContext/StoreContext into value operations.\n // Example of load->value promotion:\n // v0 = load_context +100\n // store_context +200, v0\n // v1 = load_context +100 <-- replace with v1 = v0\n // store_context +200, v1\n //\n // It'd be possible in this stage to also remove redundant context stores:\n // Example of dead store elimination:\n // store_context +100, v0 <-- removed due to following store\n // store_context +100, v1\n // This is more generally done by DSE, however if it could be done here\n // instead as it may be faster (at least on the block-level).\n\n // Promote loads to values.\n // Process each block independently, for now.\n auto block = builder->first_block();\n while (block) {\n PromoteBlock(block);\n block = block->next;\n }\n\n // Remove all dead stores.\n // This will break debugging as we can't recover this information when\n // trying to extract stack traces/register values, so we don't do that.\n if (!cvars::debug && !cvars::store_all_context_values) {\n block = builder->first_block();\n while (block) {\n RemoveDeadStoresBlock(block);\n block = block->next;\n }\n }\n\n return true;\n}\n\nvoid ContextPromotionPass::PromoteBlock(Block* block) {\n auto& validity = context_validity_;\n validity.reset();\n\n Instr* i = block->instr_head;\n while (i) {\n auto next = i->next;\n if (i->opcode->flags & OPCODE_FLAG_VOLATILE) {\n // Volatile instruction - requires all context values be flushed.\n validity.reset();\n } else if (i->opcode == &OPCODE_LOAD_CONTEXT_info) {\n size_t offset = i->src1.offset;\n if (validity.test(static_cast(offset))) {\n // Legit previous value, reuse.\n Value* previous_value = context_values_[offset];\n i->opcode = &hir::OPCODE_ASSIGN_info;\n i->set_src1(previous_value);\n } else {\n // Store the loaded value into the table.\n context_values_[offset] = i->dest;\n validity.set(static_cast(offset));\n }\n } else if (i->opcode == &OPCODE_STORE_CONTEXT_info) {\n size_t offset = i->src1.offset;\n Value* value = i->src2.value;\n // Store value into the table for later.\n context_values_[offset] = value;\n validity.set(static_cast(offset));\n }\n i = next;\n }\n}\n\nvoid ContextPromotionPass::RemoveDeadStoresBlock(Block* block) {\n auto& validity = context_validity_;\n validity.reset();\n\n // Walk backwards and mark offsets that are written to.\n // If the offset was written to earlier, ignore the store.\n Instr* i = block->instr_tail;\n while (i) {\n Instr* prev = i->prev;\n if (i->opcode->flags & (OPCODE_FLAG_VOLATILE | OPCODE_FLAG_BRANCH)) {\n // Volatile instruction - requires all context values be flushed.\n validity.reset();\n } else if (i->opcode == &OPCODE_STORE_CONTEXT_info) {\n size_t offset = i->src1.offset;\n if (!validity.test(static_cast(offset))) {\n // Offset not yet written, mark and continue.\n validity.set(static_cast(offset));\n } else {\n // Already written to. Remove this store.\n i->Remove();\n }\n }\n i = prev;\n }\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/context_promotion_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_CONTEXT_PROMOTION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_CONTEXT_PROMOTION_PASS_H_\n\n#include \n#include \n\n#include \"xenia/base/platform.h\"\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\n#if XE_COMPILER_MSVC\n#pragma warning(push)\n#pragma warning(disable : 4244)\n#pragma warning(disable : 4267)\n#include \n#pragma warning(pop)\n#else\n#include \n#endif // XE_COMPILER_MSVC\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass ContextPromotionPass : public CompilerPass {\n public:\n ContextPromotionPass();\n virtual ~ContextPromotionPass() override;\n\n bool Initialize(Compiler* compiler) override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n void PromoteBlock(hir::Block* block);\n void RemoveDeadStoresBlock(hir::Block* block);\n\n private:\n std::vector context_values_;\n llvm::BitVector context_validity_;\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_CONTEXT_PROMOTION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/control_flow_analysis_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/control_flow_analysis_pass.h\"\n\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/backend/backend.h\"\n#include \"xenia/cpu/compiler/compiler.h\"\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::Edge;\nusing xe::cpu::hir::HIRBuilder;\n\nControlFlowAnalysisPass::ControlFlowAnalysisPass() : CompilerPass() {}\n\nControlFlowAnalysisPass::~ControlFlowAnalysisPass() {}\n\nbool ControlFlowAnalysisPass::Run(HIRBuilder* builder) {\n // Reset edges for all blocks. Needed to be re-runnable.\n // Note that this wastes a bunch of arena memory, so we shouldn't\n // re-run too often.\n auto block = builder->first_block();\n while (block) {\n block->incoming_edge_head = nullptr;\n block->outgoing_edge_head = nullptr;\n block = block->next;\n }\n\n // Add edges.\n block = builder->first_block();\n while (block) {\n auto instr = block->instr_tail;\n while (instr) {\n if ((instr->opcode->flags & OPCODE_FLAG_BRANCH) == 0) {\n break;\n }\n if (instr->opcode == &OPCODE_BRANCH_info) {\n auto label = instr->src1.label;\n builder->AddEdge(block, label->block, Edge::UNCONDITIONAL);\n } else if (instr->opcode == &OPCODE_BRANCH_TRUE_info ||\n instr->opcode == &OPCODE_BRANCH_FALSE_info) {\n auto label = instr->src2.label;\n builder->AddEdge(block, label->block, 0);\n }\n instr = instr->prev;\n }\n block = block->next;\n }\n\n // Mark dominators.\n block = builder->first_block();\n while (block) {\n if (block->incoming_edge_head &&\n !block->incoming_edge_head->incoming_next) {\n block->incoming_edge_head->flags |= Edge::DOMINATES;\n }\n block = block->next;\n }\n\n return true;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/control_flow_analysis_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_CONTROL_FLOW_ANALYSIS_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_CONTROL_FLOW_ANALYSIS_PASS_H_\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass ControlFlowAnalysisPass : public CompilerPass {\n public:\n ControlFlowAnalysisPass();\n ~ControlFlowAnalysisPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_CONTROL_FLOW_ANALYSIS_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/control_flow_simplification_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/control_flow_simplification_pass.h\"\n\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/backend/backend.h\"\n#include \"xenia/cpu/compiler/compiler.h\"\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::Edge;\nusing xe::cpu::hir::HIRBuilder;\n\nControlFlowSimplificationPass::ControlFlowSimplificationPass()\n : CompilerPass() {}\n\nControlFlowSimplificationPass::~ControlFlowSimplificationPass() {}\n\nbool ControlFlowSimplificationPass::Run(HIRBuilder* builder) {\n // Walk forwards and kill any unreachable blocks.\n // Do this before merging.\n auto block = builder->first_block();\n while (block) {\n auto next_block = block->next;\n if (!block->incoming_edge_head && block->prev) {\n // Block is in the interior and has no incoming edges - kill it.\n builder->RemoveBlock(block);\n }\n block = next_block;\n }\n\n // Walk backwards and merge blocks if possible.\n bool merged_any = false;\n block = builder->last_block();\n while (block) {\n auto prev_block = block->prev;\n const uint32_t expected = Edge::DOMINATES | Edge::UNCONDITIONAL;\n if (block->incoming_edge_head &&\n (block->incoming_edge_head->flags & expected) == expected) {\n // Dominated by the incoming block.\n // If that block comes immediately before us then we can merge the\n // two blocks (assuming it's not a volatile instruction like Trap).\n if (block->prev == block->incoming_edge_head->src &&\n block->prev->instr_tail &&\n !(block->prev->instr_tail->opcode->flags & OPCODE_FLAG_VOLATILE)) {\n builder->MergeAdjacentBlocks(block->prev, block);\n merged_any = true;\n }\n }\n block = prev_block;\n }\n\n return true;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/control_flow_simplification_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_CONTROL_FLOW_SIMPLIFICATION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_CONTROL_FLOW_SIMPLIFICATION_PASS_H_\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass ControlFlowSimplificationPass : public CompilerPass {\n public:\n ControlFlowSimplificationPass();\n ~ControlFlowSimplificationPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_CONTROL_FLOW_SIMPLIFICATION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/data_flow_analysis_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/data_flow_analysis_pass.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/backend/backend.h\"\n#include \"xenia/cpu/compiler/compiler.h\"\n#include \"xenia/cpu/processor.h\"\n\n#if XE_COMPILER_MSVC\n#pragma warning(push)\n#pragma warning(disable : 4244)\n#pragma warning(disable : 4267)\n#include \n#pragma warning(pop)\n#else\n#include \n#include \n#endif // XE_COMPILER_MSVC\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::OpcodeSignatureType;\nusing xe::cpu::hir::Value;\n\nDataFlowAnalysisPass::DataFlowAnalysisPass() : CompilerPass() {}\n\nDataFlowAnalysisPass::~DataFlowAnalysisPass() {}\n\nbool DataFlowAnalysisPass::Run(HIRBuilder* builder) {\n // Linearize blocks so that we can detect cycles and propagate dependencies.\n uint32_t block_count = LinearizeBlocks(builder);\n\n // Analyze value flow and add locals as needed.\n AnalyzeFlow(builder, block_count);\n\n return true;\n}\n\nuint32_t DataFlowAnalysisPass::LinearizeBlocks(HIRBuilder* builder) {\n // TODO(benvanik): actually do this - we cheat now knowing that they are in\n // sequential order.\n uint16_t block_ordinal = 0;\n auto block = builder->first_block();\n while (block) {\n block->ordinal = block_ordinal++;\n block = block->next;\n }\n return block_ordinal;\n}\n\nvoid DataFlowAnalysisPass::AnalyzeFlow(HIRBuilder* builder,\n uint32_t block_count) {\n uint32_t max_value_estimate =\n builder->max_value_ordinal() + 1 + block_count * 4;\n\n // Stash for value map. We may want to maintain this during building.\n auto arena = builder->arena();\n auto value_map = reinterpret_cast(\n arena->Alloc(sizeof(Value*) * max_value_estimate, alignof(Value)));\n\n // Allocate incoming bitvectors for use by blocks. We don't need outgoing\n // because they are only used during the block iteration.\n // Mapped by block ordinal.\n // TODO(benvanik): cache this list, grow as needed, etc.\n auto incoming_bitvectors = (llvm::BitVector**)arena->Alloc(\n sizeof(llvm::BitVector*) * block_count, alignof(llvm::BitVector));\n for (auto n = 0u; n < block_count; n++) {\n incoming_bitvectors[n] = new llvm::BitVector(max_value_estimate);\n }\n\n // Walk blocks in reverse and calculate incoming/outgoing values.\n auto block = builder->last_block();\n while (block) {\n // Allocate bitsets based on max value number.\n block->incoming_values = incoming_bitvectors[block->ordinal];\n auto& incoming_values = *block->incoming_values;\n\n // Walk instructions and gather up incoming values.\n auto instr = block->instr_head;\n while (instr) {\n uint32_t signature = instr->opcode->signature;\n#define SET_INCOMING_VALUE(v) \\\n if (v->def && v->def->block != block) { \\\n incoming_values.set(v->ordinal); \\\n } \\\n assert_true(v->ordinal < max_value_estimate); \\\n value_map[v->ordinal] = v;\n if (GET_OPCODE_SIG_TYPE_SRC1(signature) == OPCODE_SIG_TYPE_V) {\n SET_INCOMING_VALUE(instr->src1.value);\n }\n if (GET_OPCODE_SIG_TYPE_SRC2(signature) == OPCODE_SIG_TYPE_V) {\n SET_INCOMING_VALUE(instr->src2.value);\n }\n if (GET_OPCODE_SIG_TYPE_SRC3(signature) == OPCODE_SIG_TYPE_V) {\n SET_INCOMING_VALUE(instr->src3.value);\n }\n#undef SET_INCOMING_VALUE\n instr = instr->next;\n }\n\n // Add all successor incoming values to our outgoing, as we need to\n // pass them through.\n llvm::BitVector outgoing_values(max_value_estimate);\n auto outgoing_edge = block->outgoing_edge_head;\n while (outgoing_edge) {\n if (outgoing_edge->dest->ordinal > block->ordinal) {\n outgoing_values |= *outgoing_edge->dest->incoming_values;\n }\n outgoing_edge = outgoing_edge->outgoing_next;\n }\n incoming_values |= outgoing_values;\n\n // Add stores for all outgoing values.\n auto outgoing_ordinal = outgoing_values.find_first();\n while (outgoing_ordinal != -1) {\n Value* src_value = value_map[outgoing_ordinal];\n assert_not_null(src_value);\n if (!src_value->local_slot) {\n src_value->local_slot = builder->AllocLocal(src_value->type);\n }\n builder->StoreLocal(src_value->local_slot, src_value);\n\n // If we are in the block the value was defined in:\n if (src_value->def->block == block) {\n // Move the store to right after the def, or as soon after\n // as we can (respecting PAIRED flags).\n auto def_next = src_value->def->next;\n while (def_next && def_next->opcode->flags & OPCODE_FLAG_PAIRED_PREV) {\n def_next = def_next->next;\n }\n assert_not_null(def_next);\n builder->last_instr()->MoveBefore(def_next);\n\n // We don't need it in the incoming list.\n incoming_values.reset(outgoing_ordinal);\n } else {\n // Eh, just throw at the end, before the first branch.\n auto tail = block->instr_tail;\n while (tail && tail->opcode->flags & OPCODE_FLAG_BRANCH) {\n tail = tail->prev;\n }\n assert_not_zero(tail);\n builder->last_instr()->MoveBefore(tail->next);\n }\n\n outgoing_ordinal = outgoing_values.find_next(outgoing_ordinal);\n }\n\n // Add loads for all incoming values and rename them in the block.\n auto incoming_ordinal = incoming_values.find_first();\n while (incoming_ordinal != -1) {\n Value* src_value = value_map[incoming_ordinal];\n assert_not_null(src_value);\n if (!src_value->local_slot) {\n src_value->local_slot = builder->AllocLocal(src_value->type);\n }\n Value* local_value = builder->LoadLocal(src_value->local_slot);\n builder->last_instr()->MoveBefore(block->instr_head);\n\n // Swap uses of original value with the local value.\n instr = block->instr_head;\n while (instr) {\n uint32_t signature = instr->opcode->signature;\n if (GET_OPCODE_SIG_TYPE_SRC1(signature) == OPCODE_SIG_TYPE_V) {\n if (instr->src1.value == src_value) {\n instr->set_src1(local_value);\n }\n }\n if (GET_OPCODE_SIG_TYPE_SRC2(signature) == OPCODE_SIG_TYPE_V) {\n if (instr->src2.value == src_value) {\n instr->set_src2(local_value);\n }\n }\n if (GET_OPCODE_SIG_TYPE_SRC3(signature) == OPCODE_SIG_TYPE_V) {\n if (instr->src3.value == src_value) {\n instr->set_src3(local_value);\n }\n }\n instr = instr->next;\n }\n\n incoming_ordinal = incoming_values.find_next(incoming_ordinal);\n }\n\n block = block->prev;\n }\n\n // Cleanup bitvectors.\n for (auto n = 0u; n < block_count; n++) {\n delete incoming_bitvectors[n];\n }\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/data_flow_analysis_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_DATA_FLOW_ANALYSIS_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_DATA_FLOW_ANALYSIS_PASS_H_\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass DataFlowAnalysisPass : public CompilerPass {\n public:\n DataFlowAnalysisPass();\n ~DataFlowAnalysisPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n uint32_t LinearizeBlocks(hir::HIRBuilder* builder);\n void AnalyzeFlow(hir::HIRBuilder* builder, uint32_t block_count);\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_DATA_FLOW_ANALYSIS_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/dead_code_elimination_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/dead_code_elimination_pass.h\"\n\n#include \"xenia/base/profiling.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::Instr;\nusing xe::cpu::hir::Value;\n\nDeadCodeEliminationPass::DeadCodeEliminationPass() : CompilerPass() {}\n\nDeadCodeEliminationPass::~DeadCodeEliminationPass() {}\n\nbool DeadCodeEliminationPass::Run(HIRBuilder* builder) {\n // ContextPromotion/DSE will likely leave around a lot of dead statements.\n // Code generated for comparison/testing produces many unused statements and\n // with proper use analysis it should be possible to remove most of them:\n // After context promotion/simplification:\n // v33.i8 = compare_ult v31.i32, 0\n // v34.i8 = compare_ugt v31.i32, 0\n // v35.i8 = compare_eq v31.i32, 0\n // store_context +300, v33.i8\n // store_context +301, v34.i8\n // store_context +302, v35.i8\n // branch_true v35.i8, loc_8201A484\n // After DSE:\n // v33.i8 = compare_ult v31.i32, 0\n // v34.i8 = compare_ugt v31.i32, 0\n // v35.i8 = compare_eq v31.i32, 0\n // branch_true v35.i8, loc_8201A484\n // After DCE:\n // v35.i8 = compare_eq v31.i32, 0\n // branch_true v35.i8, loc_8201A484\n\n // This also removes useless ASSIGNs:\n // v1 = v0\n // v2 = add v1, v1\n // becomes:\n // v2 = add v0, v0\n\n // We process DCE by reverse iterating over instructions and looking at the\n // use count of the dest value. If it's zero, we can safely remove the\n // instruction. Once we do that, the use counts of any of the src ops may\n // go to zero and we recursively kill up the graph. This is kind of\n // nasty in that we walk randomly and scribble all over memory, but (I think)\n // it's better than doing passes over all instructions until we quiesce.\n // To prevent our iteration from getting all messed up we just replace\n // all removed ops with NOP and then do a single pass that removes them\n // all.\n\n bool any_instr_removed = false;\n bool any_locals_removed = false;\n auto block = builder->first_block();\n while (block) {\n // Walk instructions in reverse.\n Instr* i = block->instr_tail;\n while (i) {\n auto prev = i->prev;\n\n auto opcode = i->opcode;\n if (!(opcode->flags & OPCODE_FLAG_VOLATILE) && i->dest &&\n !i->dest->use_head) {\n // Has no uses and is not volatile. This instruction can die!\n MakeNopRecursive(i);\n any_instr_removed = true;\n } else if (opcode == &OPCODE_ASSIGN_info) {\n // Assignment. These are useless, so just try to remove by completely\n // replacing the value.\n ReplaceAssignment(i);\n }\n\n i = prev;\n }\n\n // Walk instructions forward.\n i = block->instr_head;\n while (i) {\n auto next = i->next;\n\n auto opcode = i->opcode;\n if (opcode == &OPCODE_STORE_LOCAL_info) {\n // Check to see if the store has any interceeding uses after the load.\n // If not, it can be removed (as the local is just passing through the\n // function).\n // We do this after the previous pass so that removed code doesn't keep\n // the local alive.\n if (!CheckLocalUse(i)) {\n any_locals_removed = true;\n }\n }\n\n i = next;\n }\n\n block = block->next;\n }\n\n // Remove all nops.\n if (any_instr_removed) {\n block = builder->first_block();\n while (block) {\n Instr* i = block->instr_head;\n while (i) {\n Instr* next = i->next;\n if (i->opcode == &OPCODE_NOP_info) {\n // Nop - remove!\n i->Remove();\n }\n i = next;\n }\n block = block->next;\n }\n }\n\n // Remove any locals that no longer have uses.\n if (any_locals_removed) {\n // TODO(benvanik): local removal/dealloc.\n auto locals = builder->locals();\n for (auto it = locals.begin(); it != locals.end();) {\n auto next = ++it;\n auto value = *it;\n if (!value->use_head) {\n // Unused, can be removed.\n locals.erase(it);\n }\n it = next;\n }\n }\n\n return true;\n}\n\nvoid DeadCodeEliminationPass::MakeNopRecursive(Instr* i) {\n i->opcode = &hir::OPCODE_NOP_info;\n i->dest->def = NULL;\n i->dest = NULL;\n\n#define MAKE_NOP_SRC(n) \\\n if (i->src##n##_use) { \\\n Value::Use* use = i->src##n##_use; \\\n Value* value = i->src##n.value; \\\n i->src##n##_use = NULL; \\\n i->src##n.value = NULL; \\\n value->RemoveUse(use); \\\n if (!value->use_head) { \\\n /* Value is now unused, so recursively kill it. */ \\\n if (value->def && value->def != i) { \\\n MakeNopRecursive(value->def); \\\n } \\\n } \\\n }\n MAKE_NOP_SRC(1);\n MAKE_NOP_SRC(2);\n MAKE_NOP_SRC(3);\n}\n\nvoid DeadCodeEliminationPass::ReplaceAssignment(Instr* i) {\n auto src = i->src1.value;\n auto dest = i->dest;\n\n auto use = dest->use_head;\n while (use) {\n auto use_instr = use->instr;\n if (use_instr->src1.value == dest) {\n use_instr->set_src1(src);\n }\n if (use_instr->src2.value == dest) {\n use_instr->set_src2(src);\n }\n if (use_instr->src3.value == dest) {\n use_instr->set_src3(src);\n }\n use = use->next;\n }\n\n i->Remove();\n}\n\nbool DeadCodeEliminationPass::CheckLocalUse(Instr* i) {\n auto src = i->src2.value;\n\n auto use = src->use_head;\n if (use) {\n auto use_instr = use->instr;\n if (use_instr->opcode != &OPCODE_LOAD_LOCAL_info) {\n // A valid use (probably). Keep it.\n return true;\n }\n\n // Load/store are paired. They can both be removed.\n use_instr->Remove();\n }\n\n i->Remove();\n\n return false;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/dead_code_elimination_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_DEAD_CODE_ELIMINATION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_DEAD_CODE_ELIMINATION_PASS_H_\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass DeadCodeEliminationPass : public CompilerPass {\n public:\n DeadCodeEliminationPass();\n ~DeadCodeEliminationPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n void MakeNopRecursive(hir::Instr* i);\n void ReplaceAssignment(hir::Instr* i);\n bool CheckLocalUse(hir::Instr* i);\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_DEAD_CODE_ELIMINATION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/finalization_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/finalization_pass.h\"\n\n#include \"third_party/fmt/include/fmt/format.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/backend/backend.h\"\n#include \"xenia/cpu/compiler/compiler.h\"\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::HIRBuilder;\n\nFinalizationPass::FinalizationPass() : CompilerPass() {}\n\nFinalizationPass::~FinalizationPass() {}\n\nbool FinalizationPass::Run(HIRBuilder* builder) {\n // Process the HIR and prepare it for lowering.\n // After this is done the HIR should be ready for emitting.\n\n auto arena = builder->arena();\n\n uint16_t block_ordinal = 0;\n auto block = builder->first_block();\n while (block) {\n block->ordinal = block_ordinal++;\n\n // Ensure all labels have names.\n auto label = block->label_head;\n while (label) {\n if (!label->name) {\n const size_t label_len = 6 + 4;\n char* name = reinterpret_cast(arena->Alloc(label_len + 1, 1));\n assert_true(label->id <= 65535);\n auto end = fmt::format_to_n(name, label_len, \"_label{:04X}\", label->id);\n name[end.size] = '\\0';\n label->name = name;\n }\n label = label->next;\n }\n\n // Remove unneeded jumps.\n auto tail = block->instr_tail;\n if (tail && tail->opcode == &OPCODE_BRANCH_info) {\n // Jump. Check target.\n auto target = tail->src1.label;\n if (target->block == block->next) {\n // Jumping to subsequent block. Remove.\n tail->Remove();\n }\n }\n\n block = block->next;\n }\n\n return true;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/finalization_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_FINALIZATION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_FINALIZATION_PASS_H_\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass FinalizationPass : public CompilerPass {\n public:\n FinalizationPass();\n ~FinalizationPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_FINALIZATION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/memory_sequence_combination_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/memory_sequence_combination_pass.h\"\n\n#include \"xenia/base/profiling.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::Instr;\nusing xe::cpu::hir::Value;\n\nMemorySequenceCombinationPass::MemorySequenceCombinationPass()\n : CompilerPass() {}\n\nMemorySequenceCombinationPass::~MemorySequenceCombinationPass() = default;\n\nbool MemorySequenceCombinationPass::Run(HIRBuilder* builder) {\n // Run over all loads and stores and see if we can collapse sequences into the\n // fat opcodes. See the respective utility functions for examples.\n auto block = builder->first_block();\n while (block) {\n auto i = block->instr_head;\n while (i) {\n if (i->opcode == &OPCODE_LOAD_info ||\n i->opcode == &OPCODE_LOAD_OFFSET_info) {\n CombineLoadSequence(i);\n } else if (i->opcode == &OPCODE_STORE_info ||\n i->opcode == &OPCODE_STORE_OFFSET_info) {\n CombineStoreSequence(i);\n }\n i = i->next;\n }\n block = block->next;\n }\n return true;\n}\n\nvoid MemorySequenceCombinationPass::CombineLoadSequence(Instr* i) {\n // Load with swap:\n // v1.i32 = load v0\n // v2.i32 = byte_swap v1.i32\n // becomes:\n // v1.i32 = load v0, [swap]\n //\n // Load with swap and extend:\n // v1.i32 = load v0\n // v2.i32 = byte_swap v1.i32\n // v3.i64 = zero_extend v2.i32\n // becomes:\n // v1.i64 = load_convert v0, [swap|i32->i64,zero]\n\n if (!i->dest->use_head) {\n // No uses of the load result - ignore. Will be killed by DCE.\n return;\n }\n\n // Ensure all uses of the load result are BYTE_SWAP - if it's mixed we\n // shouldn't transform as we'd have to introduce new swaps!\n auto use = i->dest->use_head;\n while (use) {\n if (use->instr->opcode != &OPCODE_BYTE_SWAP_info) {\n // Not a swap.\n return;\n }\n // TODO(benvanik): allow uses by STORE (we can make that swap).\n use = use->next;\n }\n\n // Merge byte swap into load.\n // Note that we may have already been a swapped operation - this inverts that.\n i->flags ^= LoadStoreFlags::LOAD_STORE_BYTE_SWAP;\n\n // Replace use of byte swap value with loaded value.\n // It's byte_swap vN -> assign vN, so not much to do.\n use = i->dest->use_head;\n while (use) {\n auto next_use = use->next;\n use->instr->opcode = &OPCODE_ASSIGN_info;\n use->instr->flags = 0;\n use = next_use;\n }\n\n // TODO(benvanik): merge in extend/truncate.\n}\n\nvoid MemorySequenceCombinationPass::CombineStoreSequence(Instr* i) {\n // Store with swap:\n // v1.i32 = ...\n // v2.i32 = byte_swap v1.i32\n // store v0, v2.i32\n // becomes:\n // store v0, v1.i32, [swap]\n //\n // Store with truncate and swap:\n // v1.i64 = ...\n // v2.i32 = truncate v1.i64\n // v3.i32 = byte_swap v2.i32\n // store v0, v3.i32\n // becomes:\n // store_convert v0, v1.i64, [swap|i64->i32,trunc]\n\n auto src = i->src2.value;\n if (i->opcode == &OPCODE_STORE_OFFSET_info) {\n src = i->src3.value;\n }\n\n if (src->IsConstant()) {\n // Constant value write - ignore.\n return;\n }\n\n // Find source and ensure it is a byte swap.\n auto def = src->def;\n while (def && def->opcode == &OPCODE_ASSIGN_info) {\n // Skip asignments.\n def = def->src1.value->def;\n }\n if (!def || def->opcode != &OPCODE_BYTE_SWAP_info) {\n // Not a swap/not defined?\n return;\n }\n\n // Merge byte swap into store.\n // Note that we may have already been a swapped operation - this inverts\n // that.\n i->flags ^= LoadStoreFlags::LOAD_STORE_BYTE_SWAP;\n\n // Pull the original value (from before the byte swap).\n // The byte swap itself will go away in DCE.\n if (i->opcode == &OPCODE_STORE_info) {\n i->set_src2(def->src1.value);\n } else if (i->opcode == &OPCODE_STORE_OFFSET_info) {\n i->set_src3(def->src1.value);\n }\n\n // TODO(benvanik): extend/truncate.\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/memory_sequence_combination_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_MEMORY_SEQUENCE_COMBINATION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_MEMORY_SEQUENCE_COMBINATION_PASS_H_\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass MemorySequenceCombinationPass : public CompilerPass {\n public:\n MemorySequenceCombinationPass();\n ~MemorySequenceCombinationPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n void CombineMemorySequences(hir::HIRBuilder* builder);\n void CombineLoadSequence(hir::Instr* i);\n void CombineStoreSequence(hir::Instr* i);\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_MEMORY_SEQUENCE_COMBINATION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/register_allocation_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/register_allocation_pass.h\"\n\n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/profiling.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::backend::MachineInfo;\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::Instr;\nusing xe::cpu::hir::OpcodeSignatureType;\nusing xe::cpu::hir::RegAssignment;\nusing xe::cpu::hir::TypeName;\nusing xe::cpu::hir::Value;\n\n#define ASSERT_NO_CYCLES 0\n\nRegisterAllocationPass::RegisterAllocationPass(const MachineInfo* machine_info)\n : CompilerPass() {\n // Initialize register sets.\n // TODO(benvanik): rewrite in a way that makes sense - this is terrible.\n auto mi_sets = machine_info->register_sets;\n std::memset(&usage_sets_, 0, sizeof(usage_sets_));\n uint32_t n = 0;\n while (mi_sets[n].count) {\n auto& mi_set = mi_sets[n];\n auto usage_set = new RegisterSetUsage();\n usage_sets_.all_sets[n] = usage_set;\n usage_set->count = mi_set.count;\n usage_set->set = &mi_set;\n if (mi_set.types & MachineInfo::RegisterSet::INT_TYPES) {\n usage_sets_.int_set = usage_set;\n }\n if (mi_set.types & MachineInfo::RegisterSet::FLOAT_TYPES) {\n usage_sets_.float_set = usage_set;\n }\n if (mi_set.types & MachineInfo::RegisterSet::VEC_TYPES) {\n usage_sets_.vec_set = usage_set;\n }\n n++;\n }\n}\n\nRegisterAllocationPass::~RegisterAllocationPass() {\n for (size_t n = 0; n < xe::countof(usage_sets_.all_sets); n++) {\n if (!usage_sets_.all_sets[n]) {\n break;\n }\n delete usage_sets_.all_sets[n];\n }\n}\n\nbool RegisterAllocationPass::Run(HIRBuilder* builder) {\n // Simple per-block allocator that operates on SSA form.\n // Registers do not move across blocks, though this could be\n // optimized with some intra-block analysis (dominators/etc).\n // Really, it'd just be nice to have someone who knew what they\n // were doing lower SSA and do this right.\n\n uint16_t block_ordinal = 0;\n uint32_t instr_ordinal = 0;\n auto block = builder->first_block();\n while (block) {\n // Sequential block ordinals.\n block->ordinal = block_ordinal++;\n\n // Reset all state.\n PrepareBlockState();\n\n // Renumber all instructions in the block. This is required so that\n // we can sort the usage pointers below.\n auto instr = block->instr_head;\n while (instr) {\n // Sequential global instruction ordinals.\n instr->ordinal = instr_ordinal++;\n instr = instr->next;\n }\n\n instr = block->instr_head;\n while (instr) {\n const auto info = instr->opcode;\n uint32_t signature = info->signature;\n\n // Update the register use heaps.\n AdvanceUses(instr);\n\n // Check sources for retirement. If any are unused after this instruction\n // we can eagerly evict them to speed up register allocation.\n // Since X64 (and other platforms) can often take advantage of dest==src1\n // register mappings we track retired src1 so that we can attempt to\n // reuse it.\n // NOTE: these checks require that the usage list be sorted!\n bool has_preferred_reg = false;\n RegAssignment preferred_reg = {0};\n if (GET_OPCODE_SIG_TYPE_SRC1(signature) == OPCODE_SIG_TYPE_V &&\n !instr->src1.value->IsConstant()) {\n if (!instr->src1_use->next) {\n // Pull off preferred register. We will try to reuse this for the\n // dest.\n // NOTE: set may be null if this is a store local.\n if (preferred_reg.set) {\n has_preferred_reg = true;\n preferred_reg = instr->src1.value->reg;\n }\n }\n }\n\n if (GET_OPCODE_SIG_TYPE_DEST(signature) == OPCODE_SIG_TYPE_V) {\n // Must not have been set already.\n assert_null(instr->dest->reg.set);\n\n // Sort the usage list. We depend on this in future uses of this\n // variable.\n SortUsageList(instr->dest);\n\n // If we have a preferred register, use that.\n // This way we can help along the stupid X86 two opcode instructions.\n bool allocated;\n if (has_preferred_reg) {\n // Allocate with the given preferred register. If the register is in\n // the wrong set it will not be reused.\n allocated = TryAllocateRegister(instr->dest, preferred_reg);\n } else {\n // Allocate a register. This will either reserve a free one or\n // spill and reuse an active one.\n allocated = TryAllocateRegister(instr->dest);\n }\n if (!allocated) {\n // Failed to allocate register -- need to spill and try again.\n // We spill only those registers we aren't using.\n if (!SpillOneRegister(builder, block, instr->dest->type)) {\n // Unable to spill anything - this shouldn't happen.\n XELOGE(\"Unable to spill any registers\");\n assert_always();\n return false;\n }\n\n // Demand allocation.\n if (!TryAllocateRegister(instr->dest)) {\n // Boned.\n XELOGE(\"Register allocation failed\");\n assert_always();\n return false;\n }\n }\n }\n\n instr = instr->next;\n }\n block = block->next;\n }\n\n return true;\n}\n\nvoid RegisterAllocationPass::DumpUsage(const char* name) {\n#if 0\n fprintf(stdout, \"\\n%s:\\n\", name);\n for (size_t i = 0; i < xe::countof(usage_sets_.all_sets); ++i) {\n auto usage_set = usage_sets_.all_sets[i];\n if (usage_set) {\n fprintf(stdout, \"set %s:\\n\", usage_set->set->name);\n fprintf(stdout, \" avail: %s\\n\",\n usage_set->availability.to_string().c_str());\n fprintf(stdout, \" upcoming uses:\\n\");\n for (auto it = usage_set->upcoming_uses.begin();\n it != usage_set->upcoming_uses.end(); ++it) {\n fprintf(stdout, \" v%d, used at %d\\n\",\n it->value->ordinal,\n it->use->instr->ordinal);\n }\n }\n }\n fflush(stdout);\n#endif\n}\n\nvoid RegisterAllocationPass::PrepareBlockState() {\n for (size_t i = 0; i < xe::countof(usage_sets_.all_sets); ++i) {\n auto usage_set = usage_sets_.all_sets[i];\n if (usage_set) {\n usage_set->availability.set();\n usage_set->upcoming_uses.clear();\n }\n }\n DumpUsage(\"PrepareBlockState\");\n}\n\nvoid RegisterAllocationPass::AdvanceUses(Instr* instr) {\n for (size_t i = 0; i < xe::countof(usage_sets_.all_sets); ++i) {\n auto usage_set = usage_sets_.all_sets[i];\n if (!usage_set) {\n break;\n }\n auto& upcoming_uses = usage_set->upcoming_uses;\n for (size_t j = 0; j < upcoming_uses.size();) {\n auto& upcoming_use = upcoming_uses.at(j);\n if (!upcoming_use.use) {\n // No uses at all - we can remove right away.\n // This comes up from instructions where the dest is never used,\n // like the ATOMIC ops.\n MarkRegAvailable(upcoming_use.value->reg);\n upcoming_uses.erase(upcoming_uses.begin() + j);\n // i remains the same.\n continue;\n }\n if (upcoming_use.use->instr != instr) {\n // Not yet at this instruction.\n ++j;\n continue;\n }\n // The use is from this instruction.\n if (!upcoming_use.use->next) {\n // Last use of the value. We can retire it now.\n MarkRegAvailable(upcoming_use.value->reg);\n upcoming_uses.erase(upcoming_uses.begin() + j);\n // i remains the same.\n continue;\n } else {\n // Used again. Push back the next use.\n // Note that we may be used multiple times this instruction, so\n // eat those.\n auto next_use = upcoming_use.use->next;\n while (next_use->next && next_use->instr == instr) {\n next_use = next_use->next;\n }\n // Remove the iterator.\n auto value = upcoming_use.value;\n upcoming_uses.erase(upcoming_uses.begin() + j);\n assert_true(next_use->instr->block == instr->block);\n assert_true(value->def->block == instr->block);\n upcoming_uses.emplace_back(value, next_use);\n // i remains the same.\n continue;\n }\n }\n }\n DumpUsage(\"AdvanceUses\");\n}\n\nbool RegisterAllocationPass::IsRegInUse(const RegAssignment& reg) {\n RegisterSetUsage* usage_set;\n if (reg.set == usage_sets_.int_set->set) {\n usage_set = usage_sets_.int_set;\n } else if (reg.set == usage_sets_.float_set->set) {\n usage_set = usage_sets_.float_set;\n } else {\n usage_set = usage_sets_.vec_set;\n }\n return !usage_set->availability.test(reg.index);\n}\n\nRegisterAllocationPass::RegisterSetUsage* RegisterAllocationPass::MarkRegUsed(\n const RegAssignment& reg, Value* value, Value::Use* use) {\n auto usage_set = RegisterSetForValue(value);\n usage_set->availability.set(reg.index, false);\n usage_set->upcoming_uses.emplace_back(value, use);\n DumpUsage(\"MarkRegUsed\");\n return usage_set;\n}\n\nRegisterAllocationPass::RegisterSetUsage*\nRegisterAllocationPass::MarkRegAvailable(const hir::RegAssignment& reg) {\n RegisterSetUsage* usage_set;\n if (reg.set == usage_sets_.int_set->set) {\n usage_set = usage_sets_.int_set;\n } else if (reg.set == usage_sets_.float_set->set) {\n usage_set = usage_sets_.float_set;\n } else {\n usage_set = usage_sets_.vec_set;\n }\n usage_set->availability.set(reg.index, true);\n return usage_set;\n}\n\nbool RegisterAllocationPass::TryAllocateRegister(\n Value* value, const RegAssignment& preferred_reg) {\n // If the preferred register matches type and is available, use it.\n auto usage_set = RegisterSetForValue(value);\n if (usage_set->set == preferred_reg.set) {\n // Check if available.\n if (!IsRegInUse(preferred_reg)) {\n // Mark as in-use and return. Best case.\n MarkRegUsed(preferred_reg, value, value->use_head);\n value->reg = preferred_reg;\n return true;\n }\n }\n\n // Otherwise, fallback to allocating like normal.\n return TryAllocateRegister(value);\n}\n\nbool RegisterAllocationPass::TryAllocateRegister(Value* value) {\n // Get the set this register is in.\n RegisterSetUsage* usage_set = RegisterSetForValue(value);\n\n // Find the first free register, if any.\n // We have to ensure it's a valid one (in our count).\n uint32_t first_unused = 0;\n bool none_used = xe::bit_scan_forward(\n static_cast(usage_set->availability.to_ulong()), &first_unused);\n if (none_used && first_unused < usage_set->count) {\n // Available! Use it!\n value->reg.set = usage_set->set;\n value->reg.index = first_unused;\n MarkRegUsed(value->reg, value, value->use_head);\n return true;\n }\n\n // None available! Spill required.\n return false;\n}\n\nbool RegisterAllocationPass::SpillOneRegister(HIRBuilder* builder, Block* block,\n TypeName required_type) {\n // Get the set that we will be picking from.\n RegisterSetUsage* usage_set;\n if (required_type <= INT64_TYPE) {\n usage_set = usage_sets_.int_set;\n } else if (required_type <= FLOAT64_TYPE) {\n usage_set = usage_sets_.float_set;\n } else {\n usage_set = usage_sets_.vec_set;\n }\n\n DumpUsage(\"SpillOneRegister (pre)\");\n // Pick the one with the furthest next use.\n assert_true(!usage_set->upcoming_uses.empty());\n auto furthest_usage =\n std::max_element(usage_set->upcoming_uses.begin(),\n usage_set->upcoming_uses.end(), &RegisterUsage::Compare);\n assert_true(furthest_usage->value->def->block == block);\n assert_true(furthest_usage->use->instr->block == block);\n auto spill_value = furthest_usage->value;\n Value::Use* prev_use = furthest_usage->use->prev;\n Value::Use* next_use = furthest_usage->use;\n assert_not_null(next_use);\n usage_set->upcoming_uses.erase(furthest_usage);\n DumpUsage(\"SpillOneRegister (post)\");\n const auto reg = spill_value->reg;\n\n // We know the spill_value use list is sorted, so we can cut it right now.\n // This makes it easier down below.\n auto new_head_use = next_use;\n\n // Allocate local.\n if (spill_value->local_slot) {\n // Value is already assigned a slot. Since we allocate in order and this is\n // all SSA we know the stored value will be exactly what we want. Yay,\n // we can prevent the redundant store!\n // In fact, we may even want to pin this spilled value so that we always\n // use the spilled value and prevent the need for more locals.\n } else {\n // Allocate a local slot.\n spill_value->local_slot = builder->AllocLocal(spill_value->type);\n\n // Add store.\n builder->StoreLocal(spill_value->local_slot, spill_value);\n auto spill_store = builder->last_instr();\n auto spill_store_use = spill_store->src2_use;\n assert_null(spill_store_use->prev);\n if (prev_use && prev_use->instr->opcode->flags & OPCODE_FLAG_PAIRED_PREV) {\n // Instruction is paired. This is bad. We will insert the spill after the\n // paired instruction.\n assert_not_null(prev_use->instr->next);\n spill_store->MoveBefore(prev_use->instr->next);\n\n // Update last use.\n spill_value->last_use = spill_store;\n } else if (prev_use) {\n // We insert the store immediately before the previous use.\n // If we were smarter we could then re-run allocation and reuse the\n // register\n // once dropped.\n spill_store->MoveBefore(prev_use->instr);\n\n // Update last use.\n spill_value->last_use = prev_use->instr;\n } else {\n // This is the first use, so the only thing we have is the define.\n // Move the store to right after that.\n spill_store->MoveBefore(spill_value->def->next);\n\n // Update last use.\n spill_value->last_use = spill_store;\n }\n }\n\n#if ASSERT_NO_CYCLES\n builder->AssertNoCycles();\n spill_value->def->block->AssertNoCycles();\n#endif // ASSERT_NO_CYCLES\n\n // Add load.\n // Inserted immediately before the next use. Since by definition the next\n // use is after the instruction requesting the spill we know we haven't\n // done allocation for that code yet and can let that be handled\n // automatically when we get to it.\n auto new_value = builder->LoadLocal(spill_value->local_slot);\n auto spill_load = builder->last_instr();\n spill_load->MoveBefore(next_use->instr);\n // Note: implicit first use added.\n\n#if ASSERT_NO_CYCLES\n builder->AssertNoCycles();\n spill_value->def->block->AssertNoCycles();\n#endif // ASSERT_NO_CYCLES\n\n // Set the local slot of the new value to our existing one. This way we will\n // reuse that same memory if needed.\n new_value->local_slot = spill_value->local_slot;\n\n // Rename all future uses of the SSA value to the new value as loaded\n // from the local.\n // We can quickly do this by walking the use list. Because the list is\n // already sorted we know we are going to end up with a sorted list.\n auto walk_use = new_head_use;\n auto new_use_tail = walk_use;\n while (walk_use) {\n auto next_walk_use = walk_use->next;\n auto instr = walk_use->instr;\n\n uint32_t signature = instr->opcode->signature;\n if (GET_OPCODE_SIG_TYPE_SRC1(signature) == OPCODE_SIG_TYPE_V) {\n if (instr->src1.value == spill_value) {\n instr->set_src1(new_value);\n }\n }\n if (GET_OPCODE_SIG_TYPE_SRC2(signature) == OPCODE_SIG_TYPE_V) {\n if (instr->src2.value == spill_value) {\n instr->set_src2(new_value);\n }\n }\n if (GET_OPCODE_SIG_TYPE_SRC3(signature) == OPCODE_SIG_TYPE_V) {\n if (instr->src3.value == spill_value) {\n instr->set_src3(new_value);\n }\n }\n\n walk_use = next_walk_use;\n if (walk_use) {\n new_use_tail = walk_use;\n }\n }\n new_value->last_use = new_use_tail->instr;\n\n // Update tracking.\n MarkRegAvailable(reg);\n\n return true;\n}\n\nRegisterAllocationPass::RegisterSetUsage*\nRegisterAllocationPass::RegisterSetForValue(const Value* value) {\n if (value->type <= INT64_TYPE) {\n return usage_sets_.int_set;\n } else if (value->type <= FLOAT64_TYPE) {\n return usage_sets_.float_set;\n } else {\n return usage_sets_.vec_set;\n }\n}\n\nnamespace {\nint CompareValueUse(const Value::Use* a, const Value::Use* b) {\n return a->instr->ordinal - b->instr->ordinal;\n}\n} // namespace\nvoid RegisterAllocationPass::SortUsageList(Value* value) {\n // Modified in-place linked list sort from:\n // https://www.chiark.greenend.org.uk/~sgtatham/algorithms/listsort.c\n if (!value->use_head) {\n return;\n }\n Value::Use* head = value->use_head;\n Value::Use* tail = nullptr;\n int insize = 1;\n while (true) {\n auto p = head;\n head = nullptr;\n tail = nullptr;\n // count number of merges we do in this pass\n int nmerges = 0;\n while (p) {\n // there exists a merge to be done\n nmerges++;\n // step 'insize' places along from p\n auto q = p;\n int psize = 0;\n for (int i = 0; i < insize; i++) {\n psize++;\n q = q->next;\n if (!q) break;\n }\n // if q hasn't fallen off end, we have two lists to merge\n int qsize = insize;\n // now we have two lists; merge them\n while (psize > 0 || (qsize > 0 && q)) {\n // decide whether next element of merge comes from p or q\n Value::Use* e = nullptr;\n if (psize == 0) {\n // p is empty; e must come from q\n e = q;\n q = q->next;\n qsize--;\n } else if (qsize == 0 || !q) {\n // q is empty; e must come from p\n e = p;\n p = p->next;\n psize--;\n } else if (CompareValueUse(p, q) <= 0) {\n // First element of p is lower (or same); e must come from p\n e = p;\n p = p->next;\n psize--;\n } else {\n // First element of q is lower; e must come from q\n e = q;\n q = q->next;\n qsize--;\n }\n // add the next element to the merged list\n if (tail) {\n tail->next = e;\n } else {\n head = e;\n }\n // Maintain reverse pointers in a doubly linked list.\n e->prev = tail;\n tail = e;\n }\n // now p has stepped 'insize' places along, and q has too\n p = q;\n }\n if (tail) {\n tail->next = nullptr;\n }\n // If we have done only one merge, we're finished\n if (nmerges <= 1) {\n // allow for nmerges==0, the empty list case\n break;\n }\n // Otherwise repeat, merging lists twice the size\n insize *= 2;\n }\n\n value->use_head = head;\n value->last_use = tail->instr;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/register_allocation_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_REGISTER_ALLOCATION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_REGISTER_ALLOCATION_PASS_H_\n\n#include \n#include \n#include \n#include \n\n#include \"xenia/cpu/backend/machine_info.h\"\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass RegisterAllocationPass : public CompilerPass {\n public:\n explicit RegisterAllocationPass(const backend::MachineInfo* machine_info);\n ~RegisterAllocationPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n // TODO(benvanik): rewrite all this set shit -- too much indirection, the\n // complexity is not needed.\n struct RegisterUsage {\n hir::Value* value;\n hir::Value::Use* use;\n RegisterUsage() : value(nullptr), use(nullptr) {}\n RegisterUsage(hir::Value* value_, hir::Value::Use* use_)\n : value(value_), use(use_) {}\n static bool Compare(const RegisterUsage& a, const RegisterUsage& b) {\n return a.use->instr->ordinal < b.use->instr->ordinal;\n }\n };\n struct RegisterSetUsage {\n const backend::MachineInfo::RegisterSet* set = nullptr;\n uint32_t count = 0;\n std::bitset<32> availability = 0;\n // TODO(benvanik): another data type.\n std::vector upcoming_uses;\n };\n\n void DumpUsage(const char* name);\n void PrepareBlockState();\n void AdvanceUses(hir::Instr* instr);\n bool IsRegInUse(const hir::RegAssignment& reg);\n RegisterSetUsage* MarkRegUsed(const hir::RegAssignment& reg,\n hir::Value* value, hir::Value::Use* use);\n RegisterSetUsage* MarkRegAvailable(const hir::RegAssignment& reg);\n\n bool TryAllocateRegister(hir::Value* value,\n const hir::RegAssignment& preferred_reg);\n bool TryAllocateRegister(hir::Value* value);\n bool SpillOneRegister(hir::HIRBuilder* builder, hir::Block* block,\n hir::TypeName required_type);\n\n RegisterSetUsage* RegisterSetForValue(const hir::Value* value);\n\n void SortUsageList(hir::Value* value);\n\n private:\n struct {\n RegisterSetUsage* int_set = nullptr;\n RegisterSetUsage* float_set = nullptr;\n RegisterSetUsage* vec_set = nullptr;\n RegisterSetUsage* all_sets[3];\n } usage_sets_;\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_REGISTER_ALLOCATION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/simplification_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/simplification_pass.h\"\n\n#include \"xenia/base/profiling.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::Instr;\nusing xe::cpu::hir::Value;\n\nSimplificationPass::SimplificationPass() : ConditionalGroupSubpass() {}\n\nSimplificationPass::~SimplificationPass() {}\n\nbool SimplificationPass::Run(HIRBuilder* builder, bool& result) {\n result = false;\n result |= EliminateConversions(builder);\n result |= SimplifyAssignments(builder);\n return true;\n}\n\nbool SimplificationPass::EliminateConversions(HIRBuilder* builder) {\n // First, we check for truncates/extensions that can be skipped.\n // This generates some assignments which then the second step will clean up.\n // Both zero/sign extends can be skipped:\n // v1.i64 = zero/sign_extend v0.i32\n // v2.i32 = truncate v1.i64\n // becomes:\n // v1.i64 = zero/sign_extend v0.i32 (may be dead code removed later)\n // v2.i32 = v0.i32\n\n bool result = false;\n auto block = builder->first_block();\n while (block) {\n auto i = block->instr_head;\n while (i) {\n // To make things easier we check in reverse (source of truncate/extend\n // back to definition).\n if (i->opcode == &OPCODE_TRUNCATE_info) {\n // Matches zero/sign_extend + truncate.\n result |= CheckTruncate(i);\n } else if (i->opcode == &OPCODE_BYTE_SWAP_info) {\n // Matches byte swap + byte swap.\n // This is pretty rare within the same basic block, but is in the\n // memcpy hot path and (probably) worth it. Maybe.\n result |= CheckByteSwap(i);\n }\n i = i->next;\n }\n block = block->next;\n }\n return result;\n}\n\nbool SimplificationPass::CheckTruncate(Instr* i) {\n // Walk backward up src's chain looking for an extend. We may have\n // assigns, so skip those.\n auto src = i->src1.value;\n auto def = src->def;\n while (def && def->opcode == &OPCODE_ASSIGN_info) {\n // Skip asignments.\n def = def->src1.value->def;\n }\n if (def) {\n if (def->opcode == &OPCODE_SIGN_EXTEND_info) {\n // Value comes from a sign extend.\n if (def->src1.value->type == i->dest->type) {\n // Types match, use original by turning this into an assign.\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(def->src1.value);\n return true;\n }\n } else if (def->opcode == &OPCODE_ZERO_EXTEND_info) {\n // Value comes from a zero extend.\n if (def->src1.value->type == i->dest->type) {\n // Types match, use original by turning this into an assign.\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(def->src1.value);\n return true;\n }\n }\n }\n return false;\n}\n\nbool SimplificationPass::CheckByteSwap(Instr* i) {\n // Walk backward up src's chain looking for a byte swap. We may have\n // assigns, so skip those.\n auto src = i->src1.value;\n auto def = src->def;\n while (def && def->opcode == &OPCODE_ASSIGN_info) {\n // Skip asignments.\n def = def->src1.value->def;\n }\n if (def && def->opcode == &OPCODE_BYTE_SWAP_info) {\n // Value comes from a byte swap.\n if (def->src1.value->type == i->dest->type) {\n // Types match, use original by turning this into an assign.\n i->Replace(&OPCODE_ASSIGN_info, 0);\n i->set_src1(def->src1.value);\n return true;\n }\n }\n return false;\n}\n\nbool SimplificationPass::SimplifyAssignments(HIRBuilder* builder) {\n // Run over the instructions and rename assigned variables:\n // v1 = v0\n // v2 = v1\n // v3 = add v0, v2\n // becomes:\n // v1 = v0\n // v2 = v0\n // v3 = add v0, v0\n // This could be run several times, as it could make other passes faster\n // to compute (for example, ConstantPropagation). DCE will take care of\n // the useless assigns.\n //\n // We do this by walking each instruction. For each value op we\n // look at its def instr to see if it's an assign - if so, we use the src\n // of that instr. Because we may have chains, we do this recursively until\n // we find a non-assign def.\n\n bool result = false;\n auto block = builder->first_block();\n while (block) {\n auto i = block->instr_head;\n while (i) {\n uint32_t signature = i->opcode->signature;\n if (GET_OPCODE_SIG_TYPE_SRC1(signature) == OPCODE_SIG_TYPE_V) {\n bool modified = false;\n i->set_src1(CheckValue(i->src1.value, modified));\n result |= modified;\n }\n if (GET_OPCODE_SIG_TYPE_SRC2(signature) == OPCODE_SIG_TYPE_V) {\n bool modified = false;\n i->set_src2(CheckValue(i->src2.value, modified));\n result |= modified;\n }\n if (GET_OPCODE_SIG_TYPE_SRC3(signature) == OPCODE_SIG_TYPE_V) {\n bool modified = false;\n i->set_src3(CheckValue(i->src3.value, modified));\n result |= modified;\n }\n i = i->next;\n }\n block = block->next;\n }\n return result;\n}\n\nValue* SimplificationPass::CheckValue(Value* value, bool& result) {\n auto def = value->def;\n if (def && def->opcode == &OPCODE_ASSIGN_info) {\n // Value comes from an assignment - recursively find if it comes from\n // another assignment. It probably doesn't, if we already replaced it.\n auto replacement = def->src1.value;\n while (true) {\n def = replacement->def;\n if (!def || def->opcode != &OPCODE_ASSIGN_info) {\n break;\n }\n replacement = def->src1.value;\n }\n result = true;\n return replacement;\n }\n result = false;\n return value;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/simplification_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_SIMPLIFICATION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_SIMPLIFICATION_PASS_H_\n\n#include \"xenia/cpu/compiler/passes/conditional_group_subpass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass SimplificationPass : public ConditionalGroupSubpass {\n public:\n SimplificationPass();\n ~SimplificationPass() override;\n\n bool Run(hir::HIRBuilder* builder, bool& result) override;\n\n private:\n bool EliminateConversions(hir::HIRBuilder* builder);\n bool CheckTruncate(hir::Instr* i);\n bool CheckByteSwap(hir::Instr* i);\n\n bool SimplifyAssignments(hir::HIRBuilder* builder);\n hir::Value* CheckValue(hir::Value* value, bool& result);\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_SIMPLIFICATION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/validation_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/validation_pass.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/backend/backend.h\"\n#include \"xenia/cpu/compiler/compiler.h\"\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::Block;\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::Instr;\nusing xe::cpu::hir::OpcodeSignatureType;\nusing xe::cpu::hir::Value;\n\nValidationPass::ValidationPass() : CompilerPass() {}\n\nValidationPass::~ValidationPass() {}\n\nbool ValidationPass::Run(HIRBuilder* builder) {\n#if 0\n StringBuffer str;\n builder->Dump(&str);\n printf(\"%s\", str.GetString());\n fflush(stdout);\n str.Reset();\n#endif // 0\n\n auto block = builder->first_block();\n while (block) {\n auto label = block->label_head;\n while (label) {\n assert_true(label->block == block);\n if (label->block != block) {\n return false;\n }\n label = label->next;\n }\n\n auto instr = block->instr_head;\n while (instr) {\n if (!ValidateInstruction(block, instr)) {\n return false;\n }\n instr = instr->next;\n }\n\n block = block->next;\n }\n\n return true;\n}\n\nbool ValidationPass::ValidateInstruction(Block* block, Instr* instr) {\n assert_true(instr->block == block);\n if (instr->block != block) {\n return false;\n }\n\n if (instr->dest) {\n assert_true(instr->dest->def == instr);\n auto use = instr->dest->use_head;\n while (use) {\n assert_true(use->instr->block == block);\n use = use->next;\n }\n }\n\n uint32_t signature = instr->opcode->signature;\n if (GET_OPCODE_SIG_TYPE_SRC1(signature) == OPCODE_SIG_TYPE_V) {\n if (!ValidateValue(block, instr, instr->src1.value)) {\n return false;\n }\n }\n if (GET_OPCODE_SIG_TYPE_SRC2(signature) == OPCODE_SIG_TYPE_V) {\n if (!ValidateValue(block, instr, instr->src2.value)) {\n return false;\n }\n }\n if (GET_OPCODE_SIG_TYPE_SRC3(signature) == OPCODE_SIG_TYPE_V) {\n if (!ValidateValue(block, instr, instr->src3.value)) {\n return false;\n }\n }\n\n return true;\n}\n\nbool ValidationPass::ValidateValue(Block* block, Instr* instr, Value* value) {\n // if (value->def) {\n // auto def = value->def;\n // assert_true(def->block == block);\n // if (def->block != block) {\n // return false;\n // }\n //}\n return true;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/validation_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_VALIDATION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_VALIDATION_PASS_H_\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass ValidationPass : public CompilerPass {\n public:\n ValidationPass();\n ~ValidationPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n bool ValidateInstruction(hir::Block* block, hir::Instr* instr);\n bool ValidateValue(hir::Block* block, hir::Instr* instr, hir::Value* value);\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_VALIDATION_PASS_H_\n" }, { "path": "src/xenia/cpu/compiler/passes/value_reduction_pass.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/compiler/passes/value_reduction_pass.h\"\n\n#include \"xenia/base/platform.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/backend/backend.h\"\n#include \"xenia/cpu/compiler/compiler.h\"\n#include \"xenia/cpu/processor.h\"\n\n#if XE_COMPILER_MSVC\n#pragma warning(push)\n#pragma warning(disable : 4244)\n#pragma warning(disable : 4267)\n#include \n#pragma warning(pop)\n#else\n#include \n#include \n#endif // XE_COMPILER_MSVC\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\n// TODO(benvanik): remove when enums redefined.\nusing namespace xe::cpu::hir;\n\nusing xe::cpu::hir::HIRBuilder;\nusing xe::cpu::hir::OpcodeInfo;\nusing xe::cpu::hir::Value;\n\nValueReductionPass::ValueReductionPass() : CompilerPass() {}\n\nValueReductionPass::~ValueReductionPass() {}\n\nvoid ValueReductionPass::ComputeLastUse(Value* value) {\n // TODO(benvanik): compute during construction?\n // Note that this list isn't sorted (unfortunately), so we have to scan\n // them all.\n uint32_t max_ordinal = 0;\n Value::Use* last_use = nullptr;\n auto use = value->use_head;\n while (use) {\n if (!last_use || use->instr->ordinal >= max_ordinal) {\n last_use = use;\n max_ordinal = use->instr->ordinal;\n }\n use = use->next;\n }\n value->last_use = last_use ? last_use->instr : nullptr;\n}\n\nbool ValueReductionPass::Run(HIRBuilder* builder) {\n // Walk each block and reuse variable ordinals as much as possible.\n\n llvm::BitVector ordinals(builder->max_value_ordinal());\n\n auto block = builder->first_block();\n while (block) {\n // Reset used ordinals.\n ordinals.reset();\n\n // Renumber all instructions to make liveness tracking easier.\n uint32_t instr_ordinal = 0;\n auto instr = block->instr_head;\n while (instr) {\n instr->ordinal = instr_ordinal++;\n instr = instr->next;\n }\n\n instr = block->instr_head;\n while (instr) {\n const OpcodeInfo* info = instr->opcode;\n auto dest_type = GET_OPCODE_SIG_TYPE_DEST(info->signature);\n auto src1_type = GET_OPCODE_SIG_TYPE_SRC1(info->signature);\n auto src2_type = GET_OPCODE_SIG_TYPE_SRC2(info->signature);\n auto src3_type = GET_OPCODE_SIG_TYPE_SRC3(info->signature);\n if (src1_type == OPCODE_SIG_TYPE_V) {\n auto v = instr->src1.value;\n if (!v->last_use) {\n ComputeLastUse(v);\n }\n if (v->last_use == instr) {\n // Available.\n if (!instr->src1.value->IsConstant()) {\n ordinals.reset(v->ordinal);\n }\n }\n }\n if (src2_type == OPCODE_SIG_TYPE_V) {\n auto v = instr->src2.value;\n if (!v->last_use) {\n ComputeLastUse(v);\n }\n if (v->last_use == instr) {\n // Available.\n if (!instr->src2.value->IsConstant()) {\n ordinals.reset(v->ordinal);\n }\n }\n }\n if (src3_type == OPCODE_SIG_TYPE_V) {\n auto v = instr->src3.value;\n if (!v->last_use) {\n ComputeLastUse(v);\n }\n if (v->last_use == instr) {\n // Available.\n if (!instr->src3.value->IsConstant()) {\n ordinals.reset(v->ordinal);\n }\n }\n }\n if (dest_type == OPCODE_SIG_TYPE_V) {\n // Dest values are processed last, as they may be able to reuse a\n // source value ordinal.\n auto v = instr->dest;\n // Find a lower ordinal.\n for (auto n = 0u; n < ordinals.size(); n++) {\n if (!ordinals.test(n)) {\n ordinals.set(n);\n v->ordinal = n;\n break;\n }\n }\n }\n\n instr = instr->next;\n }\n\n block = block->next;\n }\n\n return true;\n}\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/compiler/passes/value_reduction_pass.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_COMPILER_PASSES_VALUE_REDUCTION_PASS_H_\n#define XENIA_CPU_COMPILER_PASSES_VALUE_REDUCTION_PASS_H_\n\n#include \"xenia/cpu/compiler/compiler_pass.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace compiler {\nnamespace passes {\n\nclass ValueReductionPass : public CompilerPass {\n public:\n ValueReductionPass();\n ~ValueReductionPass() override;\n\n bool Run(hir::HIRBuilder* builder) override;\n\n private:\n void ComputeLastUse(hir::Value* value);\n};\n\n} // namespace passes\n} // namespace compiler\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_COMPILER_PASSES_VALUE_REDUCTION_PASS_H_\n" }, { "path": "src/xenia/cpu/cpu_flags.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/cpu_flags.h\"\n\nDEFINE_string(cpu, \"any\", \"CPU backend [any, x64].\", \"CPU\");\n\nDEFINE_string(\n load_module_map, \"\",\n \"Loads a .map for symbol names and to diff with the generated symbol \"\n \"database.\",\n \"CPU\");\n\nDEFINE_bool(disassemble_functions, false,\n \"Disassemble functions during generation.\", \"CPU\");\n\nDEFINE_bool(trace_functions, false, \"Generate tracing for function statistics.\",\n \"CPU\");\nDEFINE_bool(trace_function_coverage, false,\n \"Generate tracing for function instruction coverage statistics.\",\n \"CPU\");\nDEFINE_bool(trace_function_references, false,\n \"Generate tracing for function address references.\", \"CPU\");\nDEFINE_bool(trace_function_data, false,\n \"Generate tracing for function result data.\", \"CPU\");\n\nDEFINE_bool(\n disable_global_lock, false,\n \"Disables global lock usage in guest code. Does not affect host code.\",\n \"CPU\");\n\nDEFINE_bool(validate_hir, false,\n \"Perform validation checks on the HIR during compilation.\", \"CPU\");\n\nDEFINE_uint64(\n pvr, 0x710700,\n \"Processor version and revision number.\\nBits 0 to 15 are the version \"\n \"number.\\nBits 16 to 31 are the revision number.\\nNote: Some XEXs (such as \"\n \"mfgbootlauncher.xex) may check for a value that's less than 0x710700.\",\n \"CPU\");\n\n// Breakpoints:\nDEFINE_uint64(break_on_instruction, 0,\n \"int3 before the given guest address is executed.\", \"CPU\");\nDEFINE_int32(break_condition_gpr, -1, \"GPR compared to\", \"CPU\");\nDEFINE_uint64(break_condition_value, 0, \"value compared against\", \"CPU\");\nDEFINE_string(break_condition_op, \"eq\", \"comparison operator\", \"CPU\");\nDEFINE_bool(break_condition_truncate, true, \"truncate value to 32-bits\", \"CPU\");\n\nDEFINE_bool(break_on_debugbreak, true, \"int3 on JITed __debugbreak requests.\",\n \"CPU\");\n" }, { "path": "src/xenia/cpu/cpu_flags.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_CPU_FLAGS_H_\n#define XENIA_CPU_CPU_FLAGS_H_\n#include \"xenia/base/cvar.h\"\n\nDECLARE_string(cpu);\n\nDECLARE_string(load_module_map);\n\nDECLARE_bool(disassemble_functions);\n\nDECLARE_bool(trace_functions);\nDECLARE_bool(trace_function_coverage);\nDECLARE_bool(trace_function_references);\nDECLARE_bool(trace_function_data);\n\nDECLARE_bool(disable_global_lock);\n\nDECLARE_bool(validate_hir);\n\nDECLARE_uint64(pvr);\n\n// Breakpoints:\nDECLARE_uint64(break_on_instruction);\nDECLARE_int32(break_condition_gpr);\nDECLARE_uint64(break_condition_value);\nDECLARE_string(break_condition_op);\nDECLARE_bool(break_condition_truncate);\n\nDECLARE_bool(break_on_debugbreak);\n\n#endif // XENIA_CPU_CPU_FLAGS_H_\n" }, { "path": "src/xenia/cpu/debug_listener.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2016 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_DEBUG_LISTENER_H_\n#define XENIA_CPU_DEBUG_LISTENER_H_\n\nnamespace xe {\nnamespace cpu {\n\nclass Breakpoint;\nstruct ThreadDebugInfo;\n\n// Debug event listener interface.\n// Implementations will receive calls from arbitrary threads and must marshal\n// them as required.\nclass DebugListener {\n public:\n // Handles request for debugger focus (such as when the user requests the\n // debugger be brought to the front).\n virtual void OnFocus() = 0;\n\n // Handles the debugger detaching from the target.\n // This will be called on shutdown or when the user requests the debug session\n // end.\n virtual void OnDetached() = 0;\n\n // Handles execution being interrupted and transitioning to\n // ExceutionState::kPaused.\n virtual void OnExecutionPaused() = 0;\n // Handles execution continuing when transitioning to\n // ExecutionState::kRunning or ExecutionState::kStepping.\n virtual void OnExecutionContinued() = 0;\n // Handles execution ending when transitioning to ExecutionState::kEnded.\n virtual void OnExecutionEnded() = 0;\n\n // Handles step completion events when a requested step operation completes.\n // The thread is the one that hit its step target. Note that because multiple\n // threads could be stepping simultaneously (such as a run-to-cursor) use the\n // thread passed instead of keeping any other state.\n virtual void OnStepCompleted(ThreadDebugInfo* thread_info) = 0;\n\n // Handles breakpoint events as they are hit per-thread.\n // Breakpoints may be hit during stepping.\n virtual void OnBreakpointHit(Breakpoint* breakpoint,\n ThreadDebugInfo* thread_info) = 0;\n};\n\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_DEBUG_LISTENER_H_\n" }, { "path": "src/xenia/cpu/elf_module.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/elf_module.h\"\n\n#include \n#include \n\n#include \"xenia/base/byte_order.h\"\n#include \"xenia/base/logging.h\"\n#include \"xenia/cpu/processor.h\"\n\nnamespace xe {\nnamespace cpu {\n\nElfModule::ElfModule(Processor* processor, kernel::KernelState* kernel_state)\n : Module(processor), kernel_state_(kernel_state) {}\n\nElfModule::~ElfModule() = default;\n\n// ELF structures\nstruct elf32_ehdr {\n uint8_t e_ident[16];\n xe::be e_type;\n xe::be e_machine;\n xe::be e_version;\n xe::be e_entry;\n xe::be e_phoff;\n xe::be e_shoff;\n xe::be e_flags;\n xe::be e_ehsize;\n xe::be e_phentsize;\n xe::be e_phnum;\n xe::be e_shentsize;\n xe::be e_shnum;\n xe::be e_shtrndx;\n};\n\nstruct elf32_phdr {\n xe::be p_type;\n xe::be p_offset;\n xe::be p_vaddr;\n xe::be p_paddr;\n xe::be p_filesz;\n xe::be p_memsz;\n xe::be p_flags;\n xe::be p_align;\n};\n\nbool ElfModule::is_executable() const {\n auto hdr = reinterpret_cast(elf_header_mem_.data());\n return hdr->e_entry != 0;\n}\n\nbool ElfModule::Load(const std::string_view name, const std::string_view path,\n const void* elf_addr, size_t elf_length) {\n name_ = name;\n path_ = path;\n\n uint8_t* pelf = (uint8_t*)elf_addr;\n elf32_ehdr* hdr = (elf32_ehdr*)(pelf + 0x0);\n if (hdr->e_ident[0] != 0x7F || hdr->e_ident[1] != 'E' ||\n hdr->e_ident[2] != 'L' || hdr->e_ident[3] != 'F') {\n // Not an ELF file!\n return false;\n }\n\n assert_true(hdr->e_ident[4] == 1); // 32bit\n\n if (hdr->e_type != 2 /* ET_EXEC */) {\n // Not executable (shared objects not supported yet)\n XELOGE(\"ELF: Could not load ELF because it isn't executable!\");\n return false;\n }\n\n if (hdr->e_machine != 20 /* EM_PPC */) {\n // Not a PPC ELF!\n XELOGE(\n \"ELF: Could not load ELF because target machine is not PPC! (target: \"\n \"{})\",\n uint32_t(hdr->e_machine));\n return false;\n }\n\n // Parse LOAD program headers and load into memory.\n if (!hdr->e_phoff) {\n XELOGE(\"ELF: File doesn't have a program header!\");\n return false;\n }\n\n if (!hdr->e_entry) {\n XELOGE(\"ELF: Executable has no entry point!\");\n return false;\n }\n\n // Entry point virtual address\n entry_point_ = hdr->e_entry;\n\n // Copy the ELF header\n elf_header_mem_.resize(hdr->e_ehsize);\n std::memcpy(elf_header_mem_.data(), hdr, hdr->e_ehsize);\n\n assert_true(hdr->e_phentsize == sizeof(elf32_phdr));\n elf32_phdr* phdr = (elf32_phdr*)(pelf + hdr->e_phoff);\n for (uint32_t i = 0; i < hdr->e_phnum; i++) {\n if (phdr[i].p_type == 1 /* PT_LOAD */ ||\n phdr[i].p_type == 2 /* PT_DYNAMIC */) {\n // Allocate and copy into memory.\n // Base address @ 0x80000000\n if (phdr[i].p_vaddr < 0x80000000 || phdr[i].p_vaddr > 0x9FFFFFFF) {\n XELOGE(\"ELF: Could not allocate memory for section @ address 0x{:08X}\",\n uint32_t(phdr[i].p_vaddr));\n return false;\n }\n\n uint32_t virtual_addr = phdr[i].p_vaddr & ~(phdr[i].p_align - 1);\n uint32_t virtual_size = xe::round_up(phdr[i].p_vaddr + phdr[i].p_memsz,\n uint32_t(phdr[i].p_align)) -\n virtual_addr;\n if (!memory()\n ->LookupHeap(virtual_addr)\n ->AllocFixed(\n virtual_addr, virtual_size, phdr[i].p_align,\n xe::kMemoryAllocationReserve | xe::kMemoryAllocationCommit,\n xe::kMemoryProtectRead | xe::kMemoryProtectWrite)) {\n XELOGE(\"ELF: Could not allocate memory!\");\n }\n\n auto p = memory()->TranslateVirtual(phdr[i].p_vaddr);\n std::memset(p, 0, phdr[i].p_memsz);\n std::memcpy(p, pelf + phdr[i].p_offset, phdr[i].p_filesz);\n\n // Notify backend about executable code.\n if (phdr[i].p_flags & 0x1 /* PF_X */) {\n processor_->backend()->CommitExecutableRange(\n virtual_addr, virtual_addr + virtual_size);\n }\n }\n }\n\n return true;\n}\n\nstd::unique_ptr ElfModule::CreateFunction(uint32_t address) {\n return std::unique_ptr(\n processor_->backend()->CreateGuestFunction(this, address));\n}\n\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/elf_module.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_ELF_MODULE_H_\n#define XENIA_CPU_ELF_MODULE_H_\n\n#include \n#include \n\n#include \"xenia/cpu/module.h\"\n\nnamespace xe {\nnamespace kernel {\nclass KernelState;\n} // namespace kernel\n\nnamespace cpu {\n\n// ELF module: Used to load libxenon executables.\nclass ElfModule : public xe::cpu::Module {\n public:\n ElfModule(Processor* processor, kernel::KernelState* kernel_state);\n virtual ~ElfModule();\n\n bool loaded() const { return loaded_; }\n uint32_t entry_point() const { return entry_point_; }\n const std::string& name() const override { return name_; }\n bool is_executable() const override;\n const std::string& path() const { return path_; }\n\n bool Load(const std::string_view name, const std::string_view path,\n const void* elf_addr, size_t elf_length);\n bool Unload();\n\n protected:\n std::unique_ptr CreateFunction(uint32_t address) override;\n\n private:\n std::string name_;\n std::string path_;\n kernel::KernelState* kernel_state_;\n\n bool loaded_;\n std::vector elf_header_mem_; // Holds the ELF header\n uint32_t entry_point_;\n};\n\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_ELF_MODULE_H_\n" }, { "path": "src/xenia/cpu/entry_table.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/entry_table.h\"\n\n#include \"xenia/base/profiling.h\"\n#include \"xenia/base/threading.h\"\n\nnamespace xe {\nnamespace cpu {\n\nEntryTable::EntryTable() = default;\n\nEntryTable::~EntryTable() {\n auto global_lock = global_critical_region_.Acquire();\n for (auto it : map_) {\n Entry* entry = it.second;\n delete entry;\n }\n}\n\nEntry* EntryTable::Get(uint32_t address) {\n auto global_lock = global_critical_region_.Acquire();\n const auto& it = map_.find(address);\n Entry* entry = it != map_.end() ? it->second : nullptr;\n if (entry) {\n // TODO(benvanik): wait if needed?\n if (entry->status != Entry::STATUS_READY) {\n entry = nullptr;\n }\n }\n return entry;\n}\n\nEntry::Status EntryTable::GetOrCreate(uint32_t address, Entry** out_entry) {\n // TODO(benvanik): replace with a map with wait-free for find.\n // https://github.com/facebook/folly/blob/master/folly/AtomicHashMap.h\n\n auto global_lock = global_critical_region_.Acquire();\n const auto& it = map_.find(address);\n Entry* entry = it != map_.end() ? it->second : nullptr;\n Entry::Status status;\n if (entry) {\n // If we aren't ready yet spin and wait.\n if (entry->status == Entry::STATUS_COMPILING) {\n // Still compiling, so spin.\n do {\n global_lock.unlock();\n // TODO(benvanik): sleep for less time?\n xe::threading::Sleep(std::chrono::microseconds(10));\n global_lock.lock();\n } while (entry->status == Entry::STATUS_COMPILING);\n }\n status = entry->status;\n } else {\n // Create and return for initialization.\n entry = new Entry();\n entry->address = address;\n entry->end_address = 0;\n entry->status = Entry::STATUS_COMPILING;\n entry->function = 0;\n map_[address] = entry;\n status = Entry::STATUS_NEW;\n }\n global_lock.unlock();\n *out_entry = entry;\n return status;\n}\n\nstd::vector EntryTable::FindWithAddress(uint32_t address) {\n auto global_lock = global_critical_region_.Acquire();\n std::vector fns;\n for (auto& it : map_) {\n Entry* entry = it.second;\n if (address >= entry->address && address <= entry->end_address) {\n if (entry->status == Entry::STATUS_READY) {\n fns.push_back(entry->function);\n }\n }\n }\n return fns;\n}\n\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/entry_table.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_ENTRY_TABLE_H_\n#define XENIA_CPU_ENTRY_TABLE_H_\n\n#include \n#include \n\n#include \"xenia/base/mutex.h\"\n\nnamespace xe {\nnamespace cpu {\n\nclass Function;\n\ntypedef struct Entry_t {\n typedef enum {\n STATUS_NEW = 0,\n STATUS_COMPILING,\n STATUS_READY,\n STATUS_FAILED,\n } Status;\n\n uint32_t address;\n uint32_t end_address;\n Status status;\n Function* function;\n} Entry;\n\nclass EntryTable {\n public:\n EntryTable();\n ~EntryTable();\n\n Entry* Get(uint32_t address);\n Entry::Status GetOrCreate(uint32_t address, Entry** out_entry);\n\n std::vector FindWithAddress(uint32_t address);\n\n private:\n xe::global_critical_region global_critical_region_;\n // TODO(benvanik): replace with a better data structure.\n std::unordered_map map_;\n};\n\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_ENTRY_TABLE_H_\n" }, { "path": "src/xenia/cpu/export_resolver.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/export_resolver.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/math.h\"\n#include \"xenia/base/string.h\"\n\nnamespace xe {\nnamespace cpu {\n\nExportResolver::Table::Table(const std::string_view module_name,\n const std::vector* exports_by_ordinal)\n : exports_by_ordinal_(exports_by_ordinal) {\n module_name_ = utf8::find_base_name_from_guest_path(module_name);\n\n exports_by_name_.reserve(exports_by_ordinal_->size());\n for (size_t i = 0; i < exports_by_ordinal_->size(); ++i) {\n auto export_entry = exports_by_ordinal_->at(i);\n if (export_entry) {\n exports_by_name_.push_back(export_entry);\n }\n }\n std::sort(\n exports_by_name_.begin(), exports_by_name_.end(),\n [](Export* a, Export* b) { return std::strcmp(a->name, b->name) < 0; });\n}\n\nExportResolver::ExportResolver() = default;\n\nExportResolver::~ExportResolver() = default;\n\nvoid ExportResolver::RegisterTable(\n const std::string_view module_name,\n const std::vector* exports) {\n tables_.emplace_back(module_name, exports);\n\n all_exports_by_name_.reserve(all_exports_by_name_.size() + exports->size());\n for (size_t i = 0; i < exports->size(); ++i) {\n auto export_entry = exports->at(i);\n if (export_entry) {\n all_exports_by_name_.push_back(export_entry);\n }\n }\n std::sort(\n all_exports_by_name_.begin(), all_exports_by_name_.end(),\n [](Export* a, Export* b) { return std::strcmp(a->name, b->name) < 0; });\n}\n\nExport* ExportResolver::GetExportByOrdinal(const std::string_view module_name,\n uint16_t ordinal) {\n for (const auto& table : tables_) {\n if (xe::utf8::starts_with_case(module_name, table.module_name())) {\n if (ordinal >= table.exports_by_ordinal().size()) {\n return nullptr;\n }\n return table.exports_by_ordinal().at(ordinal);\n }\n }\n return nullptr;\n}\n\nvoid ExportResolver::SetVariableMapping(const std::string_view module_name,\n uint16_t ordinal, uint32_t value) {\n auto export_entry = GetExportByOrdinal(module_name, ordinal);\n assert_not_null(export_entry);\n export_entry->tags |= ExportTag::kImplemented;\n export_entry->variable_ptr = value;\n}\n\nvoid ExportResolver::SetFunctionMapping(const std::string_view module_name,\n uint16_t ordinal,\n xe_kernel_export_shim_fn shim) {\n auto export_entry = GetExportByOrdinal(module_name, ordinal);\n assert_not_null(export_entry);\n export_entry->tags |= ExportTag::kImplemented;\n export_entry->function_data.shim = shim;\n}\n\nvoid ExportResolver::SetFunctionMapping(const std::string_view module_name,\n uint16_t ordinal,\n ExportTrampoline trampoline) {\n auto export_entry = GetExportByOrdinal(module_name, ordinal);\n assert_not_null(export_entry);\n export_entry->tags |= ExportTag::kImplemented;\n export_entry->function_data.trampoline = trampoline;\n}\n\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/export_resolver.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2020 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_EXPORT_RESOLVER_H_\n#define XENIA_CPU_EXPORT_RESOLVER_H_\n\n#include \n#include \n\n#include \"xenia/base/math.h\"\n#include \"xenia/cpu/ppc/ppc_context.h\"\n\nnamespace xe {\nnamespace cpu {\n\nenum class ExportCategory : uint8_t {\n kNone = 0,\n kAudio,\n kAvatars,\n kContent,\n kDebug,\n kFileSystem,\n kInput,\n kLocale,\n kMemory,\n kMisc,\n kModules,\n kNetworking,\n kThreading,\n kUI,\n kUserProfiles,\n kVideo,\n};\n\nstruct ExportTag {\n using type = uint32_t;\n\n // packed like so:\n // ll...... cccccccc ........ ..bihssi\n\n static const int CategoryShift = 16;\n\n // Export is implemented in some form and can be used.\n static const type kImplemented = 1u << 0;\n // Export is a stub and is probably bad.\n static const type kStub = 1u << 1;\n // Export is known to cause problems, or may not be complete.\n static const type kSketchy = 1u << 2;\n // Export is called *a lot*.\n static const type kHighFrequency = 1u << 3;\n // Export is important and should always be logged.\n static const type kImportant = 1u << 4;\n // Export blocks the calling thread\n static const type kBlocking = 1u << 5;\n\n // Export will be logged on each call.\n static const type kLog = 1u << 30;\n // Export's result will be logged on each call.\n static const type kLogResult = 1u << 31;\n};\n\n// DEPRECATED\ntypedef void (*xe_kernel_export_shim_fn)(void*, void*);\n\ntypedef void (*ExportTrampoline)(ppc::PPCContext* ppc_context);\n\nclass Export {\n public:\n enum class Type {\n kFunction = 0,\n kVariable = 1,\n };\n\n Export(uint16_t ordinal, Type type, const char* name,\n ExportTag::type tags = 0)\n : ordinal(ordinal),\n type(type),\n tags(tags),\n function_data({nullptr, nullptr, 0}) {\n std::strncpy(this->name, name, xe::countof(this->name));\n }\n\n uint16_t ordinal;\n Type type;\n char name[96];\n ExportTag::type tags;\n\n bool is_implemented() const {\n return (tags & ExportTag::kImplemented) == ExportTag::kImplemented;\n }\n\n union {\n // Variable data. Only valid when kXEKernelExportFlagVariable is set.\n // This is an address in the client memory space that the variable can\n // be found at.\n uint32_t variable_ptr;\n\n struct {\n // DEPRECATED\n xe_kernel_export_shim_fn shim;\n\n // Trampoline that is called from the guest-to-host thunk.\n // Expects only PPC context as first arg.\n ExportTrampoline trampoline;\n uint64_t call_count;\n } function_data;\n };\n};\n\nclass ExportResolver {\n public:\n class Table {\n public:\n Table(const std::string_view module_name,\n const std::vector* exports);\n\n const std::string& module_name() const { return module_name_; }\n const std::vector& exports_by_ordinal() const {\n return *exports_by_ordinal_;\n }\n const std::vector& exports_by_name() const {\n return exports_by_name_;\n }\n\n private:\n std::string module_name_;\n const std::vector* exports_by_ordinal_ = nullptr;\n std::vector exports_by_name_;\n };\n\n ExportResolver();\n ~ExportResolver();\n\n void RegisterTable(const std::string_view module_name,\n const std::vector* exports);\n const std::vector& tables() const { return tables_; }\n const std::vector& all_exports_by_name() const {\n return all_exports_by_name_;\n }\n\n Export* GetExportByOrdinal(const std::string_view module_name,\n uint16_t ordinal);\n\n void SetVariableMapping(const std::string_view module_name, uint16_t ordinal,\n uint32_t value);\n void SetFunctionMapping(const std::string_view module_name, uint16_t ordinal,\n xe_kernel_export_shim_fn shim);\n void SetFunctionMapping(const std::string_view module_name, uint16_t ordinal,\n ExportTrampoline trampoline);\n\n private:\n std::vector
tables_;\n std::vector all_exports_by_name_;\n};\n\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_EXPORT_RESOLVER_H_\n" }, { "path": "src/xenia/cpu/function.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/function.h\"\n\n#include \"xenia/base/logging.h\"\n#include \"xenia/cpu/symbol.h\"\n#include \"xenia/cpu/thread_state.h\"\n\nnamespace xe {\nnamespace cpu {\n\nFunction::Function(Module* module, uint32_t address)\n : Symbol(Symbol::Type::kFunction, module, address) {}\n\nFunction::~Function() = default;\n\nBuiltinFunction::BuiltinFunction(Module* module, uint32_t address)\n : Function(module, address) {\n behavior_ = Behavior::kBuiltin;\n}\n\nBuiltinFunction::~BuiltinFunction() = default;\n\nvoid BuiltinFunction::SetupBuiltin(Handler handler, void* arg0, void* arg1) {\n behavior_ = Behavior::kBuiltin;\n handler_ = handler;\n arg0_ = arg0;\n arg1_ = arg1;\n}\n\nbool BuiltinFunction::Call(ThreadState* thread_state, uint32_t return_address) {\n // SCOPE_profile_cpu_f(\"cpu\");\n\n ThreadState* original_thread_state = ThreadState::Get();\n if (original_thread_state != thread_state) {\n ThreadState::Bind(thread_state);\n }\n\n assert_not_null(handler_);\n handler_(thread_state->context(), arg0_, arg1_);\n\n if (original_thread_state != thread_state) {\n ThreadState::Bind(original_thread_state);\n }\n\n return true;\n}\n\nGuestFunction::GuestFunction(Module* module, uint32_t address)\n : Function(module, address) {\n behavior_ = Behavior::kDefault;\n}\n\nGuestFunction::~GuestFunction() = default;\n\nvoid GuestFunction::SetupExtern(ExternHandler handler, Export* export_data) {\n behavior_ = Behavior::kExtern;\n extern_handler_ = handler;\n export_data_ = export_data;\n}\n\nconst SourceMapEntry* GuestFunction::LookupGuestAddress(\n uint32_t guest_address) const {\n // TODO(benvanik): binary search? We know the list is sorted by code order.\n for (size_t i = 0; i < source_map_.size(); ++i) {\n const auto& entry = source_map_[i];\n if (entry.guest_address == guest_address) {\n return &entry;\n }\n }\n return nullptr;\n}\n\nconst SourceMapEntry* GuestFunction::LookupHIROffset(uint32_t offset) const {\n // TODO(benvanik): binary search? We know the list is sorted by code order.\n for (size_t i = 0; i < source_map_.size(); ++i) {\n const auto& entry = source_map_[i];\n if (entry.hir_offset >= offset) {\n return &entry;\n }\n }\n return nullptr;\n}\n\nconst SourceMapEntry* GuestFunction::LookupMachineCodeOffset(\n uint32_t offset) const {\n // TODO(benvanik): binary search? We know the list is sorted by code order.\n for (int64_t i = source_map_.size() - 1; i >= 0; --i) {\n const auto& entry = source_map_[i];\n if (entry.code_offset <= offset) {\n return &entry;\n }\n }\n return source_map_.empty() ? nullptr : &source_map_[0];\n}\n\nuint32_t GuestFunction::MapGuestAddressToMachineCodeOffset(\n uint32_t guest_address) const {\n auto entry = LookupGuestAddress(guest_address);\n return entry ? entry->code_offset : 0;\n}\n\nuintptr_t GuestFunction::MapGuestAddressToMachineCode(\n uint32_t guest_address) const {\n auto entry = LookupGuestAddress(guest_address);\n return reinterpret_cast(machine_code()) +\n (entry ? entry->code_offset : 0);\n}\n\nuint32_t GuestFunction::MapMachineCodeToGuestAddress(\n uintptr_t host_address) const {\n auto entry = LookupMachineCodeOffset(static_cast(\n host_address - reinterpret_cast(machine_code())));\n return entry ? entry->guest_address : address();\n}\n\nbool GuestFunction::Call(ThreadState* thread_state, uint32_t return_address) {\n // SCOPE_profile_cpu_f(\"cpu\");\n\n ThreadState* original_thread_state = ThreadState::Get();\n if (original_thread_state != thread_state) {\n ThreadState::Bind(thread_state);\n }\n\n bool result = CallImpl(thread_state, return_address);\n\n if (original_thread_state != thread_state) {\n ThreadState::Bind(original_thread_state);\n }\n\n return result;\n}\n\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/function.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_FUNCTION_H_\n#define XENIA_CPU_FUNCTION_H_\n\n#include \n#include \n\n#include \"xenia/cpu/function_debug_info.h\"\n#include \"xenia/cpu/function_trace_data.h\"\n#include \"xenia/cpu/ppc/ppc_context.h\"\n#include \"xenia/cpu/symbol.h\"\n#include \"xenia/cpu/thread_state.h\"\n\nnamespace xe {\nnamespace cpu {\nclass Export;\n} // namespace cpu\n\nnamespace cpu {\n\nstruct SourceMapEntry {\n uint32_t guest_address; // PPC guest address (0x82....).\n uint32_t hir_offset; // Block ordinal (16b) | Instr ordinal (16b)\n uint32_t code_offset; // Offset from emitted code start.\n};\n\nclass Function : public Symbol {\n public:\n enum class Behavior {\n kDefault = 0,\n kProlog,\n kEpilog,\n kEpilogReturn,\n kBuiltin,\n kExtern,\n };\n\n ~Function() override;\n\n uint32_t address() const { return address_; }\n bool has_end_address() const { return end_address_ > 0; }\n uint32_t end_address() const { return end_address_; }\n void set_end_address(uint32_t value) { end_address_ = value; }\n Behavior behavior() const { return behavior_; }\n void set_behavior(Behavior value) { behavior_ = value; }\n bool is_guest() const { return behavior_ != Behavior::kBuiltin; }\n\n bool ContainsAddress(uint32_t address) const {\n if (!address_ || !end_address_) {\n return false;\n }\n\n if (address >= address_ && address < end_address_) {\n return true;\n }\n\n return false;\n }\n\n virtual bool Call(ThreadState* thread_state, uint32_t return_address) = 0;\n\n protected:\n Function(Module* module, uint32_t address);\n\n uint32_t end_address_ = 0;\n Behavior behavior_ = Behavior::kDefault;\n};\n\nclass BuiltinFunction : public Function {\n public:\n typedef void (*Handler)(ppc::PPCContext* ppc_context, void* arg0, void* arg1);\n\n BuiltinFunction(Module* module, uint32_t address);\n ~BuiltinFunction() override;\n\n void SetupBuiltin(Handler handler, void* arg0, void* arg1);\n\n Handler handler() const { return handler_; }\n void* arg0() const { return arg0_; }\n void* arg1() const { return arg1_; }\n\n bool Call(ThreadState* thread_state, uint32_t return_address) override;\n\n protected:\n Handler handler_ = nullptr;\n void* arg0_ = nullptr;\n void* arg1_ = nullptr;\n};\n\nclass GuestFunction : public Function {\n public:\n typedef void (*ExternHandler)(ppc::PPCContext* ppc_context,\n kernel::KernelState* kernel_state);\n\n GuestFunction(Module* module, uint32_t address);\n ~GuestFunction() override;\n\n uint32_t address() const { return address_; }\n bool has_end_address() const { return end_address_ > 0; }\n uint32_t end_address() const { return end_address_; }\n void set_end_address(uint32_t value) { end_address_ = value; }\n\n virtual uint8_t* machine_code() const = 0;\n virtual size_t machine_code_length() const = 0;\n\n FunctionDebugInfo* debug_info() const { return debug_info_.get(); }\n void set_debug_info(std::unique_ptr debug_info) {\n debug_info_ = std::move(debug_info);\n }\n FunctionTraceData& trace_data() { return trace_data_; }\n std::vector& source_map() { return source_map_; }\n\n ExternHandler extern_handler() const { return extern_handler_; }\n Export* export_data() const { return export_data_; }\n void SetupExtern(ExternHandler handler, Export* export_data = nullptr);\n\n const SourceMapEntry* LookupGuestAddress(uint32_t guest_address) const;\n const SourceMapEntry* LookupHIROffset(uint32_t offset) const;\n const SourceMapEntry* LookupMachineCodeOffset(uint32_t offset) const;\n\n uint32_t MapGuestAddressToMachineCodeOffset(uint32_t guest_address) const;\n uintptr_t MapGuestAddressToMachineCode(uint32_t guest_address) const;\n uint32_t MapMachineCodeToGuestAddress(uintptr_t host_address) const;\n\n bool Call(ThreadState* thread_state, uint32_t return_address) override;\n\n protected:\n virtual bool CallImpl(ThreadState* thread_state, uint32_t return_address) = 0;\n\n protected:\n std::unique_ptr debug_info_;\n FunctionTraceData trace_data_;\n std::vector source_map_;\n ExternHandler extern_handler_ = nullptr;\n Export* export_data_ = nullptr;\n};\n\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_FUNCTION_H_\n" }, { "path": "src/xenia/cpu/function_debug_info.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/function_debug_info.h\"\n\n#include \n#include \n\n#include \"xenia/base/logging.h\"\n\nnamespace xe {\nnamespace cpu {\n\nFunctionDebugInfo::FunctionDebugInfo()\n : source_disasm_(nullptr),\n raw_hir_disasm_(nullptr),\n hir_disasm_(nullptr),\n machine_code_disasm_(nullptr) {}\n\nFunctionDebugInfo::~FunctionDebugInfo() {\n free(source_disasm_);\n free(raw_hir_disasm_);\n free(hir_disasm_);\n free(machine_code_disasm_);\n}\n\nvoid FunctionDebugInfo::Dump() {\n if (source_disasm_) {\n XELOGD(\"PPC:\\n{}\\n\", source_disasm_);\n }\n if (raw_hir_disasm_) {\n XELOGD(\"Unoptimized HIR:\\n{}\\n\", raw_hir_disasm_);\n }\n if (hir_disasm_) {\n XELOGD(\"Optimized HIR:\\n{}\\n\", hir_disasm_);\n }\n if (machine_code_disasm_) {\n XELOGD(\"Machine Code:\\n{}\\n\", machine_code_disasm_);\n }\n}\n\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/function_debug_info.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_FUNCTION_DEBUG_INFO_H_\n#define XENIA_CPU_FUNCTION_DEBUG_INFO_H_\n\n#include \n#include \n\nnamespace xe {\nnamespace cpu {\n\nenum DebugInfoFlags : uint32_t {\n kDebugInfoNone = 0,\n kDebugInfoDisasmSource = (1 << 1),\n kDebugInfoDisasmRawHir = (1 << 2),\n kDebugInfoDisasmHir = (1 << 3),\n kDebugInfoDisasmMachineCode = (1 << 4),\n kDebugInfoAllDisasm = kDebugInfoDisasmSource | kDebugInfoDisasmRawHir |\n kDebugInfoDisasmHir | kDebugInfoDisasmMachineCode,\n kDebugInfoTraceFunctions = (1 << 6),\n kDebugInfoTraceFunctionCoverage = (1 << 7) | kDebugInfoTraceFunctions,\n kDebugInfoTraceFunctionReferences = (1 << 8) | kDebugInfoTraceFunctions,\n kDebugInfoTraceFunctionData = (1 << 9) | kDebugInfoTraceFunctions,\n\n kDebugInfoAllTracing =\n kDebugInfoTraceFunctions | kDebugInfoTraceFunctionCoverage |\n kDebugInfoTraceFunctionReferences | kDebugInfoTraceFunctionData,\n kDebugInfoAll = 0xFFFFFFFF,\n};\n\n// DEPRECATED\n// This will be getting removed or refactored to contain only on-demand\n// disassembly data.\nclass FunctionDebugInfo {\n public:\n FunctionDebugInfo();\n ~FunctionDebugInfo();\n\n uint32_t address_reference_count() const { return address_reference_count_; }\n void set_address_reference_count(uint32_t value) {\n address_reference_count_ = value;\n }\n uint32_t instruction_result_count() const {\n return instruction_result_count_;\n }\n void set_instruction_result_count(uint32_t value) {\n instruction_result_count_ = value;\n }\n\n const char* source_disasm() const { return source_disasm_; }\n void set_source_disasm(char* value) { source_disasm_ = value; }\n const char* raw_hir_disasm() const { return raw_hir_disasm_; }\n void set_raw_hir_disasm(char* value) { raw_hir_disasm_ = value; }\n const char* hir_disasm() const { return hir_disasm_; }\n void set_hir_disasm(char* value) { hir_disasm_ = value; }\n const char* machine_code_disasm() const { return machine_code_disasm_; }\n void set_machine_code_disasm(char* value) { machine_code_disasm_ = value; }\n\n void Dump();\n\n private:\n uint32_t address_reference_count_;\n uint32_t instruction_result_count_;\n\n char* source_disasm_;\n char* raw_hir_disasm_;\n char* hir_disasm_;\n char* machine_code_disasm_;\n};\n\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_FUNCTION_DEBUG_INFO_H_\n" }, { "path": "src/xenia/cpu/function_trace_data.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2015 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_FUNCTION_TRACE_DATA_H_\n#define XENIA_CPU_FUNCTION_TRACE_DATA_H_\n\n#include \n#include \n\n#include \"xenia/base/memory.h\"\n\nnamespace xe {\nnamespace cpu {\n\nclass FunctionTraceData {\n public:\n static const int kFunctionCallerHistoryCount = 4;\n\n struct Header {\n // Format is used by tooling, changes must be made across all targets.\n // + 0 4b (data size)\n // + 4 4b start_address\n // + 8 4b end_address\n // +12 4b type (user, external, etc)\n // +16 8b function_thread_use // bitmask of thread id\n // +24 8b function_call_count\n // +32 4b+ function_caller_history[4]\n // +48 8b+ instruction_execute_count[instruction count]\n uint32_t data_size;\n uint32_t start_address;\n uint32_t end_address;\n uint32_t type;\n uint64_t function_thread_use;\n uint64_t function_call_count;\n uint32_t function_caller_history[kFunctionCallerHistoryCount];\n // uint64_t instruction_execute_count[];\n };\n\n FunctionTraceData() : header_(nullptr) {}\n\n void Reset(uint8_t* trace_data, size_t trace_data_size,\n uint32_t start_address, uint32_t end_address) {\n header_ = reinterpret_cast(trace_data);\n header_->data_size = uint32_t(trace_data_size);\n header_->start_address = start_address;\n header_->end_address = end_address;\n header_->type = 0;\n header_->function_thread_use = 0;\n header_->function_call_count = 0;\n for (int i = 0; i < kFunctionCallerHistoryCount; ++i) {\n header_->function_caller_history[i] = 0;\n }\n // Clear any remaining.\n std::memset(trace_data + sizeof(Header), 0,\n trace_data_size - sizeof(Header));\n }\n\n bool is_valid() const { return header_ != nullptr; }\n\n uint32_t start_address() const { return header_->start_address; }\n uint32_t end_address() const { return header_->end_address; }\n uint32_t instruction_count() const {\n return (header_->end_address - header_->start_address) / 4 + 1;\n }\n\n Header* header() const { return header_; }\n\n uint8_t* instruction_execute_counts() const {\n return reinterpret_cast(header_) + sizeof(Header);\n }\n\n static size_t SizeOfHeader() { return sizeof(Header); }\n\n static size_t SizeOfInstructionCounts(uint32_t start_address,\n uint32_t end_address) {\n uint32_t instruction_count = (end_address - start_address) / 4 + 1;\n return instruction_count * 8;\n }\n\n private:\n Header* header_;\n};\n\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_FUNCTION_TRACE_DATA_H_\n" }, { "path": "src/xenia/cpu/hir/block.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2014 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/hir/block.h\"\n\n#include \"xenia/base/assert.h\"\n#include \"xenia/cpu/hir/instr.h\"\n\nnamespace xe {\nnamespace cpu {\nnamespace hir {\n\nvoid Block::AssertNoCycles() {\n Instr* hare = instr_head;\n Instr* tortoise = instr_head;\n if (!hare) {\n return;\n }\n while ((hare = hare->next)) {\n if (hare == tortoise) {\n // Cycle!\n assert_always();\n }\n hare = hare->next;\n if (hare == tortoise) {\n // Cycle!\n assert_always();\n }\n tortoise = tortoise->next;\n if (!hare || !tortoise) {\n return;\n }\n }\n}\n\n} // namespace hir\n} // namespace cpu\n} // namespace xe\n" }, { "path": "src/xenia/cpu/hir/block.h", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2013 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#ifndef XENIA_CPU_HIR_BLOCK_H_\n#define XENIA_CPU_HIR_BLOCK_H_\n\n#include \"xenia/base/arena.h\"\n\nnamespace llvm {\nclass BitVector;\n} // namespace llvm\n\nnamespace xe {\nnamespace cpu {\nnamespace hir {\n\nclass Block;\nclass HIRBuilder;\nclass Instr;\nclass Label;\n\nclass Edge {\n public:\n enum EdgeFlags {\n UNCONDITIONAL = (1 << 0),\n DOMINATES = (1 << 1),\n };\n\n public:\n Edge* outgoing_next;\n Edge* outgoing_prev;\n Edge* incoming_next;\n Edge* incoming_prev;\n\n Block* src;\n Block* dest;\n\n uint32_t flags;\n};\n\nclass Block {\n public:\n Arena* arena;\n\n Block* next;\n Block* prev;\n\n Edge* incoming_edge_head;\n Edge* outgoing_edge_head;\n llvm::BitVector* incoming_values;\n\n Label* label_head;\n Label* label_tail;\n\n Instr* instr_head;\n Instr* instr_tail;\n\n uint16_t ordinal;\n\n void AssertNoCycles();\n};\n\n} // namespace hir\n} // namespace cpu\n} // namespace xe\n\n#endif // XENIA_CPU_HIR_BLOCK_H_\n" }, { "path": "src/xenia/cpu/hir/hir_builder.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \"xenia/cpu/hir/hir_builder.h\"\n\n#include \n#include \n#include \n\n#include \"xenia/base/assert.h\"\n#include \"xenia/base/profiling.h\"\n#include \"xenia/cpu/function.h\"\n#include \"xenia/cpu/hir/block.h\"\n#include \"xenia/cpu/hir/instr.h\"\n#include \"xenia/cpu/hir/label.h\"\n#include \"xenia/cpu/symbol.h\"\n\n// Will scribble arena memory to hopefully find use before clears.\n// #define SCRIBBLE_ARENA_ON_RESET\n\nnamespace xe {\nnamespace cpu {\nnamespace hir {\n\n#define ASSERT_ADDRESS_TYPE(value) \\\n assert_true((value->type) == INT32_TYPE || (value->type) == INT64_TYPE)\n#define ASSERT_CALL_ADDRESS_TYPE(value) \\\n assert_true((value->type) == INT32_TYPE || (value->type) == INT64_TYPE)\n#define ASSERT_INTEGER_TYPE(value) \\\n assert_true((value->type) == INT8_TYPE || (value->type) == INT16_TYPE || \\\n (value->type) == INT32_TYPE || (value->type) == INT64_TYPE)\n#define ASSERT_FLOAT_TYPE(value) \\\n assert_true((value->type) == FLOAT32_TYPE || (value->type) == FLOAT64_TYPE)\n#define ASSERT_NON_FLOAT_TYPE(value) \\\n assert_true((value->type) != FLOAT32_TYPE && (value->type) != FLOAT64_TYPE)\n#define ASSERT_NON_VECTOR_TYPE(value) assert_false((value->type) == VEC128_TYPE)\n#define ASSERT_VECTOR_TYPE(value) assert_true((value->type) == VEC128_TYPE)\n#define ASSERT_FLOAT_OR_VECTOR_TYPE(value) \\\n assert_true((value->type) == FLOAT32_TYPE || \\\n (value->type) == FLOAT64_TYPE || (value->type) == VEC128_TYPE)\n#define ASSERT_TYPES_EQUAL(value1, value2) \\\n assert_true((value1->type) == (value2->type))\n\nHIRBuilder::HIRBuilder() {\n arena_ = new Arena();\n Reset();\n}\n\nHIRBuilder::~HIRBuilder() {\n Reset();\n delete arena_;\n}\n\nvoid HIRBuilder::Reset() {\n attributes_ = 0;\n next_label_id_ = 0;\n next_value_ordinal_ = 0;\n locals_.clear();\n block_head_ = block_tail_ = NULL;\n current_block_ = NULL;\n#if SCRIBBLE_ARENA_ON_RESET\n arena_->DebugFill();\n#endif\n arena_->Reset();\n}\n\nbool HIRBuilder::Finalize() {\n // Scan blocks in order and add fallthrough branches. These are needed for\n // analysis passes to work. We may have also added blocks out of order and\n // need to ensure they fall through in the right order.\n for (auto block = block_head_; block != NULL; block = block->next) {\n bool needs_branch = false;\n if (block->instr_tail) {\n if (!IsUnconditionalJump(block->instr_tail)) {\n // Add tail branch to block that falls through.\n needs_branch = true;\n }\n } else {\n // Add tail branch to block with no instructions.\n // Hopefully an optimization pass will clean this up later.\n needs_branch = true;\n }\n if (needs_branch) {\n current_block_ = block;\n if (!block->next) {\n // No following block.\n // Sometimes VC++ generates functions with bl at the end even if they\n // will never return. Just add a return to satisfy things.\n // XELOGW(\"Fall-through out of the function.\");\n Trap();\n Return();\n current_block_ = NULL;\n break;\n }\n // Add branch.\n Branch(block->next, BRANCH_LIKELY);\n current_block_ = NULL;\n }\n }\n return true;\n}\n\nvoid HIRBuilder::DumpValue(StringBuffer* str, Value* value) {\n if (value->IsConstant()) {\n switch (value->type) {\n case INT8_TYPE:\n str->AppendFormat(\"{:X}\", value->constant.i8);\n break;\n case INT16_TYPE:\n str->AppendFormat(\"{:X}\", value->constant.i16);\n break;\n case INT32_TYPE:\n str->AppendFormat(\"{:X}\", value->constant.i32);\n break;\n case INT64_TYPE:\n str->AppendFormat(\"{:X}\", value->constant.i64);\n break;\n case FLOAT32_TYPE:\n str->AppendFormat(\"{:F}\", value->constant.f32);\n break;\n case FLOAT64_TYPE:\n str->AppendFormat(\"{:F}\", value->constant.f64);\n break;\n case VEC128_TYPE:\n str->AppendFormat(\"({:F},{:F},{:F},{:F})\", value->constant.v128.x,\n value->constant.v128.y, value->constant.v128.z,\n value->constant.v128.w);\n break;\n default:\n assert_always();\n break;\n }\n } else {\n static const char* type_names[] = {\n \"i8\", \"i16\", \"i32\", \"i64\", \"f32\", \"f64\", \"v128\",\n };\n str->AppendFormat(\"v{}.{}\", value->ordinal, type_names[value->type]);\n }\n if (value->reg.index != -1) {\n str->AppendFormat(\"<{}{}>\", value->reg.set->name, value->reg.index);\n }\n}\n\nvoid HIRBuilder::DumpOp(StringBuffer* str, OpcodeSignatureType sig_type,\n Instr::Op* op) {\n switch (sig_type) {\n case OPCODE_SIG_TYPE_X:\n break;\n case OPCODE_SIG_TYPE_L:\n if (op->label->name) {\n str->Append(op->label->name);\n } else {\n str->AppendFormat(\"label{}\", op->label->id);\n }\n break;\n case OPCODE_SIG_TYPE_O:\n str->AppendFormat(\"+{}\", op->offset);\n break;\n case OPCODE_SIG_TYPE_S:\n if (true) {\n auto target = op->symbol;\n str->Append(!target->name().empty() ? target->name() : \"\");\n }\n break;\n case OPCODE_SIG_TYPE_V:\n DumpValue(str, op->value);\n break;\n }\n}\n\nvoid HIRBuilder::Dump(StringBuffer* str) {\n if (attributes_) {\n str->AppendFormat(\"; attributes = {:08X}\\n\", attributes_);\n }\n\n for (auto it = locals_.begin(); it != locals_.end(); ++it) {\n auto local = *it;\n str->Append(\" ; local \");\n DumpValue(str, local);\n str->Append('\\n');\n }\n\n uint32_t block_ordinal = 0;\n Block* block = block_head_;\n while (block) {\n if (block == block_head_) {\n str->Append(\":\\n\");\n } else if (!block->label_head) {\n str->AppendFormat(\":\\n\", block_ordinal);\n }\n block_ordinal++;\n\n Label* label = block->label_head;\n while (label) {\n if (label->name) {\n str->AppendFormat(\"{}:\\n\", label->name);\n } else {\n str->AppendFormat(\"label{}:\\n\", label->id);\n }\n label = label->next;\n }\n\n Edge* incoming_edge = block->incoming_edge_head;\n while (incoming_edge) {\n auto src_label = incoming_edge->src->label_head;\n if (src_label && src_label->name) {\n str->AppendFormat(\" ; in: {}\", src_label->name);\n } else if (src_label) {\n str->AppendFormat(\" ; in: label{}\", src_label->id);\n } else {\n str->AppendFormat(\" ; in: \", incoming_edge->src->ordinal);\n }\n str->AppendFormat(\", dom:{}, uncond:{}\\n\",\n (incoming_edge->flags & Edge::DOMINATES) ? 1 : 0,\n (incoming_edge->flags & Edge::UNCONDITIONAL) ? 1 : 0);\n incoming_edge = incoming_edge->incoming_next;\n }\n Edge* outgoing_edge = block->outgoing_edge_head;\n while (outgoing_edge) {\n auto dest_label = outgoing_edge->dest->label_head;\n if (dest_label && dest_label->name) {\n str->AppendFormat(\" ; out: {}\", dest_label->name);\n } else if (dest_label) {\n str->AppendFormat(\" ; out: label{}\", dest_label->id);\n } else {\n str->AppendFormat(\" ; out: \", outgoing_edge->dest->ordinal);\n }\n str->AppendFormat(\", dom:{}, uncond:{}\\n\",\n (outgoing_edge->flags & Edge::DOMINATES) ? 1 : 0,\n (outgoing_edge->flags & Edge::UNCONDITIONAL) ? 1 : 0);\n outgoing_edge = outgoing_edge->outgoing_next;\n }\n\n Instr* i = block->instr_head;\n while (i) {\n if (i->opcode->flags & OPCODE_FLAG_HIDE) {\n i = i->next;\n continue;\n }\n if (i->opcode == &OPCODE_COMMENT_info) {\n str->AppendFormat(\" ; {}\\n\", reinterpret_cast(i->src1.offset));\n i = i->next;\n continue;\n }\n\n const OpcodeInfo* info = i->opcode;\n OpcodeSignatureType dest_type = GET_OPCODE_SIG_TYPE_DEST(info->signature);\n OpcodeSignatureType src1_type = GET_OPCODE_SIG_TYPE_SRC1(info->signature);\n OpcodeSignatureType src2_type = GET_OPCODE_SIG_TYPE_SRC2(info->signature);\n OpcodeSignatureType src3_type = GET_OPCODE_SIG_TYPE_SRC3(info->signature);\n str->Append(\" \");\n if (dest_type) {\n DumpValue(str, i->dest);\n str->Append(\" = \");\n }\n if (i->flags) {\n str->AppendFormat(\"{}.{}\", info->name, i->flags);\n } else {\n str->Append(info->name);\n }\n if (src1_type) {\n str->Append(' ');\n DumpOp(str, src1_type, &i->src1);\n }\n if (src2_type) {\n str->Append(\", \");\n DumpOp(str, src2_type, &i->src2);\n }\n if (src3_type) {\n str->Append(\", \");\n DumpOp(str, src3_type, &i->src3);\n }\n str->Append('\\n');\n i = i->next;\n }\n\n block = block->next;\n }\n}\n\nvoid HIRBuilder::AssertNoCycles() {\n Block* hare = block_head_;\n Block* tortoise = block_head_;\n if (!hare) {\n return;\n }\n while ((hare = hare->next)) {\n if (hare == tortoise) {\n // Cycle!\n assert_always();\n }\n hare = hare->next;\n if (hare == tortoise) {\n // Cycle!\n assert_always();\n }\n tortoise = tortoise->next;\n if (!hare || !tortoise) {\n return;\n }\n }\n}\n\nBlock* HIRBuilder::current_block() const { return current_block_; }\n\nInstr* HIRBuilder::last_instr() const {\n if (current_block_ && current_block_->instr_tail) {\n return current_block_->instr_tail;\n } else if (block_tail_) {\n return block_tail_->instr_tail;\n }\n return NULL;\n}\n\nLabel* HIRBuilder::NewLabel() {\n Label* label = arena_->Alloc
\\n\"\n\"\\n\"\n\"\\n\"\n\"\\n\"\n\"\\n\"\n\"
CloseClose, Never Show
\\n\"\n\"\\n\"\n\"
\\n\"\n\"
\\n\"\n\"\\n\"\n\"
  • Mode\\n\"\n\" \\n\"\n\"
    • \\n\"\n\"
  • Reference\\n\"\n\" \\n\"\n\"
    • \\n\"\n\"
  • Threads\\n\"\n\"
      \\n\"\n\"
    • All
      • \\n\"\n\"
    • ---
      • \\n\"\n\"
    \\n\"\n\"
    • \\n\"\n\"
  • Groups\\n\"\n\"
      \\n\"\n\"
    • All
      • \\n\"\n\"
    • ---
      • \\n\"\n\"
    \\n\"\n\"
    • \\n\"\n\"
  • Options      \\n\"\n\" \\n\"\n\"
    • \\n\"\n\"\\n\"\n\"\\n\"\n\"\\n\"\n\"\\n\"\n\"\\n\"\n\"\";\n\nconst size_t g_MicroProfileHtml_end_3_size = sizeof(g_MicroProfileHtml_end_3);\nconst char* g_MicroProfileHtml_end[] = {\n&g_MicroProfileHtml_end_0[0],\n&g_MicroProfileHtml_end_1[0],\n&g_MicroProfileHtml_end_2[0],\n&g_MicroProfileHtml_end_3[0],\n};\nsize_t g_MicroProfileHtml_end_sizes[] = {\nsizeof(g_MicroProfileHtml_end_0),\nsizeof(g_MicroProfileHtml_end_1),\nsizeof(g_MicroProfileHtml_end_2),\nsizeof(g_MicroProfileHtml_end_3),\n};\nsize_t g_MicroProfileHtml_end_count = 4;\n#endif //MICROPROFILE_EMBED_HTML\n\n///end file generated from microprofile.html\n" }, { "path": "third_party/microprofile/microprofileui.h", "content": "#pragma once\n// This is free and unencumbered software released into the public domain.\n// Anyone is free to copy, modify, publish, use, compile, sell, or\n// distribute this software, either in source code form or as a compiled\n// binary, for any purpose, commercial or non-commercial, and by any\n// means.\n// In jurisdictions that recognize copyright laws, the author or authors\n// of this software dedicate any and all copyright interest in the\n// software to the public domain. We make this dedication for the benefit\n// of the public at large and to the detriment of our heirs and\n// successors. We intend this dedication to be an overt act of\n// relinquishment in perpetuity of all present and future rights to this\n// software under copyright law.\n// THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF\n// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.\n// IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR\n// OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,\n// ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR\n// OTHER DEALINGS IN THE SOFTWARE.\n// For more information, please refer to \n//\n// ***********************************************************************\n//\n//\n//\n\n\n#ifndef MICROPROFILE_ENABLED\n#error \"microprofile.h must be included before including microprofileui.h\"\n#endif\n\n#ifndef MICROPROFILEUI_ENABLED\n#define MICROPROFILEUI_ENABLED MICROPROFILE_ENABLED\n#endif\n\n#ifndef MICROPROFILEUI_API\n#define MICROPROFILEUI_API\n#endif\n\n\n#if 0 == MICROPROFILEUI_ENABLED \n#define MicroProfileMouseButton(foo, bar) do{}while(0)\n#define MicroProfileMousePosition(foo, bar, z) do{}while(0)\n#define MicroProfileModKey(key) do{}while(0)\n#define MicroProfileDraw(foo, bar) do{}while(0)\n#define MicroProfileIsDrawing() 0\n#define MicroProfileToggleDisplayMode() do{}while(0)\n#define MicroProfileSetDisplayMode(f) do{}while(0)\n#else\n\n#ifndef MICROPROFILE_DRAWCURSOR\n#define MICROPROFILE_DRAWCURSOR 0\n#endif\n\n#ifndef MICROPROFILE_DETAILED_BAR_NAMES\n#define MICROPROFILE_DETAILED_BAR_NAMES 1\n#endif\n\n#ifndef MICROPROFILE_TEXT_WIDTH\n#define MICROPROFILE_TEXT_WIDTH 5\n#endif\n\n#ifndef MICROPROFILE_TEXT_HEIGHT\n#define MICROPROFILE_TEXT_HEIGHT 8\n#endif\n\n#ifndef MICROPROFILE_DETAILED_BAR_HEIGHT\n#define MICROPROFILE_DETAILED_BAR_HEIGHT 12\n#endif\n\n#ifndef MICROPROFILE_DETAILED_CONTEXT_SWITCH_HEIGHT\n#define MICROPROFILE_DETAILED_CONTEXT_SWITCH_HEIGHT 7\n#endif\n\n#ifndef MICROPROFILE_GRAPH_WIDTH\n#define MICROPROFILE_GRAPH_WIDTH 256\n#endif\n\n#ifndef MICROPROFILE_GRAPH_HEIGHT\n#define MICROPROFILE_GRAPH_HEIGHT 256\n#endif\n\n#ifndef MICROPROFILE_BORDER_SIZE \n#define MICROPROFILE_BORDER_SIZE 1\n#endif\n\n#ifndef MICROPROFILE_HELP_LEFT\n#define MICROPROFILE_HELP_LEFT \"Left-Click\"\n#endif\n\n#ifndef MICROPROFILE_HELP_RIGHT\n#define MICROPROFILE_HELP_RIGHT \"Right-Click\"\n#endif\n\n#ifndef MICROPROFILE_HELP_MOD\n#define MICROPROFILE_HELP_MOD \"Mod\"\n#endif\n\n#ifndef MICROPROFILE_BAR_WIDTH\n#define MICROPROFILE_BAR_WIDTH 100\n#endif\n\n#ifndef MICROPROFILE_CUSTOM_MAX\n#define MICROPROFILE_CUSTOM_MAX 8 \n#endif\n\n#ifndef MICROPROFILE_CUSTOM_MAX_TIMERS\n#define MICROPROFILE_CUSTOM_MAX_TIMERS 64\n#endif\n\n#ifndef MICROPROFILE_CUSTOM_PADDING\n#define MICROPROFILE_CUSTOM_PADDING 12\n#endif\n\n\n#define MICROPROFILE_FRAME_HISTORY_HEIGHT 50\n#define MICROPROFILE_FRAME_HISTORY_WIDTH 7\n#define MICROPROFILE_FRAME_HISTORY_COLOR_CPU 0xffff7f27 //255 127 39\n#define MICROPROFILE_FRAME_HISTORY_COLOR_GPU 0xff37a0ee //55 160 238\n#define MICROPROFILE_FRAME_HISTORY_COLOR_HIGHTLIGHT 0x7733bb44\n#define MICROPROFILE_FRAME_COLOR_HIGHTLIGHT 0x20009900\n#define MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU 0x20996600\n#define MICROPROFILE_NUM_FRAMES (MICROPROFILE_MAX_FRAME_HISTORY - (MICROPROFILE_GPU_FRAME_DELAY+1))\n\n#define MICROPROFILE_TOOLTIP_MAX_STRINGS (32 + MICROPROFILE_MAX_GROUPS*2)\n#define MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE (4*1024)\n#define MICROPROFILE_TOOLTIP_MAX_LOCKED 3\n\n#define MICROPROFILE_COUNTER_INDENT 4\n#define MICROPROFILE_COUNTER_WIDTH 100\n\n\n\nenum\n{\n\tMICROPROFILE_CUSTOM_BARS = 0x1,\n\tMICROPROFILE_CUSTOM_BAR_SOURCE_MAX = 0x2,\n\tMICROPROFILE_CUSTOM_BAR_SOURCE_AVG = 0,\n\tMICROPROFILE_CUSTOM_STACK = 0x4,\n\tMICROPROFILE_CUSTOM_STACK_SOURCE_MAX = 0x8,\n\tMICROPROFILE_CUSTOM_STACK_SOURCE_AVG = 0,\n};\n\n\nMICROPROFILEUI_API void MicroProfileDraw(uint32_t nWidth, uint32_t nHeight); //! call if drawing microprofilers\nMICROPROFILEUI_API bool MicroProfileIsDrawing();\nMICROPROFILEUI_API void MicroProfileToggleGraph(MicroProfileToken nToken);\nMICROPROFILEUI_API bool MicroProfileDrawGraph(uint32_t nScreenWidth, uint32_t nScreenHeight);\nMICROPROFILEUI_API void MicroProfileToggleDisplayMode(); //switch between off, bars, detailed\nMICROPROFILEUI_API void MicroProfileSetDisplayMode(int); //switch between off, bars, detailed\nMICROPROFILEUI_API void MicroProfileClearGraph();\nMICROPROFILEUI_API void MicroProfileMousePosition(uint32_t nX, uint32_t nY, int nWheelDelta);\nMICROPROFILEUI_API void MicroProfileModKey(uint32_t nKeyState);\nMICROPROFILEUI_API void MicroProfileMouseButton(uint32_t nLeft, uint32_t nRight);\nMICROPROFILEUI_API void MicroProfileDrawLineVertical(int nX, int nTop, int nBottom, uint32_t nColor);\nMICROPROFILEUI_API void MicroProfileDrawLineHorizontal(int nLeft, int nRight, int nY, uint32_t nColor);\nMICROPROFILEUI_API void MicroProfileLoadPreset(const char* pSuffix);\nMICROPROFILEUI_API void MicroProfileSavePreset(const char* pSuffix);\n\nMICROPROFILEUI_API void MicroProfileDrawText(int nX, int nY, uint32_t nColor, const char* pText, uint32_t nNumCharacters);\nMICROPROFILEUI_API void MicroProfileDrawBox(int nX, int nY, int nX1, int nY1, uint32_t nColor, MicroProfileBoxType = MicroProfileBoxTypeFlat);\nMICROPROFILEUI_API void MicroProfileDrawLine2D(uint32_t nVertices, float* pVertices, uint32_t nColor);\nMICROPROFILEUI_API void MicroProfileDumpTimers();\n\nMICROPROFILEUI_API void MicroProfileInitUI();\n\nMICROPROFILEUI_API void MicroProfileCustomGroupToggle(const char* pCustomName);\nMICROPROFILEUI_API void MicroProfileCustomGroupEnable(const char* pCustomName);\nMICROPROFILEUI_API void MicroProfileCustomGroupEnable(uint32_t nIndex);\nMICROPROFILEUI_API void MicroProfileCustomGroupDisable();\nMICROPROFILEUI_API void MicroProfileCustomGroup(const char* pCustomName, uint32_t nMaxTimers, uint32_t nAggregateFlip, float fReferenceTime, uint32_t nFlags);\nMICROPROFILEUI_API void MicroProfileCustomGroupAddTimer(const char* pCustomName, const char* pGroup, const char* pTimer);\n\n#ifdef MICROPROFILEUI_IMPL\n#ifdef _WIN32\n#define snprintf _snprintf\n#endif\n#include \n#include \n#include \n#include \n\nMICROPROFILE_DEFINE(g_MicroProfileDetailed, \"MicroProfile\", \"Detailed View\", 0x8888000);\nMICROPROFILE_DEFINE(g_MicroProfileDrawGraph, \"MicroProfile\", \"Draw Graph\", 0xff44ee00);\nMICROPROFILE_DEFINE(g_MicroProfileDrawBarView, \"MicroProfile\", \"DrawBarView\", 0x00dd77);\nMICROPROFILE_DEFINE(g_MicroProfileDraw,\"MicroProfile\", \"Draw\", 0x737373);\n\n\nstruct MicroProfileStringArray\n{\n\tconst char* ppStrings[MICROPROFILE_TOOLTIP_MAX_STRINGS];\n\tchar Buffer[MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE];\n\tchar* pBufferPos;\n\tuint32_t nNumStrings;\n};\n\nstruct MicroProfileGroupMenuItem\n{\n\tuint32_t nIsCategory;\n\tuint32_t nCategoryIndex;\n\tuint32_t nIndex;\n\tconst char* pName;\n};\n\nstruct MicroProfileCustom\n{\n\tchar pName[MICROPROFILE_NAME_MAX_LEN];\n\tuint32_t nFlags;\n\tuint32_t nAggregateFlip;\t\n\tuint32_t nNumTimers;\n\tuint32_t nMaxTimers;\n\tuint64_t nGroupMask;\n\tfloat fReference;\n\tuint64_t* pTimers;\n};\n\nstruct SOptionDesc\n{\n\tSOptionDesc(){}\n\tSOptionDesc(uint8_t nSubType, uint8_t nIndex, const char* fmt, ...):nSubType(nSubType), nIndex(nIndex)\n\t{\n\t\tva_list args;\n\t\tva_start (args, fmt);\n\t\tvsprintf(Text, fmt, args);\n\t\tva_end(args);\n\t}\n\tchar Text[32];\n\tuint8_t nSubType;\n\tuint8_t nIndex;\n\tbool bSelected;\n};\nstatic uint32_t g_MicroProfileAggregatePresets[] = {0, 10, 20, 30, 60, 120};\nstatic float g_MicroProfileReferenceTimePresets[] = {5.f, 10.f, 15.f,20.f, 33.33f, 66.66f, 100.f, 250.f, 500.f, 1000.f};\nstatic uint32_t g_MicroProfileOpacityPresets[] = {0x40, 0x80, 0xc0, 0xff};\nstatic const char* g_MicroProfilePresetNames[] = \n{\n\tMICROPROFILE_DEFAULT_PRESET,\n\t\"Render\",\n\t\"GPU\",\n\t\"Lighting\",\n\t\"AI\",\n\t\"Visibility\",\n\t\"Sound\",\n};\n\nenum\n{\n\tMICROPROFILE_NUM_REFERENCE_PRESETS = sizeof(g_MicroProfileReferenceTimePresets)/sizeof(g_MicroProfileReferenceTimePresets[0]),\n\tMICROPROFILE_NUM_OPACITY_PRESETS = sizeof(g_MicroProfileOpacityPresets)/sizeof(g_MicroProfileOpacityPresets[0]),\n#if MICROPROFILE_CONTEXT_SWITCH_TRACE\n\tMICROPROFILE_OPTION_SIZE = MICROPROFILE_NUM_REFERENCE_PRESETS + MICROPROFILE_NUM_OPACITY_PRESETS * 2 + 2 + 6,\n#else\n\tMICROPROFILE_OPTION_SIZE = MICROPROFILE_NUM_REFERENCE_PRESETS + MICROPROFILE_NUM_OPACITY_PRESETS * 2 + 2 + 3,\n#endif\n};\n\nstruct MicroProfileUI\n{\n\t//menu/mouse over stuff\n\tuint64_t nHoverToken;\n\tint64_t nHoverTime;\n\tint \t nHoverFrame;\n#if MICROPROFILE_DEBUG\n\tuint64_t nHoverAddressEnter;\n\tuint64_t nHoverAddressLeave;\n#endif\n\n\tuint32_t nWidth;\n\tuint32_t nHeight;\n\n\tint nOffsetX[MP_DRAW_SIZE];\n\tint nOffsetY[MP_DRAW_SIZE];\n\n\tfloat fDetailedOffset; //display offset relative to start of latest displayable frame.\n\tfloat fDetailedRange; //no. of ms to display\n\tfloat fDetailedOffsetTarget;\n\tfloat fDetailedRangeTarget;\n\tuint32_t nOpacityBackground;\n\tuint32_t nOpacityForeground;\n\tbool bShowSpikes;\n\n\n\n\tuint32_t \t\t\t\tnMouseX;\n\tuint32_t \t\t\t\tnMouseY;\n\tuint32_t \t\t\t\tnMouseDownX;\n\tuint32_t \t\t\t\tnMouseDownY;\n\tint\t\t\t\t\t\tnMouseWheelDelta;\n\tuint32_t\t\t\t\tnMouseDownLeft;\n\tuint32_t\t\t\t\tnMouseDownRight;\n\tuint32_t \t\t\t\tnMouseLeft;\n\tuint32_t \t\t\t\tnMouseRight;\n\tuint32_t \t\t\t\tnMouseLeftMod;\n\tuint32_t \t\t\t\tnMouseRightMod;\n\tuint32_t\t\t\t\tnModDown;\n\tuint32_t \t\t\t\tnActiveMenu;\n\n\tMicroProfileLogEntry* pDisplayMouseOver;\n\n\tint64_t\t\t\t\t\tnRangeBegin;\n\tint64_t\t\t\t\t\tnRangeEnd;\n\tint64_t\t\t\t\t\tnRangeBeginGpu;\n\tint64_t\t\t\t\t\tnRangeEndGpu;\n\tuint32_t\t\t\t\tnRangeBeginIndex;\n\tuint32_t \t\t\t\tnRangeEndIndex;\n\tMicroProfileThreadLog* \tpRangeLog;\n\tuint32_t\t\t\t\tnHoverColor;\n\tuint32_t\t\t\t\tnHoverColorShared;\n\n\tint64_t\t\t\t\t\tnTickReferenceCpu;\n\tint64_t\t\t\t\t\tnTickReferenceGpu;\n\n\tMicroProfileStringArray LockedToolTips[MICROPROFILE_TOOLTIP_MAX_LOCKED];\t\n\tuint32_t \t\t\t\tnLockedToolTipColor[MICROPROFILE_TOOLTIP_MAX_LOCKED];\t\n\tint \t\t\t\t\tLockedToolTipFront;\n\n\tMicroProfileGroupMenuItem \tGroupMenu[MICROPROFILE_MAX_GROUPS + MICROPROFILE_MAX_CATEGORIES];\n\tuint32_t \t\t\t\t\tGroupMenuCount;\n\n\n\tuint32_t\t\t\t\t\tnCustomActive;\n\tuint32_t\t\t\t\t\tnCustomTimerCount;\n\tuint32_t \t\t\t\t\tnCustomCount;\n\tMicroProfileCustom \t\t\tCustom[MICROPROFILE_CUSTOM_MAX];\n\tuint64_t\t\t\t\t\tCustomTimer[MICROPROFILE_CUSTOM_MAX_TIMERS];\n\t\n\tSOptionDesc Options[MICROPROFILE_OPTION_SIZE];\n\n\tuint32_t nCounterWidth;\n\tuint32_t nLimitWidth;\n\tuint32_t nCounterWidthTemp;\n\tuint32_t nLimitWidthTemp;\n\n\n};\n\nMicroProfileUI g_MicroProfileUI;\n#define UI g_MicroProfileUI\nstatic uint32_t g_nMicroProfileBackColors[2] = { 0x474747, 0x313131 };\n#define MICROPROFILE_NUM_CONTEXT_SWITCH_COLORS 16\nstatic uint32_t g_nMicroProfileContextSwitchThreadColors[MICROPROFILE_NUM_CONTEXT_SWITCH_COLORS] = //palette generated by http://tools.medialab.sciences-po.fr/iwanthue/index.php\n{\n\t0x63607B,\n\t0x755E2B,\n\t0x326A55,\n\t0x523135,\n\t0x904F42,\n\t0x87536B,\n\t0x346875,\n\t0x5E6046,\n\t0x35404C,\n\t0x224038,\n\t0x413D1E,\n\t0x5E3A26,\n\t0x5D6161,\n\t0x4C6234,\n\t0x7D564F,\n\t0x5C4352,\n};\n\n\nvoid MicroProfileInitUI()\n{\n\tstatic bool bInitialized = false;\n\tif(!bInitialized)\n\t{\n\t\tbInitialized = true;\n\t\tmemset(&g_MicroProfileUI, 0, sizeof(g_MicroProfileUI));\n\t\tUI.nActiveMenu = (uint32_t)-1;\n\t\tUI.fDetailedOffsetTarget = UI.fDetailedOffset = 0.f;\n\t\tUI.fDetailedRangeTarget = UI.fDetailedRange = 50.f;\n\n\t\tUI.nOpacityBackground = 0xff<<24;\n\t\tUI.nOpacityForeground = 0xff<<24;\n\n\t\tUI.bShowSpikes = false;\n\n\t\tUI.nWidth = 100;\n\t\tUI.nHeight = 100;\n\n\t\tUI.nCustomActive = (uint32_t)-1;\n\t\tUI.nCustomTimerCount = 0;\n\t\tUI.nCustomCount = 0;\n\n\t\tint nIndex = 0;\n\t\tUI.Options[nIndex++] = SOptionDesc(0xff, 0, \"%s\", \"Reference\");\n\t\tfor(int i = 0; i < MICROPROFILE_NUM_REFERENCE_PRESETS; ++i)\n\t\t{\n\t\t\tUI.Options[nIndex++] = SOptionDesc(0, i, \" %6.2fms\", g_MicroProfileReferenceTimePresets[i]);\n\t\t}\n\t\tUI.Options[nIndex++] = SOptionDesc(0xff, 0, \"%s\", \"BG Opacity\");\t\t\n\t\tfor(int i = 0; i < MICROPROFILE_NUM_OPACITY_PRESETS; ++i)\n\t\t{\n\t\t\tUI.Options[nIndex++] = SOptionDesc(1, i, \" %7d%%\", (i+1)*25);\n\t\t}\n\t\tUI.Options[nIndex++] = SOptionDesc(0xff, 0, \"%s\", \"FG Opacity\");\t\t\n\t\tfor(int i = 0; i < MICROPROFILE_NUM_OPACITY_PRESETS; ++i)\n\t\t{\n\t\t\tUI.Options[nIndex++] = SOptionDesc(2, i, \" %7d%%\", (i+1)*25);\n\t\t}\n\t\tUI.Options[nIndex++] = SOptionDesc(0xff, 0, \"%s\", \"Spike Display\");\t\t\n\t\tUI.Options[nIndex++] = SOptionDesc(3, 0, \"%s\", \" Enable\");\n\n#if MICROPROFILE_CONTEXT_SWITCH_TRACE\n\t\tUI.Options[nIndex++] = SOptionDesc(0xff, 0, \"%s\", \"CSwitch Trace\");\t\t\n\t\tUI.Options[nIndex++] = SOptionDesc(4, 0, \"%s\", \" All Threads\");\n\t\tUI.Options[nIndex++] = SOptionDesc(4, 1, \"%s\", \" No Bars\");\n#endif\n\t\tMP_ASSERT(nIndex == MICROPROFILE_OPTION_SIZE);\n\n\t\tUI.nCounterWidth = 100;\n\t\tUI.nLimitWidth = 100;\n\t\tUI.nCounterWidthTemp = 100;\n\t\tUI.nLimitWidthTemp = 100;\n\n\t}\n}\n\nvoid MicroProfileSetDisplayMode(int nValue)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tnValue = nValue >= 0 && nValue < MP_DRAW_SIZE ? nValue : S.nDisplay;\n\tS.nDisplay = nValue;\n\tUI.nOffsetY[S.nDisplay] = 0;\n}\n\nvoid MicroProfileToggleDisplayMode()\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tS.nDisplay = (S.nDisplay + 1) % MP_DRAW_SIZE;\n\tUI.nOffsetY[S.nDisplay] = 0;\n}\n\n\nvoid MicroProfileStringArrayClear(MicroProfileStringArray* pArray)\n{\n\tpArray->nNumStrings = 0;\n\tpArray->pBufferPos = &pArray->Buffer[0];\n}\n\nvoid MicroProfileStringArrayAddLiteral(MicroProfileStringArray* pArray, const char* pLiteral)\n{\n\tMP_ASSERT(pArray->nNumStrings < MICROPROFILE_TOOLTIP_MAX_STRINGS);\n\tpArray->ppStrings[pArray->nNumStrings++] = pLiteral;\n}\n\nMICROPROFILE_FORMAT(2, 3) void MicroProfileStringArrayFormat(MicroProfileStringArray* pArray, const char* fmt, ...)\n{\n\tMP_ASSERT(pArray->nNumStrings < MICROPROFILE_TOOLTIP_MAX_STRINGS);\n\tpArray->ppStrings[pArray->nNumStrings++] = pArray->pBufferPos;\n\tva_list args;\n\tva_start (args, fmt);\n\tpArray->pBufferPos += 1 + vsprintf(pArray->pBufferPos, fmt, args);\n\tva_end(args);\n\tMP_ASSERT(pArray->pBufferPos < pArray->Buffer + MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE);\n}\nvoid MicroProfileStringArrayCopy(MicroProfileStringArray* pDest, MicroProfileStringArray* pSrc)\n{\n\tmemcpy(&pDest->ppStrings[0], &pSrc->ppStrings[0], sizeof(pDest->ppStrings));\n\tmemcpy(&pDest->Buffer[0], &pSrc->Buffer[0], sizeof(pDest->Buffer));\n\tfor(uint32_t i = 0; i < MICROPROFILE_TOOLTIP_MAX_STRINGS; ++i)\n\t{\n\t\tif(i < pSrc->nNumStrings)\n\t\t{\n\t\t\tif(pSrc->ppStrings[i] >= &pSrc->Buffer[0] && pSrc->ppStrings[i] < &pSrc->Buffer[0] + MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE)\n\t\t\t{\n\t\t\t\tpDest->ppStrings[i] += &pDest->Buffer[0] - &pSrc->Buffer[0];\n\t\t\t}\n\t\t}\n\t}\n\tpDest->nNumStrings = pSrc->nNumStrings;\n}\n\nvoid MicroProfileFloatWindowSize(const char** ppStrings, uint32_t nNumStrings, uint32_t* pColors, uint32_t& nWidth, uint32_t& nHeight, uint32_t* pStringLengths = 0)\n{\n\tuint32_t* nStringLengths = pStringLengths ? pStringLengths : (uint32_t*)alloca(nNumStrings * sizeof(uint32_t));\n\tuint32_t nTextCount = nNumStrings/2;\n\tfor(uint32_t i = 0; i < nTextCount; ++i)\n\t{\n\t\tuint32_t i0 = i * 2;\n\t\tuint32_t s0, s1;\n\t\tnStringLengths[i0] = s0 = (uint32_t)strlen(ppStrings[i0]);\n\t\tnStringLengths[i0+1] = s1 = (uint32_t)strlen(ppStrings[i0+1]);\n\t\tnWidth = MicroProfileMax(s0+s1, nWidth);\n\t}\n\tnWidth = (MICROPROFILE_TEXT_WIDTH+1) * (2+nWidth) + 2 * MICROPROFILE_BORDER_SIZE;\n\tif(pColors)\n\t\tnWidth += MICROPROFILE_TEXT_WIDTH + 1;\n\tnHeight = (MICROPROFILE_TEXT_HEIGHT+1) * nTextCount + 2 * MICROPROFILE_BORDER_SIZE;\n}\n\nvoid MicroProfileDrawFloatWindow(uint32_t nX, uint32_t nY, const char** ppStrings, uint32_t nNumStrings, uint32_t nColor, uint32_t* pColors = 0)\n{\n\tuint32_t nWidth = 0, nHeight = 0;\n\tuint32_t* nStringLengths = (uint32_t*)alloca(nNumStrings * sizeof(uint32_t));\n\tMicroProfileFloatWindowSize(ppStrings, nNumStrings, pColors, nWidth, nHeight, nStringLengths);\n\tuint32_t nTextCount = nNumStrings/2;\n\tif(nX + nWidth > UI.nWidth)\n\t\tnX = UI.nWidth - nWidth;\n\tif(nY + nHeight > UI.nHeight)\n\t\tnY = UI.nHeight - nHeight;\n\tMicroProfileDrawBox(nX-1, nY-1, nX + nWidth+1, nY + nHeight+1, 0xff000000|nColor);\n\tMicroProfileDrawBox(nX, nY, nX + nWidth, nY + nHeight, 0xff000000);\n\tif(pColors)\n\t{\n\t\tnX += MICROPROFILE_TEXT_WIDTH+1;\n\t\tnWidth -= MICROPROFILE_TEXT_WIDTH+1;\n\t}\n\tfor(uint32_t i = 0; i < nTextCount; ++i)\n\t{\n\t\tint i0 = i * 2;\n\t\tif(pColors)\n\t\t{\n\t\t\tMicroProfileDrawBox(nX-MICROPROFILE_TEXT_WIDTH, nY, nX, nY + MICROPROFILE_TEXT_WIDTH, pColors[i]|0xff000000);\n\t\t}\n\t\tMicroProfileDrawText(nX + 1, nY + 1, (uint32_t)-1, ppStrings[i0], (uint32_t)strlen(ppStrings[i0]));\n\t\tMicroProfileDrawText(nX + nWidth - nStringLengths[i0+1] * (MICROPROFILE_TEXT_WIDTH+1), nY + 1, (uint32_t)-1, ppStrings[i0+1], (uint32_t)strlen(ppStrings[i0+1]));\n\t\tnY += (MICROPROFILE_TEXT_HEIGHT+1);\n\t}\n}\n\nvoid MicroProfileDrawTextBackground(uint32_t nX, uint32_t nY, uint32_t nColor, uint32_t nBgColor, const char* pString, uint32_t nStrLen)\n{\n\tuint32_t nWidth = (MICROPROFILE_TEXT_WIDTH + 1) * (nStrLen) + 2 * MICROPROFILE_BORDER_SIZE;\n\tuint32_t nHeight = (MICROPROFILE_TEXT_HEIGHT + 1) ;\n\tMicroProfileDrawBox(nX, nY, nX + nWidth, nY + nHeight, nBgColor);\n\tMicroProfileDrawText(nX, nY, nColor, pString, nStrLen);\n}\n\nvoid MicroProfileToolTipMeta(MicroProfileStringArray* pToolTip)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tif(UI.nRangeBeginIndex != UI.nRangeEndIndex && UI.pRangeLog)\n\t{\n\t\tuint64_t nMetaSum[MICROPROFILE_META_MAX] = {0};\n\t\tuint64_t nMetaSumInclusive[MICROPROFILE_META_MAX] = {0};\n\t\tint nStackDepth = 0;\n\t\tuint32_t nRange[2][2];\n\t\tMicroProfileThreadLog* pLog = UI.pRangeLog;\n\n\n\t\tMicroProfileGetRange(UI.nRangeEndIndex, UI.nRangeBeginIndex, nRange);\n\t\tfor(uint32_t i = 0; i < 2; ++i)\n\t\t{\n\t\t\tuint32_t nStart = nRange[i][0];\n\t\t\tuint32_t nEnd = nRange[i][1];\n\t\t\tfor(uint32_t j = nStart; j < nEnd; ++j)\n\t\t\t{\n\t\t\t\tMicroProfileLogEntry LE = pLog->Log[j];\n\t\t\t\tuint64_t nType = MicroProfileLogType(LE);\n\t\t\t\tswitch(nType)\n\t\t\t\t{\n\t\t\t\tcase MP_LOG_META:\n\t\t\t\t\t{\n\t\t\t\t\t\tint64_t nMetaIndex = MicroProfileLogTimerIndex(LE);\n\t\t\t\t\t\tint64_t nMetaCount = MicroProfileLogGetTick(LE);\n\t\t\t\t\t\tMP_ASSERT(nMetaIndex < MICROPROFILE_META_MAX);\n\t\t\t\t\t\tif(nStackDepth>1)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tnMetaSumInclusive[nMetaIndex] += nMetaCount;\n\t\t\t\t\t\t}\n\t\t\t\t\t\telse\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tnMetaSum[nMetaIndex] += nMetaCount;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\tbreak;\n\t\t\t\tcase MP_LOG_LEAVE:\n\t\t\t\t\tif(nStackDepth)\n\t\t\t\t\t{\n\t\t\t\t\t\tnStackDepth--;\n\t\t\t\t\t}\n\t\t\t\t\telse\n\t\t\t\t\t{\n\t\t\t\t\t\tfor(int i = 0; i < MICROPROFILE_META_MAX; ++i)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tnMetaSumInclusive[i] += nMetaSum[i];\n\t\t\t\t\t\t\tnMetaSum[i] = 0;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\tbreak;\n\t\t\t\tcase MP_LOG_ENTER:\n\t\t\t\t\tnStackDepth++;\n\t\t\t\t\tbreak;\n\t\t\t\t}\n\n\t\t\t}\n\t\t}\n\t\tbool bSpaced = false;\n\t\tfor(int i = 0; i < MICROPROFILE_META_MAX; ++i)\n\t\t{\n\t\t\tif(S.MetaCounters[i].pName && (nMetaSum[i]||nMetaSumInclusive[i]))\n\t\t\t{\n\t\t\t\tif(!bSpaced)\n\t\t\t\t{\n\t\t\t\t\tbSpaced = true;\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(pToolTip, \"\");\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(pToolTip, \"\");\t\t\t\t\t\n\t\t\t\t}\n\t\t\t\tMicroProfileStringArrayFormat(pToolTip, \"%s excl\", S.MetaCounters[i].pName);\n\t\t\t\tMicroProfileStringArrayFormat(pToolTip, \"%5lld\", (long long)nMetaSum[i]);\n\t\t\t\tMicroProfileStringArrayFormat(pToolTip, \"%s incl\", S.MetaCounters[i].pName);\n\t\t\t\tMicroProfileStringArrayFormat(pToolTip, \"%5lld\", (long long)(nMetaSum[i] + nMetaSumInclusive[i]));\n\t\t\t}\n\t\t}\n\t}\n}\n\nvoid MicroProfileToolTipLabel(MicroProfileStringArray* pToolTip)\n{\n\tif(UI.nRangeBeginIndex != UI.nRangeEndIndex && UI.pRangeLog)\n\t{\n\t\tbool bSpaced = false;\n\t\tint nStackDepth = 0;\n\t\tuint32_t nRange[2][2];\n\t\tMicroProfileThreadLog* pLog = UI.pRangeLog;\n\n\t\tMicroProfileGetRange(UI.nRangeEndIndex, UI.nRangeBeginIndex, nRange);\n\t\tfor(uint32_t i = 0; i < 2; ++i)\n\t\t{\n\t\t\tuint32_t nStart = nRange[i][0];\n\t\t\tuint32_t nEnd = nRange[i][1];\n\t\t\tfor(uint32_t j = nStart; j < nEnd; ++j)\n\t\t\t{\n\t\t\t\tMicroProfileLogEntry LE = pLog->Log[j];\n\t\t\t\tuint64_t nType = MicroProfileLogType(LE);\n\t\t\t\tswitch(nType)\n\t\t\t\t{\n\t\t\t\tcase MP_LOG_LABEL:\n\t\t\t\t\t{\n\t\t\t\t\t\tif(nStackDepth == 1)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tuint64_t nLabel = MicroProfileLogGetTick(LE);\n\t\t\t\t\t\t\tconst char* pLabelName = MicroProfileGetLabel(nLabel);\n\n\t\t\t\t\t\t\tif (!bSpaced)\n\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\tbSpaced = true;\n\t\t\t\t\t\t\t\tMicroProfileStringArrayAddLiteral(pToolTip, \"\");\n\t\t\t\t\t\t\t\tMicroProfileStringArrayAddLiteral(pToolTip, \"\");\n\t\t\t\t\t\t\t}\n\n\t\t\t\t\t\t\tif (pToolTip->nNumStrings + 2 <= MICROPROFILE_TOOLTIP_MAX_STRINGS)\n\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\tMicroProfileStringArrayAddLiteral(pToolTip, \"Label:\");\n\t\t\t\t\t\t\t\tMicroProfileStringArrayAddLiteral(pToolTip, pLabelName ? pLabelName : \"??\");\n\t\t\t\t\t\t\t}\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\tbreak;\n\t\t\t\tcase MP_LOG_LEAVE:\n\t\t\t\t\tif(nStackDepth)\n\t\t\t\t\t{\n\t\t\t\t\t\tnStackDepth--;\n\t\t\t\t\t}\n\t\t\t\t\tbreak;\n\t\t\t\tcase MP_LOG_ENTER:\n\t\t\t\t\tnStackDepth++;\n\t\t\t\t\tbreak;\n\t\t\t\t}\n\n\t\t\t}\n\t\t}\n\t}\n}\n\nvoid MicroProfileDrawFloatTooltip(uint32_t nX, uint32_t nY, uint32_t nToken, uint64_t nTime)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\n\tuint32_t nIndex = MicroProfileGetTimerIndex(nToken);\n\tuint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1;\n\tuint32_t nAggregateCount = S.Aggregate[nIndex].nCount ? S.Aggregate[nIndex].nCount : 1;\n\n\tuint32_t nGroupId = MicroProfileGetGroupIndex(nToken);\n\tuint32_t nTimerId = MicroProfileGetTimerIndex(nToken);\n\tbool bGpu = S.GroupInfo[nGroupId].Type == MicroProfileTokenTypeGpu;\n\n\tfloat fToMs = MicroProfileTickToMsMultiplier(bGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu());\n\n\tfloat fMs = fToMs * (nTime);\n\tfloat fFrameMs = fToMs * (S.Frame[nIndex].nTicks);\n\tfloat fAverage = fToMs * (S.Aggregate[nIndex].nTicks/nAggregateFrames);\n\tfloat fCallAverage = fToMs * (S.Aggregate[nIndex].nTicks / nAggregateCount);\n\tfloat fMax = fToMs * (S.AggregateMax[nIndex]);\n\tfloat fMin = fToMs * (S.AggregateMin[nIndex]);\n\n\tfloat fFrameMsExclusive = fToMs * (S.FrameExclusive[nIndex]);\n\tfloat fAverageExclusive = fToMs * (S.AggregateExclusive[nIndex]/nAggregateFrames);\n\tfloat fMaxExclusive = fToMs * (S.AggregateMaxExclusive[nIndex]);\n\n\tfloat fGroupAverage = fToMs * (S.AggregateGroup[nGroupId] / nAggregateFrames);\n\tfloat fGroupMax = fToMs * (S.AggregateGroupMax[nGroupId]);\n\tfloat fGroup = fToMs * (S.FrameGroup[nGroupId]);\n\n\n\tMicroProfileStringArray ToolTip;\n\tMicroProfileStringArrayClear(&ToolTip);\n\tconst char* pGroupName = S.GroupInfo[nGroupId].pName;\n\tconst char* pTimerName = S.TimerInfo[nTimerId].pName;\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Timer:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%s\", pTimerName);\n\n#if MICROPROFILE_DEBUG\n\tMicroProfileStringArrayFormat(&ToolTip,\"0x%p\", UI.nHoverAddressEnter);\n\tMicroProfileStringArrayFormat(&ToolTip,\"0x%p\", UI.nHoverAddressLeave);\n#endif\n\t\n\tif(nTime != (uint64_t)0)\n\t{\n\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Time:\");\n\t\tMicroProfileStringArrayFormat(&ToolTip,\"%6.3fms\", fMs);\n\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\t}\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Frame Time:\");\n\tMicroProfileStringArrayFormat(&ToolTip,\"%6.3fms\", fFrameMs);\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Average:\");\n\tMicroProfileStringArrayFormat(&ToolTip,\"%6.3fms\", fAverage);\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Max:\");\n\tMicroProfileStringArrayFormat(&ToolTip,\"%6.3fms\", fMax);\n\t\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Min:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%6.3fms\", fMin);\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Call Average:\");\n\tMicroProfileStringArrayFormat(&ToolTip,\"%6.3fms\", fCallAverage);\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Call Count:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%6.2f\", double(nAggregateCount) / nAggregateFrames);\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Exclusive Frame Time:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%6.3fms\", fFrameMsExclusive);\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Exclusive Average:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%6.3fms\", fAverageExclusive);\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Exclusive Max:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%6.3fms\", fMaxExclusive);\n\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\t\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Group:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%s\", pGroupName);\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Frame Time:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%6.3f\", fGroup);\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Frame Average:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%6.3f\", fGroupAverage);\n\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Frame Max:\");\n\tMicroProfileStringArrayFormat(&ToolTip, \"%6.3f\", fGroupMax);\n\n\n\n\n\tMicroProfileToolTipMeta(&ToolTip);\n\tMicroProfileToolTipLabel(&ToolTip);\n\n\n\tMicroProfileDrawFloatWindow(nX, nY+20, &ToolTip.ppStrings[0], ToolTip.nNumStrings, S.TimerInfo[nTimerId].nColor);\n\n\tif(UI.nMouseLeftMod)\n\t{\n\t\tint nIndex = (g_MicroProfileUI.LockedToolTipFront + MICROPROFILE_TOOLTIP_MAX_LOCKED - 1) % MICROPROFILE_TOOLTIP_MAX_LOCKED;\n\t\tg_MicroProfileUI.nLockedToolTipColor[nIndex] = S.TimerInfo[nTimerId].nColor;\n\t\tMicroProfileStringArrayCopy(&g_MicroProfileUI.LockedToolTips[nIndex], &ToolTip);\n\t\tg_MicroProfileUI.LockedToolTipFront = nIndex;\n\n\t}\n}\n\nint64_t MicroProfileGetGpuTickSync(int64_t nTickCpu, int64_t nTickGpu)\n{\n\tif(UI.nTickReferenceCpu && UI.nTickReferenceGpu)\n\t{\n\t\tint64_t nTicksPerSecondCpu = MicroProfileTicksPerSecondCpu();\n\t\tint64_t nTicksPerSecondGpu = MicroProfileTicksPerSecondGpu();\n\n\t\treturn (nTickCpu - UI.nTickReferenceCpu) * int64_t(double(nTicksPerSecondGpu) / double(nTicksPerSecondCpu)) + UI.nTickReferenceGpu;\n\t}\n\telse\n\t{\n\t\treturn nTickGpu;\n\t}\n}\n\nvoid MicroProfileZoomTo(int64_t nTickStart, int64_t nTickEnd, MicroProfileTokenType eToken)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\n\tbool bGpu = eToken == MicroProfileTokenTypeGpu;\n\tint64_t nStartCpu = S.Frames[S.nFrameCurrent].nFrameStartCpu;\n\tint64_t nStart = bGpu ? MicroProfileGetGpuTickSync(nStartCpu, S.Frames[S.nFrameCurrent].nFrameStartGpu) : nStartCpu;\n\tuint64_t nFrequency = bGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu();\n\n\tfloat fToMs = MicroProfileTickToMsMultiplier(nFrequency);\n\tUI.fDetailedOffsetTarget = MicroProfileLogTickDifference(nStart, nTickStart) * fToMs;\n\tUI.fDetailedRangeTarget = MicroProfileMax(MicroProfileLogTickDifference(nTickStart, nTickEnd) * fToMs, 0.01f); // clamp to 10us\n}\n\nvoid MicroProfileCenter(int64_t nTickCenter)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tint64_t nStart = S.Frames[S.nFrameCurrent].nFrameStartCpu;\n\tfloat fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu());\n\tfloat fCenter = MicroProfileLogTickDifference(nStart, nTickCenter) * fToMs;\n\tUI.fDetailedOffsetTarget = UI.fDetailedOffset = fCenter - 0.5f * UI.fDetailedRange;\n}\n\n#if MICROPROFILE_DEBUG\nuint64_t* g_pMicroProfileDumpStart = 0;\nuint64_t* g_pMicroProfileDumpEnd = 0;\nvoid MicroProfileDebugDumpRange()\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tif(g_pMicroProfileDumpStart != g_pMicroProfileDumpEnd)\n\t{\n\t\tuint64_t* pStart = g_pMicroProfileDumpStart;\n\t\tuint64_t* pEnd = g_pMicroProfileDumpEnd;\n\t\twhile(pStart != pEnd)\n\t\t{\n\t\t\tuint64_t nTick = MicroProfileLogGetTick(*pStart);\n\t\t\tuint64_t nToken = MicroProfileLogTimerIndex(*pStart);\n\t\t\tuint32_t nTimerId = MicroProfileGetTimerIndex(nToken);\n\t\n\t\t\tconst char* pTimerName = S.TimerInfo[nTimerId].pName;\n\t\t\tchar buffer[256];\n\t\t\tuint64_t type = MicroProfileLogType(*pStart);\n\n\t\t\tconst char* pBegin = type == MP_LOG_LEAVE ? \"END\" : \n\t\t\t\t(type == MP_LOG_ENTER ? \"BEGIN\" : \"META\");\n\t\t\tsnprintf(buffer, 255, \"DUMP 0x%p: %s :: %\" PRIx64 \": %s\\n\", pStart, pBegin, nTick, pTimerName);\n#ifdef _WIN32\n\t\t\tOutputDebugString(buffer);\n#else\n\t\t\tprintf(\"%s\", buffer);\n#endif\n\t\t\tpStart++;\n\t\t}\n\n\t\tg_pMicroProfileDumpStart = g_pMicroProfileDumpEnd;\n\t}\n}\n#define MP_DEBUG_DUMP_RANGE() MicroProfileDebugDumpRange()\n#else\n#define MP_DEBUG_DUMP_RANGE() do{} while(0)\n#endif\n\n#define MICROPROFILE_HOVER_DIST 0.5f\n\nvoid MicroProfileDrawDetailedContextSwitchBars(uint32_t nY, uint32_t nThreadId, uint32_t nContextSwitchStart, uint32_t nContextSwitchEnd, int64_t nBaseTicks, uint32_t nBaseY)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tint64_t nTickIn = -1;\n\tuint32_t nThreadBefore = -1;\n\tfloat fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu());\n\tfloat fMsToScreen = UI.nWidth / UI.fDetailedRange;\n\tfloat fMouseX = (float)UI.nMouseX;\n\tfloat fMouseY = (float)UI.nMouseY;\n\n\tint nLineDrawn = -1;\n\n\tfor(uint32_t j = nContextSwitchStart; j != nContextSwitchEnd; j = (j+1) % MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE)\n\t{\n\t\tMP_ASSERT(j < MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE);\n\t\tMicroProfileContextSwitch CS = S.ContextSwitch[j];\n\n\t\tif(nTickIn == -1)\n\t\t{\n\t\t\tif(CS.nThreadIn == nThreadId)\n\t\t\t{\n\t\t\t\tnTickIn = CS.nTicks;\n\t\t\t\tnThreadBefore = CS.nThreadOut;\n\t\t\t}\n\t\t}\n\t\telse\n\t\t{\n\t\t\tif(CS.nThreadOut == nThreadId)\n\t\t\t{\n\t\t\t\tint64_t nTickOut = CS.nTicks;\n\t\t\t\tfloat fMsStart = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickIn);\n\t\t\t\tfloat fMsEnd = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickOut);\n\t\t\t\tif(fMsStart <= fMsEnd)\n\t\t\t\t{\n\t\t\t\t\tfloat fXStart = fMsStart * fMsToScreen;\n\t\t\t\t\tfloat fXEnd = fMsEnd * fMsToScreen;\n\t\t\t\t\tfloat fYStart = (float)nY;\n\t\t\t\t\tfloat fYEnd = fYStart + (MICROPROFILE_DETAILED_CONTEXT_SWITCH_HEIGHT);\n\t\t\t\t\tuint32_t nColor = g_nMicroProfileContextSwitchThreadColors[CS.nCpu%MICROPROFILE_NUM_CONTEXT_SWITCH_COLORS];\n\t\t\t\t\tfloat fXDist = MicroProfileMax(fXStart - fMouseX, fMouseX - fXEnd);\n\t\t\t\t\tbool bHover = fXDist < MICROPROFILE_HOVER_DIST && fYStart <= fMouseY && fMouseY <= fYEnd && nBaseY < fMouseY;\n\t\t\t\t\tif(bHover)\n\t\t\t\t\t{\n\t\t\t\t\t\tUI.nRangeBegin = nTickIn;\n\t\t\t\t\t\tUI.nRangeEnd = nTickOut;\n\t\t\t\t\t\tS.nContextSwitchHoverTickIn = nTickIn;\n\t\t\t\t\t\tS.nContextSwitchHoverTickOut = nTickOut;\n\t\t\t\t\t\tS.nContextSwitchHoverThread = CS.nThreadOut;\n\t\t\t\t\t\tS.nContextSwitchHoverThreadBefore = nThreadBefore;\n\t\t\t\t\t\tS.nContextSwitchHoverThreadAfter = CS.nThreadIn;\n\t\t\t\t\t\tS.nContextSwitchHoverCpuNext = CS.nCpu;\n\t\t\t\t\t\tnColor = UI.nHoverColor;\n\t\t\t\t\t}\n\t\t\t\t\tif(CS.nCpu == S.nContextSwitchHoverCpu)\n\t\t\t\t\t{\n\t\t\t\t\t\tnColor = UI.nHoverColorShared;\n\t\t\t\t\t}\n\n\t\t\t\t\tuint32_t nIntegerWidth = (uint32_t)(fXEnd - fXStart);\n\t\t\t\t\tif(nIntegerWidth)\n\t\t\t\t\t{\n\t\t\t\t\t\tMicroProfileDrawBox((int)fXStart, (int)fYStart, (int)fXEnd, (int)fYEnd, nColor|UI.nOpacityForeground, MicroProfileBoxTypeFlat);\n\t\t\t\t\t}\n\t\t\t\t\telse\n\t\t\t\t\t{\n\t\t\t\t\t\tfloat fXAvg = 0.5f * (fXStart + fXEnd);\n\t\t\t\t\t\tint nLineX = (int)floor(fXAvg+0.5f);\n\n\t\t\t\t\t\tif(nLineDrawn != nLineX)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tnLineDrawn = nLineX;\n\t\t\t\t\t\t\tMicroProfileDrawLineVertical(nLineX, (int)(fYStart + 0.5f), (int)(fYEnd + 0.5f), nColor|UI.nOpacityForeground);\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t\tnTickIn = -1;\n\t\t\t}\n\t\t}\n\t}\n}\n\nvoid MicroProfileWriteThreadHeader(uint32_t nY, MicroProfileThreadIdType ThreadId, const char* pNamedThread, const char* pThreadModule)\n{\n\tchar Buffer[512];\n\tint nStrLen = 0;\n\tif(pThreadModule)\n\t{\n\t\tnStrLen = snprintf(Buffer, sizeof(Buffer) - 1, \"%04x: %s [%s]\", (uint32_t)ThreadId, pNamedThread ? pNamedThread : \"\", pThreadModule);\n\t}\n\telse\n\t{\n\t\tnStrLen = snprintf(Buffer, sizeof(Buffer) - 1, \"%04x: %s\", (uint32_t)ThreadId, pNamedThread ? pNamedThread : \"\");\n\t}\n\tMicroProfileDrawTextBackground(10, nY, 0xffffff, 0x88777777, Buffer, nStrLen);\n}\n\nuint32_t MicroProfileWriteProcessHeader(uint32_t nY, uint32_t nProcessId)\n{\n\tchar Name[256];\n\tconst char* pProcessName = MicroProfileGetProcessName(nProcessId, Name, sizeof(Name));\n\n\tchar Buffer[512];\n\tnY += MICROPROFILE_TEXT_HEIGHT + 1;\n\tint nStrLen = 0;\n\tif(pProcessName)\n\t{\n\t\tnStrLen = snprintf(Buffer, sizeof(Buffer) - 1, \"* %04x: %s\", nProcessId, pProcessName);\n\t}\n\telse\n\t{\n\t\tnStrLen = snprintf(Buffer, sizeof(Buffer) - 1, \"* %04x\", nProcessId);\n\t}\n\tMicroProfileDrawTextBackground(0, nY, 0xffffff, 0x88777777, Buffer, nStrLen);\n\tnY += MICROPROFILE_TEXT_HEIGHT + 1;\n\treturn nY;\n}\n\nvoid MicroProfileGetFrameRange(int64_t nTicks, int64_t nTicksEnd, int32_t nLogIndex, uint32_t* nFrameBegin, uint32_t* nFrameEnd)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\n\tbool bGpu = (nLogIndex >= 0) ? S.Pool[nLogIndex]->nGpu != 0 : false;\n\tuint32_t nPut = (nLogIndex >= 0) ? S.Pool[nLogIndex]->nPut.load(std::memory_order_relaxed) : 0;\n\n\tuint32_t nBegin = S.nFrameCurrent;\n\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_FRAME_HISTORY - MICROPROFILE_GPU_FRAME_DELAY; ++i)\n\t{\n\t\tuint32_t nFrame = (S.nFrameCurrent + MICROPROFILE_MAX_FRAME_HISTORY - i) % MICROPROFILE_MAX_FRAME_HISTORY;\n\n\t\tif(nLogIndex >= 0)\n\t\t{\n\t\t\tuint32_t nCurrStart = S.Frames[nBegin].nLogStart[nLogIndex];\n\t\t\tuint32_t nPrevStart = S.Frames[nFrame].nLogStart[nLogIndex];\n\t\t\tbool bOverflow = (nPrevStart <= nCurrStart) ? (nPut >= nPrevStart && nPut < nCurrStart) : (nPut < nCurrStart || nPut >= nPrevStart);\n\t\t\tif(bOverflow)\n\t\t\t\tbreak;\n\t\t}\n\n\t\tnBegin = nFrame;\n\t\tif((bGpu ? S.Frames[nBegin].nFrameStartGpu : S.Frames[nBegin].nFrameStartCpu) <= nTicks)\n\t\t\tbreak;\n\t}\n\n\tuint32_t nEnd = nBegin;\n\n\twhile (nEnd != S.nFrameCurrent)\n\t{\n\t\tnEnd = (nEnd + 1) % MICROPROFILE_MAX_FRAME_HISTORY;\n\t\tif((bGpu ? S.Frames[nEnd].nFrameStartGpu : S.Frames[nEnd].nFrameStartCpu) >= nTicksEnd)\n\t\t\tbreak;\n\t}\n\n\t*nFrameBegin = nBegin;\n\t*nFrameEnd = nEnd;\n}\n\nvoid MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int nBaseY, int nSelectedFrame)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tMP_DEBUG_DUMP_RANGE();\n\tint nY = nBaseY - UI.nOffsetY[MP_DRAW_DETAILED];\n\tint64_t nNumBoxes = 0;\n\tint64_t nNumLines = 0;\n\n\tUI.nRangeBegin = 0; \n\tUI.nRangeEnd = 0;\n\tUI.nRangeBeginGpu = 0;\n\tUI.nRangeEndGpu = 0;\n\tUI.nRangeBeginIndex = UI.nRangeEndIndex = 0;\n\tUI.pRangeLog = 0;\n\n\tint64_t nFrameStartCpu = S.Frames[S.nFrameCurrent].nFrameStartCpu;\n\tint64_t nFrameStartGpu = S.Frames[S.nFrameCurrent].nFrameStartGpu;\n\tint64_t nTicksPerSecondCpu = MicroProfileTicksPerSecondCpu();\n\tint64_t nTicksPerSecondGpu = MicroProfileTicksPerSecondGpu();\n\tfloat fToMsCpu = MicroProfileTickToMsMultiplier(nTicksPerSecondCpu);\n\tfloat fToMsGpu = MicroProfileTickToMsMultiplier(nTicksPerSecondGpu);\n\n\tif(!S.nRunning && UI.nTickReferenceCpu < nFrameStartCpu)\n\t{\n\t\tint64_t nRefCpu = 0, nRefGpu = 0;\n\t\tif(MicroProfileGetGpuTickReference(&nRefCpu, &nRefGpu))\n\t\t{\n\t\t\tUI.nTickReferenceCpu = nRefCpu;\n\t\t\tUI.nTickReferenceGpu = nRefGpu;\n\t\t}\n\t}\n\n\tfloat fDetailedOffset = UI.fDetailedOffset;\n\tfloat fDetailedRange = UI.fDetailedRange;\n\n\tint64_t nDetailedOffsetTicksCpu = MicroProfileMsToTick(fDetailedOffset, MicroProfileTicksPerSecondCpu());\n\tint64_t nDetailedOffsetTicksGpu = MicroProfileMsToTick(fDetailedOffset, MicroProfileTicksPerSecondGpu());\n\tint64_t nBaseTicksCpu = nDetailedOffsetTicksCpu + nFrameStartCpu;\n\tint64_t nBaseTicksGpu = MicroProfileGetGpuTickSync(nBaseTicksCpu, nDetailedOffsetTicksGpu + nFrameStartGpu);\n\tint64_t nBaseTicksEndCpu = nBaseTicksCpu + MicroProfileMsToTick(fDetailedRange, MicroProfileTicksPerSecondCpu());\n\tint64_t nBaseTicksEndGpu = nBaseTicksGpu + MicroProfileMsToTick(fDetailedRange, MicroProfileTicksPerSecondGpu());\n\n\tuint32_t nFrameBegin, nFrameEnd;\n\tMicroProfileGetFrameRange(nBaseTicksCpu, nBaseTicksEndCpu, -1, &nFrameBegin, &nFrameEnd);\n\n\tfloat fMsBase = fToMsCpu * nDetailedOffsetTicksCpu;\n\tfloat fMs = fDetailedRange;\n\tfloat fMsEnd = fMs + fMsBase;\n\tfloat fWidth = (float)nWidth;\n\tfloat fMsToScreen = fWidth / fMs;\n\n\tfor(uint32_t i = nFrameBegin; i != nFrameEnd; i = (i+1) % MICROPROFILE_MAX_FRAME_HISTORY)\n\t{\n\t\tuint64_t nTickStart = S.Frames[i].nFrameStartCpu;\n\t\tfloat fMsStart = fToMsCpu * MicroProfileLogTickDifference(nBaseTicksCpu, nTickStart);\n\t\tfloat fXStart = fMsStart * fMsToScreen;\n\n\t\tMicroProfileDrawLineVertical((int)fXStart, nBaseY, nBaseY + nHeight, UI.nOpacityForeground | 0xbbbbbb);\n\t}\n\n\t{\n\t\tfloat fRate = floor(2*(log10(fMs)-1))/2;\n\t\tfloat fStep = powf(10.f, fRate);\n\t\tfloat fRcpStep = 1.f / fStep;\n\t\tint nColorIndex = (int)(floor(fMsBase*fRcpStep));\n\t\tfloat fStart = floor(fMsBase*fRcpStep) * fStep;\n\n\t\tchar StepLabel[64] = \"\";\n\t\tif(fStep >= 0.005 && fStep <= 1000)\n\t\t{\n\t\t\tif(fStep >= 1)\n\t\t\t\tsprintf(StepLabel, \"%.3gms\", fStep);\n\t\t\telse\n\t\t\t\tsprintf(StepLabel, \"%.2fms\", fStep);\n\t\t}\n\n\t\tuint32_t nStepLabelLength = (uint32_t)strlen(StepLabel);\n\t\tfloat fStepLabelOffset = (fStep*fMsToScreen-nStepLabelLength*(MICROPROFILE_TEXT_WIDTH+1))/2;\n\n\t\tfor(float f = fStart; f < fMsEnd; )\n\t\t{\n\t\t\tfloat fStart = f;\n\t\t\tfloat fNext = f + fStep;\n\t\t\tMicroProfileDrawBox((int)((fStart-fMsBase) * fMsToScreen), nBaseY, (int)((fNext-fMsBase) * fMsToScreen+1), nBaseY + nHeight, UI.nOpacityBackground | g_nMicroProfileBackColors[nColorIndex++ & 1]);\n\n\t\t\tif(nStepLabelLength)\n\t\t\t\tMicroProfileDrawText((int)((fStart-fMsBase) * fMsToScreen + fStepLabelOffset), nBaseY, UI.nOpacityForeground | 0x808080, StepLabel, nStepLabelLength);\n\n\t\t\tf = fNext;\n\t\t}\n\t}\n\n\tnY += MICROPROFILE_TEXT_HEIGHT+1;\n\tMicroProfileLogEntry* pMouseOver = UI.pDisplayMouseOver;\n\tMicroProfileLogEntry* pMouseOverNext = 0;\n\tuint64_t nMouseOverToken = pMouseOver ? MicroProfileLogTimerIndex(*pMouseOver) : MICROPROFILE_INVALID_TOKEN;\n\tfloat fMouseX = (float)UI.nMouseX;\n\tfloat fMouseY = (float)UI.nMouseY;\n\tuint64_t nHoverToken = MICROPROFILE_INVALID_TOKEN;\n\tint64_t nHoverTime = 0;\n\n\tstatic int nHoverCounter = 155;\n\tstatic int nHoverCounterDelta = 10;\n\tnHoverCounter += nHoverCounterDelta;\n\tif(nHoverCounter >= 245)\n\t\tnHoverCounterDelta = -10;\n\telse if(nHoverCounter < 100)\n\t\tnHoverCounterDelta = 10;\n\tUI.nHoverColor = (nHoverCounter<<24)|(nHoverCounter<<16)|(nHoverCounter<<8)|nHoverCounter;\n\tuint32_t nHoverCounterShared = nHoverCounter>>2;\n\tUI.nHoverColorShared = (nHoverCounterShared<<24)|(nHoverCounterShared<<16)|(nHoverCounterShared<<8)|nHoverCounterShared;\n\n\tuint32_t nLinesDrawn[MICROPROFILE_STACK_MAX]={0};\n\n\tS.nContextSwitchHoverThread = S.nContextSwitchHoverThreadAfter = S.nContextSwitchHoverThreadBefore = -1;\n\n\tuint32_t nContextSwitchStart = -1;\n\tuint32_t nContextSwitchEnd = -1;\n\tS.nContextSwitchHoverCpuNext = 0xff;\n\tS.nContextSwitchHoverTickIn = -1;\n\tS.nContextSwitchHoverTickOut = -1;\n\tif(S.bContextSwitchRunning)\n\t{\n\t\tMicroProfileContextSwitchSearch(&nContextSwitchStart, &nContextSwitchEnd, nBaseTicksCpu, nBaseTicksEndCpu);\n\t}\n\n\tuint64_t nActiveGroup = S.nAllGroupsWanted ? S.nGroupMask : S.nActiveGroupWanted;\n\n\tbool bSkipBarView = S.bContextSwitchRunning && S.bContextSwitchNoBars;\n\n\tif(!bSkipBarView)\n\t{\n\t\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS; ++i)\n\t\t{\n\t\t\tMicroProfileThreadLog* pLog = S.Pool[i];\n\t\t\tif(!pLog)\n\t\t\t\tcontinue;\n\n\t\t\tbool bGpu = pLog->nGpu != 0;\n\t\t\tfloat fToMs = bGpu ? fToMsGpu : fToMsCpu;\n\t\t\tint64_t nBaseTicks = bGpu ? nBaseTicksGpu : nBaseTicksCpu;\n\t\t\tint64_t nBaseTicksEnd = bGpu ? nBaseTicksEndGpu : nBaseTicksEndCpu;\n\t\t\tMicroProfileThreadIdType nThreadId = pLog->nThreadId;\n\n\t\t\tint64_t nGapTime = (bGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu()) * MICROPROFILE_GAP_TIME / 1000;\n\n\t\t\tuint32_t nLogFrameBegin, nLogFrameEnd;\n\t\t\tMicroProfileGetFrameRange(nBaseTicks - nGapTime, nBaseTicksEnd + nGapTime, i, &nLogFrameBegin, &nLogFrameEnd);\n\n\t\t\tuint32_t nGet = S.Frames[nLogFrameBegin].nLogStart[i];\n\t\t\tuint32_t nPut = nLogFrameEnd == S.nFrameCurrent ? pLog->nPut.load(std::memory_order_relaxed) : S.Frames[nLogFrameEnd].nLogStart[i];\n\t\t\tif(nPut == nGet)\n\t\t\t\tcontinue;\n\n\t\t\tuint32_t nRange[2][2] = { {0, 0}, {0, 0}, };\n\t\t\tMicroProfileGetRange(nPut, nGet, nRange);\n\n\t\t\tuint32_t nMaxStackDepth = 0;\n\n\t\t\tnY += 3;\n\t\t\tMicroProfileWriteThreadHeader(nY, nThreadId, &pLog->ThreadName[0], nullptr);\n\t\t\tnY += 3;\n\t\t\tnY += MICROPROFILE_TEXT_HEIGHT + 1;\n\n\t\t\tif(S.bContextSwitchRunning)\n\t\t\t{\n\t\t\t\tMicroProfileDrawDetailedContextSwitchBars(nY, pLog->nThreadId, nContextSwitchStart, nContextSwitchEnd, nBaseTicks, nBaseY);\n\t\t\t\tnY -= MICROPROFILE_DETAILED_BAR_HEIGHT;\n\t\t\t\tnY += MICROPROFILE_DETAILED_CONTEXT_SWITCH_HEIGHT+1;\n\t\t\t}\n\n\t\t\tuint32_t nYDelta = MICROPROFILE_DETAILED_BAR_HEIGHT;\n\t\t\tuint32_t nStack[MICROPROFILE_STACK_MAX];\n\t\t\tuint32_t nStackPos = 0;\n\t\t\tfor(uint32_t j = 0; j < 2; ++j)\n\t\t\t{\n\t\t\t\tuint32_t nStart = nRange[j][0];\n\t\t\t\tuint32_t nEnd = nRange[j][1];\n\t\t\t\tfor(uint32_t k = nStart; k < nEnd; ++k)\n\t\t\t\t{\n\t\t\t\t\tMicroProfileLogEntry* pEntry = pLog->Log + k;\n\t\t\t\t\tuint64_t nType = MicroProfileLogType(*pEntry);\n\t\t\t\t\tif(MP_LOG_ENTER == nType)\n\t\t\t\t\t{\n\t\t\t\t\t\tMP_ASSERT(nStackPos < MICROPROFILE_STACK_MAX);\n\t\t\t\t\t\tnStack[nStackPos++] = k;\n\t\t\t\t\t}\n\t\t\t\t\telse if(MP_LOG_META == nType)\n\t\t\t\t\t{\n\n\t\t\t\t\t}\n\t\t\t\t\telse if(MP_LOG_LEAVE == nType)\n\t\t\t\t\t{\n\t\t\t\t\t\tif(0 == nStackPos)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tcontinue;\n\t\t\t\t\t\t}\n\n\t\t\t\t\t\tMicroProfileLogEntry* pEntryEnter = pLog->Log + nStack[nStackPos-1];\n\t\t\t\t\t\tif(MicroProfileLogTimerIndex(*pEntryEnter) != MicroProfileLogTimerIndex(*pEntry))\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\t//uprintf(\"mismatch %llx %llx\\n\", pEntryEnter->nToken, pEntry->nToken);\n\t\t\t\t\t\t\tcontinue;\n\t\t\t\t\t\t}\n\t\t\t\t\t\tint64_t nTickStart = MicroProfileLogGetTick(*pEntryEnter);\n\t\t\t\t\t\tint64_t nTickEnd = MicroProfileLogGetTick(*pEntry);\n\t\t\t\t\t\tuint64_t nTimerIndex = MicroProfileLogTimerIndex(*pEntry);\n\t\t\t\t\t\tuint32_t nColor = S.TimerInfo[nTimerIndex].nColor;\n\t\t\t\t\t\tif(!(nActiveGroup & (1ull << S.TimerInfo[nTimerIndex].nGroupIndex)))\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tnStackPos--;\n\t\t\t\t\t\t\tcontinue;\n\t\t\t\t\t\t}\n\t\t\t\t\t\tif(nMouseOverToken == nTimerIndex)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tif(pEntry == pMouseOver)\n\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\tnColor = UI.nHoverColor;\n\t\t\t\t\t\t\t\tif(bGpu)\n\t\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\t\tUI.nRangeBeginGpu = *pEntryEnter;\n\t\t\t\t\t\t\t\t\tUI.nRangeEndGpu = *pEntry;\n\t\t\t\t\t\t\t\t\tuint32_t nCpuBegin = (nStack[nStackPos-1] + 1) % MICROPROFILE_BUFFER_SIZE;\n\t\t\t\t\t\t\t\t\tuint32_t nCpuEnd = (k + 1) % MICROPROFILE_BUFFER_SIZE;\n\t\t\t\t\t\t\t\t\tMicroProfileLogEntry LogCpuBegin = pLog->Log[nCpuBegin];\n\t\t\t\t\t\t\t\t\tMicroProfileLogEntry LogCpuEnd = pLog->Log[nCpuEnd];\n\t\t\t\t\t\t\t\t\tif(MicroProfileLogType(LogCpuBegin) == MP_LOG_GPU_EXTRA && MicroProfileLogType(LogCpuEnd) == MP_LOG_GPU_EXTRA)\n\t\t\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\t\t\tUI.nRangeBegin = LogCpuBegin;\n\t\t\t\t\t\t\t\t\t\tUI.nRangeEnd = LogCpuEnd;\n\t\t\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\t\t\tUI.nRangeBeginIndex = nStack[nStackPos-1];\n\t\t\t\t\t\t\t\t\tUI.nRangeEndIndex = k;\n\t\t\t\t\t\t\t\t\tUI.pRangeLog = pLog;\n\t\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\t\telse\n\t\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\t\tUI.nRangeBegin = *pEntryEnter;\n\t\t\t\t\t\t\t\t\tUI.nRangeEnd = *pEntry;\n\t\t\t\t\t\t\t\t\tUI.nRangeBeginIndex = nStack[nStackPos-1];\n\t\t\t\t\t\t\t\t\tUI.nRangeEndIndex = k;\n\t\t\t\t\t\t\t\t\tUI.pRangeLog = pLog;\n\n\t\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\telse\n\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\tnColor = UI.nHoverColorShared;\n\t\t\t\t\t\t\t}\n\t\t\t\t\t\t}\n\n\t\t\t\t\t\tconst char* pName = S.TimerInfo[nTimerIndex].pName;\n\t\t\t\t\t\tuint32_t nNameLen = S.TimerInfo[nTimerIndex].nNameLen;\n\n\t\t\t\t\t\tif (pName[0] == '$' && pEntryEnter < pEntry && MicroProfileLogType(pEntryEnter[1 + bGpu]) == MP_LOG_LABEL)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tconst char* pLabel = MicroProfileGetLabel(MicroProfileLogGetTick(pEntryEnter[1 + bGpu]));\n\n\t\t\t\t\t\t\tif (pLabel)\n\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\tpName = pLabel;\n\t\t\t\t\t\t\t\tnNameLen = (uint32_t)strlen(pLabel);\n\t\t\t\t\t\t\t}\n\t\t\t\t\t\t}\n\n\t\t\t\t\t\tnMaxStackDepth = MicroProfileMax(nMaxStackDepth, nStackPos);\n\t\t\t\t\t\tfloat fMsStart = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickStart);\n\t\t\t\t\t\tfloat fMsEnd = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickEnd);\n\t\t\t\t\t\tfloat fXStart = fMsStart * fMsToScreen;\n\t\t\t\t\t\tfloat fXEnd = fMsEnd * fMsToScreen;\n\t\t\t\t\t\tfloat fYStart = (float)(nY + nStackPos * nYDelta);\n\t\t\t\t\t\tfloat fYEnd = fYStart + (MICROPROFILE_DETAILED_BAR_HEIGHT);\n\t\t\t\t\t\tfloat fXDist = MicroProfileMax(fXStart - fMouseX, fMouseX - fXEnd);\n\t\t\t\t\t\tbool bHover = fXDist < MICROPROFILE_HOVER_DIST && fYStart <= fMouseY && fMouseY <= fYEnd && nBaseY < fMouseY;\n\t\t\t\t\t\tuint32_t nIntegerWidth = (uint32_t)(fXEnd - fXStart);\n\t\t\t\t\t\tif(nIntegerWidth)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tif(bHover && UI.nActiveMenu == (uint32_t)-1)\n\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\tnHoverToken = MicroProfileLogTimerIndex(*pEntry);\n\t#if MICROPROFILE_DEBUG\n\t\t\t\t\t\t\t\tUI.nHoverAddressEnter = (uint64_t)pEntryEnter;\n\t\t\t\t\t\t\t\tUI.nHoverAddressLeave = (uint64_t)pEntry;\n\t#endif\n\t\t\t\t\t\t\t\tnHoverTime = MicroProfileLogTickDifference(nTickStart, nTickEnd);\n\t\t\t\t\t\t\t\tpMouseOverNext = pEntry;\n\t\t\t\t\t\t\t}\n\n\t\t\t\t\t\t\tMicroProfileDrawBox((int)fXStart, (int)fYStart, (int)fXEnd, (int)fYEnd, nColor|UI.nOpacityForeground, MicroProfileBoxTypeBar);\n#if MICROPROFILE_DETAILED_BAR_NAMES\n\t\t\t\t\t\t\tif(nIntegerWidth>3*MICROPROFILE_TEXT_WIDTH)\n\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\tfloat fXStartText = MicroProfileMax(fXStart, 0.f);\n\t\t\t\t\t\t\t\tint nTextWidth = (int)(fXEnd - fXStartText);\n\t\t\t\t\t\t\t\tint nCharacters = (nTextWidth - MICROPROFILE_TEXT_WIDTH) / (MICROPROFILE_TEXT_WIDTH+1);\n\t\t\t\t\t\t\t\tif(nCharacters>0)\n\t\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\t\tMicroProfileDrawText((int)(fXStartText+1), (int)(fYStart+1), -1, pName, MicroProfileMin(nNameLen, nCharacters));\n\t\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\t}\n#endif\n\t\t\t\t\t\t\t++nNumBoxes;\n\t\t\t\t\t\t}\n\t\t\t\t\t\telse\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tfloat fXAvg = 0.5f * (fXStart + fXEnd);\n\t\t\t\t\t\t\tint nLineX = (int)floor(fXAvg+0.5f);\n\t\t\t\t\t\t\tif(nLineX != (int)nLinesDrawn[nStackPos])\n\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\tif(bHover && UI.nActiveMenu == (uint32_t)-1)\n\t\t\t\t\t\t\t\t{\n\t\t\t\t\t\t\t\t\tnHoverToken = (uint32_t)MicroProfileLogTimerIndex(*pEntry);\n\t\t\t\t\t\t\t\t\tnHoverTime = MicroProfileLogTickDifference(nTickStart, nTickEnd);\n\t\t\t\t\t\t\t\t\tpMouseOverNext = pEntry;\n\t\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\t\tnLinesDrawn[nStackPos] = nLineX;\n\t\t\t\t\t\t\t\tMicroProfileDrawLineVertical(nLineX, (int)(fYStart + 0.5f), (int)(fYEnd + 0.5f), nColor|UI.nOpacityForeground);\n\t\t\t\t\t\t\t\t++nNumLines;\n\t\t\t\t\t\t\t}\n\t\t\t\t\t\t}\n\t\t\t\t\t\tnStackPos--;\n\n\t\t\t\t\t\tif(0 == nStackPos && MicroProfileLogTickDifference(nTickEnd, nBaseTicksEnd) < 0)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tbreak;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\t\t\tnY += nMaxStackDepth * nYDelta + MICROPROFILE_DETAILED_BAR_HEIGHT+1;\n\t\t}\n\t}\n\tif(S.bContextSwitchRunning && (S.bContextSwitchAllThreads||S.bContextSwitchNoBars))\n\t{\n\t\tuint32_t nContextSwitchSearchEnd = S.bContextSwitchAllThreads ? nContextSwitchEnd : nContextSwitchStart;\n\n\t\tMicroProfileThreadInfo Threads[MICROPROFILE_MAX_CONTEXT_SWITCH_THREADS];\n\t\tuint32_t nNumThreadsBase = 0;\n\t\tuint32_t nNumThreads = MicroProfileContextSwitchGatherThreads(nContextSwitchStart, nContextSwitchSearchEnd, Threads, &nNumThreadsBase);\n\n\t\tstd::sort(&Threads[nNumThreadsBase], &Threads[nNumThreads],\n\t\t\t\t\t[](const MicroProfileThreadInfo& l, const MicroProfileThreadInfo& r)\n\t\t{\n\t\t\treturn l.nProcessId == r.nProcessId ? l.nThreadId < r.nThreadId : l.nProcessId > r.nProcessId;\n\t\t});\n\n\t\tuint32_t nStart = nNumThreadsBase;\n\t\tif(S.bContextSwitchNoBars)\n\t\t\tnStart = 0;\n\t\tMicroProfileProcessIdType nLastProcessId = MP_GETCURRENTPROCESSID();\n\t\tfor(uint32_t i = nStart; i < nNumThreads; ++i)\n\t\t{\n\t\t\tMicroProfileThreadInfo tt = Threads[i];\n\t\t\tif(tt.nThreadId)\n\t\t\t{\n\t\t\t\tif (nLastProcessId != tt.nProcessId)\n\t\t\t\t{\n\t\t\t\t\tnY = MicroProfileWriteProcessHeader(nY, (uint32_t)tt.nProcessId);\n\t\t\t\t\tnLastProcessId = tt.nProcessId;\n\t\t\t\t}\n\n\t\t\t\tMicroProfileDrawDetailedContextSwitchBars(nY + 2, tt.nThreadId, nContextSwitchStart, nContextSwitchEnd, nBaseTicksCpu, nBaseY);\n\n\t\t\t\tMicroProfileWriteThreadHeader(nY, tt.nThreadId, i < nNumThreadsBase ? &S.Pool[i]->ThreadName[0] : nullptr, nullptr);\n\t\t\t\tnY += MICROPROFILE_TEXT_HEIGHT + 1;\n\t\t\t}\n\t\t}\n\t}\n\n\tS.nContextSwitchHoverCpu = S.nContextSwitchHoverCpuNext;\n\n\n\n\n\tUI.pDisplayMouseOver = pMouseOverNext;\n\n\tif(!S.nRunning)\n\t{\n\t\tif(nHoverToken != MICROPROFILE_INVALID_TOKEN && nHoverTime)\n\t\t{\n\t\t\tUI.nHoverToken = nHoverToken;\n\t\t\tUI.nHoverTime = nHoverTime;\n\t\t}\n\n\t\tif(nSelectedFrame != -1)\n\t\t{\n\t\t\tUI.nRangeBegin = S.Frames[nSelectedFrame].nFrameStartCpu;\n\t\t\tUI.nRangeEnd = S.Frames[(nSelectedFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY].nFrameStartCpu;\n\t\t\tUI.nRangeBeginGpu = S.Frames[nSelectedFrame].nFrameStartGpu;\n\t\t\tUI.nRangeEndGpu = S.Frames[(nSelectedFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY].nFrameStartGpu;\n\t\t}\n\t\tif(UI.nRangeBegin != UI.nRangeEnd)\n\t\t{\n\t\t\tfloat fMsStart = fToMsCpu * MicroProfileLogTickDifference(nBaseTicksCpu, UI.nRangeBegin);\n\t\t\tfloat fMsEnd = fToMsCpu * MicroProfileLogTickDifference(nBaseTicksCpu, UI.nRangeEnd);\n\t\t\tfloat fXStart = fMsStart * fMsToScreen;\n\t\t\tfloat fXEnd = fMsEnd * fMsToScreen;\t\n\t\t\tMicroProfileDrawBox((int)fXStart, nBaseY, (int)fXEnd, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT, MicroProfileBoxTypeFlat);\n\t\t\tMicroProfileDrawLineVertical((int)fXStart, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT | 0x44000000);\n\t\t\tMicroProfileDrawLineVertical((int)fXEnd, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT | 0x44000000);\n\n\t\t\tfMsStart += fDetailedOffset;\n\t\t\tfMsEnd += fDetailedOffset;\n\t\t\tchar sBuffer[32];\n\t\t\tuint32_t nLenStart = snprintf(sBuffer, sizeof(sBuffer)-1, \"%.2fms\", fMsStart);\n\t\t\tfloat fStartTextWidth = (float)((1+MICROPROFILE_TEXT_WIDTH) * nLenStart);\n\t\t\tfloat fStartTextX = fXStart - fStartTextWidth - 2;\n\t\t\tMicroProfileDrawBox((int)fStartTextX, nBaseY, (int)(fStartTextX + fStartTextWidth + 2), MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat);\n\t\t\tMicroProfileDrawText((int)fStartTextX+1, nBaseY, (uint32_t)-1, sBuffer, nLenStart);\n\t\t\tuint32_t nLenEnd = snprintf(sBuffer, sizeof(sBuffer)-1, \"%.2fms\", fMsEnd);\n\t\t\tMicroProfileDrawBox((int)(fXEnd+1), nBaseY, (int)(fXEnd+1+(1+MICROPROFILE_TEXT_WIDTH) * nLenEnd + 3), MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat);\n\t\t\tMicroProfileDrawText((int)(fXEnd+2), nBaseY+1, (uint32_t)-1, sBuffer, nLenEnd);\n\n\t\t\tif(UI.nMouseRight)\n\t\t\t{\n\t\t\t\tMicroProfileZoomTo(UI.nRangeBegin, UI.nRangeEnd, MicroProfileTokenTypeCpu);\n\t\t\t}\n\t\t}\n\n\t\tif(UI.nRangeBeginGpu != UI.nRangeEndGpu)\n\t\t{\n\t\t\tfloat fMsStart = fToMsGpu * MicroProfileLogTickDifference(nBaseTicksGpu, UI.nRangeBeginGpu);\n\t\t\tfloat fMsEnd = fToMsGpu * MicroProfileLogTickDifference(nBaseTicksGpu, UI.nRangeEndGpu);\n\t\t\tfloat fXStart = fMsStart * fMsToScreen;\n\t\t\tfloat fXEnd = fMsEnd * fMsToScreen;\t\n\t\t\tMicroProfileDrawBox((int)fXStart, nBaseY, (int)fXEnd, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU, MicroProfileBoxTypeFlat);\n\t\t\tMicroProfileDrawLineVertical((int)fXStart, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU | 0x44000000);\n\t\t\tMicroProfileDrawLineVertical((int)fXEnd, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU | 0x44000000);\n\n\t\t\tnBaseY += MICROPROFILE_TEXT_HEIGHT+1;\n\n\t\t\tfMsStart += fDetailedOffset;\n\t\t\tfMsEnd += fDetailedOffset;\n\t\t\tchar sBuffer[32];\n\t\t\tuint32_t nLenStart = snprintf(sBuffer, sizeof(sBuffer)-1, \"%.2fms\", fMsStart);\n\t\t\tfloat fStartTextWidth = (float)((1+MICROPROFILE_TEXT_WIDTH) * nLenStart);\n\t\t\tfloat fStartTextX = fXStart - fStartTextWidth - 2;\n\t\t\tMicroProfileDrawBox((int)fStartTextX, nBaseY, (int)(fStartTextX + fStartTextWidth + 2), MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat);\n\t\t\tMicroProfileDrawText((int)(fStartTextX+1), nBaseY, (uint32_t)-1, sBuffer, nLenStart);\n\t\t\tuint32_t nLenEnd = snprintf(sBuffer, sizeof(sBuffer)-1, \"%.2fms\", fMsEnd);\n\t\t\tMicroProfileDrawBox((int)(fXEnd+1), nBaseY, (int)(fXEnd+1+(1+MICROPROFILE_TEXT_WIDTH) * nLenEnd + 3), MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat);\n\t\t\tMicroProfileDrawText((int)(fXEnd+2), nBaseY+1, (uint32_t)-1, sBuffer, nLenEnd);\n\n\t\t\tif(UI.nMouseRight)\n\t\t\t{\n\t\t\t\tMicroProfileZoomTo(UI.nRangeBeginGpu, UI.nRangeEndGpu, MicroProfileTokenTypeGpu);\n\t\t\t}\n\t\t}\n\t}\n}\n\n\nvoid MicroProfileDrawDetailedFrameHistory(uint32_t nWidth, uint32_t nHeight, uint32_t nBaseY, uint32_t nSelectedFrame)\n{\n\t(void)nHeight;\n\n\tMicroProfile& S = *MicroProfileGet();\n\n\tconst uint32_t nBarHeight = MICROPROFILE_FRAME_HISTORY_HEIGHT;\n\tfloat fBaseX = (float)nWidth;\n\tfloat fDx = fBaseX / MICROPROFILE_NUM_FRAMES;\n\n\tuint32_t nLastIndex = (S.nFrameCurrent+1) % MICROPROFILE_MAX_FRAME_HISTORY;\n\tMicroProfileDrawBox(0, nBaseY, nWidth, nBaseY+MICROPROFILE_FRAME_HISTORY_HEIGHT, 0xff000000 | g_nMicroProfileBackColors[0], MicroProfileBoxTypeFlat);\n\tfloat fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()) * S.fRcpReferenceTime;\n\tfloat fToMsGpu = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondGpu()) * S.fRcpReferenceTime;\n\n\t\n\tMicroProfileFrameState* pFrameCurrent = &S.Frames[S.nFrameCurrent];\n\tuint64_t nFrameStartCpu = pFrameCurrent->nFrameStartCpu;\n\tint64_t nDetailedOffsetTicksCpu = MicroProfileMsToTick(UI.fDetailedOffset, MicroProfileTicksPerSecondCpu());\n\tint64_t nCpuStart = nDetailedOffsetTicksCpu + nFrameStartCpu;\n\tint64_t nCpuEnd = nCpuStart + MicroProfileMsToTick(UI.fDetailedRange, MicroProfileTicksPerSecondCpu());;\n\n\n\tfloat fSelectionStart = (float)nWidth;\n\tfloat fSelectionEnd = 0.f;\n\tfor(uint32_t i = 0; i < MICROPROFILE_NUM_FRAMES; ++i)\n\t{\n\t\tuint32_t nIndex = (S.nFrameCurrent + MICROPROFILE_MAX_FRAME_HISTORY - i) % MICROPROFILE_MAX_FRAME_HISTORY;\n\t\tMicroProfileFrameState* pCurrent = &S.Frames[nIndex];\n\t\tMicroProfileFrameState* pNext = &S.Frames[nLastIndex];\n\t\t\n\t\tint64_t nTicks = pNext->nFrameStartCpu - pCurrent->nFrameStartCpu;\n\t\tint64_t nTicksGpu = pNext->nFrameStartGpu - pCurrent->nFrameStartGpu;\n\t\tfloat fScale = fToMs * nTicks;\n\t\tfloat fScaleGpu = fToMsGpu * nTicksGpu;\n\t\tfScale = fScale > 1.f ? 0.f : 1.f - fScale;\n\t\tfScaleGpu = fScaleGpu > 1.f ? 0.f : 1.f - fScaleGpu;\n\t\tfloat fXEnd = fBaseX;\n\t\tfloat fXStart = fBaseX - fDx;\n\t\tfBaseX = fXStart;\n\t\tuint32_t nColor = MICROPROFILE_FRAME_HISTORY_COLOR_CPU;\n\t\tif(nIndex == nSelectedFrame)\n\t\t\tnColor = (uint32_t)-1;\n\t\tMicroProfileDrawBox((int)fXStart, (int)(nBaseY + fScale * nBarHeight), (int)fXEnd, nBaseY+MICROPROFILE_FRAME_HISTORY_HEIGHT, nColor, MicroProfileBoxTypeBar);\n\t\tif(pNext->nFrameStartCpu > nCpuStart)\n\t\t{\n\t\t\tfSelectionStart = fXStart;\n\t\t}\n\t\tif(pCurrent->nFrameStartCpu < nCpuEnd && fSelectionEnd == 0.f)\n\t\t{\n\t\t\tfSelectionEnd = fXEnd;\n\t\t}\n\t\tnLastIndex = nIndex;\n\t}\n\tMicroProfileDrawBox((int)fSelectionStart, nBaseY, (int)fSelectionEnd, nBaseY+MICROPROFILE_FRAME_HISTORY_HEIGHT, MICROPROFILE_FRAME_HISTORY_COLOR_HIGHTLIGHT, MicroProfileBoxTypeFlat);\n}\nvoid MicroProfileDrawDetailedView(uint32_t nWidth, uint32_t nHeight, bool bDrawBars)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\n\tMICROPROFILE_SCOPE(g_MicroProfileDetailed);\n\tuint32_t nBaseY = MICROPROFILE_TEXT_HEIGHT + 1;\n\n\tint nSelectedFrame = -1;\n\tif(UI.nMouseY > nBaseY && UI.nMouseY <= nBaseY + MICROPROFILE_FRAME_HISTORY_HEIGHT && UI.nActiveMenu == (uint32_t)-1)\n\t{\n\n\t\tnSelectedFrame = ((MICROPROFILE_NUM_FRAMES) * (UI.nWidth-UI.nMouseX) / UI.nWidth);\n\t\tnSelectedFrame = (S.nFrameCurrent + MICROPROFILE_MAX_FRAME_HISTORY - nSelectedFrame) % MICROPROFILE_MAX_FRAME_HISTORY;\n\t\tUI.nHoverFrame = nSelectedFrame;\n\t\tif(UI.nMouseRight)\n\t\t{\n\t\t\tint64_t nRangeBegin = S.Frames[nSelectedFrame].nFrameStartCpu;\n\t\t\tint64_t nRangeEnd = S.Frames[(nSelectedFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY].nFrameStartCpu;\n\t\t\tMicroProfileZoomTo(nRangeBegin, nRangeEnd, MicroProfileTokenTypeCpu);\n\t\t}\n\t\tif(UI.nMouseDownLeft)\n\t\t{\n\t\t\tuint64_t nFrac = (1024 * (MICROPROFILE_NUM_FRAMES) * (UI.nMouseX) / UI.nWidth) % 1024;\n\t\t\tint64_t nRangeBegin = S.Frames[nSelectedFrame].nFrameStartCpu;\n\t\t\tint64_t nRangeEnd = S.Frames[(nSelectedFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY].nFrameStartCpu;\n\t\t\tMicroProfileCenter(nRangeBegin + (nRangeEnd-nRangeBegin) * nFrac / 1024);\n\t\t}\n\t}\n\telse\n\t{\n\t\tUI.nHoverFrame = -1;\n\t}\n\n\tif (bDrawBars)\n\t{\n\t\tMicroProfileDrawDetailedBars(nWidth, nHeight, nBaseY + MICROPROFILE_FRAME_HISTORY_HEIGHT, nSelectedFrame);\n\t}\n\n\tMicroProfileDrawDetailedFrameHistory(nWidth, nHeight, nBaseY, nSelectedFrame);\n}\nvoid MicroProfileDrawTextRight(uint32_t nX, uint32_t nY, uint32_t nColor, const char* pStr, uint32_t nStrLen)\n{\n\tMicroProfileDrawText(nX - nStrLen * (MICROPROFILE_TEXT_WIDTH+1), nY, nColor, pStr, nStrLen);\n}\nvoid MicroProfileDrawHeader(int32_t nX, uint32_t nWidth, const char* pName)\n{\n\tif(pName)\n\t{\n\t\tMicroProfileDrawBox(nX-8, MICROPROFILE_TEXT_HEIGHT + 2, nX + nWidth+5, MICROPROFILE_TEXT_HEIGHT + 2 + (MICROPROFILE_TEXT_HEIGHT+1), 0xff000000|g_nMicroProfileBackColors[1]);\n\t\tMicroProfileDrawText(nX, MICROPROFILE_TEXT_HEIGHT + 2, (uint32_t)-1, pName, (uint32_t)strlen(pName));\n\t}\n}\n\n\ntypedef void (*MicroProfileLoopGroupCallback)(uint32_t nTimer, uint32_t nIdx, uint32_t nX, uint32_t nY, void* pData);\n\nvoid MicroProfileLoopActiveGroupsDraw(int32_t nX, int32_t nY, MicroProfileLoopGroupCallback CB, void* pData)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tnY += MICROPROFILE_TEXT_HEIGHT + 2;\n\tuint64_t nGroup = S.nAllGroupsWanted ? S.nGroupMask : S.nActiveGroupWanted;\n\tuint32_t nCount = 0;\n\tfor(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j)\n\t{\n\t\tuint64_t nMask = 1ll << j;\n\t\tif(nMask & nGroup)\n\t\t{\n\t\t\tnY += MICROPROFILE_TEXT_HEIGHT + 1;\n\t\t\tfor(uint32_t i = 0; i < S.nTotalTimers;++i)\n\t\t\t{\n\t\t\t\tuint64_t nTokenMask = MicroProfileGetGroupMask(S.TimerInfo[i].nToken);\n\t\t\t\tif(nTokenMask & nMask)\n\t\t\t\t{\n\t\t\t\t\tif(nY >= 0)\n\t\t\t\t\t\tCB(i, nCount, nX, nY, pData);\n\t\t\t\t\t\n\t\t\t\t\tnCount += 2;\n\t\t\t\t\tnY += MICROPROFILE_TEXT_HEIGHT + 1;\n\n\t\t\t\t\tif(nY > (int)UI.nHeight)\n\t\t\t\t\t\treturn;\n\t\t\t\t}\n\t\t\t}\n\t\t\t\n\t\t}\n\t}\n}\n\n\nvoid MicroProfileCalcTimers(float* pTimers, float* pAverage, float* pMax, float* pMin, float* pCallAverage, float* pExclusive, float* pAverageExclusive, float* pMaxExclusive, uint64_t nGroup, uint32_t nSize)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\n\tuint32_t nCount = 0;\n\tuint64_t nMask = 1;\n\n\tfor(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j)\n\t{\n\t\tif(nMask & nGroup)\n\t\t{\n\t\t\tconst float fToMs = MicroProfileTickToMsMultiplier(S.GroupInfo[j].Type == MicroProfileTokenTypeGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu());\n\t\t\tfor(uint32_t i = 0; i < S.nTotalTimers;++i)\n\t\t\t{\n\t\t\t\tuint64_t nTokenMask = MicroProfileGetGroupMask(S.TimerInfo[i].nToken);\n\t\t\t\tif(nTokenMask & nMask)\n\t\t\t\t{\n\t\t\t\t\tMP_ASSERT(nCount + 2 <= nSize);\n\t\t\t\t\t{\n\t\t\t\t\t\tuint32_t nTimer = i;\n\t\t\t\t\t\tuint32_t nIdx = nCount;\n\t\t\t\t\t\tuint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1;\n\t\t\t\t\t\tuint32_t nAggregateCount = S.Aggregate[nTimer].nCount ? S.Aggregate[nTimer].nCount : 1;\n\t\t\t\t\t\tfloat fToPrc = S.fRcpReferenceTime;\n\t\t\t\t\t\tfloat fMs = fToMs * (S.Frame[nTimer].nTicks);\n\t\t\t\t\t\tfloat fPrc = MicroProfileMin(fMs * fToPrc, 1.f);\n\t\t\t\t\t\tfloat fAverageMs = fToMs * (S.Aggregate[nTimer].nTicks / nAggregateFrames);\n\t\t\t\t\t\tfloat fAveragePrc = MicroProfileMin(fAverageMs * fToPrc, 1.f);\n\t\t\t\t\t\tfloat fMaxMs = fToMs * (S.AggregateMax[nTimer]);\n\t\t\t\t\t\tfloat fMaxPrc = MicroProfileMin(fMaxMs * fToPrc, 1.f);\n\t\t\t\t\t\tfloat fMinMs = fToMs * (S.AggregateMin[nTimer] != uint64_t(-1) ? S.AggregateMin[nTimer] : 0);\n\t\t\t\t\t\tfloat fMinPrc = MicroProfileMin(fMinMs * fToPrc, 1.f);\n\t\t\t\t\t\tfloat fCallAverageMs = fToMs * (S.Aggregate[nTimer].nTicks / nAggregateCount);\n\t\t\t\t\t\tfloat fCallAveragePrc = MicroProfileMin(fCallAverageMs * fToPrc, 1.f);\n\t\t\t\t\t\tfloat fMsExclusive = fToMs * (S.FrameExclusive[nTimer]);\n\t\t\t\t\t\tfloat fPrcExclusive = MicroProfileMin(fMsExclusive * fToPrc, 1.f);\n\t\t\t\t\t\tfloat fAverageMsExclusive = fToMs * (S.AggregateExclusive[nTimer] / nAggregateFrames);\n\t\t\t\t\t\tfloat fAveragePrcExclusive = MicroProfileMin(fAverageMsExclusive * fToPrc, 1.f);\n\t\t\t\t\t\tfloat fMaxMsExclusive = fToMs * (S.AggregateMaxExclusive[nTimer]);\n\t\t\t\t\t\tfloat fMaxPrcExclusive = MicroProfileMin(fMaxMsExclusive * fToPrc, 1.f);\n\t\t\t\t\t\tpTimers[nIdx] = fMs;\n\t\t\t\t\t\tpTimers[nIdx+1] = fPrc;\n\t\t\t\t\t\tpAverage[nIdx] = fAverageMs;\n\t\t\t\t\t\tpAverage[nIdx+1] = fAveragePrc;\n\t\t\t\t\t\tpMax[nIdx] = fMaxMs;\n\t\t\t\t\t\tpMax[nIdx+1] = fMaxPrc;\n\t\t\t\t\t\tpMin[nIdx] = fMinMs;\n\t\t\t\t\t\tpMin[nIdx + 1] = fMinPrc;\n\t\t\t\t\t\tpCallAverage[nIdx] = fCallAverageMs;\n\t\t\t\t\t\tpCallAverage[nIdx+1] = fCallAveragePrc;\n\t\t\t\t\t\tpExclusive[nIdx] = fMsExclusive;\n\t\t\t\t\t\tpExclusive[nIdx+1] = fPrcExclusive;\n\t\t\t\t\t\tpAverageExclusive[nIdx] = fAverageMsExclusive;\n\t\t\t\t\t\tpAverageExclusive[nIdx+1] = fAveragePrcExclusive;\n\t\t\t\t\t\tpMaxExclusive[nIdx] = fMaxMsExclusive;\n\t\t\t\t\t\tpMaxExclusive[nIdx+1] = fMaxPrcExclusive;\n\t\t\t\t\t}\n\t\t\t\t\tnCount += 2;\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t\tnMask <<= 1ll;\n\t}\n}\n\n#define SBUF_MAX 32\n\nvoid MicroProfileDrawBarArrayCallback(uint32_t nTimer, uint32_t nIdx, uint32_t nX, uint32_t nY, void* pExtra)\n{\n\tconst uint32_t nHeight = MICROPROFILE_TEXT_HEIGHT;\t\n\tconst uint32_t nTextWidth = 6 * (1+MICROPROFILE_TEXT_WIDTH);\n\tconst float fWidth = (float)MICROPROFILE_BAR_WIDTH;\n\n\tfloat* pTimers = ((float**)pExtra)[0];\n\tfloat* pTimers2 = ((float**)pExtra)[1];\n\tMicroProfile& S = *MicroProfileGet();\n\tchar sBuffer[SBUF_MAX];\n\tif (pTimers2 && pTimers2[nIdx] > 0.1f)\n\t\tsnprintf(sBuffer, SBUF_MAX-1, \"%5.2f %3.1fx\", pTimers[nIdx], pTimers[nIdx] / pTimers2[nIdx]);\n\telse\n\t\tsnprintf(sBuffer, SBUF_MAX-1, \"%5.2f\", pTimers[nIdx]);\n\tif (!pTimers2)\n\t\tMicroProfileDrawBox(nX + nTextWidth, nY, (int)(nX + nTextWidth + fWidth * pTimers[nIdx+1]), nY + nHeight, UI.nOpacityForeground|S.TimerInfo[nTimer].nColor, MicroProfileBoxTypeBar);\n\tMicroProfileDrawText(nX, nY, (uint32_t)-1, sBuffer, (uint32_t)strlen(sBuffer));\t\t\n}\n\n\nuint32_t MicroProfileDrawBarArray(int32_t nX, int32_t nY, float* pTimers, const char* pName, uint32_t nTotalHeight, float* pTimers2 = NULL)\n{\n\tconst uint32_t nTextWidth = 6 * (1+MICROPROFILE_TEXT_WIDTH);\n\tconst uint32_t nWidth = MICROPROFILE_BAR_WIDTH;\n\n\tMicroProfileDrawLineVertical(nX-5, 0, nTotalHeight+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);\n\tfloat* pTimersArray[2] = {pTimers, pTimers2};\n\tMicroProfileLoopActiveGroupsDraw(nX, nY, MicroProfileDrawBarArrayCallback, pTimersArray);\n\tMicroProfileDrawHeader(nX, nTextWidth + nWidth, pName);\n\treturn nWidth + 5 + nTextWidth;\n\n}\nvoid MicroProfileDrawBarCallCountCallback(uint32_t nTimer, uint32_t nIdx, uint32_t nX, uint32_t nY, void* pExtra)\n{\n\t(void)nIdx;\n\t(void)pExtra;\n\n\tMicroProfile& S = *MicroProfileGet();\n\tchar sBuffer[SBUF_MAX];\n\tint nLen = snprintf(sBuffer, SBUF_MAX-1, \"%5d\", S.Frame[nTimer].nCount);//fix\n\tMicroProfileDrawText(nX, nY, (uint32_t)-1, sBuffer, nLen);\n}\n\nuint32_t MicroProfileDrawBarCallCount(int32_t nX, int32_t nY, const char* pName)\n{\n\tMicroProfileLoopActiveGroupsDraw(nX, nY, MicroProfileDrawBarCallCountCallback, 0);\n\tconst uint32_t nTextWidth = 6 * MICROPROFILE_TEXT_WIDTH;\n\tMicroProfileDrawHeader(nX, 5 + nTextWidth, pName);\n\treturn 5 + nTextWidth;\n}\n\nstruct MicroProfileMetaAverageArgs\n{\n\tuint64_t* pCounters;\n\tfloat fRcpFrames;\n};\n\nvoid MicroProfileDrawBarMetaAverageCallback(uint32_t nTimer, uint32_t nIdx, uint32_t nX, uint32_t nY, void* pExtra)\n{\n\t(void)nIdx;\n\n\tMicroProfileMetaAverageArgs* pArgs = (MicroProfileMetaAverageArgs*)pExtra;\n\tuint64_t* pCounters = pArgs->pCounters;\n\tfloat fRcpFrames = pArgs->fRcpFrames;\n\tchar sBuffer[SBUF_MAX];\n\tint nLen = snprintf(sBuffer, SBUF_MAX-1, \"%5.2f\", pCounters[nTimer] * fRcpFrames);\n\tMicroProfileDrawText(nX - nLen * (MICROPROFILE_TEXT_WIDTH+1), nY, (uint32_t)-1, sBuffer, nLen);\n}\n\nuint32_t MicroProfileDrawBarMetaAverage(int32_t nX, int32_t nY, uint64_t* pCounters, const char* pName, uint32_t nTotalHeight)\n{\n\tif(!pName)\n\t\treturn 0;\n\tMicroProfileDrawLineVertical(nX-5, 0, nTotalHeight+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);\n\tuint32_t nTextWidth = (1+MICROPROFILE_TEXT_WIDTH) * MicroProfileMax(6, (uint32_t)strlen(pName));\n\tfloat fRcpFrames = 1.f / (MicroProfileGet()->nAggregateFrames ? MicroProfileGet()->nAggregateFrames : 1);\n\tMicroProfileMetaAverageArgs Args = {pCounters, fRcpFrames};\n\tMicroProfileLoopActiveGroupsDraw(nX + nTextWidth, nY, MicroProfileDrawBarMetaAverageCallback, &Args);\n\tMicroProfileDrawHeader(nX, 5 + nTextWidth, pName);\t\n\treturn 5 + nTextWidth;\n}\n\n\nvoid MicroProfileDrawBarMetaCountCallback(uint32_t nTimer, uint32_t nIdx, uint32_t nX, uint32_t nY, void* pExtra)\n{\n\t(void)nIdx;\n\n\tuint64_t* pCounters = (uint64_t*)pExtra;\n\tchar sBuffer[SBUF_MAX];\n\tint nLen = snprintf(sBuffer, SBUF_MAX-1, \"%5lld\", (long long)pCounters[nTimer]);\n\tMicroProfileDrawText(nX - nLen * (MICROPROFILE_TEXT_WIDTH+1), nY, (uint32_t)-1, sBuffer, nLen);\t\n}\n\nuint32_t MicroProfileDrawBarMetaCount(int32_t nX, int32_t nY, uint64_t* pCounters, const char* pName, uint32_t nTotalHeight)\n{\n\tif(!pName)\n\t\treturn 0;\n\n\tMicroProfileDrawLineVertical(nX-5, 0, nTotalHeight+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);\n\tuint32_t nTextWidth = (1+MICROPROFILE_TEXT_WIDTH) * MicroProfileMax(6, (uint32_t)strlen(pName));\n\tMicroProfileLoopActiveGroupsDraw(nX + nTextWidth, nY, MicroProfileDrawBarMetaCountCallback, pCounters);\n\tMicroProfileDrawHeader(nX, 5 + nTextWidth, pName);\n\treturn 5 + nTextWidth;\n}\n\nvoid MicroProfileDrawBarLegendCallback(uint32_t nTimer, uint32_t nIdx, uint32_t nX, uint32_t nY, void* pExtra)\n{\n\t(void)nIdx;\n\t(void)pExtra;\n\n\tMicroProfile& S = *MicroProfileGet();\n\tif (S.TimerInfo[nTimer].bGraph)\n\t{\n\t\tMicroProfileDrawText(nX, nY, S.TimerInfo[nTimer].nColor, \">\", 1);\n\t}\n\tMicroProfileDrawTextRight(nX, nY, S.TimerInfo[nTimer].nColor, S.TimerInfo[nTimer].pName, (uint32_t)strlen(S.TimerInfo[nTimer].pName));\n\tif(UI.nMouseY >= nY && UI.nMouseY < nY + MICROPROFILE_TEXT_HEIGHT+1)\n\t{\n\t\tUI.nHoverToken = nTimer;\n\t\tUI.nHoverTime = 0;\n\t}\n}\n\nuint32_t MicroProfileDrawBarLegend(int32_t nX, int32_t nY, uint32_t nTotalHeight, uint32_t nMaxWidth)\n{\n\tMicroProfileDrawLineVertical(nX-5, nY, nTotalHeight, UI.nOpacityBackground | g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);\n\tMicroProfileLoopActiveGroupsDraw(nMaxWidth, nY, MicroProfileDrawBarLegendCallback, 0);\n\treturn nX;\n}\n\nbool MicroProfileDrawGraph(uint32_t nScreenWidth, uint32_t nScreenHeight)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\n\tMICROPROFILE_SCOPE(g_MicroProfileDrawGraph);\n\tbool bEnabled = false;\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t\tif(S.Graph[i].nToken != MICROPROFILE_INVALID_TOKEN)\n\t\t\tbEnabled = true;\n\tif(!bEnabled)\n\t\treturn false;\n\t\n\tuint32_t nX = nScreenWidth - MICROPROFILE_GRAPH_WIDTH;\n\tuint32_t nY = nScreenHeight - MICROPROFILE_GRAPH_HEIGHT;\n\tMicroProfileDrawBox(nX, nY, nX + MICROPROFILE_GRAPH_WIDTH, nY + MICROPROFILE_GRAPH_HEIGHT, 0x88000000 | g_nMicroProfileBackColors[0]);\n\tbool bMouseOver = UI.nMouseX >= nX && UI.nMouseY >= nY;\n\tfloat fMouseXPrc =(float(UI.nMouseX - nX)) / MICROPROFILE_GRAPH_WIDTH;\n\tif(bMouseOver)\n\t{\n\t\tfloat fXAvg = fMouseXPrc * MICROPROFILE_GRAPH_WIDTH + nX;\n\t\tMicroProfileDrawLineVertical((int)fXAvg, nY, nY + MICROPROFILE_GRAPH_HEIGHT, (uint32_t)-1);\n\t}\n\n\t\n\tfloat fY = (float)nScreenHeight;\n\tfloat fDX = MICROPROFILE_GRAPH_WIDTH * 1.f / MICROPROFILE_GRAPH_HISTORY; \n\tfloat fDY = MICROPROFILE_GRAPH_HEIGHT;\n\tuint32_t nPut = S.nGraphPut;\n\tfloat* pGraphData = (float*)alloca(sizeof(float)* MICROPROFILE_GRAPH_HISTORY*2);\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t{\n\t\tif(S.Graph[i].nToken != MICROPROFILE_INVALID_TOKEN)\n\t\t{\n\t\t\tuint32_t nGroupId = MicroProfileGetGroupIndex(S.Graph[i].nToken);\n\t\t\tbool bGpu = S.GroupInfo[nGroupId].Type == MicroProfileTokenTypeGpu;\n\t\t\tfloat fToMs = MicroProfileTickToMsMultiplier(bGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu());\n\t\t\tfloat fToPrc = fToMs * S.fRcpReferenceTime * 3 / 4;\n\n\t\t\tfloat fX = (float)nX;\n\t\t\tfor(uint32_t j = 0; j < MICROPROFILE_GRAPH_HISTORY; ++j)\n\t\t\t{\n\t\t\t\tfloat fWeigth = MicroProfileMin(fToPrc * (S.Graph[i].nHistory[(j+nPut)%MICROPROFILE_GRAPH_HISTORY]), 1.f);\n\t\t\t\tpGraphData[(j*2)] = fX;\n\t\t\t\tpGraphData[(j*2)+1] = fY - fDY * fWeigth;\n\t\t\t\tfX += fDX;\n\t\t\t}\n\t\t\tMicroProfileDrawLine2D(MICROPROFILE_GRAPH_HISTORY, pGraphData, UI.nOpacityForeground|S.TimerInfo[MicroProfileGetTimerIndex(S.Graph[i].nToken)].nColor);\n\t\t}\n\t}\n\t{\n\t\tfloat fY1 = 0.25f * MICROPROFILE_GRAPH_HEIGHT + nY;\n\t\tfloat fY2 = 0.50f * MICROPROFILE_GRAPH_HEIGHT + nY;\n\t\tfloat fY3 = 0.75f * MICROPROFILE_GRAPH_HEIGHT + nY;\n\t\tMicroProfileDrawLineHorizontal(nX, nX + MICROPROFILE_GRAPH_WIDTH, (int)fY1, 0xffdd4444);\n\t\tMicroProfileDrawLineHorizontal(nX, nX + MICROPROFILE_GRAPH_WIDTH, (int)fY2, 0xff000000 | g_nMicroProfileBackColors[0]);\n\t\tMicroProfileDrawLineHorizontal(nX, nX + MICROPROFILE_GRAPH_WIDTH, (int)fY3, 0xff000000 | g_nMicroProfileBackColors[0]);\n\n\t\tchar buf[32];\n\t\tint nLen = snprintf(buf, sizeof(buf)-1, \"%5.2fms\", S.fReferenceTime);\n\t\tMicroProfileDrawText(nX+1, (int)(fY1 - (2+MICROPROFILE_TEXT_HEIGHT)), (uint32_t)-1, buf, nLen);\n\t}\n\n\n\n\tif(bMouseOver)\n\t{\n\t\tuint32_t pColors[MICROPROFILE_MAX_GRAPHS];\n\t\tMicroProfileStringArray Strings;\n\t\tMicroProfileStringArrayClear(&Strings);\n\t\tuint32_t nTextCount = 0;\n\t\tuint32_t nGraphIndex = (S.nGraphPut + MICROPROFILE_GRAPH_HISTORY - int(MICROPROFILE_GRAPH_HISTORY*(1.f - fMouseXPrc))) % MICROPROFILE_GRAPH_HISTORY;\n\n\t\tuint32_t nX = UI.nMouseX;\n\t\tuint32_t nY = UI.nMouseY + 20;\n\n\t\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t\t{\n\t\t\tif(S.Graph[i].nToken != MICROPROFILE_INVALID_TOKEN)\n\t\t\t{\n\t\t\t\tuint32_t nGroupId = MicroProfileGetGroupIndex(S.Graph[i].nToken);\n\t\t\t\tbool bGpu = S.GroupInfo[nGroupId].Type == MicroProfileTokenTypeGpu;\n\t\t\t\tfloat fToMs = MicroProfileTickToMsMultiplier(bGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu());\n\t\t\t\tuint32_t nIndex = MicroProfileGetTimerIndex(S.Graph[i].nToken);\n\t\t\t\tuint32_t nColor = S.TimerInfo[nIndex].nColor;\n\t\t\t\tconst char* pName = S.TimerInfo[nIndex].pName;\n\t\t\t\tpColors[nTextCount++] = nColor;\n\t\t\t\tMicroProfileStringArrayAddLiteral(&Strings, pName);\n\t\t\t\tMicroProfileStringArrayFormat(&Strings, \"%5.2fms\", fToMs * (S.Graph[i].nHistory[nGraphIndex]));\n\t\t\t}\n\t\t}\n\t\tif(nTextCount)\n\t\t{\n\t\t\tMicroProfileDrawFloatWindow(nX, nY, Strings.ppStrings, Strings.nNumStrings, 0, pColors);\n\t\t}\n\n\t\tif(UI.nMouseRight)\n\t\t{\n\t\t\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t\t\t{\n\t\t\t\tS.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN;\n\t\t\t}\n\t\t}\n\t}\n\n\treturn bMouseOver;\n}\n\nvoid MicroProfileDumpTimers()\n{\n\tMicroProfile& S = *MicroProfileGet();\n\n\tuint64_t nActiveGroup = S.nGroupMask;\n\n\tuint32_t nNumTimers = S.nTotalTimers;\n\tuint32_t nBlockSize = 2 * nNumTimers;\n\tfloat* pTimers = (float*)alloca(nBlockSize * 8 * sizeof(float));\n\tfloat* pAverage = pTimers + nBlockSize;\n\tfloat* pMax = pTimers + 2 * nBlockSize;\n\tfloat* pMin = pTimers + 3 * nBlockSize;\n\tfloat* pCallAverage = pTimers + 4 * nBlockSize;\n\tfloat* pTimersExclusive = pTimers + 5 * nBlockSize;\n\tfloat* pAverageExclusive = pTimers + 6 * nBlockSize;\n\tfloat* pMaxExclusive = pTimers + 7 * nBlockSize;\n\tMicroProfileCalcTimers(pTimers, pAverage, pMax, pMin, pCallAverage, pTimersExclusive, pAverageExclusive, pMaxExclusive, nActiveGroup, nBlockSize);\n\n\tMICROPROFILE_PRINTF(\"%11s, \", \"Time\");\n\tMICROPROFILE_PRINTF(\"%11s, \", \"Average\");\n\tMICROPROFILE_PRINTF(\"%11s, \", \"Max\");\n\tMICROPROFILE_PRINTF(\"%11s, \", \"Min\");\n\tMICROPROFILE_PRINTF(\"%11s, \", \"Call Avg\");\n\tMICROPROFILE_PRINTF(\"%9s, \", \"Count\");\n\tMICROPROFILE_PRINTF(\"%11s, \", \"Excl\");\n\tMICROPROFILE_PRINTF(\"%11s, \", \"Avg Excl\");\n\tMICROPROFILE_PRINTF(\"%11s, \\n\", \"Max Excl\");\n\n\tfor(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j)\n\t{\n\t\tuint64_t nMask = 1ll << j;\n\t\tif(nMask & nActiveGroup)\n\t\t{\n\t\t\tMICROPROFILE_PRINTF(\"%s\\n\", S.GroupInfo[j].pName);\n\t\t\tfor(uint32_t i = 0; i < S.nTotalTimers;++i)\n\t\t\t{\n\t\t\t\tuint64_t nTokenMask = MicroProfileGetGroupMask(S.TimerInfo[i].nToken);\n\t\t\t\tif(nTokenMask & nMask)\n\t\t\t\t{\n\t\t\t\t\tuint32_t nIdx = i * 2;\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9.2fms, \", pTimers[nIdx]);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9.2fms, \", pAverage[nIdx]);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9.2fms, \", pMax[nIdx]);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9.2fms, \", pMin[nIdx]);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9.2fms, \", pCallAverage[nIdx]);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9d, \", S.Frame[i].nCount);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9.2fms, \", pTimersExclusive[nIdx]);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9.2fms, \", pAverageExclusive[nIdx]);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%9.2fms, \", pMaxExclusive[nIdx]);\n\t\t\t\t\tMICROPROFILE_PRINTF(\"%s\\n\", S.TimerInfo[i].pName);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n}\n\n\n\nuint32_t MicroProfileDrawCounterRecursive(uint32_t nIndex, uint32_t nY, uint32_t nOffset, uint32_t nTimerWidth)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tuint8_t bGraphDetailed = 0 != (S.CounterInfo[nIndex].nFlags & MICROPROFILE_COUNTER_FLAG_DETAILED_GRAPH);\n\tuint32_t nRows = bGraphDetailed ? 5 : 1;\n\tconst uint32_t nHeight = MICROPROFILE_TEXT_HEIGHT;\n\tconst uint32_t nCounterWidth = UI.nCounterWidth;\n\tconst uint32_t nLimitWidth = UI.nLimitWidth;\n\n\tuint32_t nY0 = nY + nOffset * (nHeight+1);\n\tuint32_t nBackHeight = (nHeight+1) * nRows;\n\n\tMicroProfileCounterInfo& CI = S.CounterInfo[nIndex];\n\tbool bInside = (UI.nActiveMenu == (uint32_t)-1) && ((UI.nMouseY >= nY0) && (UI.nMouseY < (nY0 + nBackHeight)));\n\tuint32_t nTotalWidth = nTimerWidth + nCounterWidth * 3 + MICROPROFILE_COUNTER_WIDTH + nLimitWidth + 4 * (MICROPROFILE_TEXT_WIDTH+1)\n\t+ 4 + MICROPROFILE_GRAPH_HISTORY;\n\tuint32_t nBackColor = 0xff000000 | (g_nMicroProfileBackColors[nOffset & 1] + ((bInside) ? 0x002c2c2c : 0));\n\tMicroProfileDrawBox(0, nY0, nTotalWidth, nY0 + nBackHeight +1, nBackColor);\n\tuint32_t nIndent = MICROPROFILE_COUNTER_INDENT*CI.nLevel * (MICROPROFILE_TEXT_WIDTH+1);\n\tif(CI.nFirstChild != -1 && 0 != (CI.nFlags & MICROPROFILE_COUNTER_FLAG_CLOSED))\n\t{\n\t\tMicroProfileDrawText(nIndent, nY0, 0xffffffff, \"*\", 1);\n\t}\n\n\tMicroProfileDrawText(nIndent + MICROPROFILE_TEXT_WIDTH+1, nY0, 0xffffffff, CI.pName, CI.nNameLen);\n\tchar buffer[64];\n\tint64_t nCounterValue = S.Counters[nIndex].load();\n\tuint32_t nX = nTimerWidth + nCounterWidth;\n\tint nLen = MicroProfileFormatCounter(S.CounterInfo[nIndex].eFormat, nCounterValue, buffer, sizeof(buffer));\n\tUI.nCounterWidthTemp = MicroProfileMax((uint32_t)nLen, UI.nCounterWidthTemp);\n\tif(0 != nCounterValue || 0 != (CI.nFlags & MICROPROFILE_COUNTER_FLAG_LEAF))\n\t{\n\t\tMicroProfileDrawTextRight(nX, nY0, 0xffffffff, buffer, nLen);\n\t}\n\tint64_t nLimit = S.CounterInfo[nIndex].nLimit;\n\tif(nLimit)\n\t{\n\t\tnX += MICROPROFILE_TEXT_WIDTH+1;\n\t\tMicroProfileDrawText(nX, nY0, 0xffffffff, \"/\", 1);\n\t\tnX += 2 * (MICROPROFILE_TEXT_WIDTH+1);\n\t\tint nLen = MicroProfileFormatCounter(S.CounterInfo[nIndex].eFormat, nLimit, buffer, sizeof(buffer));\n\t\tUI.nLimitWidthTemp = MicroProfileMax(UI.nLimitWidthTemp, (uint32_t)nLen);\n\t\tMicroProfileDrawText(nX, nY0, 0xffffffff, buffer, nLen);\n\t\tnX += nLimitWidth;\n\t\tnY0 += 1;\n\n\t\tfloat fCounterPrc = (float)nCounterValue / nLimit;\n\t\tfCounterPrc = MicroProfileMax(fCounterPrc, 0.f);\n\t\tfloat fBoxPrc = 1.f;\n\t\tif(fCounterPrc>1.f)\n\t\t{\n\t\t\tfBoxPrc = 1.f / fCounterPrc;\n\t\t\tfCounterPrc = 1.f;\n\t\t}\n\n\t\tMicroProfileDrawBox(nX, nY0, nX + (int)(fBoxPrc * MICROPROFILE_COUNTER_WIDTH), nY0 + nHeight, 0xffffffff, MicroProfileBoxTypeFlat);\n\t\tMicroProfileDrawBox(nX+1, nY0+1, nX + MICROPROFILE_COUNTER_WIDTH - 1, nY0 + nHeight - 1, nBackColor, MicroProfileBoxTypeFlat);\n\t\tMicroProfileDrawBox(nX+1, nY0+1, nX + (int)(fCounterPrc * (MICROPROFILE_COUNTER_WIDTH - 1)), nY0 + nHeight - 1, 0xff0088ff, MicroProfileBoxTypeFlat);\n\t\tnX += MICROPROFILE_COUNTER_WIDTH + 5;\n\t}\n\telse\n\t{\n\t\tnX += MICROPROFILE_TEXT_WIDTH+1;\n\t\tnX += 2 * (MICROPROFILE_TEXT_WIDTH+1);\n\t\tnX += nLimitWidth;\n\t\tnX += MICROPROFILE_COUNTER_WIDTH + 5;\n\t}\n\n\tif(bInside && (UI.nMouseLeft || UI.nMouseRight))\n\t{\n\t\tif(UI.nMouseX > nX)\n\t\t{\n\t\t\tif(UI.nMouseRight)\n\t\t\t{\n\t\t\t\tCI.nFlags &= ~MICROPROFILE_COUNTER_FLAG_DETAILED;\n\t\t\t}\n\t\t\telse\n\t\t\t{\n\t\t\t\t// toggle through detailed & detailed graph\n\t\t\t\tif(CI.nFlags & MICROPROFILE_COUNTER_FLAG_DETAILED)\n\t\t\t\t{\n\t\t\t\t\tCI.nFlags ^= MICROPROFILE_COUNTER_FLAG_DETAILED_GRAPH;\n\t\t\t\t}\n\t\t\t\telse\n\t\t\t\t{\n\t\t\t\t\tCI.nFlags |= MICROPROFILE_COUNTER_FLAG_DETAILED;\n\t\t\t\t}\n\t\t\t}\n\t\t\tif(0 == (CI.nFlags & MICROPROFILE_COUNTER_FLAG_DETAILED))\n\t\t\t{\n\t\t\t\tCI.nFlags &= ~MICROPROFILE_COUNTER_FLAG_DETAILED_GRAPH;\n\t\t\t}\n\t\t}\n\t\telse if(UI.nMouseLeft)\n\t\t{\n\t\t\tCI.nFlags ^= MICROPROFILE_COUNTER_FLAG_CLOSED;\n\t\t}\n\t}\n\n#if MICROPROFILE_COUNTER_HISTORY\n\tif(0 != (CI.nFlags & MICROPROFILE_COUNTER_FLAG_DETAILED))\n\t{\n\t\tstatic float pGraphData[MICROPROFILE_GRAPH_HISTORY*2];\n\t\tstatic float pGraphFillData[MICROPROFILE_GRAPH_HISTORY*4];\n\n\t\tint32_t nMouseGraph = UI.nMouseX - nX;\n\n\n\t\tint64_t nCounterMax = S.nCounterMax[nIndex];\n\t\tint64_t nCounterMin = S.nCounterMin[nIndex];\n\t\tuint32_t nBaseIndex = S.nCounterHistoryPut;\n\t\tfloat fX = (float)nX;\n\n\t\tint64_t nCounterHeightBase = nCounterMax;\n\t\tint64_t nCounterOffset = 0;\n\t\tif(nCounterMin < 0)\n\t\t{\n\t\t\tnCounterHeightBase = nCounterMax - nCounterMin;\n\t\t\tnCounterOffset = -nCounterMin;\n\t\t}\n\t\tconst int32_t nGraphHeight = nRows * nHeight;\n\t\tdouble fRcpMax = nGraphHeight * 1.0 / nCounterHeightBase;\n\t\tconst int32_t nYOffset = nY0 + (bGraphDetailed ? 3 : 1);\n\t\tconst int32_t nYBottom = nGraphHeight + nYOffset;\n\t\tfor(uint32_t i = 0; i < MICROPROFILE_GRAPH_HISTORY; ++i)\n\t\t{\n\t\t\tuint32_t nHistoryIndex = (nBaseIndex + i) % MICROPROFILE_GRAPH_HISTORY;\n\t\t\tint64_t nValue = MicroProfileClamp(S.nCounterHistory[nHistoryIndex][nIndex], nCounterMin, nCounterMax);\n\t\t\tfloat fPrc = nGraphHeight - (float)((double)(nValue+nCounterOffset) * fRcpMax);\n\t\t\tpGraphData[(i*2)] = fX;\n\t\t\tpGraphData[(i*2)+1] = nYOffset + fPrc;\n\n\t\t\tpGraphFillData[(i*4) + 0] = fX;\n\t\t\tpGraphFillData[(i*4) + 1] = nYOffset + fPrc;\n\t\t\tpGraphFillData[(i*4) + 2] = fX;\n\t\t\tpGraphFillData[(i*4) + 3] = (float)nYBottom;\n\n\t\t\tfX += 1;\n\t\t}\n\t\tMicroProfileDrawLine2D(MICROPROFILE_GRAPH_HISTORY*2, pGraphFillData, 0x330088ff);\n\t\tMicroProfileDrawLine2D(MICROPROFILE_GRAPH_HISTORY, pGraphData, 0xff0088ff);\n\n\t\tif(nMouseGraph < MICROPROFILE_GRAPH_HISTORY && bInside && nCounterMin <= nCounterMax)\n\t\t{\n\t\t\tuint32_t nMouseX = nX + nMouseGraph;\n\t\t\tfloat fMouseX = (float)nMouseX;\n\t\t\tuint32_t nHistoryIndex = (nBaseIndex + nMouseGraph) % MICROPROFILE_GRAPH_HISTORY;\n\t\t\tint64_t nValue = MicroProfileClamp(S.nCounterHistory[nHistoryIndex][nIndex], nCounterMin, nCounterMax);\n\t\t\tfloat fPrc = nGraphHeight - (float)((double)(nValue+nCounterOffset) * fRcpMax);\n\t\t\tfloat fCursor[4];\n\t\t\tfCursor[0] = fMouseX-2.f;\n\t\t\tfCursor[1] = nYOffset + fPrc + 2.f;\n\t\t\tfCursor[2] = fMouseX+2.f;\n\t\t\tfCursor[3] = nYOffset + fPrc - 2.f;\n\t\t\tMicroProfileDrawLine2D(2, fCursor, 0xddff8800);\n\t\t\tfCursor[0] = fMouseX+2.f;\n\t\t\tfCursor[1] = nYOffset + fPrc + 2.f;\n\t\t\tfCursor[2] = fMouseX-2.f;\n\t\t\tfCursor[3] = nYOffset + fPrc - 2.f;\n\t\t\tMicroProfileDrawLine2D(2, fCursor, 0xddff8800);\n\t\t\tint nLen = MicroProfileFormatCounter(S.CounterInfo[nIndex].eFormat, nValue, buffer, sizeof(buffer));\n\t\t\tMicroProfileDrawText(nX, nY0, 0xffffffff, buffer, nLen);\n\t\t}\n\n\n\t\tnX += MICROPROFILE_GRAPH_HISTORY + 5;\n\t\tif(nCounterMin <= nCounterMax)\n\t\t{\n\t\t\tint nLen = MicroProfileFormatCounter(S.CounterInfo[nIndex].eFormat, nCounterMin, buffer, sizeof(buffer));\n\t\t\tMicroProfileDrawText(nX, nY0, 0xffffffff, buffer, nLen);\n\t\t\tnX += nCounterWidth;\n\t\t\tnLen = MicroProfileFormatCounter(S.CounterInfo[nIndex].eFormat, nCounterMax, buffer, sizeof(buffer));\n\t\t\tMicroProfileDrawText(nX, nY0, 0xffffffff, buffer, nLen);\n\t\t}\n\n\t}\n#endif\n\n\tnOffset += nRows;\n\tif(0 == (CI.nFlags & MICROPROFILE_COUNTER_FLAG_CLOSED))\n\t{\n\t\tint nChild = CI.nFirstChild;\t\n\t\twhile(nChild != -1)\n\t\t{\n\t\t\tnOffset = MicroProfileDrawCounterRecursive(nChild, nY, nOffset, nTimerWidth);\n\t\t\tnChild = S.CounterInfo[nChild].nSibling;\n\t\t}\n\t}\n\n\n\treturn nOffset;\n}\n\nvoid MicroProfileDrawCounterView(uint32_t nScreenWidth, uint32_t nScreenHeight)\n{\n\t(void)nScreenWidth;\n\t(void)nScreenHeight;\n\n\tMicroProfile& S = *MicroProfileGet();\n\tMICROPROFILE_SCOPE(g_MicroProfileDrawBarView);\n\n\tUI.nCounterWidthTemp = 7;\n\tUI.nLimitWidthTemp = 7;\n\tconst uint32_t nHeight = MICROPROFILE_TEXT_HEIGHT;\n\tuint32_t nTimerWidth = 7 * (MICROPROFILE_TEXT_WIDTH+1);\n\tfor(uint32_t i = 0; i < S.nNumCounters; ++i)\n\t{\n\t\tuint32_t nWidth = (2+S.CounterInfo[i].nNameLen + MICROPROFILE_COUNTER_INDENT*S.CounterInfo[i].nLevel) * (MICROPROFILE_TEXT_WIDTH+1);\n\t\tnTimerWidth = MicroProfileMax(nTimerWidth, nWidth);\n\t}\n\tuint32_t nX = nTimerWidth + UI.nOffsetX[MP_DRAW_COUNTERS];\n\tuint32_t nY = nHeight + 3 - UI.nOffsetY[MP_DRAW_COUNTERS];\t\n\tuint32_t nNumCounters = S.nNumCounters;\n\tnX = 0;\n\tnY = (2*nHeight) + 3 - UI.nOffsetY[MP_DRAW_COUNTERS];\t\n\tuint32_t nOffset = 0;\n\tfor(uint32_t i = 0; i < nNumCounters; ++i)\n\t{\n\t\tif(S.CounterInfo[i].nParent == -1)\n\t\t{\n\t\t\tnOffset = MicroProfileDrawCounterRecursive(i, nY, nOffset, nTimerWidth);\n\t\t}\n\t}\n\tnX = 0;\n\tMicroProfileDrawHeader(nX, nTimerWidth, \"Name\");\n\tnX += nTimerWidth;\n\tMicroProfileDrawHeader(nX, UI.nCounterWidth + 1 * (MICROPROFILE_TEXT_WIDTH*3), \"Value\");\n\tnX += UI.nCounterWidth;\n\tnX += 1 * (MICROPROFILE_TEXT_WIDTH*3);\n\tMicroProfileDrawHeader(nX, UI.nLimitWidth + MICROPROFILE_COUNTER_WIDTH, \"Limit\");\n\tnX += UI.nLimitWidth + MICROPROFILE_COUNTER_WIDTH + 4;\n\tMicroProfileDrawHeader(nX, MICROPROFILE_GRAPH_HISTORY, \"Graph\");\n\tnX += MICROPROFILE_GRAPH_HISTORY + 4;\n\tMicroProfileDrawHeader(nX, UI.nCounterWidth , \"Min\");\n\tnX += UI.nCounterWidth;\n\tMicroProfileDrawHeader(nX, UI.nCounterWidth , \"Max\");\n\tnX += UI.nCounterWidth;\n\tuint32_t nTotalWidth = nX;//nTimerWidth + UI.nCounterWidth + MICROPROFILE_COUNTER_WIDTH + UI.nLimitWidth + 3 * (MICROPROFILE_TEXT_WIDTH+1);\n\n\n\n\tMicroProfileDrawLineVertical(nTimerWidth-2, 0, nOffset*(nHeight+1)+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);\n\tMicroProfileDrawLineHorizontal(0, nTotalWidth, 2*MICROPROFILE_TEXT_HEIGHT + 3, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);\n\n\tUI.nCounterWidth = (1+UI.nCounterWidthTemp) * (MICROPROFILE_TEXT_WIDTH+1);\n\tUI.nLimitWidth = (1+UI.nLimitWidthTemp) * (MICROPROFILE_TEXT_WIDTH+1);\n\n}\n\n\n\nvoid MicroProfileDrawBarView(uint32_t nScreenWidth, uint32_t nScreenHeight)\n{\n\t(void)nScreenWidth;\n\t(void)nScreenHeight;\n\n\tMicroProfile& S = *MicroProfileGet();\n\n\tuint64_t nActiveGroup = S.nAllGroupsWanted ? S.nGroupMask : S.nActiveGroupWanted;\n\tif(!nActiveGroup)\n\t\treturn;\n\tMICROPROFILE_SCOPE(g_MicroProfileDrawBarView);\n\n\tconst uint32_t nHeight = MICROPROFILE_TEXT_HEIGHT;\n\tint nColorIndex = 0;\n\tuint32_t nMaxTimerNameLen = 1;\n\tuint32_t nNumTimers = 0;\n\tuint32_t nNumGroups = 0;\n\tfor(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j)\n\t{\n\t\tif(nActiveGroup & (1ll << j))\n\t\t{\n\t\t\tnNumTimers += S.GroupInfo[j].nNumTimers;\n\t\t\tnNumGroups += 1;\n\t\t\tnMaxTimerNameLen = MicroProfileMax(nMaxTimerNameLen, S.GroupInfo[j].nMaxTimerNameLen);\n\t\t}\n\t}\n\tuint32_t nTimerWidth = 2+(4+nMaxTimerNameLen) * (MICROPROFILE_TEXT_WIDTH+1);\n\tuint32_t nX = nTimerWidth + UI.nOffsetX[MP_DRAW_BARS];\n\tuint32_t nY = nHeight + 3 - UI.nOffsetY[MP_DRAW_BARS];\t\n\tuint32_t nBlockSize = 2 * nNumTimers;\n\tfloat* pTimers = (float*)alloca(nBlockSize * 8 * sizeof(float));\n\tfloat* pAverage = pTimers + nBlockSize;\n\tfloat* pMax = pTimers + 2 * nBlockSize;\n\tfloat* pMin = pTimers + 3 * nBlockSize;\n\tfloat* pCallAverage = pTimers + 4 * nBlockSize;\n\tfloat* pTimersExclusive = pTimers + 5 * nBlockSize;\n\tfloat* pAverageExclusive = pTimers + 6 * nBlockSize;\n\tfloat* pMaxExclusive = pTimers + 7 * nBlockSize;\n\tMicroProfileCalcTimers(pTimers, pAverage, pMax, pMin, pCallAverage, pTimersExclusive, pAverageExclusive, pMaxExclusive, nActiveGroup, nBlockSize);\n\tuint32_t nWidth = 0;\n\t{\n\t\tuint32_t nMetaIndex = 0;\n\t\tfor(uint32_t i = 1; i ; i <<= 1)\n\t\t{\n\t\t\tif(S.nBars & i)\n\t\t\t{\n\t\t\t\tif(i >= MP_DRAW_META_FIRST)\n\t\t\t\t{\n\t\t\t\t\tif(nMetaIndex < MICROPROFILE_META_MAX && S.MetaCounters[nMetaIndex].pName)\n\t\t\t\t\t{\n\t\t\t\t\t\tuint32_t nStrWidth = (uint32_t)strlen(S.MetaCounters[nMetaIndex].pName);\n\t\t\t\t\t\tif(S.nBars & MP_DRAW_TIMERS)\n\t\t\t\t\t\t\tnWidth += 6 + (1+MICROPROFILE_TEXT_WIDTH) * (nStrWidth);\n\t\t\t\t\t\tif(S.nBars & MP_DRAW_AVERAGE)\n\t\t\t\t\t\t\tnWidth += 6 + (1+MICROPROFILE_TEXT_WIDTH) * (nStrWidth + 4);\n\t\t\t\t\t\tif(S.nBars & MP_DRAW_MAX)\n\t\t\t\t\t\t\tnWidth += 6 + (1+MICROPROFILE_TEXT_WIDTH) * (nStrWidth + 4);\n\t\t\t\t\t\tif (S.nBars & MP_DRAW_MIN)\n\t\t\t\t\t\t\tnWidth += 6 + (1 + MICROPROFILE_TEXT_WIDTH) * (nStrWidth + 4);\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t\telse\n\t\t\t\t{\n\t\t\t\t\tnWidth += MICROPROFILE_BAR_WIDTH + 6 + 6 * (1+MICROPROFILE_TEXT_WIDTH);\n\t\t\t\t\tif(i & MP_DRAW_CALL_COUNT)\n\t\t\t\t\t\tnWidth += 6 + 6 * MICROPROFILE_TEXT_WIDTH;\n\t\t\t\t}\n\t\t\t}\n\t\t\tif(i >= MP_DRAW_META_FIRST)\n\t\t\t{\n\t\t\t\t++nMetaIndex;\n\t\t\t}\n\t\t}\n\t\tnWidth += (1+nMaxTimerNameLen) * (MICROPROFILE_TEXT_WIDTH+1);\n\t\tfor(uint32_t i = 0; i < nNumTimers+nNumGroups+1; ++i)\n\t\t{\n\t\t\tuint32_t nY0 = nY + i * (nHeight + 1);\n\t\t\tbool bInside = (UI.nActiveMenu == (uint32_t)-1) && ((UI.nMouseY >= nY0) && (UI.nMouseY < (nY0 + nHeight + 1)));\n\t\t\tMicroProfileDrawBox(nX, nY0, nWidth+nX, nY0 + (nHeight+1)+1, UI.nOpacityBackground | (g_nMicroProfileBackColors[nColorIndex++ & 1] + ((bInside) ? 0x002c2c2c : 0)));\n\t\t}\n\t\tnX += 10;\n\t}\n\tint nTotalHeight = (nNumTimers+nNumGroups+1) * (nHeight+1);\n\tuint32_t nLegendOffset = 1;\n\tif(S.nBars & MP_DRAW_TIMERS)\t\t\n\t\tnX += MicroProfileDrawBarArray(nX, nY, pTimers, \"Time\", nTotalHeight) + 1;\n\tif(S.nBars & MP_DRAW_AVERAGE)\t\t\n\t\tnX += MicroProfileDrawBarArray(nX, nY, pAverage, \"Average\", nTotalHeight) + 1;\n\tif(S.nBars & MP_DRAW_MAX)\t\t\n\t\tnX += MicroProfileDrawBarArray(nX, nY, pMax, (!UI.bShowSpikes) ? \"Max Time\" : \"Max Time, Spike\", nTotalHeight, UI.bShowSpikes ? pAverage : NULL) + 1;\n\tif (S.nBars & MP_DRAW_MIN)\n\t\tnX += MicroProfileDrawBarArray(nX, nY, pMin, (!UI.bShowSpikes) ? \"Min Time\" : \"Min Time, Spike\", nTotalHeight, UI.bShowSpikes ? pAverage : NULL) + 1;\n\tif(S.nBars & MP_DRAW_CALL_COUNT)\t\t\n\t{\n\t\tnX += MicroProfileDrawBarArray(nX, nY, pCallAverage, \"Call Average\", nTotalHeight) + 1;\n\t\tnX += MicroProfileDrawBarCallCount(nX, nY, \"Count\") + 1; \n\t}\n\tif(S.nBars & MP_DRAW_TIMERS_EXCLUSIVE)\t\t\n\t\tnX += MicroProfileDrawBarArray(nX, nY, pTimersExclusive, \"Exclusive Time\", nTotalHeight) + 1;\n\tif(S.nBars & MP_DRAW_AVERAGE_EXCLUSIVE)\t\t\n\t\tnX += MicroProfileDrawBarArray(nX, nY, pAverageExclusive, \"Exclusive Average\", nTotalHeight) + 1;\n\tif(S.nBars & MP_DRAW_MAX_EXCLUSIVE)\t\t\n\t\tnX += MicroProfileDrawBarArray(nX, nY, pMaxExclusive, (!UI.bShowSpikes) ? \"Exclusive Max Time\" :\"Excl Max Time, Spike\", nTotalHeight, UI.bShowSpikes ? pAverageExclusive : NULL) + 1;\n\n\tfor(int i = 0; i < MICROPROFILE_META_MAX; ++i)\n\t{\n\t\tif(0 != (S.nBars & (MP_DRAW_META_FIRST<= nY0) && (UI.nMouseY < (nY0 + nHeight + 1)));\n\t\tMicroProfileDrawBox(nX, nY0, nTimerWidth, nY0 + (nHeight+1)+1, 0xff000000 | (g_nMicroProfileBackColors[nColorIndex++ & 1] + ((bInside) ? 0x002c2c2c : 0)));\n\t}\n\tnX += MicroProfileDrawBarLegend(nX, nY, nTotalHeight, nTimerWidth-5) + 1;\n\n\tfor(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j)\n\t{\n\t\tif(nActiveGroup & (1ll << j))\n\t\t{\n\t\t\tMicroProfileDrawText(nX, nY + (1+nHeight) * nLegendOffset, (uint32_t)-1, S.GroupInfo[j].pName, S.GroupInfo[j].nNameLen);\n\t\t\tnLegendOffset += S.GroupInfo[j].nNumTimers+1;\n\t\t}\n\t}\n\tMicroProfileDrawHeader(nX, nTimerWidth-5, \"Group\");\n\tMicroProfileDrawTextRight(nTimerWidth-3, MICROPROFILE_TEXT_HEIGHT + 2, (uint32_t)-1, \"Timer\", 5);\n\tMicroProfileDrawLineVertical(nTimerWidth, 0, nTotalHeight+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);\n\tMicroProfileDrawLineHorizontal(0, nWidth, 2*MICROPROFILE_TEXT_HEIGHT + 3, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]);\n}\n\ntypedef const char* (*MicroProfileSubmenuCallback)(int, bool* bSelected); \ntypedef void (*MicroProfileClickCallback)(int);\n\nconst char* MicroProfileUIMenuMode(int nIndex, bool* bSelected)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tswitch(nIndex)\n\t{\n\t\tcase 0: \n\t\t\t*bSelected = S.nDisplay == MP_DRAW_DETAILED;\n\t\t\treturn \"Detailed\";\n\t\tcase 1:\n\t\t\t*bSelected = S.nDisplay == MP_DRAW_BARS;\n\t\t\treturn \"Timers\";\n\t\tcase 2:\n\t\t\t*bSelected = S.nDisplay == MP_DRAW_COUNTERS; \n\t\t\treturn \"Counters\";\n\t\tcase 3:\n\t\t\t*bSelected = S.nDisplay == MP_DRAW_FRAME; \n\t\t\treturn \"Frame\";\n\t\tcase 4:\n\t\t\t*bSelected = S.nDisplay == MP_DRAW_HIDDEN; \n\t\t\treturn \"Hidden\";\n\t\tcase 5:\n\t\t\t*bSelected = false; \n\t\t\treturn \"Off\";\n\t\tcase 6:\n\t\t\t*bSelected = false;\n\t\t\treturn \"------\";\n\t\tcase 7:\n\t\t\t*bSelected = S.nForceEnable != 0;\n\t\t\treturn \"Force Enable\";\n\n\t\tdefault: return 0;\n\t}\n}\n\nconst char* MicroProfileUIMenuGroups(int nIndex, bool* bSelected)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\t*bSelected = false;\n\tif(nIndex == 0)\n\t{\n\t\t*bSelected = S.nAllGroupsWanted != 0;\n\t\treturn \"[ALL]\";\n\t}\n\telse\n\t{\n\t\tnIndex = nIndex-1;\n\t\tif(nIndex < (int)UI.GroupMenuCount)\n\t\t{\n\t\t\tMicroProfileGroupMenuItem& Item = UI.GroupMenu[nIndex];\n\t\t\tstatic char buffer[MICROPROFILE_NAME_MAX_LEN+32];\n\t\t\tif(Item.nIsCategory)\n\t\t\t{\n\t\t\t\tuint64_t nGroupMask = S.CategoryInfo[Item.nIndex].nGroupMask;\n\t\t\t\t*bSelected = nGroupMask == (nGroupMask & S.nActiveGroupWanted);\n\t\t\t\tsnprintf(buffer, sizeof(buffer)-1, \"[%s]\", Item.pName);\n\t\t\t}\n\t\t\telse\n\t\t\t{\n\t\t\t\t*bSelected = 0 != (S.nActiveGroupWanted & (1ll << Item.nIndex));\n\t\t\t\tsnprintf(buffer, sizeof(buffer)-1, \" %s\", Item.pName);\n\t\t\t}\n\t\t\treturn buffer;\n\t\t}\n\t\treturn 0;\n\t}\t\n}\n\nconst char* MicroProfileUIMenuAggregate(int nIndex, bool* bSelected)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tint nNumPresets = (int)sizeof(g_MicroProfileAggregatePresets) / (int)sizeof(g_MicroProfileAggregatePresets[0]);\n\tif(nIndex < nNumPresets)\n\t{\n\t\tint val = g_MicroProfileAggregatePresets[nIndex];\n\t\t*bSelected = (int)S.nAggregateFlip == val;\n\t\tif(0 == val)\n\t\t\treturn \"Infinite\";\n\t\telse\n\t\t{\n\t\t\tstatic char buf[128];\n\t\t\tsnprintf(buf, sizeof(buf)-1, \"%7d\", val);\n\t\t\treturn buf;\n\t\t}\n\t}\n\treturn 0;\n\n}\n\nconst char* MicroProfileUIMenuTimers(int nIndex, bool* bSelected)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\n\tif(nIndex < 8)\n\t{\n\t\tstatic const char* kNames[] = { \"Time\", \"Average\", \"Max\", \"Min\", \"Call Count\", \"Exclusive Timers\", \"Exclusive Average\", \"Exclusive Max\" };\n\n\t\t*bSelected = 0 != (S.nBars & (1 << nIndex));\n\t\treturn kNames[nIndex];\n\t}\n\telse if(nIndex == 8)\n\t{\n\t\t*bSelected = false;\n\t\treturn \"------\";\n\t}\n\telse\n\t{\n\t\tint nMetaIndex = nIndex - 9;\n\t\tif(nMetaIndex < MICROPROFILE_META_MAX)\n\t\t{\n\t\t\t*bSelected = 0 != (S.nBars & (MP_DRAW_META_FIRST << nMetaIndex));\n\t\t\treturn S.MetaCounters[nMetaIndex].pName;\n\t\t}\n\t}\n\treturn 0;\t\n}\n\nconst char* MicroProfileUIMenuOptions(int nIndex, bool* bSelected)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tif(nIndex >= MICROPROFILE_OPTION_SIZE) return 0;\n\tswitch(UI.Options[nIndex].nSubType)\n\t{\n\tcase 0:\n\t\t*bSelected = S.fReferenceTime == g_MicroProfileReferenceTimePresets[UI.Options[nIndex].nIndex];\n\t\tbreak;\n\tcase 1:\n\t\t*bSelected = UI.nOpacityBackground>>24 == g_MicroProfileOpacityPresets[UI.Options[nIndex].nIndex];\n\t\tbreak;\n\tcase 2:\n\t\t*bSelected = UI.nOpacityForeground>>24 == g_MicroProfileOpacityPresets[UI.Options[nIndex].nIndex];\t\t\t\t\n\t\tbreak;\n\tcase 3:\n\t\t*bSelected = UI.bShowSpikes;\n\t\tbreak;\n#if MICROPROFILE_CONTEXT_SWITCH_TRACE\n\tcase 4:\n\t\t{\n\t\t\tswitch(UI.Options[nIndex].nIndex)\n\t\t\t{\n\t\t\tcase 0:\n\t\t\t\t*bSelected = S.bContextSwitchAllThreads;\n\t\t\t\tbreak;\n\t\t\tcase 1: \n\t\t\t\t*bSelected = S.bContextSwitchNoBars;\n\t\t\t\tbreak;\n\t\t\t}\n\t\t}\n\t\tbreak;\n#endif\n\t}\n\treturn UI.Options[nIndex].Text;\n}\n\nconst char* MicroProfileUIMenuPreset(int nIndex, bool* bSelected)\n{\n\tstatic char buf[128];\n\t*bSelected = false;\n\tint nNumPresets = sizeof(g_MicroProfilePresetNames) / sizeof(g_MicroProfilePresetNames[0]);\n\tint nIndexSave = nIndex - nNumPresets - 1;\n\tif(nIndex == nNumPresets)\n\t\treturn \"--\";\n\telse if(nIndexSave >=0 && nIndexSave = 0 && nIndexSave < nNumPresets)\n\t{\n\t\tMicroProfileSavePreset(g_MicroProfilePresetNames[nIndexSave]);\n\t}\n\telse if(nIndex >= 0 && nIndex < nNumPresets)\n\t{\n\t\tMicroProfileLoadPreset(g_MicroProfilePresetNames[nIndex]);\n\t}\n}\n\nvoid MicroProfileUIClickCustom(int nIndex)\n{\n\tif(nIndex == 0)\n\t{\n\t\tMicroProfileCustomGroupDisable();\n\t}\n\telse\n\t{\n\t\tMicroProfileCustomGroupEnable(nIndex-2);\n\t}\n}\n\nvoid MicroProfileUIClickDump(int nIndex)\n{\n\ttime_t t = time(0);\n\n\tchar Name[128] = {};\n\tstrftime(Name, sizeof(Name), \"microprofile-%Y%m%d-%H%M%S.html\", localtime(&t));\n\n\tchar Path[512] = {};\n\tconst char* pHome = getenv(\"HOME\");\n\tconst char* pHomeDrive = getenv(\"HOMEDRIVE\");\n\tconst char* pHomePath = getenv(\"HOMEPATH\");\n\tif(pHome)\n\t{\n\t\tsnprintf(Path, sizeof(Path)-1, \"%s/%s\", pHome, Name);\n\t}\n\telse if(pHomeDrive && pHomePath)\n\t{\n\t\tsnprintf(Path, sizeof(Path)-1, \"%s%s/%s\", pHomeDrive, pHomePath, Name);\n\t}\n\telse\n\t{\n\t\tsnprintf(Path, sizeof(Path)-1, \"%s\", Name);\n\t}\n\n\tMicroProfileDumpFile(Path, MicroProfileDumpTypeHtml, 32 << nIndex);\n}\n\nvoid MicroProfileDrawMenu(uint32_t nWidth, uint32_t nHeight)\n{\n\t(void)nWidth;\n\t(void)nHeight;\n\n\tMicroProfile& S = *MicroProfileGet();\n\n\tuint32_t nX = 0;\n\tuint32_t nY = 0;\n#define SBUF_SIZE 256\n\tchar buffer[256];\n\tMicroProfileDrawBox(nX, nY, nX + nWidth, nY + (MICROPROFILE_TEXT_HEIGHT+1)+1, 0xff000000|g_nMicroProfileBackColors[1]);\n\n#define MICROPROFILE_MENU_MAX 16\n\tconst char* pMenuText[MICROPROFILE_MENU_MAX] = {0};\n\tuint32_t \tnMenuX[MICROPROFILE_MENU_MAX] = {0};\n\tuint32_t nNumMenuItems = 0;\n\n\tint nLen = snprintf(buffer, 127, \"MicroProfile\");\n\tMicroProfileDrawText(nX, nY, (uint32_t)-1, buffer, nLen);\n\tnX += (sizeof(\"MicroProfile\")+2) * (MICROPROFILE_TEXT_WIDTH+1);\n\tpMenuText[nNumMenuItems++] = \"Mode\";\n\tpMenuText[nNumMenuItems++] = \"Groups\";\n\tchar AggregateText[64];\n\tsnprintf(AggregateText, sizeof(AggregateText)-1, \"Aggregate[%d]\", S.nAggregateFlip ? S.nAggregateFlip : S.nAggregateFlipCount);\n\tpMenuText[nNumMenuItems++] = &AggregateText[0];\n\tpMenuText[nNumMenuItems++] = \"Timers\";\n\tpMenuText[nNumMenuItems++] = \"Options\";\n\tpMenuText[nNumMenuItems++] = \"Preset\";\n\tpMenuText[nNumMenuItems++] = \"Custom\";\n\tpMenuText[nNumMenuItems++] = \"Dump\";\n\tconst int nPauseIndex = nNumMenuItems;\n\tpMenuText[nNumMenuItems++] = S.nRunning ? \"Pause\" : \"Unpause\";\n\tpMenuText[nNumMenuItems++] = \"Help\";\n\n\tif(S.nOverflow)\n\t{\n\t\tpMenuText[nNumMenuItems++] = \"!BUFFERSFULL!\";\n\t}\n\n\n\tif(UI.GroupMenuCount != S.nGroupCount + S.nCategoryCount)\n\t{\n\t\tUI.GroupMenuCount = S.nGroupCount + S.nCategoryCount;\n\t\tfor(uint32_t i = 0; i < S.nCategoryCount; ++i)\n\t\t{\n\t\t\tUI.GroupMenu[i].nIsCategory = 1;\n\t\t\tUI.GroupMenu[i].nCategoryIndex = i;\n\t\t\tUI.GroupMenu[i].nIndex = i;\n\t\t\tUI.GroupMenu[i].pName = S.CategoryInfo[i].pName;\n\t\t}\n\t\tfor(uint32_t i = 0; i < S.nGroupCount; ++i)\n\t\t{\n\t\t\tuint32_t idx = i + S.nCategoryCount;\n\t\t\tUI.GroupMenu[idx].nIsCategory = 0;\n\t\t\tUI.GroupMenu[idx].nCategoryIndex = S.GroupInfo[i].nCategory;\n\t\t\tUI.GroupMenu[idx].nIndex = i;\n\t\t\tUI.GroupMenu[idx].pName = S.GroupInfo[i].pName;\n\t\t}\n\t\tstd::sort(&UI.GroupMenu[0], &UI.GroupMenu[UI.GroupMenuCount], \n\t\t\t[] (const MicroProfileGroupMenuItem& l, const MicroProfileGroupMenuItem& r) -> bool\n\t\t\t{\n\t\t\t\tif(l.nCategoryIndex < r.nCategoryIndex)\n\t\t\t\t{\n\t\t\t\t\treturn true;\n\t\t\t\t}\n\t\t\t\telse if(r.nCategoryIndex < l.nCategoryIndex)\n\t\t\t\t{\n\t\t\t\t\treturn false;\n\t\t\t\t}\n\t\t\t\tif(r.nIsCategory || l.nIsCategory)\n\t\t\t\t{\n\t\t\t\t\treturn l.nIsCategory > r.nIsCategory;\n\t\t\t\t}\n\t\t\t\treturn MP_STRCASECMP(l.pName, r.pName)<0;\n\t\t\t}\n\t\t);\n\t}\n\n\tMicroProfileSubmenuCallback GroupCallback[MICROPROFILE_MENU_MAX] = \n\t{\n\t\tMicroProfileUIMenuMode,\n\t\tMicroProfileUIMenuGroups,\n\t\tMicroProfileUIMenuAggregate,\n\t\tMicroProfileUIMenuTimers,\n\t\tMicroProfileUIMenuOptions,\n\t\tMicroProfileUIMenuPreset,\n\t\tMicroProfileUIMenuCustom,\n\t\tMicroProfileUIMenuDump,\n\t};\n\n\tMicroProfileClickCallback CBClick[MICROPROFILE_MENU_MAX] =\n\t{\n\t\tMicroProfileUIClickMode,\n\t\tMicroProfileUIClickGroups,\n\t\tMicroProfileUIClickAggregate,\n\t\tMicroProfileUIClickTimers,\n\t\tMicroProfileUIClickOptions,\n\t\tMicroProfileUIClickPreset,\n\t\tMicroProfileUIClickCustom,\n\t\tMicroProfileUIClickDump,\n\t};\n\n\n\tuint32_t nSelectMenu = (uint32_t)-1;\n\tfor(uint32_t i = 0; i < nNumMenuItems; ++i)\n\t{\n\t\tnMenuX[i] = nX;\n\t\tuint32_t nLen = (uint32_t)strlen(pMenuText[i]);\n\t\tuint32_t nEnd = nX + nLen * (MICROPROFILE_TEXT_WIDTH+1);\n\t\tif(UI.nMouseY <= MICROPROFILE_TEXT_HEIGHT && UI.nMouseX <= nEnd && UI.nMouseX >= nX)\n\t\t{\n\t\t\tMicroProfileDrawBox(nX-1, nY, nX + nLen * (MICROPROFILE_TEXT_WIDTH+1), nY +(MICROPROFILE_TEXT_HEIGHT+1)+1, 0xff888888);\n\t\t\tnSelectMenu = i;\n\t\t\tif((UI.nMouseLeft || UI.nMouseRight) && (int)i == nPauseIndex)\n\t\t\t{\n\t\t\t\tS.nToggleRunning = 1;\n\t\t\t}\n\t\t}\n\t\tMicroProfileDrawText(nX, nY, (uint32_t)-1, pMenuText[i], (uint32_t)strlen(pMenuText[i]));\n\t\tnX += (nLen+1) * (MICROPROFILE_TEXT_WIDTH+1);\n\t}\n\tuint32_t nMenu = nSelectMenu != (uint32_t)-1 ? nSelectMenu : UI.nActiveMenu;\n\tUI.nActiveMenu = nSelectMenu;\n\tif((uint32_t)-1 != nMenu && GroupCallback[nMenu])\n\t{\n\t\tnX = nMenuX[nMenu];\n\t\tnY += MICROPROFILE_TEXT_HEIGHT+1;\n\t\tMicroProfileSubmenuCallback CB = GroupCallback[nMenu];\n\t\tint nNumLines = 0;\n\t\tbool bSelected = false;\n\t\tconst char* pString = CB(nNumLines, &bSelected);\n\t\tuint32_t nWidth = 0, nHeight = 0;\n\t\twhile(pString)\n\t\t{\n\t\t\tnWidth = MicroProfileMax(nWidth, (int)strlen(pString));\n\t\t\tnNumLines++;\n\t\t\tpString = CB(nNumLines, &bSelected);\n\t\t}\n\t\tnWidth = (2+nWidth) * (MICROPROFILE_TEXT_WIDTH+1);\n\t\tnHeight = nNumLines * (MICROPROFILE_TEXT_HEIGHT+1);\n\t\tif(UI.nMouseY <= nY + nHeight+0 && UI.nMouseY >= nY-0 && UI.nMouseX <= nX + nWidth + 0 && UI.nMouseX >= nX - 0)\n\t\t{\n\t\t\tUI.nActiveMenu = nMenu;\n\t\t}\n\t\tMicroProfileDrawBox(nX, nY, nX + nWidth, nY + nHeight, 0xff000000|g_nMicroProfileBackColors[1]);\n\t\tfor(int i = 0; i < nNumLines; ++i)\n\t\t{\n\t\t\tbool bSelected = false;\n\t\t\tconst char* pString = CB(i, &bSelected);\n\t\t\tif(UI.nMouseY >= nY && UI.nMouseY < nY + MICROPROFILE_TEXT_HEIGHT + 1)\n\t\t\t{\n\t\t\t\tif((UI.nMouseLeft || UI.nMouseRight) && CBClick[nMenu])\n\t\t\t\t{\n\t\t\t\t\tCBClick[nMenu](i);\n\t\t\t\t}\n\t\t\t\tMicroProfileDrawBox(nX, nY, nX + nWidth, nY + MICROPROFILE_TEXT_HEIGHT + 1, 0xff888888);\n\t\t\t}\n\t\t\tint nLen = snprintf(buffer, SBUF_SIZE-1, \"%c %s\", bSelected ? '*' : ' ' ,pString);\n\t\t\tMicroProfileDrawText(nX, nY, (uint32_t)-1, buffer, nLen);\n\t\t\tnY += MICROPROFILE_TEXT_HEIGHT+1;\n\t\t}\n\t}\n\n\t{\n\t\tstatic char FrameTimeMessage[64];\n\t\tfloat fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu());\n\t\tuint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1;\n\t\tfloat fMs = fToMs * (S.nFlipTicks);\n\t\tfloat fAverageMs = fToMs * (S.nFlipAggregateDisplay / nAggregateFrames);\n\t\tfloat fMaxMs = fToMs * S.nFlipMaxDisplay;\n\t\tint nLen = snprintf(FrameTimeMessage, sizeof(FrameTimeMessage)-1, \"Time[%6.2f] Avg[%6.2f] Max[%6.2f]\", fMs, fAverageMs, fMaxMs);\n\t\tpMenuText[nNumMenuItems++] = &FrameTimeMessage[0];\n\t\tMicroProfileDrawText(nWidth - nLen * (MICROPROFILE_TEXT_WIDTH+1), 0, -1, FrameTimeMessage, nLen);\n\t}\n}\n\n\nvoid MicroProfileMoveGraph()\n{\n\n\tint nZoom = UI.nMouseWheelDelta;\n\tint nPanX = 0;\n\tint nPanY = 0;\n\tstatic int X = 0, Y = 0;\n\tif(UI.nMouseDownLeft && !UI.nModDown)\n\t{\n\t\tnPanX = UI.nMouseX - X;\n\t\tnPanY = UI.nMouseY - Y;\n\t}\n\tX = UI.nMouseX;\n\tY = UI.nMouseY;\n\n\tif(nZoom)\n\t{\n\t\tfloat fOldRange = UI.fDetailedRange;\n\t\tif(nZoom>0)\n\t\t{\n\t\t\tUI.fDetailedRangeTarget = UI.fDetailedRange *= UI.nModDown ? 1.40f : 1.05f;\n\t\t}\n\t\telse\n\t\t{\n\t\t\tfloat fNewDetailedRange = UI.fDetailedRange / (UI.nModDown ? 1.40f : 1.05f);\n\t\t\tif(fNewDetailedRange < 1e-4f) //100ns\n\t\t\t\tfNewDetailedRange = 1e-4f;\n\t\t\tUI.fDetailedRangeTarget = UI.fDetailedRange = fNewDetailedRange;\n\t\t}\n\n\t\tfloat fDiff = fOldRange - UI.fDetailedRange;\n\t\tfloat fMousePrc = MicroProfileMax((float)UI.nMouseX / UI.nWidth ,0.f);\n\t\tUI.fDetailedOffsetTarget = UI.fDetailedOffset += fDiff * fMousePrc;\n\n\t}\n\tif(nPanX)\n\t{\n\t\tUI.fDetailedOffsetTarget = UI.fDetailedOffset += -nPanX * UI.fDetailedRange / UI.nWidth;\n\t}\n\tint nMode = MicroProfileGet()->nDisplay;\n\tif(nMode < MP_DRAW_SIZE)\n\t{\n\t\tUI.nOffsetY[nMode] -= nPanY;\n\t\tUI.nOffsetX[nMode] += nPanX;\n\t\tif(UI.nOffsetX[nMode] > 0)\n\t\t\tUI.nOffsetX[nMode] = 0;\n\t\tif(UI.nOffsetY[nMode] < 0)\n\t\t\tUI.nOffsetY[nMode] = 0;\n\t}\n}\n\nvoid MicroProfileDrawCustom(uint32_t nWidth, uint32_t nHeight)\n{\n\t(void)nWidth;\n\n\tif((uint32_t)-1 != UI.nCustomActive)\n\t{\n\t\tMicroProfile& S = *MicroProfileGet();\n\t\tMP_ASSERT(UI.nCustomActive < MICROPROFILE_CUSTOM_MAX);\n\t\tMicroProfileCustom* pCustom = &UI.Custom[UI.nCustomActive];\n\t\tuint32_t nCount = pCustom->nNumTimers;\n\t\tuint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1;\n\t\tuint32_t nExtraOffset = 1 + ((pCustom->nFlags & MICROPROFILE_CUSTOM_STACK) != 0 ? 3 : 0);\n\t\tuint32_t nOffsetYBase = nHeight - (nExtraOffset+nCount)* (1+MICROPROFILE_TEXT_HEIGHT) - MICROPROFILE_CUSTOM_PADDING;\n\t\tuint32_t nOffsetY = nOffsetYBase;\n\t\tfloat fReference = pCustom->fReference;\n\t\tfloat fRcpReference = 1.f / fReference;\n\t\tuint32_t nReducedWidth = UI.nWidth - 2*MICROPROFILE_CUSTOM_PADDING - MICROPROFILE_GRAPH_WIDTH;\n\n\t\tchar Buffer[MICROPROFILE_NAME_MAX_LEN*2+1];\n\t\tfloat* pTime = (float*)alloca(sizeof(float)*nCount);\n\t\tfloat* pTimeAvg = (float*)alloca(sizeof(float)*nCount);\n\t\tfloat* pTimeMax = (float*)alloca(sizeof(float)*nCount);\n\t\tuint32_t* pColors = (uint32_t*)alloca(sizeof(uint32_t)*nCount);\n\t\tuint32_t nMaxOffsetX = 0;\n\t\tMicroProfileDrawBox(MICROPROFILE_CUSTOM_PADDING-1, nOffsetY-1, MICROPROFILE_CUSTOM_PADDING+nReducedWidth+1, UI.nHeight - MICROPROFILE_CUSTOM_PADDING+1, 0x88000000|g_nMicroProfileBackColors[0]);\n\n\t\tfor(uint32_t i = 0; i < nCount; ++i)\n\t\t{\n\t\t\tuint16_t nTimerIndex = MicroProfileGetTimerIndex(pCustom->pTimers[i]);\n\t\t\tuint16_t nGroupIndex = MicroProfileGetGroupIndex(pCustom->pTimers[i]);\n\t\t\tfloat fToMs = MicroProfileTickToMsMultiplier(S.GroupInfo[nGroupIndex].Type == MicroProfileTokenTypeGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu());\n\t\t\tpTime[i] = S.Frame[nTimerIndex].nTicks * fToMs;\n\t\t\tpTimeAvg[i] = fToMs * (S.Aggregate[nTimerIndex].nTicks / nAggregateFrames);\n\t\t\tpTimeMax[i] = fToMs * (S.AggregateMax[nTimerIndex]);\n\t\t\tpColors[i] = S.TimerInfo[nTimerIndex].nColor;\n\t\t}\n\n\t\tMicroProfileDrawText(MICROPROFILE_CUSTOM_PADDING + 3*MICROPROFILE_TEXT_WIDTH, nOffsetY, (uint32_t)-1, \"Avg\", sizeof(\"Avg\")-1);\n\t\tMicroProfileDrawText(MICROPROFILE_CUSTOM_PADDING + 13*MICROPROFILE_TEXT_WIDTH, nOffsetY, (uint32_t)-1, \"Max\", sizeof(\"Max\")-1);\n\t\tfor(uint32_t i = 0; i < nCount; ++i)\n\t\t{\n\t\t\tnOffsetY += (1+MICROPROFILE_TEXT_HEIGHT);\n\t\t\tuint16_t nTimerIndex = MicroProfileGetTimerIndex(pCustom->pTimers[i]);\n\t\t\tuint16_t nGroupIndex = MicroProfileGetGroupIndex(pCustom->pTimers[i]);\n\t\t\tMicroProfileTimerInfo* pTimerInfo = &S.TimerInfo[nTimerIndex];\n\t\t\tint nSize;\n\t\t\tuint32_t nOffsetX = MICROPROFILE_CUSTOM_PADDING;\n\t\t\tnSize = snprintf(Buffer, sizeof(Buffer)-1, \"%6.2f\", pTimeAvg[i]);\n\t\t\tMicroProfileDrawText(nOffsetX, nOffsetY, (uint32_t)-1, Buffer, nSize);\n\t\t\tnOffsetX += (nSize+2) * (MICROPROFILE_TEXT_WIDTH+1);\n\t\t\tnSize = snprintf(Buffer, sizeof(Buffer)-1, \"%6.2f\", pTimeMax[i]);\n\t\t\tMicroProfileDrawText(nOffsetX, nOffsetY, (uint32_t)-1, Buffer, nSize);\n\t\t\tnOffsetX += (nSize+2) * (MICROPROFILE_TEXT_WIDTH+1);\n\t\t\tnSize = snprintf(Buffer, sizeof(Buffer)-1, \"%s:%s\", S.GroupInfo[nGroupIndex].pName, pTimerInfo->pName);\n\t\t\tMicroProfileDrawText(nOffsetX, nOffsetY, pTimerInfo->nColor, Buffer, nSize);\n\t\t\tnOffsetX += (nSize+2) * (MICROPROFILE_TEXT_WIDTH+1);\n\t\t\tnMaxOffsetX = MicroProfileMax(nMaxOffsetX, nOffsetX);\n\t\t}\n\t\tuint32_t nMaxWidth = nReducedWidth- nMaxOffsetX;\n\n\t\tif(pCustom->nFlags & MICROPROFILE_CUSTOM_BARS)\n\t\t{\n\t\t\tnOffsetY = nOffsetYBase;\n\t\t\tfloat* pMs = pCustom->nFlags & MICROPROFILE_CUSTOM_BAR_SOURCE_MAX ? pTimeMax : pTimeAvg;\n\t\t\tconst char* pString = pCustom->nFlags & MICROPROFILE_CUSTOM_BAR_SOURCE_MAX ? \"Max\" : \"Avg\";\n\t\t\tMicroProfileDrawText(nMaxOffsetX, nOffsetY, (uint32_t)-1, pString, (uint32_t)strlen(pString));\n\t\t\tint nSize = snprintf(Buffer, sizeof(Buffer)-1, \"%6.2fms\", fReference);\n\t\t\tMicroProfileDrawText(nReducedWidth - (1+nSize) * (MICROPROFILE_TEXT_WIDTH+1), nOffsetY, (uint32_t)-1, Buffer, nSize);\n\t\t\tfor(uint32_t i = 0; i < nCount; ++i)\n\t\t\t{\n\t\t\t\tnOffsetY += (1+MICROPROFILE_TEXT_HEIGHT);\n\t\t\t\tuint32_t nWidth = MicroProfileMin(nMaxWidth, (uint32_t)(nMaxWidth * pMs[i] * fRcpReference));\n\t\t\t\tMicroProfileDrawBox(nMaxOffsetX, nOffsetY, nMaxOffsetX+nWidth, nOffsetY+MICROPROFILE_TEXT_HEIGHT, pColors[i]|0xff000000);\n\t\t\t}\n\t\t}\n\t\tif(pCustom->nFlags & MICROPROFILE_CUSTOM_STACK)\n\t\t{\n\t\t\tnOffsetY += 2*(1+MICROPROFILE_TEXT_HEIGHT);\n\t\t\tconst char* pString = pCustom->nFlags & MICROPROFILE_CUSTOM_STACK_SOURCE_MAX ? \"Max\" : \"Avg\";\n\t\t\tMicroProfileDrawText(MICROPROFILE_CUSTOM_PADDING, nOffsetY, (uint32_t)-1, pString, (uint32_t)strlen(pString));\n\t\t\tint nSize = snprintf(Buffer, sizeof(Buffer)-1, \"%6.2fms\", fReference);\n\t\t\tMicroProfileDrawText(nReducedWidth - (1+nSize) * (MICROPROFILE_TEXT_WIDTH+1), nOffsetY, (uint32_t)-1, Buffer, nSize);\n\t\t\tnOffsetY += (1+MICROPROFILE_TEXT_HEIGHT);\n\t\t\tfloat fPosX = MICROPROFILE_CUSTOM_PADDING;\n\t\t\tfloat* pMs = pCustom->nFlags & MICROPROFILE_CUSTOM_STACK_SOURCE_MAX ? pTimeMax : pTimeAvg;\n\t\t\tfor(uint32_t i = 0; i < nCount; ++i)\n\t\t\t{\n\t\t\t\tfloat fWidth = pMs[i] * fRcpReference * nReducedWidth;\n\t\t\t\tuint32_t nX = (uint32_t)fPosX;\n\t\t\t\tfPosX += fWidth;\n\t\t\t\tuint32_t nXEnd = (uint32_t)fPosX;\n\t\t\t\tif(nX < nXEnd)\n\t\t\t\t{\n\t\t\t\t\tMicroProfileDrawBox(nX, nOffsetY, nXEnd, nOffsetY+MICROPROFILE_TEXT_HEIGHT, pColors[i]|0xff000000);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n}\nvoid MicroProfileDraw(uint32_t nWidth, uint32_t nHeight)\n{\n\tMICROPROFILE_SCOPE(g_MicroProfileDraw);\n\tMicroProfile& S = *MicroProfileGet();\n\n\t{\n\t\tstatic int once = 0;\n\t\tif(0 == once)\n\t\t{\t\n\t\t\tstd::recursive_mutex& m = MicroProfileGetMutex();\n\t\t\tm.lock();\n\t\t\tMicroProfileInitUI();\n\t\t\tuint32_t nDisplay = S.nDisplay;\n\t\t\tMicroProfileLoadPreset(MICROPROFILE_DEFAULT_PRESET);\n\t\t\tonce++;\n\t\t\tS.nDisplay = nDisplay;// dont load display, just state\n\t\t\tm.unlock();\n\n\t\t}\n\t}\n\n\n\tif(S.nDisplay)\n\t{\n\t\tstd::recursive_mutex& m = MicroProfileGetMutex();\n\t\tm.lock();\n\t\tUI.nWidth = nWidth;\n\t\tUI.nHeight = nHeight;\n\t\tUI.nHoverToken = MICROPROFILE_INVALID_TOKEN;\n\t\tUI.nHoverTime = 0;\n\t\tUI.nHoverFrame = -1;\n\t\tif(S.nDisplay != MP_DRAW_DETAILED)\n\t\t\tS.nContextSwitchHoverThread = S.nContextSwitchHoverThreadAfter = S.nContextSwitchHoverThreadBefore = -1;\n\t\tMicroProfileMoveGraph();\n\n\n\t\tif(S.nDisplay == MP_DRAW_DETAILED || S.nDisplay == MP_DRAW_FRAME)\n\t\t{\n\t\t\tMicroProfileDrawDetailedView(nWidth, nHeight, /* bDrawBars= */ S.nDisplay == MP_DRAW_DETAILED);\n\t\t}\n\t\telse if(S.nDisplay == MP_DRAW_BARS && S.nBars)\n\t\t{\n\t\t\tMicroProfileDrawBarView(nWidth, nHeight);\n\t\t}\n\t\telse if(S.nDisplay == MP_DRAW_COUNTERS)\n\t\t{\n\t\t\tMicroProfileDrawCounterView(nWidth, nHeight);\n\t\t}\n\t\t\n\t\tMicroProfileDrawMenu(nWidth, nHeight);\n\t\tbool bMouseOverGraph = MicroProfileDrawGraph(nWidth, nHeight);\n\t\tMicroProfileDrawCustom(nWidth, nHeight);\n\t\tbool bHidden = S.nDisplay == MP_DRAW_HIDDEN;\n\t\tif(!bHidden)\n\t\t{\n\t\t\tuint32_t nLockedToolTipX = 3;\n\t\t\tbool bDeleted = false;\n\t\t\tfor(int i = 0; i < MICROPROFILE_TOOLTIP_MAX_LOCKED; ++i)\n\t\t\t{\n\t\t\t\tint nIndex = (g_MicroProfileUI.LockedToolTipFront + i) % MICROPROFILE_TOOLTIP_MAX_LOCKED;\n\t\t\t\tif(g_MicroProfileUI.LockedToolTips[nIndex].ppStrings[0])\n\t\t\t\t{\n\t\t\t\t\tuint32_t nToolTipWidth = 0, nToolTipHeight = 0;\n\t\t\t\t\tMicroProfileFloatWindowSize(g_MicroProfileUI.LockedToolTips[nIndex].ppStrings, g_MicroProfileUI.LockedToolTips[nIndex].nNumStrings, 0, nToolTipWidth, nToolTipHeight, 0);\n\t\t\t\t\tuint32_t nStartY = nHeight - nToolTipHeight - 2;\n\t\t\t\t\tif(!bDeleted && UI.nMouseY > nStartY && UI.nMouseX > nLockedToolTipX && UI.nMouseX <= nLockedToolTipX + nToolTipWidth && (UI.nMouseLeft || UI.nMouseRight) )\n\t\t\t\t\t{\n\t\t\t\t\t\tbDeleted = true;\n\t\t\t\t\t\tint j = i;\n\t\t\t\t\t\tfor(; j < MICROPROFILE_TOOLTIP_MAX_LOCKED-1; ++j)\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tint nIndex0 = (g_MicroProfileUI.LockedToolTipFront + j) % MICROPROFILE_TOOLTIP_MAX_LOCKED;\n\t\t\t\t\t\t\tint nIndex1 = (g_MicroProfileUI.LockedToolTipFront + j+1) % MICROPROFILE_TOOLTIP_MAX_LOCKED;\n\t\t\t\t\t\t\tMicroProfileStringArrayCopy(&g_MicroProfileUI.LockedToolTips[nIndex0], &g_MicroProfileUI.LockedToolTips[nIndex1]);\n\t\t\t\t\t\t}\n\t\t\t\t\t\tMicroProfileStringArrayClear(&g_MicroProfileUI.LockedToolTips[(g_MicroProfileUI.LockedToolTipFront + j) % MICROPROFILE_TOOLTIP_MAX_LOCKED]);\n\t\t\t\t\t}\n\t\t\t\t\telse\n\t\t\t\t\t{\n\t\t\t\t\t\tMicroProfileDrawFloatWindow(nLockedToolTipX, nHeight-nToolTipHeight-2, &g_MicroProfileUI.LockedToolTips[nIndex].ppStrings[0], g_MicroProfileUI.LockedToolTips[nIndex].nNumStrings, g_MicroProfileUI.nLockedToolTipColor[nIndex]);\n\t\t\t\t\t\tnLockedToolTipX += nToolTipWidth + 4;\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tif(UI.nActiveMenu == 9)\n\t\t\t{\n\t\t\t\tif(S.nDisplay & MP_DRAW_DETAILED)\n\t\t\t\t{\n\t\t\t\t\tMicroProfileStringArray DetailedHelp;\n\t\t\t\t\tMicroProfileStringArrayClear(&DetailedHelp);\n\t\t\t\t\tMicroProfileStringArrayFormat(&DetailedHelp, \"%s\", MICROPROFILE_HELP_LEFT);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHelp, \"Toggle Graph\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&DetailedHelp, \"%s\", MICROPROFILE_HELP_RIGHT);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHelp, \"Zoom\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&DetailedHelp, \"%s + %s\", MICROPROFILE_HELP_MOD, MICROPROFILE_HELP_LEFT);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHelp, \"Lock Tooltip\");\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHelp, \"Drag\");\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHelp, \"Pan View\");\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHelp, \"Mouse Wheel\");\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHelp, \"Zoom\");\n\t\t\t\t\tMicroProfileDrawFloatWindow(nWidth, MICROPROFILE_FRAME_HISTORY_HEIGHT+20, DetailedHelp.ppStrings, DetailedHelp.nNumStrings, 0xff777777);\n\n\t\t\t\t\tMicroProfileStringArray DetailedHistoryHelp;\n\t\t\t\t\tMicroProfileStringArrayClear(&DetailedHistoryHelp);\n\t\t\t\t\tMicroProfileStringArrayFormat(&DetailedHistoryHelp, \"%s\", MICROPROFILE_HELP_LEFT);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHistoryHelp, \"Center View\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&DetailedHistoryHelp, \"%s\", MICROPROFILE_HELP_RIGHT);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&DetailedHistoryHelp, \"Zoom to frame\");\n\t\t\t\t\tMicroProfileDrawFloatWindow(nWidth, 20, DetailedHistoryHelp.ppStrings, DetailedHistoryHelp.nNumStrings, 0xff777777);\n\n\n\n\t\t\t\t}\n\t\t\t\telse if(0 != (S.nDisplay & MP_DRAW_BARS) && S.nBars)\n\t\t\t\t{\n\t\t\t\t\tMicroProfileStringArray BarHelp;\n\t\t\t\t\tMicroProfileStringArrayClear(&BarHelp);\n\t\t\t\t\tMicroProfileStringArrayFormat(&BarHelp, \"%s\", MICROPROFILE_HELP_LEFT);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&BarHelp, \"Toggle Graph\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&BarHelp, \"%s + %s\", MICROPROFILE_HELP_MOD, MICROPROFILE_HELP_LEFT);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&BarHelp, \"Lock Tooltip\");\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&BarHelp, \"Drag\");\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&BarHelp, \"Pan View\");\n\t\t\t\t\tMicroProfileDrawFloatWindow(nWidth, MICROPROFILE_FRAME_HISTORY_HEIGHT+20, BarHelp.ppStrings, BarHelp.nNumStrings, 0xff777777);\n\n\t\t\t\t}\n\t\t\t\tMicroProfileStringArray Debug;\n\t\t\t\tMicroProfileStringArrayClear(&Debug);\n\t\t\t\tMicroProfileStringArrayAddLiteral(&Debug, \"Memory Usage\");\n\t\t\t\tMicroProfileStringArrayFormat(&Debug, \"%4.2fmb\", S.nMemUsage / (1024.f * 1024.f));\n#if MICROPROFILE_WEBSERVER\n\t\t\t\tMicroProfileStringArrayAddLiteral(&Debug, \"Web Server Port\");\n\t\t\t\tMicroProfileStringArrayFormat(&Debug, \"%d\", MicroProfileWebServerPort());\n#endif\n\t\t\t\tuint32_t nFrameNext = (S.nFrameCurrent+1) % MICROPROFILE_MAX_FRAME_HISTORY;\n\t\t\t\tMicroProfileFrameState* pFrameCurrent = &S.Frames[S.nFrameCurrent];\n\t\t\t\tMicroProfileFrameState* pFrameNext = &S.Frames[nFrameNext];\n\n\n\t\t\t\tMicroProfileStringArrayAddLiteral(&Debug, \"\");\n\t\t\t\tMicroProfileStringArrayAddLiteral(&Debug, \"\");\n\t\t\t\tMicroProfileStringArrayAddLiteral(&Debug, \"Usage\");\n\t\t\t\tMicroProfileStringArrayAddLiteral(&Debug, \"markers [frames] \");\n\n#if MICROPROFILE_CONTEXT_SWITCH_TRACE\n\t\t\t\tMicroProfileStringArrayAddLiteral(&Debug, \"Context Switch\");\n\t\t\t\tMicroProfileStringArrayFormat(&Debug, \"%9d [%7d]\", S.nContextSwitchUsage, MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE / S.nContextSwitchUsage );\n#endif\n\n\t\t\t\tfor(int i = 0; i < MICROPROFILE_MAX_THREADS; ++i)\n\t\t\t\t{\n\t\t\t\t\tif(pFrameCurrent->nLogStart[i] && S.Pool[i])\n\t\t\t\t\t{\n\t\t\t\t\t\tuint32_t nEnd = pFrameNext->nLogStart[i];\n\t\t\t\t\t\tuint32_t nStart = pFrameCurrent->nLogStart[i];\n\t\t\t\t\t\tuint32_t nUsage = nStart <= nEnd ? (nEnd - nStart) : (nEnd + MICROPROFILE_BUFFER_SIZE - nStart);\n\t\t\t\t\t\tuint32_t nFrameSupport = (nUsage == 0) ? MICROPROFILE_BUFFER_SIZE : MICROPROFILE_BUFFER_SIZE / nUsage;\n\t\t\t\t\t\tMicroProfileStringArrayFormat(&Debug, \"%s\", &S.Pool[i]->ThreadName[0]);\n\t\t\t\t\t\tMicroProfileStringArrayFormat(&Debug, \"%9d [%7d]\", nUsage, nFrameSupport);\n\t\t\t\t\t}\n\t\t\t\t}\n\n\t\t\t\tMicroProfileDrawFloatWindow(0, nHeight-10, Debug.ppStrings, Debug.nNumStrings, 0xff777777);\n\t\t\t}\n\n\n\n\t\t\tif(UI.nActiveMenu == (uint32_t)-1 && !bMouseOverGraph)\n\t\t\t{\n\t\t\t\tif(UI.nHoverToken != MICROPROFILE_INVALID_TOKEN)\n\t\t\t\t{\n\t\t\t\t\tMicroProfileDrawFloatTooltip(UI.nMouseX, UI.nMouseY, (uint32_t)UI.nHoverToken, UI.nHoverTime);\n\t\t\t\t}\n\t\t\t\telse if(S.nContextSwitchHoverThreadAfter != (uint32_t)-1 && S.nContextSwitchHoverThreadBefore != (uint32_t)-1)\n\t\t\t\t{\n\t\t\t\t\tfloat fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu());\n\t\t\t\t\tMicroProfileStringArray ToolTip;\n\t\t\t\t\tMicroProfileStringArrayClear(&ToolTip);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Context Switch\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%04x\", S.nContextSwitchHoverThread);\t\t\t\t\t\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Before\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%04x\", S.nContextSwitchHoverThreadBefore);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"After\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%04x\", S.nContextSwitchHoverThreadAfter);\t\t\t\t\t\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"Duration\");\n\t\t\t\t\tint64_t nDifference = MicroProfileLogTickDifference(S.nContextSwitchHoverTickIn, S.nContextSwitchHoverTickOut);\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%6.2fms\", fToMs * nDifference );\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"CPU\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%d\", S.nContextSwitchHoverCpu);\n\t\t\t\t\tMicroProfileDrawFloatWindow(UI.nMouseX, UI.nMouseY+20, &ToolTip.ppStrings[0], ToolTip.nNumStrings, -1);\n\n\n\t\t\t\t}\n\t\t\t\telse if(UI.nHoverFrame != -1)\n\t\t\t\t{\n\t\t\t\t\tuint32_t nNextFrame = (UI.nHoverFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY;\n\t\t\t\t\tint64_t nTick = S.Frames[UI.nHoverFrame].nFrameStartCpu;\n\t\t\t\t\tint64_t nTickNext = S.Frames[nNextFrame].nFrameStartCpu;\n\t\t\t\t\tint64_t nTickGpu = S.Frames[UI.nHoverFrame].nFrameStartGpu;\n\t\t\t\t\tint64_t nTickNextGpu = S.Frames[nNextFrame].nFrameStartGpu;\n\n\t\t\t\t\tfloat fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu());\n\t\t\t\t\tfloat fToMsGpu = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondGpu());\n\t\t\t\t\tfloat fMs = fToMs * (nTickNext - nTick);\n\t\t\t\t\tfloat fMsGpu = fToMsGpu * (nTickNextGpu - nTickGpu);\n\t\t\t\t\tMicroProfileStringArray ToolTip;\n\t\t\t\t\tMicroProfileStringArrayClear(&ToolTip);\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"Frame %d\", UI.nHoverFrame);\n\t#if MICROPROFILE_DEBUG\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%p\", &S.Frames[UI.nHoverFrame]);\n\t#else\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"\");\n\t#endif\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"CPU Time\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%6.2fms\", fMs);\n\t\t\t\t\tMicroProfileStringArrayAddLiteral(&ToolTip, \"GPU Time\");\n\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%6.2fms\", fMsGpu);\n\t\t\t\t\t#if MICROPROFILE_DEBUG\n\t\t\t\t\tfor(int i = 0; i < MICROPROFILE_MAX_THREADS; ++i)\n\t\t\t\t\t{\n\t\t\t\t\t\tif(S.Frames[UI.nHoverFrame].nLogStart[i])\n\t\t\t\t\t\t{\n\t\t\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%d\", i);\n\t\t\t\t\t\t\tMicroProfileStringArrayFormat(&ToolTip, \"%d\", S.Frames[UI.nHoverFrame].nLogStart[i]);\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\t#endif\n\t\t\t\t\tMicroProfileDrawFloatWindow(UI.nMouseX, UI.nMouseY+20, &ToolTip.ppStrings[0], ToolTip.nNumStrings, -1);\n\t\t\t\t}\n\t\t\t\tif(UI.nMouseLeft)\n\t\t\t\t{\n\t\t\t\t\tif(UI.nHoverToken != MICROPROFILE_INVALID_TOKEN)\n\t\t\t\t\t\tMicroProfileToggleGraph(UI.nHoverToken);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\n#if MICROPROFILE_DRAWCURSOR\n\t\t{\n\t\t\tfloat fCursor[8] = \n\t\t\t{\n\t\t\t\tfloat(MicroProfileMax(0, (int)UI.nMouseX-3)), float(UI.nMouseY),\n\t\t\t\tfloat(MicroProfileMin(nWidth, UI.nMouseX+3)), float(UI.nMouseY),\n\t\t\t\tfloat(UI.nMouseX), float(MicroProfileMax((int)UI.nMouseY-3, 0)),\n\t\t\t\tfloat(UI.nMouseX), float(MicroProfileMin(nHeight, UI.nMouseY+3)),\n\t\t\t};\n\t\t\tMicroProfileDrawLine2D(2, &fCursor[0], 0xff00ff00);\n\t\t\tMicroProfileDrawLine2D(2, &fCursor[4], 0xff00ff00);\n\t\t}\n#endif\n\t\tm.unlock();\n\t}\n\telse if(UI.nCustomActive != (uint32_t)-1)\n\t{\n\t\tstd::recursive_mutex& m = MicroProfileGetMutex();\n\t\tm.lock();\n\t\tMicroProfileDrawGraph(nWidth, nHeight);\n\t\tMicroProfileDrawCustom(nWidth, nHeight);\n\t\tm.unlock();\n\n\t}\n\tUI.nMouseLeft = UI.nMouseRight = 0;\n\tUI.nMouseLeftMod = UI.nMouseRightMod = 0;\n\tUI.nMouseWheelDelta = 0;\n\tif(S.nOverflow)\n\t\tS.nOverflow--;\n\n\tUI.fDetailedOffset = UI.fDetailedOffset + (UI.fDetailedOffsetTarget - UI.fDetailedOffset) * MICROPROFILE_ANIM_DELAY_PRC;\n\tUI.fDetailedRange = UI.fDetailedRange + (UI.fDetailedRangeTarget - UI.fDetailedRange) * MICROPROFILE_ANIM_DELAY_PRC;\n\n\n}\n\nbool MicroProfileIsDrawing()\n{\n\tMicroProfile& S = *MicroProfileGet();\t\n\treturn S.nDisplay != 0;\n}\n\nvoid MicroProfileToggleGraph(MicroProfileToken nToken)\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tuint32_t nTimerId = MicroProfileGetTimerIndex(nToken);\n\tnToken &= 0xffff;\n\tint32_t nMinSort = 0x7fffffff;\n\tint32_t nFreeIndex = -1;\n\tint32_t nMinIndex = 0;\n\tint32_t nMaxSort = 0x80000000;\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t{\n\t\tif(S.Graph[i].nToken == MICROPROFILE_INVALID_TOKEN)\n\t\t\tnFreeIndex = i;\n\t\tif(S.Graph[i].nToken == nToken)\n\t\t{\n\t\t\tS.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN;\n\t\t\tS.TimerInfo[nTimerId].bGraph = false;\n\t\t\treturn;\n\t\t}\n\t\tif(S.Graph[i].nKey < nMinSort)\n\t\t{\n\t\t\tnMinSort = S.Graph[i].nKey;\n\t\t\tnMinIndex = i;\n\t\t}\n\t\tif(S.Graph[i].nKey > nMaxSort)\n\t\t{\n\t\t\tnMaxSort = S.Graph[i].nKey;\n\t\t}\n\t}\n\tint nIndex = nFreeIndex > -1 ? nFreeIndex : nMinIndex;\n\tif (nFreeIndex == -1)\n\t{\n\t\tuint32_t idx = MicroProfileGetTimerIndex(S.Graph[nIndex].nToken);\n\t\tS.TimerInfo[idx].bGraph = false;\n\t}\n\tS.Graph[nIndex].nToken = nToken;\n\tS.Graph[nIndex].nKey = nMaxSort+1;\n\tmemset(&S.Graph[nIndex].nHistory[0], 0, sizeof(S.Graph[nIndex].nHistory));\n\tS.TimerInfo[nTimerId].bGraph = true;\n}\n\n\nvoid MicroProfileMousePosition(uint32_t nX, uint32_t nY, int nWheelDelta)\n{\n\tUI.nMouseX = nX;\n\tUI.nMouseY = nY;\n\tUI.nMouseWheelDelta = nWheelDelta;\n}\n\nvoid MicroProfileModKey(uint32_t nKeyState)\n{\n\tUI.nModDown = nKeyState ? 1 : 0;\n}\n\nvoid MicroProfileClearGraph()\n{\n\tMicroProfile& S = *MicroProfileGet();\t\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t{\n\t\tif(S.Graph[i].nToken != 0)\n\t\t{\n\t\t\tS.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN;\n\t\t}\n\t}\n}\n\nvoid MicroProfileMouseButton(uint32_t nLeft, uint32_t nRight)\n{\n\tbool bCanRelease = abs((int)(UI.nMouseDownX - UI.nMouseX)) + abs((int)(UI.nMouseDownY - UI.nMouseY)) < 3;\n\n\tif(0 == nLeft && UI.nMouseDownLeft && bCanRelease)\n\t{\n\t\tif(UI.nModDown)\n\t\t\tUI.nMouseLeftMod = 1;\n\t\telse\n\t\t\tUI.nMouseLeft = 1;\n\t}\n\n\tif(0 == nRight && UI.nMouseDownRight && bCanRelease)\n\t{\n\t\tif(UI.nModDown)\n\t\t\tUI.nMouseRightMod = 1;\n\t\telse\n\t\t\tUI.nMouseRight = 1;\n\t}\n\tif((nLeft || nRight) && !(UI.nMouseDownLeft || UI.nMouseDownRight))\n\t{\n\t\tUI.nMouseDownX = UI.nMouseX;\n\t\tUI.nMouseDownY = UI.nMouseY;\n\t}\n\n\tUI.nMouseDownLeft = nLeft;\n\tUI.nMouseDownRight = nRight;\n\t\n}\n\nvoid MicroProfileDrawLineVertical(int nX, int nTop, int nBottom, uint32_t nColor)\n{\n\tMicroProfileDrawBox(nX, nTop, nX + 1, nBottom, nColor);\n}\n\nvoid MicroProfileDrawLineHorizontal(int nLeft, int nRight, int nY, uint32_t nColor)\n{\n\tMicroProfileDrawBox(nLeft, nY, nRight, nY + 1, nColor);\n}\n\n\n\n#include \n\n#define MICROPROFILE_PRESET_HEADER_MAGIC 0x28586813\n#define MICROPROFILE_PRESET_HEADER_VERSION 0x00000102\nstruct MicroProfilePresetHeader\n{\n\tuint32_t nMagic;\n\tuint32_t nVersion;\n\t//groups, threads, aggregate, reference frame, graphs timers\n\tuint32_t nGroups[MICROPROFILE_MAX_GROUPS];\n\tuint32_t nThreads[MICROPROFILE_MAX_THREADS];\n\tuint32_t nGraphName[MICROPROFILE_MAX_GRAPHS];\n\tuint32_t nGraphGroupName[MICROPROFILE_MAX_GRAPHS];\n\tuint32_t nAllGroupsWanted;\n\tuint32_t nAllThreadsWanted;\n\tuint32_t nAggregateFlip;\n\tfloat fReferenceTime;\n\tuint32_t nBars;\n\tuint32_t nDisplay;\n\tuint32_t nOpacityBackground;\n\tuint32_t nOpacityForeground;\n\tuint32_t nShowSpikes;\n};\n\n#ifndef MICROPROFILE_PRESET_FILENAME_FUNC\n#define MICROPROFILE_PRESET_FILENAME_FUNC MicroProfilePresetFilename\nstatic const char* MicroProfilePresetFilename(const char* pSuffix)\n{\n\tstatic char filename[512];\n\tsnprintf(filename, sizeof(filename)-1, \".microprofilepreset.%s\", pSuffix);\n\treturn filename;\n}\n#endif\n\nvoid MicroProfileSavePreset(const char* pPresetName)\n{\n\tstd::lock_guard Lock(MicroProfileGetMutex());\n\tFILE* F = fopen(MICROPROFILE_PRESET_FILENAME_FUNC(pPresetName), \"wb\");\n\tif(!F) return;\n\n\tMicroProfile& S = *MicroProfileGet();\n\n\tMicroProfilePresetHeader Header;\n\tmemset(&Header, 0, sizeof(Header));\n\tHeader.nAggregateFlip = S.nAggregateFlip;\n\tHeader.nBars = S.nBars;\n\tHeader.fReferenceTime = S.fReferenceTime;\n\tHeader.nAllGroupsWanted = S.nAllGroupsWanted;\n\tHeader.nAllThreadsWanted = S.nAllThreadsWanted;\n\tHeader.nMagic = MICROPROFILE_PRESET_HEADER_MAGIC;\n\tHeader.nVersion = MICROPROFILE_PRESET_HEADER_VERSION;\n\tHeader.nDisplay = S.nDisplay;\n\tHeader.nOpacityBackground = UI.nOpacityBackground;\n\tHeader.nOpacityForeground = UI.nOpacityForeground;\n\tHeader.nShowSpikes = UI.bShowSpikes ? 1 : 0;\n\tfwrite(&Header, sizeof(Header), 1, F);\n\tuint64_t nMask = 1;\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GROUPS; ++i)\n\t{\n\t\tif(S.nActiveGroupWanted & nMask)\n\t\t{\n\t\t\tuint32_t offset = ftell(F);\n\t\t\tconst char* pName = S.GroupInfo[i].pName;\n\t\t\tint nLen = (int)strlen(pName)+1;\n\t\t\tfwrite(pName, nLen, 1, F);\n\t\t\tHeader.nGroups[i] = offset;\n\t\t}\n\t\tnMask <<= 1;\n\t}\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS; ++i)\n\t{\n\t\tMicroProfileThreadLog* pLog = S.Pool[i];\n\t\tif(pLog && S.nThreadActive[i])\n\t\t{\n\t\t\tuint32_t nOffset = ftell(F);\n\t\t\tconst char* pName = &pLog->ThreadName[0];\n\t\t\tint nLen = (int)strlen(pName)+1;\n\t\t\tfwrite(pName, nLen, 1, F);\n\t\t\tHeader.nThreads[i] = nOffset;\n\t\t}\n\t}\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t{\n\t\tMicroProfileToken nToken = S.Graph[i].nToken;\n\t\tif(nToken != MICROPROFILE_INVALID_TOKEN)\n\t\t{\n\t\t\tuint32_t nGroupIndex = MicroProfileGetGroupIndex(nToken);\n\t\t\tuint32_t nTimerIndex = MicroProfileGetTimerIndex(nToken);\n\t\t\tconst char* pGroupName = S.GroupInfo[nGroupIndex].pName;\n\t\t\tconst char* pTimerName = S.TimerInfo[nTimerIndex].pName;\n\t\t\tMP_ASSERT(pGroupName);\n\t\t\tMP_ASSERT(pTimerName);\n\t\t\tint nGroupLen = (int)strlen(pGroupName)+1;\n\t\t\tint nTimerLen = (int)strlen(pTimerName)+1;\n\n\t\t\tuint32_t nOffsetGroup = ftell(F);\n\t\t\tfwrite(pGroupName, nGroupLen, 1, F);\n\t\t\tuint32_t nOffsetTimer = ftell(F);\n\t\t\tfwrite(pTimerName, nTimerLen, 1, F);\n\t\t\tHeader.nGraphName[i] = nOffsetTimer;\n\t\t\tHeader.nGraphGroupName[i] = nOffsetGroup;\n\t\t}\n\t}\n\tfseek(F, 0, SEEK_SET);\n\tfwrite(&Header, sizeof(Header), 1, F);\n\n\tfclose(F);\n\n}\n\n\n\nvoid MicroProfileLoadPreset(const char* pSuffix)\n{\n\tstd::lock_guard Lock(MicroProfileGetMutex());\n\tFILE* F = fopen(MICROPROFILE_PRESET_FILENAME_FUNC(pSuffix), \"rb\");\n\tif(!F)\n\t{\n\t \treturn;\n\t}\n\tfseek(F, 0, SEEK_END);\n\tint nSize = ftell(F);\n\tchar* const pBuffer = (char*)alloca(nSize);\n\tfseek(F, 0, SEEK_SET);\n\tint nRead = (int)fread(pBuffer, nSize, 1, F);\n\tfclose(F);\n\tif(1 != nRead)\n\t\treturn;\n\n\tMicroProfile& S = *MicroProfileGet();\n\t\n\tMicroProfilePresetHeader& Header = *(MicroProfilePresetHeader*)pBuffer;\n\n\tif(Header.nMagic != MICROPROFILE_PRESET_HEADER_MAGIC || Header.nVersion != MICROPROFILE_PRESET_HEADER_VERSION)\n\t{\n\t\treturn;\n\t}\n\n\tS.nAggregateFlip = Header.nAggregateFlip;\n\tS.nBars = Header.nBars;\n\tS.fReferenceTime = Header.fReferenceTime;\n\tS.fRcpReferenceTime = 1.f / Header.fReferenceTime;\n\tS.nAllGroupsWanted = Header.nAllGroupsWanted;\n\tS.nAllThreadsWanted = Header.nAllThreadsWanted;\n\tS.nDisplay = Header.nDisplay;\n\tS.nActiveGroupWanted = 0;\n\tUI.nOpacityBackground = Header.nOpacityBackground;\n\tUI.nOpacityForeground = Header.nOpacityForeground;\n\tUI.bShowSpikes = Header.nShowSpikes == 1;\n\n\tmemset(&S.nThreadActive[0], 0, sizeof(S.nThreadActive));\n\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GROUPS; ++i)\n\t{\n\t\tif(Header.nGroups[i])\n\t\t{\n\t\t\tconst char* pGroupName = pBuffer + Header.nGroups[i];\t\n\t\t\tfor(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j)\n\t\t\t{\n\t\t\t\tif(0 == MP_STRCASECMP(pGroupName, S.GroupInfo[j].pName))\n\t\t\t\t{\n\t\t\t\t\tS.nActiveGroupWanted |= (1ll << j);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS; ++i)\n\t{\n\t\tif(Header.nThreads[i])\n\t\t{\n\t\t\tconst char* pThreadName = pBuffer + Header.nThreads[i];\n\t\t\tfor(uint32_t j = 0; j < MICROPROFILE_MAX_THREADS; ++j)\n\t\t\t{\n\t\t\t\tMicroProfileThreadLog* pLog = S.Pool[j];\n\t\t\t\tif(pLog && 0 == MP_STRCASECMP(pThreadName, &pLog->ThreadName[0]))\n\t\t\t\t{\n\t\t\t\t\tS.nThreadActive[j] = 1;\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t{\n\t\tMicroProfileToken nPrevToken = S.Graph[i].nToken;\n\t\tS.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN;\n\t\tif(Header.nGraphName[i] && Header.nGraphGroupName[i])\n\t\t{\n\t\t\tconst char* pGraphName = pBuffer + Header.nGraphName[i];\n\t\t\tconst char* pGraphGroupName = pBuffer + Header.nGraphGroupName[i];\n\t\t\tfor(uint32_t j = 0; j < S.nTotalTimers; ++j)\n\t\t\t{\n\t\t\t\tuint64_t nGroupIndex = S.TimerInfo[j].nGroupIndex;\n\t\t\t\tif(0 == MP_STRCASECMP(pGraphName, S.TimerInfo[j].pName) && 0 == MP_STRCASECMP(pGraphGroupName, S.GroupInfo[nGroupIndex].pName))\n\t\t\t\t{\n\t\t\t\t\tMicroProfileToken nToken = MicroProfileMakeToken(1ll << nGroupIndex, (uint16_t)j);\n\t\t\t\t\tS.Graph[i].nToken = nToken;\t\t\t// note: group index is stored here but is checked without in MicroProfileToggleGraph()!\n\t\t\t\t\tS.TimerInfo[j].bGraph = true;\n\t\t\t\t\tif(nToken != nPrevToken)\n\t\t\t\t\t{\n\t\t\t\t\t\tmemset(&S.Graph[i].nHistory, 0, sizeof(S.Graph[i].nHistory));\n\t\t\t\t\t}\n\t\t\t\t\tbreak;\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}\n}\n\nuint32_t MicroProfileCustomGroupFind(const char* pCustomName)\n{\n\tfor(uint32_t i = 0; i < UI.nCustomCount; ++i)\n\t{\n\t\tif(!MP_STRCASECMP(pCustomName, UI.Custom[i].pName))\n\t\t{\n\t\t\treturn i;\n\t\t}\n\t}\n\treturn (uint32_t)-1;\n}\n\nuint32_t MicroProfileCustomGroup(const char* pCustomName)\n{\n\tfor(uint32_t i = 0; i < UI.nCustomCount; ++i)\n\t{\n\t\tif(!MP_STRCASECMP(pCustomName, UI.Custom[i].pName))\n\t\t{\n\t\t\treturn i;\n\t\t}\n\t}\n\tMP_ASSERT(UI.nCustomCount < MICROPROFILE_CUSTOM_MAX);\n\tuint32_t nIndex = UI.nCustomCount;\n\tUI.nCustomCount++;\n\tmemset(&UI.Custom[nIndex], 0, sizeof(UI.Custom[nIndex]));\n\tuint32_t nLen = (uint32_t)strlen(pCustomName);\n\tif(nLen > MICROPROFILE_NAME_MAX_LEN-1)\n\t\tnLen = MICROPROFILE_NAME_MAX_LEN-1;\n\tmemcpy(&UI.Custom[nIndex].pName[0], pCustomName, nLen);\n\tUI.Custom[nIndex].pName[nLen] = '\\0';\n\treturn nIndex;\n}\nvoid MicroProfileCustomGroup(const char* pCustomName, uint32_t nMaxTimers, uint32_t nAggregateFlip, float fReferenceTime, uint32_t nFlags)\n{\n\tuint32_t nIndex = MicroProfileCustomGroup(pCustomName);\n\tMP_ASSERT(UI.Custom[nIndex].pTimers == 0);//only call once!\n\tUI.Custom[nIndex].pTimers = &UI.CustomTimer[UI.nCustomTimerCount];\n\tUI.Custom[nIndex].nMaxTimers = nMaxTimers;\n\tUI.Custom[nIndex].fReference = fReferenceTime;\n\tUI.nCustomTimerCount += nMaxTimers;\t\n\tMP_ASSERT(UI.nCustomTimerCount <= MICROPROFILE_CUSTOM_MAX_TIMERS); //bump MICROPROFILE_CUSTOM_MAX_TIMERS\n\tUI.Custom[nIndex].nFlags = nFlags;\n\tUI.Custom[nIndex].nAggregateFlip = nAggregateFlip;\n}\n\nvoid MicroProfileCustomGroupEnable(uint32_t nIndex)\n{\n\tif(nIndex < UI.nCustomCount)\n\t{\n\t\tMicroProfile& S = *MicroProfileGet();\n\t\tS.nForceGroupUI = UI.Custom[nIndex].nGroupMask;\n\t\tMicroProfileSetAggregateFrames(UI.Custom[nIndex].nAggregateFlip);\n\t\tS.fReferenceTime = UI.Custom[nIndex].fReference;\n\t\tS.fRcpReferenceTime = 1.f / UI.Custom[nIndex].fReference;\n\t\tUI.nCustomActive = nIndex;\n\n\t\tfor(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)\n\t\t{\n\t\t\tif(S.Graph[i].nToken != MICROPROFILE_INVALID_TOKEN)\n\t\t\t{\n\t\t\t\tuint32_t nTimerId = MicroProfileGetTimerIndex(S.Graph[i].nToken);\n\t\t\t\tS.TimerInfo[nTimerId].bGraph = false;\n\t\t\t\tS.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN;\n\t\t\t}\n\t\t}\n\n\t\tfor(uint32_t i = 0; i < UI.Custom[nIndex].nNumTimers; ++i)\n\t\t{\n\t\t\tif(i == MICROPROFILE_MAX_GRAPHS)\n\t\t\t{\n\t\t\t\tbreak;\n\t\t\t}\n\t\t\tS.Graph[i].nToken = UI.Custom[nIndex].pTimers[i];\n\t\t\tS.Graph[i].nKey = i;\n\t\t\tuint32_t nTimerId = MicroProfileGetTimerIndex(S.Graph[i].nToken);\n\t\t\tS.TimerInfo[nTimerId].bGraph = true;\n\t\t}\n\t}\n}\n\nvoid MicroProfileCustomGroupToggle(const char* pCustomName)\n{\n\tuint32_t nIndex = MicroProfileCustomGroupFind(pCustomName);\n\tif(nIndex == (uint32_t)-1 || nIndex == UI.nCustomActive)\n\t{\n\t\tMicroProfileCustomGroupDisable();\n\t}\n\telse\n\t{\n\t\tMicroProfileCustomGroupEnable(nIndex);\n\t}\n}\n\nvoid MicroProfileCustomGroupEnable(const char* pCustomName)\n{\n\tuint32_t nIndex = MicroProfileCustomGroupFind(pCustomName);\n\tMicroProfileCustomGroupEnable(nIndex);\n}\nvoid MicroProfileCustomGroupDisable()\n{\n\tMicroProfile& S = *MicroProfileGet();\n\tS.nForceGroupUI = 0;\n\tUI.nCustomActive = (uint32_t)-1;\n}\n\nvoid MicroProfileCustomGroupAddTimer(const char* pCustomName, const char* pGroup, const char* pTimer)\n{\n\tuint32_t nIndex = MicroProfileCustomGroupFind(pCustomName);\n\tif((uint32_t)-1 == nIndex)\n\t{\n\t\treturn;\n\t}\n\tuint32_t nTimerIndex = UI.Custom[nIndex].nNumTimers;\n\tMP_ASSERT(nTimerIndex < UI.Custom[nIndex].nMaxTimers);\n\tuint64_t nToken = MicroProfileFindToken(pGroup, pTimer);\n\tMP_ASSERT(nToken != MICROPROFILE_INVALID_TOKEN); //Timer must be registered first.\n\tUI.Custom[nIndex].pTimers[nTimerIndex] = nToken;\t\n\tuint16_t nGroup = MicroProfileGetGroupIndex(nToken);\n\tUI.Custom[nIndex].nGroupMask |= (1ll << nGroup);\n\tUI.Custom[nIndex].nNumTimers++;\n}\n\n#undef UI\n\n#endif\n#endif\n" }, { "path": "third_party/mspack/COPYING.LIB", "content": "\t\t GNU LESSER GENERAL PUBLIC LICENSE\n\t\t Version 2.1, February 1999\n\n Copyright (C) 1991, 1999 Free Software Foundation, Inc.\n 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA\n Everyone is permitted to copy and distribute verbatim copies\n of this license document, but changing it is not allowed.\n\n[This is the first released version of the Lesser GPL. It also counts\n as the successor of the GNU Library Public License, version 2, hence\n the version number 2.1.]\n\n\t\t\t Preamble\n\n The licenses for most software are designed to take away your\nfreedom to share and change it. By contrast, the GNU General Public\nLicenses are intended to guarantee your freedom to share and change\nfree software--to make sure the software is free for all its users.\n\n This license, the Lesser General Public License, applies to some\nspecially designated software packages--typically libraries--of the\nFree Software Foundation and other authors who decide to use it. You\ncan use it too, but we suggest you first think carefully about whether\nthis license or the ordinary General Public License is the better\nstrategy to use in any particular case, based on the explanations below.\n\n When we speak of free software, we are referring to freedom of use,\nnot price. Our General Public Licenses are designed to make sure that\nyou have the freedom to distribute copies of free software (and charge\nfor this service if you wish); that you receive source code or can get\nit if you want it; that you can change the software and use pieces of\nit in new free programs; and that you are informed that you can do\nthese things.\n\n To protect your rights, we need to make restrictions that forbid\ndistributors to deny you these rights or to ask you to surrender these\nrights. These restrictions translate to certain responsibilities for\nyou if you distribute copies of the library or if you modify it.\n\n For example, if you distribute copies of the library, whether gratis\nor for a fee, you must give the recipients all the rights that we gave\nyou. You must make sure that they, too, receive or can get the source\ncode. If you link other code with the library, you must provide\ncomplete object files to the recipients, so that they can relink them\nwith the library after making changes to the library and recompiling\nit. And you must show them these terms so they know their rights.\n\n We protect your rights with a two-step method: (1) we copyright the\nlibrary, and (2) we offer you this license, which gives you legal\npermission to copy, distribute and/or modify the library.\n\n To protect each distributor, we want to make it very clear that\nthere is no warranty for the free library. Also, if the library is\nmodified by someone else and passed on, the recipients should know\nthat what they have is not the original version, so that the original\nauthor's reputation will not be affected by problems that might be\nintroduced by others.\n\f\n Finally, software patents pose a constant threat to the existence of\nany free program. We wish to make sure that a company cannot\neffectively restrict the users of a free program by obtaining a\nrestrictive license from a patent holder. Therefore, we insist that\nany patent license obtained for a version of the library must be\nconsistent with the full freedom of use specified in this license.\n\n Most GNU software, including some libraries, is covered by the\nordinary GNU General Public License. This license, the GNU Lesser\nGeneral Public License, applies to certain designated libraries, and\nis quite different from the ordinary General Public License. We use\nthis license for certain libraries in order to permit linking those\nlibraries into non-free programs.\n\n When a program is linked with a library, whether statically or using\na shared library, the combination of the two is legally speaking a\ncombined work, a derivative of the original library. The ordinary\nGeneral Public License therefore permits such linking only if the\nentire combination fits its criteria of freedom. The Lesser General\nPublic License permits more lax criteria for linking other code with\nthe library.\n\n We call this license the \"Lesser\" General Public License because it\ndoes Less to protect the user's freedom than the ordinary General\nPublic License. It also provides other free software developers Less\nof an advantage over competing non-free programs. These disadvantages\nare the reason we use the ordinary General Public License for many\nlibraries. However, the Lesser license provides advantages in certain\nspecial circumstances.\n\n For example, on rare occasions, there may be a special need to\nencourage the widest possible use of a certain library, so that it becomes\na de-facto standard. To achieve this, non-free programs must be\nallowed to use the library. A more frequent case is that a free\nlibrary does the same job as widely used non-free libraries. In this\ncase, there is little to gain by limiting the free library to free\nsoftware only, so we use the Lesser General Public License.\n\n In other cases, permission to use a particular library in non-free\nprograms enables a greater number of people to use a large body of\nfree software. For example, permission to use the GNU C Library in\nnon-free programs enables many more people to use the whole GNU\noperating system, as well as its variant, the GNU/Linux operating\nsystem.\n\n Although the Lesser General Public License is Less protective of the\nusers' freedom, it does ensure that the user of a program that is\nlinked with the Library has the freedom and the wherewithal to run\nthat program using a modified version of the Library.\n\n The precise terms and conditions for copying, distribution and\nmodification follow. Pay close attention to the difference between a\n\"work based on the library\" and a \"work that uses the library\". The\nformer contains code derived from the library, whereas the latter must\nbe combined with the library in order to run.\n\f\n\t\t GNU LESSER GENERAL PUBLIC LICENSE\n TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION\n\n 0. This License Agreement applies to any software library or other\nprogram which contains a notice placed by the copyright holder or\nother authorized party saying it may be distributed under the terms of\nthis Lesser General Public License (also called \"this License\").\nEach licensee is addressed as \"you\".\n\n A \"library\" means a collection of software functions and/or data\nprepared so as to be conveniently linked with application programs\n(which use some of those functions and data) to form executables.\n\n The \"Library\", below, refers to any such software library or work\nwhich has been distributed under these terms. A \"work based on the\nLibrary\" means either the Library or any derivative work under\ncopyright law: that is to say, a work containing the Library or a\nportion of it, either verbatim or with modifications and/or translated\nstraightforwardly into another language. (Hereinafter, translation is\nincluded without limitation in the term \"modification\".)\n\n \"Source code\" for a work means the preferred form of the work for\nmaking modifications to it. For a library, complete source code means\nall the source code for all modules it contains, plus any associated\ninterface definition files, plus the scripts used to control compilation\nand installation of the library.\n\n Activities other than copying, distribution and modification are not\ncovered by this License; they are outside its scope. The act of\nrunning a program using the Library is not restricted, and output from\nsuch a program is covered only if its contents constitute a work based\non the Library (independent of the use of the Library in a tool for\nwriting it). Whether that is true depends on what the Library does\nand what the program that uses the Library does.\n \n 1. You may copy and distribute verbatim copies of the Library's\ncomplete source code as you receive it, in any medium, provided that\nyou conspicuously and appropriately publish on each copy an\nappropriate copyright notice and disclaimer of warranty; keep intact\nall the notices that refer to this License and to the absence of any\nwarranty; and distribute a copy of this License along with the\nLibrary.\n\n You may charge a fee for the physical act of transferring a copy,\nand you may at your option offer warranty protection in exchange for a\nfee.\n\f\n 2. You may modify your copy or copies of the Library or any portion\nof it, thus forming a work based on the Library, and copy and\ndistribute such modifications or work under the terms of Section 1\nabove, provided that you also meet all of these conditions:\n\n a) The modified work must itself be a software library.\n\n b) You must cause the files modified to carry prominent notices\n stating that you changed the files and the date of any change.\n\n c) You must cause the whole of the work to be licensed at no\n charge to all third parties under the terms of this License.\n\n d) If a facility in the modified Library refers to a function or a\n table of data to be supplied by an application program that uses\n the facility, other than as an argument passed when the facility\n is invoked, then you must make a good faith effort to ensure that,\n in the event an application does not supply such function or\n table, the facility still operates, and performs whatever part of\n its purpose remains meaningful.\n\n (For example, a function in a library to compute square roots has\n a purpose that is entirely well-defined independent of the\n application. Therefore, Subsection 2d requires that any\n application-supplied function or table used by this function must\n be optional: if the application does not supply it, the square\n root function must still compute square roots.)\n\nThese requirements apply to the modified work as a whole. If\nidentifiable sections of that work are not derived from the Library,\nand can be reasonably considered independent and separate works in\nthemselves, then this License, and its terms, do not apply to those\nsections when you distribute them as separate works. But when you\ndistribute the same sections as part of a whole which is a work based\non the Library, the distribution of the whole must be on the terms of\nthis License, whose permissions for other licensees extend to the\nentire whole, and thus to each and every part regardless of who wrote\nit.\n\nThus, it is not the intent of this section to claim rights or contest\nyour rights to work written entirely by you; rather, the intent is to\nexercise the right to control the distribution of derivative or\ncollective works based on the Library.\n\nIn addition, mere aggregation of another work not based on the Library\nwith the Library (or with a work based on the Library) on a volume of\na storage or distribution medium does not bring the other work under\nthe scope of this License.\n\n 3. You may opt to apply the terms of the ordinary GNU General Public\nLicense instead of this License to a given copy of the Library. To do\nthis, you must alter all the notices that refer to this License, so\nthat they refer to the ordinary GNU General Public License, version 2,\ninstead of to this License. (If a newer version than version 2 of the\nordinary GNU General Public License has appeared, then you can specify\nthat version instead if you wish.) Do not make any other change in\nthese notices.\n\f\n Once this change is made in a given copy, it is irreversible for\nthat copy, so the ordinary GNU General Public License applies to all\nsubsequent copies and derivative works made from that copy.\n\n This option is useful when you wish to copy part of the code of\nthe Library into a program that is not a library.\n\n 4. You may copy and distribute the Library (or a portion or\nderivative of it, under Section 2) in object code or executable form\nunder the terms of Sections 1 and 2 above provided that you accompany\nit with the complete corresponding machine-readable source code, which\nmust be distributed under the terms of Sections 1 and 2 above on a\nmedium customarily used for software interchange.\n\n If distribution of object code is made by offering access to copy\nfrom a designated place, then offering equivalent access to copy the\nsource code from the same place satisfies the requirement to\ndistribute the source code, even though third parties are not\ncompelled to copy the source along with the object code.\n\n 5. A program that contains no derivative of any portion of the\nLibrary, but is designed to work with the Library by being compiled or\nlinked with it, is called a \"work that uses the Library\". Such a\nwork, in isolation, is not a derivative work of the Library, and\ntherefore falls outside the scope of this License.\n\n However, linking a \"work that uses the Library\" with the Library\ncreates an executable that is a derivative of the Library (because it\ncontains portions of the Library), rather than a \"work that uses the\nlibrary\". The executable is therefore covered by this License.\nSection 6 states terms for distribution of such executables.\n\n When a \"work that uses the Library\" uses material from a header file\nthat is part of the Library, the object code for the work may be a\nderivative work of the Library even though the source code is not.\nWhether this is true is especially significant if the work can be\nlinked without the Library, or if the work is itself a library. The\nthreshold for this to be true is not precisely defined by law.\n\n If such an object file uses only numerical parameters, data\nstructure layouts and accessors, and small macros and small inline\nfunctions (ten lines or less in length), then the use of the object\nfile is unrestricted, regardless of whether it is legally a derivative\nwork. (Executables containing this object code plus portions of the\nLibrary will still fall under Section 6.)\n\n Otherwise, if the work is a derivative of the Library, you may\ndistribute the object code for the work under the terms of Section 6.\nAny executables containing that work also fall under Section 6,\nwhether or not they are linked directly with the Library itself.\n\f\n 6. As an exception to the Sections above, you may also combine or\nlink a \"work that uses the Library\" with the Library to produce a\nwork containing portions of the Library, and distribute that work\nunder terms of your choice, provided that the terms permit\nmodification of the work for the customer's own use and reverse\nengineering for debugging such modifications.\n\n You must give prominent notice with each copy of the work that the\nLibrary is used in it and that the Library and its use are covered by\nthis License. You must supply a copy of this License. If the work\nduring execution displays copyright notices, you must include the\ncopyright notice for the Library among them, as well as a reference\ndirecting the user to the copy of this License. Also, you must do one\nof these things:\n\n a) Accompany the work with the complete corresponding\n machine-readable source code for the Library including whatever\n changes were used in the work (which must be distributed under\n Sections 1 and 2 above); and, if the work is an executable linked\n with the Library, with the complete machine-readable \"work that\n uses the Library\", as object code and/or source code, so that the\n user can modify the Library and then relink to produce a modified\n executable containing the modified Library. (It is understood\n that the user who changes the contents of definitions files in the\n Library will not necessarily be able to recompile the application\n to use the modified definitions.)\n\n b) Use a suitable shared library mechanism for linking with the\n Library. A suitable mechanism is one that (1) uses at run time a\n copy of the library already present on the user's computer system,\n rather than copying library functions into the executable, and (2)\n will operate properly with a modified version of the library, if\n the user installs one, as long as the modified version is\n interface-compatible with the version that the work was made with.\n\n c) Accompany the work with a written offer, valid for at\n least three years, to give the same user the materials\n specified in Subsection 6a, above, for a charge no more\n than the cost of performing this distribution.\n\n d) If distribution of the work is made by offering access to copy\n from a designated place, offer equivalent access to copy the above\n specified materials from the same place.\n\n e) Verify that the user has already received a copy of these\n materials or that you have already sent this user a copy.\n\n For an executable, the required form of the \"work that uses the\nLibrary\" must include any data and utility programs needed for\nreproducing the executable from it. However, as a special exception,\nthe materials to be distributed need not include anything that is\nnormally distributed (in either source or binary form) with the major\ncomponents (compiler, kernel, and so on) of the operating system on\nwhich the executable runs, unless that component itself accompanies\nthe executable.\n\n It may happen that this requirement contradicts the license\nrestrictions of other proprietary libraries that do not normally\naccompany the operating system. Such a contradiction means you cannot\nuse both them and the Library together in an executable that you\ndistribute.\n\f\n 7. You may place library facilities that are a work based on the\nLibrary side-by-side in a single library together with other library\nfacilities not covered by this License, and distribute such a combined\nlibrary, provided that the separate distribution of the work based on\nthe Library and of the other library facilities is otherwise\npermitted, and provided that you do these two things:\n\n a) Accompany the combined library with a copy of the same work\n based on the Library, uncombined with any other library\n facilities. This must be distributed under the terms of the\n Sections above.\n\n b) Give prominent notice with the combined library of the fact\n that part of it is a work based on the Library, and explaining\n where to find the accompanying uncombined form of the same work.\n\n 8. You may not copy, modify, sublicense, link with, or distribute\nthe Library except as expressly provided under this License. Any\nattempt otherwise to copy, modify, sublicense, link with, or\ndistribute the Library is void, and will automatically terminate your\nrights under this License. However, parties who have received copies,\nor rights, from you under this License will not have their licenses\nterminated so long as such parties remain in full compliance.\n\n 9. You are not required to accept this License, since you have not\nsigned it. However, nothing else grants you permission to modify or\ndistribute the Library or its derivative works. These actions are\nprohibited by law if you do not accept this License. Therefore, by\nmodifying or distributing the Library (or any work based on the\nLibrary), you indicate your acceptance of this License to do so, and\nall its terms and conditions for copying, distributing or modifying\nthe Library or works based on it.\n\n 10. Each time you redistribute the Library (or any work based on the\nLibrary), the recipient automatically receives a license from the\noriginal licensor to copy, distribute, link with or modify the Library\nsubject to these terms and conditions. You may not impose any further\nrestrictions on the recipients' exercise of the rights granted herein.\nYou are not responsible for enforcing compliance by third parties with\nthis License.\n\f\n 11. If, as a consequence of a court judgment or allegation of patent\ninfringement or for any other reason (not limited to patent issues),\nconditions are imposed on you (whether by court order, agreement or\notherwise) that contradict the conditions of this License, they do not\nexcuse you from the conditions of this License. If you cannot\ndistribute so as to satisfy simultaneously your obligations under this\nLicense and any other pertinent obligations, then as a consequence you\nmay not distribute the Library at all. For example, if a patent\nlicense would not permit royalty-free redistribution of the Library by\nall those who receive copies directly or indirectly through you, then\nthe only way you could satisfy both it and this License would be to\nrefrain entirely from distribution of the Library.\n\nIf any portion of this section is held invalid or unenforceable under any\nparticular circumstance, the balance of the section is intended to apply,\nand the section as a whole is intended to apply in other circumstances.\n\nIt is not the purpose of this section to induce you to infringe any\npatents or other property right claims or to contest validity of any\nsuch claims; this section has the sole purpose of protecting the\nintegrity of the free software distribution system which is\nimplemented by public license practices. Many people have made\ngenerous contributions to the wide range of software distributed\nthrough that system in reliance on consistent application of that\nsystem; it is up to the author/donor to decide if he or she is willing\nto distribute software through any other system and a licensee cannot\nimpose that choice.\n\nThis section is intended to make thoroughly clear what is believed to\nbe a consequence of the rest of this License.\n\n 12. If the distribution and/or use of the Library is restricted in\ncertain countries either by patents or by copyrighted interfaces, the\noriginal copyright holder who places the Library under this License may add\nan explicit geographical distribution limitation excluding those countries,\nso that distribution is permitted only in or among countries not thus\nexcluded. In such case, this License incorporates the limitation as if\nwritten in the body of this License.\n\n 13. The Free Software Foundation may publish revised and/or new\nversions of the Lesser General Public License from time to time.\nSuch new versions will be similar in spirit to the present version,\nbut may differ in detail to address new problems or concerns.\n\nEach version is given a distinguishing version number. If the Library\nspecifies a version number of this License which applies to it and\n\"any later version\", you have the option of following the terms and\nconditions either of that version or of any later version published by\nthe Free Software Foundation. If the Library does not specify a\nlicense version number, you may choose any version ever published by\nthe Free Software Foundation.\n\f\n 14. If you wish to incorporate parts of the Library into other free\nprograms whose distribution conditions are incompatible with these,\nwrite to the author to ask for permission. For software which is\ncopyrighted by the Free Software Foundation, write to the Free\nSoftware Foundation; we sometimes make exceptions for this. Our\ndecision will be guided by the two goals of preserving the free status\nof all derivatives of our free software and of promoting the sharing\nand reuse of software generally.\n\n\t\t\t NO WARRANTY\n\n 15. BECAUSE THE LIBRARY IS LICENSED FREE OF CHARGE, THERE IS NO\nWARRANTY FOR THE LIBRARY, TO THE EXTENT PERMITTED BY APPLICABLE LAW.\nEXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR\nOTHER PARTIES PROVIDE THE LIBRARY \"AS IS\" WITHOUT WARRANTY OF ANY\nKIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE\nIMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR\nPURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE\nLIBRARY IS WITH YOU. SHOULD THE LIBRARY PROVE DEFECTIVE, YOU ASSUME\nTHE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.\n\n 16. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN\nWRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY\nAND/OR REDISTRIBUTE THE LIBRARY AS PERMITTED ABOVE, BE LIABLE TO YOU\nFOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR\nCONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE\nLIBRARY (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING\nRENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A\nFAILURE OF THE LIBRARY TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF\nSUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH\nDAMAGES.\n\n\t\t END OF TERMS AND CONDITIONS\n\f\n How to Apply These Terms to Your New Libraries\n\n If you develop a new library, and you want it to be of the greatest\npossible use to the public, we recommend making it free software that\neveryone can redistribute and change. You can do so by permitting\nredistribution under these terms (or, alternatively, under the terms of the\nordinary General Public License).\n\n To apply these terms, attach the following notices to the library. It is\nsafest to attach them to the start of each source file to most effectively\nconvey the exclusion of warranty; and each file should have at least the\n\"copyright\" line and a pointer to where the full notice is found.\n\n \n Copyright (C) \n\n This library is free software; you can redistribute it and/or\n modify it under the terms of the GNU Lesser General Public\n License as published by the Free Software Foundation; either\n version 2.1 of the License, or (at your option) any later version.\n\n This library is distributed in the hope that it will be useful,\n but WITHOUT ANY WARRANTY; without even the implied warranty of\n MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU\n Lesser General Public License for more details.\n\n You should have received a copy of the GNU Lesser General Public\n License along with this library; if not, write to the Free Software\n Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA\n\nAlso add information on how to contact you by electronic and paper mail.\n\nYou should also get your employer (if you work as a programmer) or your\nschool, if any, to sign a \"copyright disclaimer\" for the library, if\nnecessary. Here is a sample; alter the names:\n\n Yoyodyne, Inc., hereby disclaims all copyright interest in the\n library `Frob' (a library for tweaking knobs) written by James Random Hacker.\n\n , 1 April 1990\n Ty Coon, President of Vice\n\nThat's all there is to it!\n\n\n" }, { "path": "third_party/mspack/config.h", "content": "/* config.h.in. Generated from configure.ac by autoheader. */\n\n/* Turn debugging mode on? */\n#undef DEBUG\n\n/* Define to 1 if you have the header file. */\n#undef HAVE_DLFCN_H\n\n/* Define to 1 if fseeko (and presumably ftello) exists and is declared. */\n#undef HAVE_FSEEKO\n\n/* Define to 1 if you have the header file. */\n#define HAVE_INTTYPES_H 1\n\n/* Define to 1 if you have the header file. */\n#undef HAVE_MEMORY_H\n\n/* Define to 1 if you have the `mkdir' function. */\n#undef HAVE_MKDIR\n\n/* Define to 1 if you have the header file. */\n#define HAVE_STDINT_H 1\n\n/* Define to 1 if you have the header file. */\n#define HAVE_STDLIB_H 1\n\n/* Define to 1 if you have the header file. */\n#undef HAVE_STRINGS_H\n\n/* Define to 1 if you have the header file. */\n#define HAVE_STRING_H 1\n\n/* Define to 1 if you have the header file. */\n#undef HAVE_SYS_STAT_H\n\n/* Define to 1 if you have the header file. */\n#undef HAVE_SYS_TYPES_H\n\n/* Define to 1 if you have the `towlower' function. */\n#undef HAVE_TOWLOWER\n\n/* Define to 1 if you have the header file. */\n#undef HAVE_UNISTD_H\n\n/* Define to 1 if you have the `_mkdir' function. */\n#undef HAVE__MKDIR\n\n/* Define to the sub-directory where libtool stores uninstalled libraries. */\n#undef LT_OBJDIR\n\n/* Define if mkdir takes only one argument. */\n#undef MKDIR_TAKES_ONE_ARG\n\n/* Name of package */\n#undef PACKAGE\n\n/* Define to the address where bug reports for this package should be sent. */\n#undef PACKAGE_BUGREPORT\n\n/* Define to the full name of this package. */\n#undef PACKAGE_NAME\n\n/* Define to the full name and version of this package. */\n#undef PACKAGE_STRING\n\n/* Define to the one symbol short name of this package. */\n#undef PACKAGE_TARNAME\n\n/* Define to the home page for this package. */\n#undef PACKAGE_URL\n\n/* Define to the version of this package. */\n#undef PACKAGE_VERSION\n\n/* The size of `off_t', as computed by sizeof. */\n#undef SIZEOF_OFF_T\n\n/* Define to 1 if you have the ANSI C header files. */\n#undef STDC_HEADERS\n\n/* Version number of package */\n#undef VERSION\n\n/* Enable large inode numbers on Mac OS X 10.5. */\n#ifndef _DARWIN_USE_64_BIT_INODE\n# define _DARWIN_USE_64_BIT_INODE 1\n#endif\n\n/* Number of bits in a file offset, on hosts where this is settable. */\n#undef _FILE_OFFSET_BITS\n\n/* Define to 1 to make fseeko visible on some hosts (e.g. glibc 2.2). */\n#undef _LARGEFILE_SOURCE\n\n/* Define for large files, on AIX-style hosts. */\n#undef _LARGE_FILES\n\n/* Define to empty if `const' does not conform to ANSI C. */\n#undef const\n\n/* Define to `__inline__' or `__inline' if that's what the C compiler\n calls it, or to nothing if 'inline' is not supported under any name. */\n#ifndef __cplusplus\n#undef inline\n#endif\n\n/* Define to `int' if does not define. */\n#undef mode_t\n\n/* Define to `long int' if does not define. */\n#undef off_t\n\n/* Define to `unsigned int' if does not define. */\n#undef size_t\n" }, { "path": "third_party/mspack/logging.cc", "content": "#include \"xenia/base/logging.h\"\n\n#include \n\nextern \"C\" void xenia_log(const char* fmt, ...) {\n char buffer[128];\n va_list args;\n va_start(args, fmt);\n vsnprintf(buffer, sizeof(buffer), fmt, args);\n va_end(args);\n XELOGW(\"mspack: {}\", buffer);\n}\n" }, { "path": "third_party/mspack/lzx.h", "content": "/* This file is part of libmspack.\n * (C) 2003-2013 Stuart Caie.\n *\n * The LZX method was created by Jonathan Forbes and Tomi Poutanen, adapted\n * by Microsoft Corporation.\n *\n * libmspack is free software; you can redistribute it and/or modify it under\n * the terms of the GNU Lesser General Public License (LGPL) version 2.1\n *\n * For further details, see the file COPYING.LIB distributed with libmspack\n */\n\n#ifndef MSPACK_LZX_H\n#define MSPACK_LZX_H 1\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\n/* LZX compression / decompression definitions */\n\n/* some constants defined by the LZX specification */\n#define LZX_MIN_MATCH (2)\n#define LZX_MAX_MATCH (257)\n#define LZX_NUM_CHARS (256)\n#define LZX_BLOCKTYPE_INVALID (0) /* also blocktypes 4-7 invalid */\n#define LZX_BLOCKTYPE_VERBATIM (1)\n#define LZX_BLOCKTYPE_ALIGNED (2)\n#define LZX_BLOCKTYPE_UNCOMPRESSED (3)\n#define LZX_PRETREE_NUM_ELEMENTS (20)\n#define LZX_ALIGNED_NUM_ELEMENTS (8) /* aligned offset tree #elements */\n#define LZX_NUM_PRIMARY_LENGTHS (7) /* this one missing from spec! */\n#define LZX_NUM_SECONDARY_LENGTHS (249) /* length tree #elements */\n\n/* LZX huffman defines: tweak tablebits as desired */\n#define LZX_PRETREE_MAXSYMBOLS (LZX_PRETREE_NUM_ELEMENTS)\n#define LZX_PRETREE_TABLEBITS (6)\n#define LZX_MAINTREE_MAXSYMBOLS (LZX_NUM_CHARS + 290*8)\n#define LZX_MAINTREE_TABLEBITS (12)\n#define LZX_LENGTH_MAXSYMBOLS (LZX_NUM_SECONDARY_LENGTHS+1)\n#define LZX_LENGTH_TABLEBITS (12)\n#define LZX_ALIGNED_MAXSYMBOLS (LZX_ALIGNED_NUM_ELEMENTS)\n#define LZX_ALIGNED_TABLEBITS (7)\n#define LZX_LENTABLE_SAFETY (64) /* table decoding overruns are allowed */\n\n#define LZX_FRAME_SIZE (32768) /* the size of a frame in LZX */\n\nstruct lzxd_stream {\n struct mspack_system *sys; /* I/O routines */\n struct mspack_file *input; /* input file handle */\n struct mspack_file *output; /* output file handle */\n\n off_t offset; /* number of bytes actually output */\n off_t length; /* overall decompressed length of stream */\n\n unsigned char *window; /* decoding window */\n unsigned int window_size; /* window size */\n unsigned int ref_data_size; /* LZX DELTA reference data size */\n unsigned int num_offsets; /* number of match_offset entries in table */\n unsigned int window_posn; /* decompression offset within window */\n unsigned int frame_posn; /* current frame offset within in window */\n unsigned int frame; /* the number of 32kb frames processed */\n unsigned int reset_interval; /* which frame do we reset the compressor? */\n\n unsigned int R0, R1, R2; /* for the LRU offset system */\n unsigned int block_length; /* uncompressed length of this LZX block */\n unsigned int block_remaining; /* uncompressed bytes still left to decode */\n\n signed int intel_filesize; /* magic header value used for transform */\n signed int intel_curpos; /* current offset in transform space */\n\n unsigned char intel_started; /* has intel E8 decoding started? */\n unsigned char block_type; /* type of the current block */\n unsigned char header_read; /* have we started decoding at all yet? */\n unsigned char input_end; /* have we reached the end of input? */\n unsigned char is_delta; /* does stream follow LZX DELTA spec? */\n\n int error;\n\n /* I/O buffering */\n unsigned char *inbuf, *i_ptr, *i_end, *o_ptr, *o_end;\n unsigned int bit_buffer, bits_left, inbuf_size;\n\n /* huffman code lengths */\n unsigned char PRETREE_len [LZX_PRETREE_MAXSYMBOLS + LZX_LENTABLE_SAFETY];\n unsigned char MAINTREE_len [LZX_MAINTREE_MAXSYMBOLS + LZX_LENTABLE_SAFETY];\n unsigned char LENGTH_len [LZX_LENGTH_MAXSYMBOLS + LZX_LENTABLE_SAFETY];\n unsigned char ALIGNED_len [LZX_ALIGNED_MAXSYMBOLS + LZX_LENTABLE_SAFETY];\n\n /* huffman decoding tables */\n unsigned short PRETREE_table [(1 << LZX_PRETREE_TABLEBITS) +\n (LZX_PRETREE_MAXSYMBOLS * 2)];\n unsigned short MAINTREE_table[(1 << LZX_MAINTREE_TABLEBITS) +\n (LZX_MAINTREE_MAXSYMBOLS * 2)];\n unsigned short LENGTH_table [(1 << LZX_LENGTH_TABLEBITS) +\n (LZX_LENGTH_MAXSYMBOLS * 2)];\n unsigned short ALIGNED_table [(1 << LZX_ALIGNED_TABLEBITS) +\n (LZX_ALIGNED_MAXSYMBOLS * 2)];\n unsigned char LENGTH_empty;\n\n /* this is used purely for doing the intel E8 transform */\n unsigned char e8_buf[LZX_FRAME_SIZE];\n};\n\n/**\n * Allocates and initialises LZX decompression state for decoding an LZX\n * stream.\n *\n * This routine uses system->alloc() to allocate memory. If memory\n * allocation fails, or the parameters to this function are invalid,\n * NULL is returned.\n *\n * @param system an mspack_system structure used to read from\n * the input stream and write to the output\n * stream, also to allocate and free memory.\n * @param input an input stream with the LZX data.\n * @param output an output stream to write the decoded data to.\n * @param window_bits the size of the decoding window, which must be\n * between 15 and 21 inclusive for regular LZX\n * data, or between 17 and 25 inclusive for\n * LZX DELTA data.\n * @param reset_interval the interval at which the LZX bitstream is\n * reset, in multiples of LZX frames (32678\n * bytes), e.g. a value of 2 indicates the input\n * stream resets after every 65536 output bytes.\n * A value of 0 indicates that the bitstream never\n * resets, such as in CAB LZX streams.\n * @param input_buffer_size the number of bytes to use as an input\n * bitstream buffer.\n * @param output_length the length in bytes of the entirely\n * decompressed output stream, if known in\n * advance. It is used to correctly perform the\n * Intel E8 transformation, which must stop 6\n * bytes before the very end of the\n * decompressed stream. It is not otherwise used\n * or adhered to. If the full decompressed\n * length is known in advance, set it here.\n * If it is NOT known, use the value 0, and call\n * lzxd_set_output_length() once it is\n * known. If never set, 4 of the final 6 bytes\n * of the output stream may be incorrect.\n * @param is_delta should be zero for all regular LZX data,\n * non-zero for LZX DELTA encoded data.\n * @return a pointer to an initialised lzxd_stream structure, or NULL if\n * there was not enough memory or parameters to the function were wrong.\n */\nextern struct lzxd_stream *lzxd_init(struct mspack_system *system,\n struct mspack_file *input,\n struct mspack_file *output,\n int window_bits,\n int reset_interval,\n int input_buffer_size,\n off_t output_length,\n char is_delta);\n\n/* see description of output_length in lzxd_init() */\nextern void lzxd_set_output_length(struct lzxd_stream *lzx,\n off_t output_length);\n\n/**\n * Reads LZX DELTA reference data into the window and allows\n * lzxd_decompress() to reference it.\n *\n * Call this before the first call to lzxd_decompress().\n\n * @param lzx the LZX stream to apply this reference data to\n * @param system an mspack_system implementation to use with the\n * input param. Only read() will be called.\n * @param input an input file handle to read reference data using\n * system->read().\n * @param length the length of the reference data. Cannot be longer\n * than the LZX window size.\n * @return an error code, or MSPACK_ERR_OK if successful\n */\nextern int lzxd_set_reference_data(struct lzxd_stream *lzx,\n struct mspack_system *system,\n struct mspack_file *input,\n unsigned int length);\n\n/**\n * Decompresses entire or partial LZX streams.\n *\n * The number of bytes of data that should be decompressed is given as the\n * out_bytes parameter. If more bytes are decoded than are needed, they\n * will be kept over for a later invocation.\n *\n * The output bytes will be passed to the system->write() function given in\n * lzxd_init(), using the output file handle given in lzxd_init(). More than\n * one call may be made to system->write().\n\n * Input bytes will be read in as necessary using the system->read()\n * function given in lzxd_init(), using the input file handle given in\n * lzxd_init(). This will continue until system->read() returns 0 bytes,\n * or an error. Errors will be passed out of the function as\n * MSPACK_ERR_READ errors. Input streams should convey an \"end of input\n * stream\" by refusing to supply all the bytes that LZX asks for when they\n * reach the end of the stream, rather than return an error code.\n *\n * If any error code other than MSPACK_ERR_OK is returned, the stream\n * should be considered unusable and lzxd_decompress() should not be\n * called again on this stream.\n *\n * @param lzx LZX decompression state, as allocated by lzxd_init().\n * @param out_bytes the number of bytes of data to decompress.\n * @return an error code, or MSPACK_ERR_OK if successful\n */\nextern int lzxd_decompress(struct lzxd_stream *lzx, off_t out_bytes);\n\n/**\n * Frees all state associated with an LZX data stream. This will call\n * system->free() using the system pointer given in lzxd_init().\n *\n * @param lzx LZX decompression state to free.\n */\nvoid lzxd_free(struct lzxd_stream *lzx);\n\n#ifdef __cplusplus\n}\n#endif\n\n#endif\n" }, { "path": "third_party/mspack/lzxd.c", "content": "/* This file is part of libmspack.\n * (C) 2003-2013 Stuart Caie.\n *\n * The LZX method was created by Jonathan Forbes and Tomi Poutanen, adapted\n * by Microsoft Corporation.\n *\n * libmspack is free software; you can redistribute it and/or modify it under\n * the terms of the GNU Lesser General Public License (LGPL) version 2.1\n *\n * For further details, see the file COPYING.LIB distributed with libmspack\n */\n\n/* LZX decompression implementation */\n\n#include \n#include \n\nextern void xenia_log(const char*, ...);\n\n#undef D\n#define D(x) do { xenia_log x; } while (0);\n\n/* Microsoft's LZX document (in cab-sdk.exe) and their implementation\n * of the com.ms.util.cab Java package do not concur.\n *\n * In the LZX document, there is a table showing the correlation between\n * window size and the number of position slots. It states that the 1MB\n * window = 40 slots and the 2MB window = 42 slots. In the implementation,\n * 1MB = 42 slots, 2MB = 50 slots. The actual calculation is 'find the\n * first slot whose position base is equal to or more than the required\n * window size'. This would explain why other tables in the document refer\n * to 50 slots rather than 42.\n *\n * The constant NUM_PRIMARY_LENGTHS used in the decompression pseudocode\n * is not defined in the specification.\n *\n * The LZX document does not state the uncompressed block has an\n * uncompressed length field. Where does this length field come from, so\n * we can know how large the block is? The implementation has it as the 24\n * bits following after the 3 blocktype bits, before the alignment\n * padding.\n *\n * The LZX document states that aligned offset blocks have their aligned\n * offset huffman tree AFTER the main and length trees. The implementation\n * suggests that the aligned offset tree is BEFORE the main and length\n * trees.\n *\n * The LZX document decoding algorithm states that, in an aligned offset\n * block, if an extra_bits value is 1, 2 or 3, then that number of bits\n * should be read and the result added to the match offset. This is\n * correct for 1 and 2, but not 3, where just a huffman symbol (using the\n * aligned tree) should be read.\n *\n * Regarding the E8 preprocessing, the LZX document states 'No translation\n * may be performed on the last 6 bytes of the input block'. This is\n * correct. However, the pseudocode provided checks for the *E8 leader*\n * up to the last 6 bytes. If the leader appears between -10 and -7 bytes\n * from the end, this would cause the next four bytes to be modified, at\n * least one of which would be in the last 6 bytes, which is not allowed\n * according to the spec.\n *\n * The specification states that the huffman trees must always contain at\n * least one element. However, many CAB files contain blocks where the\n * length tree is completely empty (because there are no matches), and\n * this is expected to succeed.\n *\n * The errors in LZX documentation appear have been corrected in the\n * new documentation for the LZX DELTA format.\n *\n * http://msdn.microsoft.com/en-us/library/cc483133.aspx\n *\n * However, this is a different format, an extension of regular LZX.\n * I have noticed the following differences, there may be more:\n *\n * The maximum window size has increased from 2MB to 32MB. This also\n * increases the maximum number of position slots, etc.\n *\n * If the match length is 257 (the maximum possible), this signals\n * a further length decoding step, that allows for matches up to\n * 33024 bytes long.\n *\n * The format now allows for \"reference data\", supplied by the caller.\n * If match offsets go further back than the number of bytes\n * decompressed so far, that is them accessing the reference data.\n */\n\n/* import bit-reading macros and code */\n#define BITS_TYPE struct lzxd_stream\n#define BITS_VAR lzx\n#define BITS_ORDER_MSB\n#define READ_BYTES do { \\\n unsigned char b0, b1; \\\n READ_IF_NEEDED; b0 = *i_ptr++; \\\n READ_IF_NEEDED; b1 = *i_ptr++; \\\n INJECT_BITS((b1 << 8) | b0, 16); \\\n} while (0)\n#include \n\n/* import huffman-reading macros and code */\n#define TABLEBITS(tbl) LZX_##tbl##_TABLEBITS\n#define MAXSYMBOLS(tbl) LZX_##tbl##_MAXSYMBOLS\n#define HUFF_TABLE(tbl,idx) lzx->tbl##_table[idx]\n#define HUFF_LEN(tbl,idx) lzx->tbl##_len[idx]\n#define HUFF_ERROR return lzx->error = MSPACK_ERR_DECRUNCH\n#include \n\n/* BUILD_TABLE(tbl) builds a huffman lookup table from code lengths */\n#define BUILD_TABLE(tbl) \\\n if (make_decode_table(MAXSYMBOLS(tbl), TABLEBITS(tbl), \\\n &HUFF_LEN(tbl,0), &HUFF_TABLE(tbl,0))) \\\n { \\\n D((\"failed to build %s table\", #tbl)) \\\n return lzx->error = MSPACK_ERR_DECRUNCH; \\\n }\n\n#define BUILD_TABLE_MAYBE_EMPTY(tbl) do { \\\n lzx->tbl##_empty = 0; \\\n if (make_decode_table(MAXSYMBOLS(tbl), TABLEBITS(tbl), \\\n &HUFF_LEN(tbl,0), &HUFF_TABLE(tbl,0))) \\\n { \\\n for (i = 0; i < MAXSYMBOLS(tbl); i++) { \\\n if (HUFF_LEN(tbl, i) > 0) { \\\n D((\"failed to build %s table\", #tbl)) \\\n return lzx->error = MSPACK_ERR_DECRUNCH; \\\n } \\\n } \\\n /* empty tree - allow it, but don't decode symbols with it */ \\\n lzx->tbl##_empty = 1; \\\n } \\\n} while (0)\n\n/* READ_LENGTHS(tablename, first, last) reads in code lengths for symbols\n * first to last in the given table. The code lengths are stored in their\n * own special LZX way.\n */\n#define READ_LENGTHS(tbl, first, last) do { \\\n STORE_BITS; \\\n if (lzxd_read_lens(lzx, &HUFF_LEN(tbl, 0), (first), \\\n (unsigned int)(last))) return lzx->error; \\\n RESTORE_BITS; \\\n} while (0)\n\nstatic int lzxd_read_lens(struct lzxd_stream *lzx, unsigned char *lens,\n unsigned int first, unsigned int last)\n{\n /* bit buffer and huffman symbol decode variables */\n unsigned int bit_buffer;\n int bits_left, i;\n unsigned short sym;\n unsigned char *i_ptr, *i_end;\n\n unsigned int x, y;\n int z;\n\n RESTORE_BITS;\n \n /* read lengths for pretree (20 symbols, lengths stored in fixed 4 bits) */\n for (x = 0; x < 20; x++) {\n READ_BITS(y, 4);\n lzx->PRETREE_len[x] = y;\n }\n BUILD_TABLE(PRETREE);\n\n for (x = first; x < last; ) {\n READ_HUFFSYM(PRETREE, z);\n if (z == 17) {\n /* code = 17, run of ([read 4 bits]+4) zeros */\n READ_BITS(y, 4); y += 4;\n while (y--) lens[x++] = 0;\n }\n else if (z == 18) {\n /* code = 18, run of ([read 5 bits]+20) zeros */\n READ_BITS(y, 5); y += 20;\n while (y--) lens[x++] = 0;\n }\n else if (z == 19) {\n /* code = 19, run of ([read 1 bit]+4) [read huffman symbol] */\n READ_BITS(y, 1); y += 4;\n READ_HUFFSYM(PRETREE, z);\n z = lens[x] - z; if (z < 0) z += 17;\n while (y--) lens[x++] = z;\n }\n else {\n /* code = 0 to 16, delta current length entry */\n z = lens[x] - z; if (z < 0) z += 17;\n lens[x++] = z;\n }\n }\n\n STORE_BITS;\n\n return MSPACK_ERR_OK;\n}\n\n/* LZX static data tables:\n *\n * LZX uses 'position slots' to represent match offsets. For every match,\n * a small 'position slot' number and a small offset from that slot are\n * encoded instead of one large offset.\n *\n * The number of slots is decided by how many are needed to encode the\n * largest offset for a given window size. This is easy when the gap between\n * slots is less than 128Kb, it's a linear relationship. But when extra_bits\n * reaches its limit of 17 (because LZX can only ensure reading 17 bits of\n * data at a time), we can only jump 128Kb at a time and have to start\n * using more and more position slots as each window size doubles.\n *\n * position_base[] is an index to the position slot bases\n *\n * extra_bits[] states how many bits of offset-from-base data is needed.\n *\n * They are calculated as follows:\n * extra_bits[i] = 0 where i < 4\n * extra_bits[i] = floor(i/2)-1 where i >= 4 && i < 36\n * extra_bits[i] = 17 where i >= 36\n * position_base[0] = 0\n * position_base[i] = position_base[i-1] + (1 << extra_bits[i-1])\n */\nstatic const unsigned int position_slots[11] = {\n 30, 32, 34, 36, 38, 42, 50, 66, 98, 162, 290\n};\nstatic const unsigned char extra_bits[36] = {\n 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8,\n 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14, 14, 15, 15, 16, 16\n};\nstatic const unsigned int position_base[290] = {\n 0, 1, 2, 3, 4, 6, 8, 12, 16, 24, 32, 48, 64, 96, 128, 192, 256, 384, 512,\n 768, 1024, 1536, 2048, 3072, 4096, 6144, 8192, 12288, 16384, 24576, 32768,\n 49152, 65536, 98304, 131072, 196608, 262144, 393216, 524288, 655360,\n 786432, 917504, 1048576, 1179648, 1310720, 1441792, 1572864, 1703936,\n 1835008, 1966080, 2097152, 2228224, 2359296, 2490368, 2621440, 2752512,\n 2883584, 3014656, 3145728, 3276800, 3407872, 3538944, 3670016, 3801088,\n 3932160, 4063232, 4194304, 4325376, 4456448, 4587520, 4718592, 4849664,\n 4980736, 5111808, 5242880, 5373952, 5505024, 5636096, 5767168, 5898240,\n 6029312, 6160384, 6291456, 6422528, 6553600, 6684672, 6815744, 6946816,\n 7077888, 7208960, 7340032, 7471104, 7602176, 7733248, 7864320, 7995392,\n 8126464, 8257536, 8388608, 8519680, 8650752, 8781824, 8912896, 9043968,\n 9175040, 9306112, 9437184, 9568256, 9699328, 9830400, 9961472, 10092544,\n 10223616, 10354688, 10485760, 10616832, 10747904, 10878976, 11010048,\n 11141120, 11272192, 11403264, 11534336, 11665408, 11796480, 11927552,\n 12058624, 12189696, 12320768, 12451840, 12582912, 12713984, 12845056,\n 12976128, 13107200, 13238272, 13369344, 13500416, 13631488, 13762560,\n 13893632, 14024704, 14155776, 14286848, 14417920, 14548992, 14680064,\n 14811136, 14942208, 15073280, 15204352, 15335424, 15466496, 15597568,\n 15728640, 15859712, 15990784, 16121856, 16252928, 16384000, 16515072,\n 16646144, 16777216, 16908288, 17039360, 17170432, 17301504, 17432576,\n 17563648, 17694720, 17825792, 17956864, 18087936, 18219008, 18350080,\n 18481152, 18612224, 18743296, 18874368, 19005440, 19136512, 19267584,\n 19398656, 19529728, 19660800, 19791872, 19922944, 20054016, 20185088,\n 20316160, 20447232, 20578304, 20709376, 20840448, 20971520, 21102592,\n 21233664, 21364736, 21495808, 21626880, 21757952, 21889024, 22020096,\n 22151168, 22282240, 22413312, 22544384, 22675456, 22806528, 22937600,\n 23068672, 23199744, 23330816, 23461888, 23592960, 23724032, 23855104,\n 23986176, 24117248, 24248320, 24379392, 24510464, 24641536, 24772608,\n 24903680, 25034752, 25165824, 25296896, 25427968, 25559040, 25690112,\n 25821184, 25952256, 26083328, 26214400, 26345472, 26476544, 26607616,\n 26738688, 26869760, 27000832, 27131904, 27262976, 27394048, 27525120,\n 27656192, 27787264, 27918336, 28049408, 28180480, 28311552, 28442624,\n 28573696, 28704768, 28835840, 28966912, 29097984, 29229056, 29360128,\n 29491200, 29622272, 29753344, 29884416, 30015488, 30146560, 30277632,\n 30408704, 30539776, 30670848, 30801920, 30932992, 31064064, 31195136,\n 31326208, 31457280, 31588352, 31719424, 31850496, 31981568, 32112640,\n 32243712, 32374784, 32505856, 32636928, 32768000, 32899072, 33030144,\n 33161216, 33292288, 33423360\n};\n\nstatic void lzxd_reset_state(struct lzxd_stream *lzx) {\n int i;\n\n lzx->R0 = 1;\n lzx->R1 = 1;\n lzx->R2 = 1;\n lzx->header_read = 0;\n lzx->block_remaining = 0;\n lzx->block_type = LZX_BLOCKTYPE_INVALID;\n\n /* initialise tables to 0 (because deltas will be applied to them) */\n for (i = 0; i < LZX_MAINTREE_MAXSYMBOLS; i++) lzx->MAINTREE_len[i] = 0;\n for (i = 0; i < LZX_LENGTH_MAXSYMBOLS; i++) lzx->LENGTH_len[i] = 0;\n}\n\n/*-------- main LZX code --------*/\n\nstruct lzxd_stream *lzxd_init(struct mspack_system *system,\n struct mspack_file *input,\n struct mspack_file *output,\n int window_bits,\n int reset_interval,\n int input_buffer_size,\n off_t output_length,\n char is_delta)\n{\n unsigned int window_size = 1 << window_bits;\n struct lzxd_stream *lzx;\n\n if (!system) return NULL;\n\n /* LZX DELTA window sizes are between 2^17 (128KiB) and 2^25 (32MiB),\n * regular LZX windows are between 2^15 (32KiB) and 2^21 (2MiB)\n */\n if (is_delta) {\n if (window_bits < 17 || window_bits > 25) return NULL;\n }\n else {\n if (window_bits < 15 || window_bits > 21) return NULL;\n }\n\n if (reset_interval < 0 || output_length < 0) {\n D((\"reset interval or output length < 0\"))\n return NULL;\n }\n\n /* round up input buffer size to multiple of two */\n input_buffer_size = (input_buffer_size + 1) & -2;\n if (input_buffer_size < 2) return NULL;\n\n /* allocate decompression state */\n if (!(lzx = (struct lzxd_stream *) system->alloc(system, sizeof(struct lzxd_stream)))) {\n return NULL;\n }\n\n /* allocate decompression window and input buffer */\n lzx->window = (unsigned char *) system->alloc(system, (size_t) window_size);\n lzx->inbuf = (unsigned char *) system->alloc(system, (size_t) input_buffer_size);\n if (!lzx->window || !lzx->inbuf) {\n system->free(lzx->window);\n system->free(lzx->inbuf);\n system->free(lzx);\n return NULL;\n }\n\n /* initialise decompression state */\n lzx->sys = system;\n lzx->input = input;\n lzx->output = output;\n lzx->offset = 0;\n lzx->length = output_length;\n\n lzx->inbuf_size = input_buffer_size;\n lzx->window_size = 1 << window_bits;\n lzx->ref_data_size = 0;\n lzx->window_posn = 0;\n lzx->frame_posn = 0;\n lzx->frame = 0;\n lzx->reset_interval = reset_interval;\n lzx->intel_filesize = 0;\n lzx->intel_curpos = 0;\n lzx->intel_started = 0;\n lzx->error = MSPACK_ERR_OK;\n lzx->num_offsets = position_slots[window_bits - 15] << 3;\n lzx->is_delta = is_delta;\n\n lzx->o_ptr = lzx->o_end = &lzx->e8_buf[0];\n lzxd_reset_state(lzx);\n INIT_BITS;\n return lzx;\n}\n\nint lzxd_set_reference_data(struct lzxd_stream *lzx,\n struct mspack_system *system,\n struct mspack_file *input,\n unsigned int length)\n{\n if (!lzx) return MSPACK_ERR_ARGS;\n\n if (!lzx->is_delta) {\n D((\"only LZX DELTA streams support reference data\"))\n return MSPACK_ERR_ARGS;\n }\n if (lzx->offset) {\n D((\"too late to set reference data after decoding starts\"))\n return MSPACK_ERR_ARGS;\n }\n if (length > lzx->window_size) {\n D((\"reference length (%u) is longer than the window\", length))\n return MSPACK_ERR_ARGS;\n }\n if (length > 0 && (!system || !input)) {\n D((\"length > 0 but no system or input\"))\n return MSPACK_ERR_ARGS;\n }\n\n lzx->ref_data_size = length;\n if (length > 0) {\n /* copy reference data */\n unsigned char *pos = &lzx->window[lzx->window_size - length];\n int bytes = system->read(input, pos, length);\n /* length can't be more than 2^25, so no signedness problem */\n if (bytes < (int)length) return MSPACK_ERR_READ;\n }\n lzx->ref_data_size = length;\n return MSPACK_ERR_OK;\n}\n\nvoid lzxd_set_output_length(struct lzxd_stream *lzx, off_t out_bytes) {\n if (lzx && out_bytes > 0) lzx->length = out_bytes;\n}\n\nint lzxd_decompress(struct lzxd_stream *lzx, off_t out_bytes) {\n /* bitstream and huffman reading variables */\n unsigned int bit_buffer;\n int bits_left, i=0;\n unsigned char *i_ptr, *i_end;\n unsigned short sym;\n\n int match_length, length_footer, extra, verbatim_bits, bytes_todo;\n int this_run, main_element, aligned_bits, j, warned = 0;\n unsigned char *window, *runsrc, *rundest, buf[12];\n unsigned int frame_size=0, end_frame, match_offset, window_posn;\n unsigned int R0, R1, R2;\n\n /* easy answers */\n if (!lzx || (out_bytes < 0)) return MSPACK_ERR_ARGS;\n if (lzx->error) return lzx->error;\n\n /* flush out any stored-up bytes before we begin */\n i = (int)(lzx->o_end - lzx->o_ptr);\n if ((off_t) i > out_bytes) i = (int) out_bytes;\n if (i) {\n if (lzx->sys->write(lzx->output, lzx->o_ptr, i) != i) {\n return lzx->error = MSPACK_ERR_WRITE;\n }\n lzx->o_ptr += i;\n lzx->offset += i;\n out_bytes -= i;\n }\n if (out_bytes == 0) return MSPACK_ERR_OK;\n\n /* restore local state */\n RESTORE_BITS;\n window = lzx->window;\n window_posn = lzx->window_posn;\n R0 = lzx->R0;\n R1 = lzx->R1;\n R2 = lzx->R2;\n\n end_frame = (unsigned int)((lzx->offset + out_bytes) / LZX_FRAME_SIZE) + 1;\n\n while (lzx->frame < end_frame) {\n /* have we reached the reset interval? (if there is one?) */\n if (lzx->reset_interval && ((lzx->frame % lzx->reset_interval) == 0)) {\n if (lzx->block_remaining) {\n /* this is a file format error, we can make a best effort to extract what we can */\n D((\"%d bytes remaining at reset interval\", lzx->block_remaining))\n if (!warned) {\n lzx->sys->message(NULL, \"WARNING; invalid reset interval detected during LZX decompression\");\n warned++;\n }\n }\n\n /* re-read the intel header and reset the huffman lengths */\n lzxd_reset_state(lzx);\n R0 = lzx->R0;\n R1 = lzx->R1;\n R2 = lzx->R2;\n }\n\n /* LZX DELTA format has chunk_size, not present in LZX format */\n if (lzx->is_delta) {\n ENSURE_BITS(16);\n REMOVE_BITS(16);\n }\n\n /* read header if necessary */\n if (!lzx->header_read) {\n /* read 1 bit. if bit=0, intel filesize = 0.\n * if bit=1, read intel filesize (32 bits) */\n j = 0; READ_BITS(i, 1); if (i) { READ_BITS(i, 16); READ_BITS(j, 16); }\n lzx->intel_filesize = (i << 16) | j;\n lzx->header_read = 1;\n } \n\n /* calculate size of frame: all frames are 32k except the final frame\n * which is 32kb or less. this can only be calculated when lzx->length\n * has been filled in. */\n frame_size = LZX_FRAME_SIZE;\n if (lzx->length && (lzx->length - lzx->offset) < (off_t)frame_size) {\n frame_size = lzx->length - lzx->offset;\n }\n\n /* decode until one more frame is available */\n bytes_todo = lzx->frame_posn + frame_size - window_posn;\n while (bytes_todo > 0) {\n /* initialise new block, if one is needed */\n if (lzx->block_remaining == 0) {\n /* realign if previous block was an odd-sized UNCOMPRESSED block */\n if ((lzx->block_type == LZX_BLOCKTYPE_UNCOMPRESSED) &&\n (lzx->block_length & 1))\n {\n READ_IF_NEEDED;\n i_ptr++;\n }\n\n /* read block type (3 bits) and block length (24 bits) */\n READ_BITS(lzx->block_type, 3);\n READ_BITS(i, 16); READ_BITS(j, 8);\n lzx->block_remaining = lzx->block_length = (i << 8) | j;\n /*D((\"new block t%d len %u\", lzx->block_type, lzx->block_length))*/\n\n /* read individual block headers */\n switch (lzx->block_type) {\n case LZX_BLOCKTYPE_ALIGNED:\n /* read lengths of and build aligned huffman decoding tree */\n for (i = 0; i < 8; i++) { READ_BITS(j, 3); lzx->ALIGNED_len[i] = j; }\n BUILD_TABLE(ALIGNED);\n /* rest of aligned header is same as verbatim */ /*@fallthrough@*/\n case LZX_BLOCKTYPE_VERBATIM:\n /* read lengths of and build main huffman decoding tree */\n READ_LENGTHS(MAINTREE, 0, 256);\n READ_LENGTHS(MAINTREE, 256, LZX_NUM_CHARS + lzx->num_offsets);\n BUILD_TABLE(MAINTREE);\n /* if the literal 0xE8 is anywhere in the block... */\n if (lzx->MAINTREE_len[0xE8] != 0) lzx->intel_started = 1;\n /* read lengths of and build lengths huffman decoding tree */\n READ_LENGTHS(LENGTH, 0, LZX_NUM_SECONDARY_LENGTHS);\n BUILD_TABLE_MAYBE_EMPTY(LENGTH);\n break;\n\n case LZX_BLOCKTYPE_UNCOMPRESSED:\n /* because we can't assume otherwise */\n lzx->intel_started = 1;\n\n /* read 1-16 (not 0-15) bits to align to bytes */\n if (bits_left == 0) ENSURE_BITS(16);\n bits_left = 0; bit_buffer = 0;\n\n /* read 12 bytes of stored R0 / R1 / R2 values */\n for (rundest = &buf[0], i = 0; i < 12; i++) {\n READ_IF_NEEDED;\n *rundest++ = *i_ptr++;\n }\n R0 = buf[0] | (buf[1] << 8) | (buf[2] << 16) | (buf[3] << 24);\n R1 = buf[4] | (buf[5] << 8) | (buf[6] << 16) | (buf[7] << 24);\n R2 = buf[8] | (buf[9] << 8) | (buf[10] << 16) | (buf[11] << 24);\n break;\n\n default:\n D((\"bad block type\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n }\n\n /* decode more of the block:\n * run = min(what's available, what's needed) */\n this_run = lzx->block_remaining;\n if (this_run > bytes_todo) this_run = bytes_todo;\n\n /* assume we decode exactly this_run bytes, for now */\n bytes_todo -= this_run;\n lzx->block_remaining -= this_run;\n\n /* decode at least this_run bytes */\n switch (lzx->block_type) {\n case LZX_BLOCKTYPE_VERBATIM:\n while (this_run > 0) {\n READ_HUFFSYM(MAINTREE, main_element);\n if (main_element < LZX_NUM_CHARS) {\n /* literal: 0 to LZX_NUM_CHARS-1 */\n window[window_posn++] = main_element;\n this_run--;\n }\n else {\n /* match: LZX_NUM_CHARS + ((slot<<3) | length_header (3 bits)) */\n main_element -= LZX_NUM_CHARS;\n\n /* get match length */\n match_length = main_element & LZX_NUM_PRIMARY_LENGTHS;\n if (match_length == LZX_NUM_PRIMARY_LENGTHS) {\n if (lzx->LENGTH_empty) {\n D((\"LENGTH symbol needed but tree is empty\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n READ_HUFFSYM(LENGTH, length_footer);\n match_length += length_footer;\n }\n match_length += LZX_MIN_MATCH;\n\n /* get match offset */\n switch ((match_offset = (main_element >> 3))) {\n case 0: match_offset = R0; break;\n case 1: match_offset = R1; R1=R0; R0 = match_offset; break;\n case 2: match_offset = R2; R2=R0; R0 = match_offset; break;\n case 3: match_offset = 1; R2=R1; R1=R0; R0 = match_offset; break;\n default:\n extra = (match_offset >= 36) ? 17 : extra_bits[match_offset];\n READ_BITS(verbatim_bits, extra);\n match_offset = position_base[match_offset] - 2 + verbatim_bits;\n R2 = R1; R1 = R0; R0 = match_offset;\n }\n\n /* LZX DELTA uses max match length to signal even longer match */\n if (match_length == LZX_MAX_MATCH && lzx->is_delta) {\n int extra_len = 0;\n ENSURE_BITS(3); /* 4 entry huffman tree */\n if (PEEK_BITS(1) == 0) {\n REMOVE_BITS(1); /* '0' -> 8 extra length bits */\n READ_BITS(extra_len, 8);\n }\n else if (PEEK_BITS(2) == 2) {\n REMOVE_BITS(2); /* '10' -> 10 extra length bits + 0x100 */\n READ_BITS(extra_len, 10);\n extra_len += 0x100;\n }\n else if (PEEK_BITS(3) == 6) {\n REMOVE_BITS(3); /* '110' -> 12 extra length bits + 0x500 */\n READ_BITS(extra_len, 12);\n extra_len += 0x500;\n }\n else {\n REMOVE_BITS(3); /* '111' -> 15 extra length bits */\n READ_BITS(extra_len, 15);\n }\n match_length += extra_len;\n }\n\n if ((window_posn + match_length) > lzx->window_size) {\n D((\"match ran over window wrap\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n \n /* copy match */\n rundest = &window[window_posn];\n i = match_length;\n /* does match offset wrap the window? */\n if (match_offset > window_posn) {\n if ((off_t)match_offset > lzx->offset &&\n (match_offset - window_posn) > lzx->ref_data_size)\n {\n D((\"match offset beyond LZX stream\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n /* j = length from match offset to end of window */\n j = match_offset - window_posn;\n if (j > (int) lzx->window_size) {\n D((\"match offset beyond window boundaries\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n runsrc = &window[lzx->window_size - j];\n if (j < i) {\n /* if match goes over the window edge, do two copy runs */\n i -= j; while (j-- > 0) *rundest++ = *runsrc++;\n runsrc = window;\n }\n while (i-- > 0) *rundest++ = *runsrc++;\n }\n else {\n runsrc = rundest - match_offset;\n while (i-- > 0) *rundest++ = *runsrc++;\n }\n\n this_run -= match_length;\n window_posn += match_length;\n }\n } /* while (this_run > 0) */\n break;\n\n case LZX_BLOCKTYPE_ALIGNED:\n while (this_run > 0) {\n READ_HUFFSYM(MAINTREE, main_element);\n if (main_element < LZX_NUM_CHARS) {\n /* literal: 0 to LZX_NUM_CHARS-1 */\n window[window_posn++] = main_element;\n this_run--;\n }\n else {\n /* match: LZX_NUM_CHARS + ((slot<<3) | length_header (3 bits)) */\n main_element -= LZX_NUM_CHARS;\n\n /* get match length */\n match_length = main_element & LZX_NUM_PRIMARY_LENGTHS;\n if (match_length == LZX_NUM_PRIMARY_LENGTHS) {\n if (lzx->LENGTH_empty) {\n D((\"LENGTH symbol needed but tree is empty\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n } \n READ_HUFFSYM(LENGTH, length_footer);\n match_length += length_footer;\n }\n match_length += LZX_MIN_MATCH;\n\n /* get match offset */\n switch ((match_offset = (main_element >> 3))) {\n case 0: match_offset = R0; break;\n case 1: match_offset = R1; R1 = R0; R0 = match_offset; break;\n case 2: match_offset = R2; R2 = R0; R0 = match_offset; break;\n default:\n extra = (match_offset >= 36) ? 17 : extra_bits[match_offset];\n match_offset = position_base[match_offset] - 2;\n if (extra > 3) {\n /* verbatim and aligned bits */\n extra -= 3;\n READ_BITS(verbatim_bits, extra);\n match_offset += (verbatim_bits << 3);\n READ_HUFFSYM(ALIGNED, aligned_bits);\n match_offset += aligned_bits;\n }\n else if (extra == 3) {\n /* aligned bits only */\n READ_HUFFSYM(ALIGNED, aligned_bits);\n match_offset += aligned_bits;\n }\n else if (extra > 0) { /* extra==1, extra==2 */\n /* verbatim bits only */\n READ_BITS(verbatim_bits, extra);\n match_offset += verbatim_bits;\n }\n else /* extra == 0 */ {\n /* ??? not defined in LZX specification! */\n match_offset = 1;\n }\n /* update repeated offset LRU queue */\n R2 = R1; R1 = R0; R0 = match_offset;\n }\n\n /* LZX DELTA uses max match length to signal even longer match */\n if (match_length == LZX_MAX_MATCH && lzx->is_delta) {\n int extra_len = 0;\n ENSURE_BITS(3); /* 4 entry huffman tree */\n if (PEEK_BITS(1) == 0) {\n REMOVE_BITS(1); /* '0' -> 8 extra length bits */\n READ_BITS(extra_len, 8);\n }\n else if (PEEK_BITS(2) == 2) {\n REMOVE_BITS(2); /* '10' -> 10 extra length bits + 0x100 */\n READ_BITS(extra_len, 10);\n extra_len += 0x100;\n }\n else if (PEEK_BITS(3) == 6) {\n REMOVE_BITS(3); /* '110' -> 12 extra length bits + 0x500 */\n READ_BITS(extra_len, 12);\n extra_len += 0x500;\n }\n else {\n REMOVE_BITS(3); /* '111' -> 15 extra length bits */\n READ_BITS(extra_len, 15);\n }\n match_length += extra_len;\n }\n\n if ((window_posn + match_length) > lzx->window_size) {\n D((\"match ran over window wrap\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n\n /* copy match */\n rundest = &window[window_posn];\n i = match_length;\n /* does match offset wrap the window? */\n if (match_offset > window_posn) {\n if ((off_t)match_offset > lzx->offset &&\n (match_offset - window_posn) > lzx->ref_data_size)\n {\n D((\"match offset beyond LZX stream\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n /* j = length from match offset to end of window */\n j = match_offset - window_posn;\n if (j > (int) lzx->window_size) {\n D((\"match offset beyond window boundaries\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n runsrc = &window[lzx->window_size - j];\n if (j < i) {\n /* if match goes over the window edge, do two copy runs */\n i -= j; while (j-- > 0) *rundest++ = *runsrc++;\n runsrc = window;\n }\n while (i-- > 0) *rundest++ = *runsrc++;\n }\n else {\n runsrc = rundest - match_offset;\n while (i-- > 0) *rundest++ = *runsrc++;\n }\n\n this_run -= match_length;\n window_posn += match_length;\n }\n } /* while (this_run > 0) */\n break;\n\n case LZX_BLOCKTYPE_UNCOMPRESSED:\n /* as this_run is limited not to wrap a frame, this also means it\n * won't wrap the window (as the window is a multiple of 32k) */\n rundest = &window[window_posn];\n window_posn += this_run;\n while (this_run > 0) {\n if ((i = (int)(i_end - i_ptr)) == 0) {\n READ_IF_NEEDED;\n }\n else {\n if (i > this_run) i = this_run;\n lzx->sys->copy(i_ptr, rundest, (size_t) i);\n rundest += i;\n i_ptr += i;\n this_run -= i;\n }\n }\n break;\n\n default:\n return lzx->error = MSPACK_ERR_DECRUNCH; /* might as well */\n }\n\n /* did the final match overrun our desired this_run length? */\n if (this_run < 0) {\n if ((unsigned int)(-this_run) > lzx->block_remaining) {\n D((\"overrun went past end of block by %d (%d remaining)\",\n -this_run, lzx->block_remaining ))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n lzx->block_remaining -= -this_run;\n }\n } /* while (bytes_todo > 0) */\n\n /* streams don't extend over frame boundaries */\n if ((window_posn - lzx->frame_posn) != frame_size) {\n D((\"decode beyond output frame limits! %d != %d\",\n window_posn - lzx->frame_posn, frame_size))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n\n /* re-align input bitstream */\n if (bits_left > 0) ENSURE_BITS(16);\n if (bits_left & 15) REMOVE_BITS(bits_left & 15);\n\n /* check that we've used all of the previous frame first */\n if (lzx->o_ptr != lzx->o_end) {\n D((\"%ld avail bytes, new %d frame\",\n (long)(lzx->o_end - lzx->o_ptr), frame_size))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n\n /* does this intel block _really_ need decoding? */\n if (lzx->intel_started && lzx->intel_filesize &&\n (lzx->frame <= 32768) && (frame_size > 10))\n {\n unsigned char *data = &lzx->e8_buf[0];\n unsigned char *dataend = &lzx->e8_buf[frame_size - 10];\n signed int curpos = lzx->intel_curpos;\n signed int filesize = lzx->intel_filesize;\n signed int abs_off, rel_off;\n\n /* copy e8 block to the e8 buffer and tweak if needed */\n lzx->o_ptr = data;\n lzx->sys->copy(&lzx->window[lzx->frame_posn], data, frame_size);\n\n while (data < dataend) {\n if (*data++ != 0xE8) { curpos++; continue; }\n abs_off = data[0] | (data[1]<<8) | (data[2]<<16) | (data[3]<<24);\n if ((abs_off >= -curpos) && (abs_off < filesize)) {\n rel_off = (abs_off >= 0) ? abs_off - curpos : abs_off + filesize;\n data[0] = (unsigned char) rel_off;\n data[1] = (unsigned char) (rel_off >> 8);\n data[2] = (unsigned char) (rel_off >> 16);\n data[3] = (unsigned char) (rel_off >> 24);\n }\n data += 4;\n curpos += 5;\n }\n lzx->intel_curpos += frame_size;\n }\n else {\n lzx->o_ptr = &lzx->window[lzx->frame_posn];\n if (lzx->intel_filesize) lzx->intel_curpos += frame_size;\n }\n lzx->o_end = &lzx->o_ptr[frame_size];\n\n /* write a frame */\n i = (out_bytes < (off_t)frame_size) ? (unsigned int)out_bytes : frame_size;\n if (lzx->sys->write(lzx->output, lzx->o_ptr, i) != i) {\n return lzx->error = MSPACK_ERR_WRITE;\n }\n lzx->o_ptr += i;\n lzx->offset += i;\n out_bytes -= i;\n\n /* advance frame start position */\n lzx->frame_posn += frame_size;\n lzx->frame++;\n\n /* wrap window / frame position pointers */\n if (window_posn == lzx->window_size) window_posn = 0;\n if (lzx->frame_posn == lzx->window_size) lzx->frame_posn = 0;\n\n } /* while (lzx->frame < end_frame) */\n\n if (out_bytes) {\n D((\"bytes left to output\"))\n return lzx->error = MSPACK_ERR_DECRUNCH;\n }\n\n /* store local state */\n STORE_BITS;\n lzx->window_posn = window_posn;\n lzx->R0 = R0;\n lzx->R1 = R1;\n lzx->R2 = R2;\n\n return MSPACK_ERR_OK;\n}\n\nvoid lzxd_free(struct lzxd_stream *lzx) {\n struct mspack_system *sys;\n if (lzx) {\n sys = lzx->sys;\n sys->free(lzx->inbuf);\n sys->free(lzx->window);\n sys->free(lzx);\n }\n}\n" }, { "path": "third_party/mspack/mspack.h", "content": "/* libmspack -- a library for working with Microsoft compression formats.\n * (C) 2003-2016 Stuart Caie \n *\n * libmspack is free software; you can redistribute it and/or modify it under\n * the terms of the GNU Lesser General Public License (LGPL) version 2.1\n *\n * This program is distributed in the hope that it will be useful,\n * but WITHOUT ANY WARRANTY; without even the implied warranty of\n * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the\n * GNU Lesser General Public License for more details.\n *\n * You should have received a copy of the GNU Lesser General Public License\n * along with this program; if not, write to the Free Software\n * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.\n */\n\n/** \\mainpage\n *\n * \\section intro Introduction\n *\n * libmspack is a library which provides compressors and decompressors,\n * archivers and dearchivers for Microsoft compression formats.\n *\n * \\section formats Formats supported\n *\n * The following file formats are supported:\n * - SZDD files, which use LZSS compression\n * - KWAJ files, which use LZSS, LZSS+Huffman or deflate compression\n * - .HLP (MS Help) files, which use LZSS compression\n * - .CAB (MS Cabinet) files, which use deflate, LZX or Quantum compression\n * - .CHM (HTML Help) files, which use LZX compression\n * - .LIT (MS EBook) files, which use LZX compression and DES encryption\n * - .LZX (Exchange Offline Addressbook) files, which use LZX compression\n *\n * To determine the capabilities of the library, and the binary\n * compatibility version of any particular compressor or decompressor, use\n * the mspack_version() function. The UNIX library interface version is\n * defined as the highest-versioned library component.\n *\n * \\section starting Getting started\n *\n * The macro MSPACK_SYS_SELFTEST() should be used to ensure the library can\n * be used. In particular, it checks if the caller is using 32-bit file I/O\n * when the library is compiled for 64-bit file I/O and vice versa.\n *\n * If compiled normally, the library includes basic file I/O and memory\n * management functionality using the standard C library. This can be\n * customised and replaced entirely by creating a mspack_system structure.\n *\n * A compressor or decompressor for the required format must be\n * instantiated before it can be used. Each construction function takes\n * one parameter, which is either a pointer to a custom mspack_system\n * structure, or NULL to use the default. The instantiation returned, if\n * not NULL, contains function pointers (methods) to work with the given\n * file format.\n * \n * For compression:\n * - mspack_create_cab_compressor() creates a mscab_compressor\n * - mspack_create_chm_compressor() creates a mschm_compressor\n * - mspack_create_lit_compressor() creates a mslit_compressor\n * - mspack_create_hlp_compressor() creates a mshlp_compressor\n * - mspack_create_szdd_compressor() creates a msszdd_compressor\n * - mspack_create_kwaj_compressor() creates a mskwaj_compressor\n * - mspack_create_oab_compressor() creates a msoab_compressor\n *\n * For decompression:\n * - mspack_create_cab_decompressor() creates a mscab_decompressor\n * - mspack_create_chm_decompressor() creates a mschm_decompressor\n * - mspack_create_lit_decompressor() creates a mslit_decompressor\n * - mspack_create_hlp_decompressor() creates a mshlp_decompressor\n * - mspack_create_szdd_decompressor() creates a msszdd_decompressor\n * - mspack_create_kwaj_decompressor() creates a mskwaj_decompressor\n * - mspack_create_oab_decompressor() creates a msoab_decompressor\n *\n * Once finished working with a format, each kind of\n * compressor/decompressor has its own specific destructor:\n * - mspack_destroy_cab_compressor()\n * - mspack_destroy_cab_decompressor()\n * - mspack_destroy_chm_compressor()\n * - mspack_destroy_chm_decompressor()\n * - mspack_destroy_lit_compressor()\n * - mspack_destroy_lit_decompressor()\n * - mspack_destroy_hlp_compressor()\n * - mspack_destroy_hlp_decompressor()\n * - mspack_destroy_szdd_compressor()\n * - mspack_destroy_szdd_decompressor()\n * - mspack_destroy_kwaj_compressor()\n * - mspack_destroy_kwaj_decompressor()\n * - mspack_destroy_oab_compressor()\n * - mspack_destroy_oab_decompressor()\n *\n * Destroying a compressor or decompressor does not destroy any objects,\n * structures or handles that have been created using that compressor or\n * decompressor. Ensure that everything created or opened is destroyed or\n * closed before compressor/decompressor is itself destroyed.\n *\n * \\section errors Error codes\n *\n * All compressors and decompressors use the same set of error codes. Most\n * methods return an error code directly. For methods which do not\n * return error codes directly, the error code can be obtained with the\n * last_error() method.\n *\n * - #MSPACK_ERR_OK is used to indicate success. This error code is defined\n * as zero, all other code are non-zero.\n * - #MSPACK_ERR_ARGS indicates that a method was called with inappropriate\n * arguments.\n * - #MSPACK_ERR_OPEN indicates that mspack_system::open() failed.\n * - #MSPACK_ERR_READ indicates that mspack_system::read() failed.\n * - #MSPACK_ERR_WRITE indicates that mspack_system::write() failed.\n * - #MSPACK_ERR_SEEK indicates that mspack_system::seek() failed.\n * - #MSPACK_ERR_NOMEMORY indicates that mspack_system::alloc() failed.\n * - #MSPACK_ERR_SIGNATURE indicates that the file being read does not\n * have the correct \"signature\". It is probably not a valid file for\n * whatever format is being read.\n * - #MSPACK_ERR_DATAFORMAT indicates that the file being used or read\n * is corrupt.\n * - #MSPACK_ERR_CHECKSUM indicates that a data checksum has failed.\n * - #MSPACK_ERR_CRUNCH indicates an error occured during compression.\n * - #MSPACK_ERR_DECRUNCH indicates an error occured during decompression.\n *\n * \\section threading Multi-threading\n *\n * libmspack methods are reentrant and multithreading-safe when each\n * thread has its own compressor or decompressor.\n\n * You should not call multiple methods simultaneously on a single\n * compressor or decompressor instance.\n *\n * If this may happen, you can either use one compressor or\n * decompressor per thread, or you can use your preferred lock,\n * semaphore or mutex library to ensure no more than one method on a\n * compressor/decompressor is called simultaneously. libmspack will\n * not do this locking for you.\n *\n * Example of incorrect behaviour:\n * - thread 1 calls mspack_create_cab_decompressor()\n * - thread 1 calls open()\n * - thread 1 calls extract() for one file\n * - thread 2 simultaneously calls extract() for another file\n *\n * Correct behaviour:\n * - thread 1 calls mspack_create_cab_decompressor()\n * - thread 2 calls mspack_create_cab_decompressor()\n * - thread 1 calls its own open() / extract()\n * - thread 2 simultaneously calls its own open() / extract()\n *\n * Also correct behaviour:\n * - thread 1 calls mspack_create_cab_decompressor()\n * - thread 1 locks a mutex for with the decompressor before\n * calling any methods on it, and unlocks the mutex after each\n * method returns.\n * - thread 1 can share the results of open() with thread 2, and both\n * can call extract(), provided they both guard against simultaneous\n * use of extract(), and any other methods, with the mutex\n */\n\n#ifndef LIB_MSPACK_H\n#define LIB_MSPACK_H 1\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\n#include \n#include \n\n/**\n * System self-test function, to ensure both library and calling program\n * can use one another.\n *\n * A result of MSPACK_ERR_OK means the library and caller are\n * compatible. Any other result indicates that the library and caller are\n * not compatible and should not be used. In particular, a value of\n * MSPACK_ERR_SEEK means the library and caller use different off_t\n * datatypes.\n *\n * It should be used like so:\n *\n * @code\n * int selftest_result;\n * MSPACK_SYS_SELFTEST(selftest_result);\n * if (selftest_result != MSPACK_ERR_OK) {\n * fprintf(stderr, \"incompatible with this build of libmspack\\n\");\n * exit(0);\n * }\n * @endcode\n *\n * @param result an int variable to store the result of the self-test\n */\n#define MSPACK_SYS_SELFTEST(result) do { \\\n (result) = mspack_sys_selftest_internal(sizeof(off_t)); \\\n} while (0)\n\n/** Part of the MSPACK_SYS_SELFTEST() macro, must not be used directly. */\nextern int mspack_sys_selftest_internal(int);\n\n/**\n * Enquire about the binary compatibility version of a specific interface in\n * the library. Currently, the following interfaces are defined:\n *\n * - #MSPACK_VER_LIBRARY: the overall library\n * - #MSPACK_VER_SYSTEM: the mspack_system interface\n * - #MSPACK_VER_MSCABD: the mscab_decompressor interface\n * - #MSPACK_VER_MSCABC: the mscab_compressor interface\n * - #MSPACK_VER_MSCHMD: the mschm_decompressor interface\n * - #MSPACK_VER_MSCHMC: the mschm_compressor interface\n * - #MSPACK_VER_MSLITD: the mslit_decompressor interface\n * - #MSPACK_VER_MSLITC: the mslit_compressor interface\n * - #MSPACK_VER_MSHLPD: the mshlp_decompressor interface\n * - #MSPACK_VER_MSHLPC: the mshlp_compressor interface\n * - #MSPACK_VER_MSSZDDD: the msszdd_decompressor interface\n * - #MSPACK_VER_MSSZDDC: the msszdd_compressor interface\n * - #MSPACK_VER_MSKWAJD: the mskwaj_decompressor interface\n * - #MSPACK_VER_MSKWAJC: the mskwaj_compressor interface\n * - #MSPACK_VER_MSOABD: the msoab_decompressor interface\n * - #MSPACK_VER_MSOABC: the msoab_compressor interface\n *\n * The result of the function should be interpreted as follows:\n * - -1: this interface is completely unknown to the library\n * - 0: this interface is known, but non-functioning\n * - 1: this interface has all basic functionality\n * - 2, 3, ...: this interface has additional functionality, clearly marked\n * in the documentation as \"version 2\", \"version 3\" and so on.\n *\n * @param entity the interface to request current version of\n * @return the version of the requested interface\n */\nextern int mspack_version(int entity);\n\n/** Pass to mspack_version() to get the overall library version */\n#define MSPACK_VER_LIBRARY (0)\n/** Pass to mspack_version() to get the mspack_system version */\n#define MSPACK_VER_SYSTEM (1)\n/** Pass to mspack_version() to get the mscab_decompressor version */\n#define MSPACK_VER_MSCABD (2)\n/** Pass to mspack_version() to get the mscab_compressor version */\n#define MSPACK_VER_MSCABC (3)\n/** Pass to mspack_version() to get the mschm_decompressor version */\n#define MSPACK_VER_MSCHMD (4)\n/** Pass to mspack_version() to get the mschm_compressor version */\n#define MSPACK_VER_MSCHMC (5)\n/** Pass to mspack_version() to get the mslit_decompressor version */\n#define MSPACK_VER_MSLITD (6)\n/** Pass to mspack_version() to get the mslit_compressor version */\n#define MSPACK_VER_MSLITC (7)\n/** Pass to mspack_version() to get the mshlp_decompressor version */\n#define MSPACK_VER_MSHLPD (8)\n/** Pass to mspack_version() to get the mshlp_compressor version */\n#define MSPACK_VER_MSHLPC (9)\n/** Pass to mspack_version() to get the msszdd_decompressor version */\n#define MSPACK_VER_MSSZDDD (10)\n/** Pass to mspack_version() to get the msszdd_compressor version */\n#define MSPACK_VER_MSSZDDC (11)\n/** Pass to mspack_version() to get the mskwaj_decompressor version */\n#define MSPACK_VER_MSKWAJD (12)\n/** Pass to mspack_version() to get the mskwaj_compressor version */\n#define MSPACK_VER_MSKWAJC (13)\n/** Pass to mspack_version() to get the msoab_decompressor version */\n#define MSPACK_VER_MSOABD (14)\n/** Pass to mspack_version() to get the msoab_compressor version */\n#define MSPACK_VER_MSOABC (15)\n\n/* --- file I/O abstraction ------------------------------------------------ */\n\n/**\n * A structure which abstracts file I/O and memory management.\n *\n * The library always uses the mspack_system structure for interaction\n * with the file system and to allocate, free and copy all memory. It also\n * uses it to send literal messages to the library user.\n *\n * When the library is compiled normally, passing NULL to a compressor or\n * decompressor constructor will result in a default mspack_system being\n * used, where all methods are implemented with the standard C library.\n * However, all constructors support being given a custom created\n * mspack_system structure, with the library user's own methods. This\n * allows for more abstract interaction, such as reading and writing files\n * directly to memory, or from a network socket or pipe.\n *\n * Implementors of an mspack_system structure should read all\n * documentation entries for every structure member, and write methods\n * which conform to those standards.\n */\nstruct mspack_system {\n /**\n * Opens a file for reading, writing, appending or updating.\n *\n * @param self a self-referential pointer to the mspack_system\n * structure whose open() method is being called. If\n * this pointer is required by close(), read(), write(),\n * seek() or tell(), it should be stored in the result\n * structure at this time.\n * @param filename the file to be opened. It is passed directly from the\n * library caller without being modified, so it is up to\n * the caller what this parameter actually represents.\n * @param mode one of #MSPACK_SYS_OPEN_READ (open an existing file\n * for reading), #MSPACK_SYS_OPEN_WRITE (open a new file\n * for writing), #MSPACK_SYS_OPEN_UPDATE (open an existing\n * file for reading/writing from the start of the file) or\n * #MSPACK_SYS_OPEN_APPEND (open an existing file for\n * reading/writing from the end of the file)\n * @return a pointer to a mspack_file structure. This structure officially\n * contains no members, its true contents are up to the\n * mspack_system implementor. It should contain whatever is needed\n * for other mspack_system methods to operate. Returning the NULL\n * pointer indicates an error condition.\n * @see close(), read(), write(), seek(), tell(), message()\n */\n struct mspack_file * (*open)(struct mspack_system *self,\n const char *filename,\n int mode);\n\n /**\n * Closes a previously opened file. If any memory was allocated for this\n * particular file handle, it should be freed at this time.\n * \n * @param file the file to close\n * @see open()\n */\n void (*close)(struct mspack_file *file);\n\n /**\n * Reads a given number of bytes from an open file.\n *\n * @param file the file to read from\n * @param buffer the location where the read bytes should be stored\n * @param bytes the number of bytes to read from the file.\n * @return the number of bytes successfully read (this can be less than\n * the number requested), zero to mark the end of file, or less\n * than zero to indicate an error. The library does not \"retry\"\n * reads and assumes short reads are due to EOF, so you should\n * avoid returning short reads because of transient errors.\n * @see open(), write()\n */\n int (*read)(struct mspack_file *file,\n void *buffer,\n int bytes);\n\n /**\n * Writes a given number of bytes to an open file.\n *\n * @param file the file to write to\n * @param buffer the location where the written bytes should be read from\n * @param bytes the number of bytes to write to the file.\n * @return the number of bytes successfully written, this can be less\n * than the number requested. Zero or less can indicate an error\n * where no bytes at all could be written. All cases where less\n * bytes were written than requested are considered by the library\n * to be an error.\n * @see open(), read()\n */\n int (*write)(struct mspack_file *file,\n void *buffer,\n int bytes);\n\n /**\n * Seeks to a specific file offset within an open file.\n *\n * Sometimes the library needs to know the length of a file. It does\n * this by seeking to the end of the file with seek(file, 0,\n * MSPACK_SYS_SEEK_END), then calling tell(). Implementations may want\n * to make a special case for this.\n *\n * Due to the potentially varying 32/64 bit datatype off_t on some\n * architectures, the #MSPACK_SYS_SELFTEST macro MUST be used before\n * using the library. If not, the error caused by the library passing an\n * inappropriate stackframe to seek() is subtle and hard to trace.\n *\n * @param file the file to be seeked\n * @param offset an offset to seek, measured in bytes\n * @param mode one of #MSPACK_SYS_SEEK_START (the offset should be\n * measured from the start of the file), #MSPACK_SYS_SEEK_CUR\n * (the offset should be measured from the current file offset)\n * or #MSPACK_SYS_SEEK_END (the offset should be measured from\n * the end of the file)\n * @return zero for success, non-zero for an error\n * @see open(), tell()\n */\n int (*seek)(struct mspack_file *file,\n off_t offset,\n int mode);\n\n /**\n * Returns the current file position (in bytes) of the given file.\n *\n * @param file the file whose file position is wanted\n * @return the current file position of the file\n * @see open(), seek()\n */\n off_t (*tell)(struct mspack_file *file);\n \n /**\n * Used to send messages from the library to the user.\n *\n * Occasionally, the library generates warnings or other messages in\n * plain english to inform the human user. These are informational only\n * and can be ignored if not wanted.\n *\n * @param file may be a file handle returned from open() if this message\n * pertains to a specific open file, or NULL if not related to\n * a specific file.\n * @param format a printf() style format string. It does NOT include a\n * trailing newline.\n * @see open()\n */\n void (*message)(struct mspack_file *file,\n const char *format,\n ...);\n\n /**\n * Allocates memory.\n *\n * @param self a self-referential pointer to the mspack_system\n * structure whose alloc() method is being called.\n * @param bytes the number of bytes to allocate\n * @result a pointer to the requested number of bytes, or NULL if\n * not enough memory is available\n * @see free()\n */\n void * (*alloc)(struct mspack_system *self,\n size_t bytes);\n \n /**\n * Frees memory.\n * \n * @param ptr the memory to be freed. NULL is accepted and ignored.\n * @see alloc()\n */\n void (*free)(void *ptr);\n\n /**\n * Copies from one region of memory to another.\n * \n * The regions of memory are guaranteed not to overlap, are usually less\n * than 256 bytes, and may not be aligned. Please note that the source\n * parameter comes before the destination parameter, unlike the standard\n * C function memcpy().\n *\n * @param src the region of memory to copy from\n * @param dest the region of memory to copy to\n * @param bytes the size of the memory region, in bytes\n */\n void (*copy)(void *src,\n void *dest,\n size_t bytes);\n\n /**\n * A null pointer to mark the end of mspack_system. It must equal NULL.\n *\n * Should the mspack_system structure extend in the future, this NULL\n * will be seen, rather than have an invalid method pointer called.\n */\n void *null_ptr;\n};\n\n/** mspack_system::open() mode: open existing file for reading. */\n#define MSPACK_SYS_OPEN_READ (0)\n/** mspack_system::open() mode: open new file for writing */\n#define MSPACK_SYS_OPEN_WRITE (1)\n/** mspack_system::open() mode: open existing file for writing */\n#define MSPACK_SYS_OPEN_UPDATE (2)\n/** mspack_system::open() mode: open existing file for writing */\n#define MSPACK_SYS_OPEN_APPEND (3)\n\n/** mspack_system::seek() mode: seek relative to start of file */\n#define MSPACK_SYS_SEEK_START (0)\n/** mspack_system::seek() mode: seek relative to current offset */\n#define MSPACK_SYS_SEEK_CUR (1)\n/** mspack_system::seek() mode: seek relative to end of file */\n#define MSPACK_SYS_SEEK_END (2)\n\n/** \n * A structure which represents an open file handle. The contents of this\n * structure are determined by the implementation of the\n * mspack_system::open() method.\n */\nstruct mspack_file {\n int dummy;\n};\n\n/* --- error codes --------------------------------------------------------- */\n\n/** Error code: no error */\n#define MSPACK_ERR_OK (0)\n/** Error code: bad arguments to method */\n#define MSPACK_ERR_ARGS (1)\n/** Error code: error opening file */\n#define MSPACK_ERR_OPEN (2)\n/** Error code: error reading file */\n#define MSPACK_ERR_READ (3)\n/** Error code: error writing file */\n#define MSPACK_ERR_WRITE (4)\n/** Error code: seek error */\n#define MSPACK_ERR_SEEK (5)\n/** Error code: out of memory */\n#define MSPACK_ERR_NOMEMORY (6)\n/** Error code: bad \"magic id\" in file */\n#define MSPACK_ERR_SIGNATURE (7)\n/** Error code: bad or corrupt file format */\n#define MSPACK_ERR_DATAFORMAT (8)\n/** Error code: bad checksum or CRC */\n#define MSPACK_ERR_CHECKSUM (9)\n/** Error code: error during compression */\n#define MSPACK_ERR_CRUNCH (10)\n/** Error code: error during decompression */\n#define MSPACK_ERR_DECRUNCH (11)\n\n/* --- functions available in library -------------------------------------- */\n\n/** Creates a new CAB compressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mscab_compressor or NULL\n */\nextern struct mscab_compressor *\n mspack_create_cab_compressor(struct mspack_system *sys);\n\n/** Creates a new CAB decompressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mscab_decompressor or NULL\n */\nextern struct mscab_decompressor *\n mspack_create_cab_decompressor(struct mspack_system *sys);\n\n/** Destroys an existing CAB compressor.\n * @param self the #mscab_compressor to destroy\n */\nextern void mspack_destroy_cab_compressor(struct mscab_compressor *self);\n\n/** Destroys an existing CAB decompressor.\n * @param self the #mscab_decompressor to destroy\n */\nextern void mspack_destroy_cab_decompressor(struct mscab_decompressor *self);\n\n\n/** Creates a new CHM compressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mschm_compressor or NULL\n */\nextern struct mschm_compressor *\n mspack_create_chm_compressor(struct mspack_system *sys);\n\n/** Creates a new CHM decompressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mschm_decompressor or NULL\n */\nextern struct mschm_decompressor *\n mspack_create_chm_decompressor(struct mspack_system *sys);\n\n/** Destroys an existing CHM compressor.\n * @param self the #mschm_compressor to destroy\n */\nextern void mspack_destroy_chm_compressor(struct mschm_compressor *self);\n\n/** Destroys an existing CHM decompressor.\n * @param self the #mschm_decompressor to destroy\n */\nextern void mspack_destroy_chm_decompressor(struct mschm_decompressor *self);\n\n\n/** Creates a new LIT compressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mslit_compressor or NULL\n */\nextern struct mslit_compressor *\n mspack_create_lit_compressor(struct mspack_system *sys);\n\n/** Creates a new LIT decompressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mslit_decompressor or NULL\n */\nextern struct mslit_decompressor *\n mspack_create_lit_decompressor(struct mspack_system *sys);\n\n/** Destroys an existing LIT compressor.\n * @param self the #mslit_compressor to destroy\n */\nextern void mspack_destroy_lit_compressor(struct mslit_compressor *self);\n\n/** Destroys an existing LIT decompressor.\n * @param self the #mslit_decompressor to destroy\n */\nextern void mspack_destroy_lit_decompressor(struct mslit_decompressor *self);\n\n\n/** Creates a new HLP compressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mshlp_compressor or NULL\n */\nextern struct mshlp_compressor *\n mspack_create_hlp_compressor(struct mspack_system *sys);\n\n/** Creates a new HLP decompressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mshlp_decompressor or NULL\n */\nextern struct mshlp_decompressor *\n mspack_create_hlp_decompressor(struct mspack_system *sys);\n\n/** Destroys an existing hlp compressor.\n * @param self the #mshlp_compressor to destroy\n */\nextern void mspack_destroy_hlp_compressor(struct mshlp_compressor *self);\n\n/** Destroys an existing hlp decompressor.\n * @param self the #mshlp_decompressor to destroy\n */\nextern void mspack_destroy_hlp_decompressor(struct mshlp_decompressor *self);\n\n\n/** Creates a new SZDD compressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #msszdd_compressor or NULL\n */\nextern struct msszdd_compressor *\n mspack_create_szdd_compressor(struct mspack_system *sys);\n\n/** Creates a new SZDD decompressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #msszdd_decompressor or NULL\n */\nextern struct msszdd_decompressor *\n mspack_create_szdd_decompressor(struct mspack_system *sys);\n\n/** Destroys an existing SZDD compressor.\n * @param self the #msszdd_compressor to destroy\n */\nextern void mspack_destroy_szdd_compressor(struct msszdd_compressor *self);\n\n/** Destroys an existing SZDD decompressor.\n * @param self the #msszdd_decompressor to destroy\n */\nextern void mspack_destroy_szdd_decompressor(struct msszdd_decompressor *self);\n\n\n/** Creates a new KWAJ compressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mskwaj_compressor or NULL\n */\nextern struct mskwaj_compressor *\n mspack_create_kwaj_compressor(struct mspack_system *sys);\n\n/** Creates a new KWAJ decompressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #mskwaj_decompressor or NULL\n */\nextern struct mskwaj_decompressor *\n mspack_create_kwaj_decompressor(struct mspack_system *sys);\n\n/** Destroys an existing KWAJ compressor.\n * @param self the #mskwaj_compressor to destroy\n */\nextern void mspack_destroy_kwaj_compressor(struct mskwaj_compressor *self);\n\n/** Destroys an existing KWAJ decompressor.\n * @param self the #mskwaj_decompressor to destroy\n */\nextern void mspack_destroy_kwaj_decompressor(struct mskwaj_decompressor *self);\n\n\n/** Creates a new OAB compressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #msoab_compressor or NULL\n */\nextern struct msoab_compressor *\n mspack_create_oab_compressor(struct mspack_system *sys);\n\n/** Creates a new OAB decompressor.\n * @param sys a custom mspack_system structure, or NULL to use the default\n * @return a #msoab_decompressor or NULL\n */\nextern struct msoab_decompressor *\n mspack_create_oab_decompressor(struct mspack_system *sys);\n\n/** Destroys an existing OAB compressor.\n * @param self the #msoab_compressor to destroy\n */\nextern void mspack_destroy_oab_compressor(struct msoab_compressor *self);\n\n/** Destroys an existing OAB decompressor.\n * @param self the #msoab_decompressor to destroy\n */\nextern void mspack_destroy_oab_decompressor(struct msoab_decompressor *self);\n\n\n/* --- support for .CAB (MS Cabinet) file format --------------------------- */\n\n/**\n * A structure which represents a single cabinet file.\n *\n * All fields are READ ONLY.\n *\n * If this cabinet is part of a merged cabinet set, the #files and #folders\n * fields are common to all cabinets in the set, and will be identical.\n *\n * @see mscab_decompressor::open(), mscab_decompressor::close(),\n * mscab_decompressor::search()\n */\nstruct mscabd_cabinet {\n /**\n * The next cabinet in a chained list, if this cabinet was opened with\n * mscab_decompressor::search(). May be NULL to mark the end of the\n * list.\n */\n struct mscabd_cabinet *next;\n\n /**\n * The filename of the cabinet. More correctly, the filename of the\n * physical file that the cabinet resides in. This is given by the\n * library user and may be in any format.\n */\n const char *filename;\n \n /** The file offset of cabinet within the physical file it resides in. */\n off_t base_offset;\n\n /** The length of the cabinet file in bytes. */\n unsigned int length;\n\n /** The previous cabinet in a cabinet set, or NULL. */\n struct mscabd_cabinet *prevcab;\n\n /** The next cabinet in a cabinet set, or NULL. */\n struct mscabd_cabinet *nextcab;\n\n /** The filename of the previous cabinet in a cabinet set, or NULL. */\n char *prevname;\n\n /** The filename of the next cabinet in a cabinet set, or NULL. */\n char *nextname;\n\n /** The name of the disk containing the previous cabinet in a cabinet\n * set, or NULL.\n */\n char *previnfo;\n\n /** The name of the disk containing the next cabinet in a cabinet set,\n * or NULL.\n */\n char *nextinfo;\n\n /** A list of all files in the cabinet or cabinet set. */\n struct mscabd_file *files;\n\n /** A list of all folders in the cabinet or cabinet set. */\n struct mscabd_folder *folders;\n\n /** \n * The set ID of the cabinet. All cabinets in the same set should have\n * the same set ID.\n */\n unsigned short set_id;\n\n /**\n * The index number of the cabinet within the set. Numbering should\n * start from 0 for the first cabinet in the set, and increment by 1 for\n * each following cabinet.\n */\n unsigned short set_index;\n\n /**\n * The number of bytes reserved in the header area of the cabinet.\n *\n * If this is non-zero and flags has MSCAB_HDR_RESV set, this data can\n * be read by the calling application. It is of the given length,\n * located at offset (base_offset + MSCAB_HDR_RESV_OFFSET) in the\n * cabinet file.\n *\n * @see flags\n */\n unsigned short header_resv;\n\n /**\n * Header flags.\n *\n * - MSCAB_HDR_PREVCAB indicates the cabinet is part of a cabinet set, and\n * has a predecessor cabinet.\n * - MSCAB_HDR_NEXTCAB indicates the cabinet is part of a cabinet set, and\n * has a successor cabinet.\n * - MSCAB_HDR_RESV indicates the cabinet has reserved header space.\n *\n * @see prevname, previnfo, nextname, nextinfo, header_resv\n */\n int flags;\n};\n\n/** Offset from start of cabinet to the reserved header data (if present). */\n#define MSCAB_HDR_RESV_OFFSET (0x28)\n\n/** Cabinet header flag: cabinet has a predecessor */\n#define MSCAB_HDR_PREVCAB (0x01)\n/** Cabinet header flag: cabinet has a successor */\n#define MSCAB_HDR_NEXTCAB (0x02)\n/** Cabinet header flag: cabinet has reserved header space */\n#define MSCAB_HDR_RESV (0x04)\n\n/**\n * A structure which represents a single folder in a cabinet or cabinet set.\n *\n * All fields are READ ONLY.\n *\n * A folder is a single compressed stream of data. When uncompressed, it\n * holds the data of one or more files. A folder may be split across more\n * than one cabinet.\n */\nstruct mscabd_folder {\n /**\n * A pointer to the next folder in this cabinet or cabinet set, or NULL\n * if this is the final folder.\n */\n struct mscabd_folder *next;\n\n /** \n * The compression format used by this folder.\n *\n * The macro MSCABD_COMP_METHOD() should be used on this field to get\n * the algorithm used. The macro MSCABD_COMP_LEVEL() should be used to get\n * the \"compression level\".\n *\n * @see MSCABD_COMP_METHOD(), MSCABD_COMP_LEVEL()\n */\n int comp_type;\n\n /**\n * The total number of data blocks used by this folder. This includes\n * data blocks present in other files, if this folder spans more than\n * one cabinet.\n */\n unsigned int num_blocks;\n};\n\n/**\n * Returns the compression method used by a folder.\n *\n * @param comp_type a mscabd_folder::comp_type value\n * @return one of #MSCAB_COMP_NONE, #MSCAB_COMP_MSZIP, #MSCAB_COMP_QUANTUM\n * or #MSCAB_COMP_LZX\n */\n#define MSCABD_COMP_METHOD(comp_type) ((comp_type) & 0x0F)\n/**\n * Returns the compression level used by a folder.\n *\n * @param comp_type a mscabd_folder::comp_type value\n * @return the compression level. This is only defined by LZX and Quantum\n * compression\n */\n#define MSCABD_COMP_LEVEL(comp_type) (((comp_type) >> 8) & 0x1F)\n\n/** Compression mode: no compression. */\n#define MSCAB_COMP_NONE (0)\n/** Compression mode: MSZIP (deflate) compression. */\n#define MSCAB_COMP_MSZIP (1)\n/** Compression mode: Quantum compression */\n#define MSCAB_COMP_QUANTUM (2)\n/** Compression mode: LZX compression */\n#define MSCAB_COMP_LZX (3)\n\n/**\n * A structure which represents a single file in a cabinet or cabinet set.\n *\n * All fields are READ ONLY.\n */\nstruct mscabd_file {\n /**\n * The next file in the cabinet or cabinet set, or NULL if this is the\n * final file.\n */\n struct mscabd_file *next;\n\n /**\n * The filename of the file.\n *\n * A null terminated string of up to 255 bytes in length, it may be in\n * either ISO-8859-1 or UTF8 format, depending on the file attributes.\n *\n * @see attribs\n */\n char *filename;\n\n /** The uncompressed length of the file, in bytes. */\n unsigned int length;\n\n /**\n * File attributes.\n *\n * The following attributes are defined:\n * - #MSCAB_ATTRIB_RDONLY indicates the file is write protected.\n * - #MSCAB_ATTRIB_HIDDEN indicates the file is hidden.\n * - #MSCAB_ATTRIB_SYSTEM indicates the file is a operating system file.\n * - #MSCAB_ATTRIB_ARCH indicates the file is \"archived\".\n * - #MSCAB_ATTRIB_EXEC indicates the file is an executable program.\n * - #MSCAB_ATTRIB_UTF_NAME indicates the filename is in UTF8 format rather\n * than ISO-8859-1.\n */\n int attribs;\n\n /** File's last modified time, hour field. */\n char time_h;\n /** File's last modified time, minute field. */\n char time_m;\n /** File's last modified time, second field. */\n char time_s;\n\n /** File's last modified date, day field. */\n char date_d;\n /** File's last modified date, month field. */\n char date_m;\n /** File's last modified date, year field. */\n int date_y;\n\n /** A pointer to the folder that contains this file. */\n struct mscabd_folder *folder;\n\n /** The uncompressed offset of this file in its folder. */\n unsigned int offset;\n};\n\n/** mscabd_file::attribs attribute: file is read-only. */\n#define MSCAB_ATTRIB_RDONLY (0x01)\n/** mscabd_file::attribs attribute: file is hidden. */\n#define MSCAB_ATTRIB_HIDDEN (0x02)\n/** mscabd_file::attribs attribute: file is an operating system file. */\n#define MSCAB_ATTRIB_SYSTEM (0x04)\n/** mscabd_file::attribs attribute: file is \"archived\". */\n#define MSCAB_ATTRIB_ARCH (0x20)\n/** mscabd_file::attribs attribute: file is an executable program. */\n#define MSCAB_ATTRIB_EXEC (0x40)\n/** mscabd_file::attribs attribute: filename is UTF8, not ISO-8859-1. */\n#define MSCAB_ATTRIB_UTF_NAME (0x80)\n\n/** mscab_decompressor::set_param() parameter: search buffer size. */\n#define MSCABD_PARAM_SEARCHBUF (0)\n/** mscab_decompressor::set_param() parameter: repair MS-ZIP streams? */\n#define MSCABD_PARAM_FIXMSZIP (1)\n/** mscab_decompressor::set_param() parameter: size of decompression buffer */\n#define MSCABD_PARAM_DECOMPBUF (2)\n/** mscab_decompressor::set_param() parameter: salvage data from bad cabinets?\n * If enabled, open() will skip file with bad folder indices or filenames\n * rather than reject the whole cabinet, and extract() will limit rather than\n * reject files with invalid offsets and lengths, and bad data block checksums\n * will be ignored. Available only in CAB decoder version 2 and above.\n */\n#define MSCABD_PARAM_SALVAGE (3)\n\n/** TODO */\nstruct mscab_compressor {\n int dummy; \n};\n\n/**\n * A decompressor for .CAB (Microsoft Cabinet) files\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_cab_decompressor(), mspack_destroy_cab_decompressor()\n */\nstruct mscab_decompressor {\n /**\n * Opens a cabinet file and reads its contents.\n *\n * If the file opened is a valid cabinet file, all headers will be read\n * and a mscabd_cabinet structure will be returned, with a full list of\n * folders and files.\n *\n * In the case of an error occuring, NULL is returned and the error code\n * is available from last_error().\n *\n * The filename pointer should be considered \"in use\" until close() is\n * called on the cabinet.\n *\n * @param self a self-referential pointer to the mscab_decompressor\n * instance being called\n * @param filename the filename of the cabinet file. This is passed\n * directly to mspack_system::open().\n * @return a pointer to a mscabd_cabinet structure, or NULL on failure\n * @see close(), search(), last_error()\n */\n struct mscabd_cabinet * (*open) (struct mscab_decompressor *self,\n const char *filename);\n\n /**\n * Closes a previously opened cabinet or cabinet set.\n *\n * This closes a cabinet, all cabinets associated with it via the\n * mscabd_cabinet::next, mscabd_cabinet::prevcab and\n * mscabd_cabinet::nextcab pointers, and all folders and files. All\n * memory used by these entities is freed.\n *\n * The cabinet pointer is now invalid and cannot be used again. All\n * mscabd_folder and mscabd_file pointers from that cabinet or cabinet\n * set are also now invalid, and cannot be used again.\n *\n * If the cabinet pointer given was created using search(), it MUST be\n * the cabinet pointer returned by search() and not one of the later\n * cabinet pointers further along the mscabd_cabinet::next chain.\n\n * If extra cabinets have been added using append() or prepend(), these\n * will all be freed, even if the cabinet pointer given is not the first\n * cabinet in the set. Do NOT close() more than one cabinet in the set.\n *\n * The mscabd_cabinet::filename is not freed by the library, as it is\n * not allocated by the library. The caller should free this itself if\n * necessary, before it is lost forever.\n *\n * @param self a self-referential pointer to the mscab_decompressor\n * instance being called\n * @param cab the cabinet to close\n * @see open(), search(), append(), prepend()\n */\n void (*close)(struct mscab_decompressor *self,\n struct mscabd_cabinet *cab);\n\n /**\n * Searches a regular file for embedded cabinets.\n *\n * This opens a normal file with the given filename and will search the\n * entire file for embedded cabinet files\n *\n * If any cabinets are found, the equivalent of open() is called on each\n * potential cabinet file at the offset it was found. All successfully\n * open()ed cabinets are kept in a list.\n *\n * The first cabinet found will be returned directly as the result of\n * this method. Any further cabinets found will be chained in a list\n * using the mscabd_cabinet::next field.\n *\n * In the case of an error occuring anywhere other than the simulated\n * open(), NULL is returned and the error code is available from\n * last_error().\n *\n * If no error occurs, but no cabinets can be found in the file, NULL is\n * returned and last_error() returns MSPACK_ERR_OK.\n *\n * The filename pointer should be considered in use until close() is\n * called on the cabinet.\n *\n * close() should only be called on the result of search(), not on any\n * subsequent cabinets in the mscabd_cabinet::next chain.\n *\n * @param self a self-referential pointer to the mscab_decompressor\n * instance being called\n * @param filename the filename of the file to search for cabinets. This\n * is passed directly to mspack_system::open().\n * @return a pointer to a mscabd_cabinet structure, or NULL\n * @see close(), open(), last_error()\n */\n struct mscabd_cabinet * (*search) (struct mscab_decompressor *self,\n const char *filename);\n\n /**\n * Appends one mscabd_cabinet to another, forming or extending a cabinet\n * set.\n *\n * This will attempt to append one cabinet to another such that\n * (cab->nextcab == nextcab) && (nextcab->prevcab == cab) and\n * any folders split between the two cabinets are merged.\n *\n * The cabinets MUST be part of a cabinet set -- a cabinet set is a\n * cabinet that spans more than one physical cabinet file on disk -- and\n * must be appropriately matched.\n *\n * It can be determined if a cabinet has further parts to load by\n * examining the mscabd_cabinet::flags field:\n *\n * - if (flags & MSCAB_HDR_PREVCAB) is non-zero, there is a\n * predecessor cabinet to open() and prepend(). Its MS-DOS\n * case-insensitive filename is mscabd_cabinet::prevname\n * - if (flags & MSCAB_HDR_NEXTCAB) is non-zero, there is a\n * successor cabinet to open() and append(). Its MS-DOS case-insensitive\n * filename is mscabd_cabinet::nextname\n *\n * If the cabinets do not match, an error code will be returned. Neither\n * cabinet has been altered, and both should be closed seperately.\n *\n * Files and folders in a cabinet set are a single entity. All cabinets\n * in a set use the same file list, which is updated as cabinets in the\n * set are added. All pointers to mscabd_folder and mscabd_file\n * structures in either cabinet must be discarded and re-obtained after\n * merging.\n *\n * @param self a self-referential pointer to the mscab_decompressor\n * instance being called\n * @param cab the cabinet which will be appended to,\n * predecessor of nextcab\n * @param nextcab the cabinet which will be appended,\n * successor of cab\n * @return an error code, or MSPACK_ERR_OK if successful\n * @see prepend(), open(), close()\n */\n int (*append) (struct mscab_decompressor *self,\n struct mscabd_cabinet *cab,\n struct mscabd_cabinet *nextcab);\n\n /**\n * Prepends one mscabd_cabinet to another, forming or extending a\n * cabinet set.\n *\n * This will attempt to prepend one cabinet to another, such that\n * (cab->prevcab == prevcab) && (prevcab->nextcab == cab). In\n * all other respects, it is identical to append(). See append() for the\n * full documentation.\n *\n * @param self a self-referential pointer to the mscab_decompressor\n * instance being called\n * @param cab the cabinet which will be prepended to,\n * successor of prevcab\n * @param prevcab the cabinet which will be prepended,\n * predecessor of cab\n * @return an error code, or MSPACK_ERR_OK if successful\n * @see append(), open(), close()\n */\n int (*prepend) (struct mscab_decompressor *self,\n struct mscabd_cabinet *cab,\n struct mscabd_cabinet *prevcab);\n\n /**\n * Extracts a file from a cabinet or cabinet set.\n *\n * This extracts a compressed file in a cabinet and writes it to the given\n * filename.\n *\n * The MS-DOS filename of the file, mscabd_file::filename, is NOT USED\n * by extract(). The caller must examine this MS-DOS filename, copy and\n * change it as necessary, create directories as necessary, and provide\n * the correct filename as a parameter, which will be passed unchanged\n * to the decompressor's mspack_system::open()\n *\n * If the file belongs to a split folder in a multi-part cabinet set,\n * and not enough parts of the cabinet set have been loaded and appended\n * or prepended, an error will be returned immediately.\n *\n * @param self a self-referential pointer to the mscab_decompressor\n * instance being called\n * @param file the file to be decompressed\n * @param filename the filename of the file being written to\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*extract)(struct mscab_decompressor *self,\n struct mscabd_file *file,\n const char *filename);\n\n /**\n * Sets a CAB decompression engine parameter.\n *\n * The following parameters are defined:\n * - #MSCABD_PARAM_SEARCHBUF: How many bytes should be allocated as a\n * buffer when using search()? The minimum value is 4. The default\n * value is 32768.\n * - #MSCABD_PARAM_FIXMSZIP: If non-zero, extract() will ignore bad\n * checksums and recover from decompression errors in MS-ZIP\n * compressed folders. The default value is 0 (don't recover).\n * - #MSCABD_PARAM_DECOMPBUF: How many bytes should be used as an input\n * bit buffer by decompressors? The minimum value is 4. The default\n * value is 4096.\n *\n * @param self a self-referential pointer to the mscab_decompressor\n * instance being called\n * @param param the parameter to set\n * @param value the value to set the parameter to\n * @return MSPACK_ERR_OK if all is OK, or MSPACK_ERR_ARGS if there\n * is a problem with either parameter or value.\n * @see search(), extract()\n */\n int (*set_param)(struct mscab_decompressor *self,\n int param,\n int value);\n\n /**\n * Returns the error code set by the most recently called method.\n *\n * This is useful for open() and search(), which do not return an error\n * code directly.\n *\n * @param self a self-referential pointer to the mscab_decompressor\n * instance being called\n * @return the most recent error code\n * @see open(), search()\n */\n int (*last_error)(struct mscab_decompressor *self);\n};\n\n/* --- support for .CHM (HTMLHelp) file format ----------------------------- */\n\n/**\n * A structure which represents a file to be placed in a CHM helpfile.\n *\n * A contiguous array of these structures should be passed to\n * mschm_compressor::generate(). The array list is terminated with an\n * entry whose mschmc_file::section field is set to #MSCHMC_ENDLIST, the\n * other fields in this entry are ignored.\n */\nstruct mschmc_file {\n /** One of #MSCHMC_ENDLIST, #MSCHMC_UNCOMP or #MSCHMC_MSCOMP. */\n int section;\n\n /** The filename of the source file that will be added to the CHM. This\n * is passed directly to mspack_system::open(). */\n const char *filename;\n\n /** The full path and filename of the file within the CHM helpfile, a\n * UTF-1 encoded null-terminated string. */\n char *chm_filename;\n\n /** The length of the file, in bytes. This will be adhered to strictly\n * and a read error will be issued if this many bytes cannot be read\n * from the real file at CHM generation time. */\n off_t length;\n};\n\n/**\n * A structure which represents a section of a CHM helpfile.\n *\n * All fields are READ ONLY.\n *\n * Not used directly, but used as a generic base type for\n * mschmd_sec_uncompressed and mschmd_sec_mscompressed.\n */\nstruct mschmd_section {\n /** A pointer to the CHM helpfile that contains this section. */\n struct mschmd_header *chm;\n\n /**\n * The section ID. Either 0 for the uncompressed section\n * mschmd_sec_uncompressed, or 1 for the LZX compressed section\n * mschmd_sec_mscompressed. No other section IDs are known.\n */\n unsigned int id;\n};\n\n/**\n * A structure which represents the uncompressed section of a CHM helpfile.\n * \n * All fields are READ ONLY.\n */\nstruct mschmd_sec_uncompressed {\n /** Generic section data. */\n struct mschmd_section base;\n\n /** The file offset of where this section begins in the CHM helpfile. */\n off_t offset;\n};\n\n/**\n * A structure which represents the LZX compressed section of a CHM helpfile. \n * \n * All fields are READ ONLY.\n */\nstruct mschmd_sec_mscompressed {\n /** Generic section data. */\n struct mschmd_section base;\n\n /** A pointer to the meta-file which represents all LZX compressed data. */\n struct mschmd_file *content;\n\n /** A pointer to the file which contains the LZX control data. */\n struct mschmd_file *control;\n\n /** A pointer to the file which contains the LZX reset table. */\n struct mschmd_file *rtable;\n\n /** A pointer to the file which contains the LZX span information.\n * Available only in CHM decoder version 2 and above.\n */\n struct mschmd_file *spaninfo;\n};\n\n/**\n * A structure which represents a CHM helpfile.\n * \n * All fields are READ ONLY.\n */\nstruct mschmd_header {\n /** The version of the CHM file format used in this file. */\n unsigned int version;\n\n /**\n * The \"timestamp\" of the CHM helpfile. \n *\n * It is the lower 32 bits of a 64-bit value representing the number of\n * centiseconds since 1601-01-01 00:00:00 UTC, plus 42. It is not useful\n * as a timestamp, but it is useful as a semi-unique ID.\n */\n unsigned int timestamp;\n \n /**\n * The default Language and Country ID (LCID) of the user who ran the\n * HTMLHelp Compiler. This is not the language of the CHM file itself.\n */\n unsigned int language;\n\n /**\n * The filename of the CHM helpfile. This is given by the library user\n * and may be in any format.\n */\n const char *filename;\n\n /** The length of the CHM helpfile, in bytes. */\n off_t length;\n\n /** A list of all non-system files in the CHM helpfile. */\n struct mschmd_file *files;\n\n /**\n * A list of all system files in the CHM helpfile.\n *\n * System files are files which begin with \"::\". They are meta-files\n * generated by the CHM creation process.\n */\n struct mschmd_file *sysfiles;\n\n /** The section 0 (uncompressed) data in this CHM helpfile. */\n struct mschmd_sec_uncompressed sec0;\n\n /** The section 1 (MSCompressed) data in this CHM helpfile. */\n struct mschmd_sec_mscompressed sec1;\n\n /** The file offset of the first PMGL/PMGI directory chunk. */\n off_t dir_offset;\n\n /** The number of PMGL/PMGI directory chunks in this CHM helpfile. */\n unsigned int num_chunks;\n\n /** The size of each PMGL/PMGI chunk, in bytes. */\n unsigned int chunk_size;\n\n /** The \"density\" of the quick-reference section in PMGL/PMGI chunks. */\n unsigned int density;\n\n /** The depth of the index tree.\n *\n * - if 1, there are no PMGI chunks, only PMGL chunks.\n * - if 2, there is 1 PMGI chunk. All chunk indices point to PMGL chunks.\n * - if 3, the root PMGI chunk points to secondary PMGI chunks, which in\n * turn point to PMGL chunks.\n * - and so on...\n */\n unsigned int depth;\n\n /**\n * The number of the root PMGI chunk.\n *\n * If there is no index in the CHM helpfile, this will be 0xFFFFFFFF.\n */\n unsigned int index_root;\n\n /**\n * The number of the first PMGL chunk. Usually zero.\n * Available only in CHM decoder version 2 and above.\n */\n unsigned int first_pmgl;\n\n /**\n * The number of the last PMGL chunk. Usually num_chunks-1.\n * Available only in CHM decoder version 2 and above.\n */\n unsigned int last_pmgl;\n\n /**\n * A cache of loaded chunks, filled in by mschm_decoder::fast_find().\n * Available only in CHM decoder version 2 and above.\n */\n unsigned char **chunk_cache;\n};\n\n/**\n * A structure which represents a file stored in a CHM helpfile.\n * \n * All fields are READ ONLY.\n */\nstruct mschmd_file {\n /**\n * A pointer to the next file in the list, or NULL if this is the final\n * file.\n */\n struct mschmd_file *next;\n\n /**\n * A pointer to the section that this file is located in. Indirectly,\n * it also points to the CHM helpfile the file is located in.\n */\n struct mschmd_section *section;\n\n /** The offset within the section data that this file is located at. */\n off_t offset;\n\n /** The length of this file, in bytes */\n off_t length;\n\n /** The filename of this file -- a null terminated string in UTF-8. */\n char *filename;\n};\n\n/** mschmc_file::section value: end of CHM file list */\n#define MSCHMC_ENDLIST (0)\n/** mschmc_file::section value: this file is in the Uncompressed section */\n#define MSCHMC_UNCOMP (1)\n/** mschmc_file::section value: this file is in the MSCompressed section */\n#define MSCHMC_MSCOMP (2)\n \n/** mschm_compressor::set_param() parameter: \"timestamp\" header */\n#define MSCHMC_PARAM_TIMESTAMP (0)\n/** mschm_compressor::set_param() parameter: \"language\" header */\n#define MSCHMC_PARAM_LANGUAGE (1)\n/** mschm_compressor::set_param() parameter: LZX window size */\n#define MSCHMC_PARAM_LZXWINDOW (2)\n/** mschm_compressor::set_param() parameter: intra-chunk quickref density */\n#define MSCHMC_PARAM_DENSITY (3)\n/** mschm_compressor::set_param() parameter: whether to create indices */\n#define MSCHMC_PARAM_INDEX (4)\n\n/**\n * A compressor for .CHM (Microsoft HTMLHelp) files.\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_chm_compressor(), mspack_destroy_chm_compressor()\n */\nstruct mschm_compressor {\n /**\n * Generates a CHM help file.\n *\n * The help file will contain up to two sections, an Uncompressed\n * section and potentially an MSCompressed (LZX compressed)\n * section.\n *\n * While the contents listing of a CHM file is always in lexical order,\n * the file list passed in will be taken as the correct order for files\n * within the sections. It is in your interest to place similar files\n * together for better compression.\n *\n * There are two modes of generation, to use a temporary file or not to\n * use one. See use_temporary_file() for the behaviour of generate() in\n * these two different modes.\n *\n * @param self a self-referential pointer to the mschm_compressor\n * instance being called\n * @param file_list an array of mschmc_file structures, terminated\n * with an entry whose mschmc_file::section field is\n * #MSCHMC_ENDLIST. The order of the list is\n * preserved within each section. The length of any\n * mschmc_file::chm_filename string cannot exceed\n * roughly 4096 bytes. Each source file must be able\n * to supply as many bytes as given in the\n * mschmc_file::length field.\n * @param output_file the file to write the generated CHM helpfile to.\n * This is passed directly to mspack_system::open()\n * @return an error code, or MSPACK_ERR_OK if successful\n * @see use_temporary_file() set_param()\n */\n int (*generate)(struct mschm_compressor *self,\n struct mschmc_file file_list[],\n const char *output_file);\n\n /**\n * Specifies whether a temporary file is used during CHM generation.\n *\n * The CHM file format includes data about the compressed section (such\n * as its overall size) that is stored in the output CHM file prior to\n * the compressed section itself. This unavoidably requires that the\n * compressed section has to be generated, before these details can be\n * set. There are several ways this can be handled. Firstly, the\n * compressed section could be generated entirely in memory before\n * writing any of the output CHM file. This approach is not used in\n * libmspack, as the compressed section can exceed the addressable\n * memory space on most architectures.\n *\n * libmspack has two options, either to write these unknowable sections\n * with blank data, generate the compressed section, then re-open the\n * output file for update once the compressed section has been\n * completed, or to write the compressed section to a temporary file,\n * then write the entire output file at once, performing a simple\n * file-to-file copy for the compressed section.\n *\n * The simple solution of buffering the entire compressed section in\n * memory can still be used, if desired. As the temporary file's\n * filename is passed directly to mspack_system::open(), it is possible\n * for a custom mspack_system implementation to hold this file in memory,\n * without writing to a disk.\n *\n * If a temporary file is set, generate() performs the following\n * sequence of events: the temporary file is opened for writing, the\n * compression algorithm writes to the temporary file, the temporary\n * file is closed. Then the output file is opened for writing and the\n * temporary file is re-opened for reading. The output file is written\n * and the temporary file is read from. Both files are then closed. The\n * temporary file itself is not deleted. If that is desired, the\n * temporary file should be deleted after the completion of generate(),\n * if it exists.\n *\n * If a temporary file is set not to be used, generate() performs the\n * following sequence of events: the output file is opened for writing,\n * then it is written and closed. The output file is then re-opened for\n * update, the appropriate sections are seek()ed to and re-written, then\n * the output file is closed.\n *\n * @param self a self-referential pointer to the\n * mschm_compressor instance being called\n * @param use_temp_file non-zero if the temporary file should be used,\n * zero if the temporary file should not be used.\n * @param temp_file a file to temporarily write compressed data to,\n * before opening it for reading and copying the\n * contents to the output file. This is passed\n * directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n * @see generate()\n */\n int (*use_temporary_file)(struct mschm_compressor *self,\n int use_temp_file,\n const char *temp_file);\n /**\n * Sets a CHM compression engine parameter.\n *\n * The following parameters are defined:\n\n * - #MSCHMC_PARAM_TIMESTAMP: Sets the \"timestamp\" of the CHM file\n * generated. This is not a timestamp, see mschmd_header::timestamp\n * for a description. If this timestamp is 0, generate() will use its\n * own algorithm for making a unique ID, based on the lengths and\n * names of files in the CHM itself. Defaults to 0, any value between\n * 0 and (2^32)-1 is valid.\n * - #MSCHMC_PARAM_LANGUAGE: Sets the \"language\" of the CHM file\n * generated. This is not the language used in the CHM file, but the\n * language setting of the user who ran the HTMLHelp compiler. It\n * defaults to 0x0409. The valid range is between 0x0000 and 0x7F7F.\n * - #MSCHMC_PARAM_LZXWINDOW: Sets the size of the LZX history window,\n * which is also the interval at which the compressed data stream can be\n * randomly accessed. The value is not a size in bytes, but a power of\n * two. The default value is 16 (which makes the window 2^16 bytes, or\n * 64 kilobytes), the valid range is from 15 (32 kilobytes) to 21 (2\n * megabytes).\n * - #MSCHMC_PARAM_DENSITY: Sets the \"density\" of quick reference\n * entries stored at the end of directory listing chunk. Each chunk is\n * 4096 bytes in size, and contains as many file entries as there is\n * room for. At the other end of the chunk, a list of \"quick reference\"\n * pointers is included. The offset of every 'N'th file entry is given a\n * quick reference, where N = (2^density) + 1. The default density is\n * 2. The smallest density is 0 (N=2), the maximum is 10 (N=1025). As\n * each file entry requires at least 5 bytes, the maximum number of\n * entries in a single chunk is roughly 800, so the maximum value 10\n * can be used to indicate there are no quickrefs at all.\n * - #MSCHMC_PARAM_INDEX: Sets whether or not to include quick lookup\n * index chunk(s), in addition to normal directory listing chunks. A\n * value of zero means no index chunks will be created, a non-zero value\n * means index chunks will be created. The default is zero, \"don't\n * create an index\".\n *\n * @param self a self-referential pointer to the mschm_compressor\n * instance being called\n * @param param the parameter to set\n * @param value the value to set the parameter to\n * @return MSPACK_ERR_OK if all is OK, or MSPACK_ERR_ARGS if there\n * is a problem with either parameter or value.\n * @see generate()\n */\n int (*set_param)(struct mschm_compressor *self,\n int param,\n unsigned int value);\n\n /**\n * Returns the error code set by the most recently called method.\n *\n * @param self a self-referential pointer to the mschm_compressor\n * instance being called\n * @return the most recent error code\n * @see set_param(), generate()\n */\n int (*last_error)(struct mschm_compressor *self);\n};\n\n/**\n * A decompressor for .CHM (Microsoft HTMLHelp) files\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_chm_decompressor(), mspack_destroy_chm_decompressor()\n */\nstruct mschm_decompressor {\n /**\n * Opens a CHM helpfile and reads its contents.\n *\n * If the file opened is a valid CHM helpfile, all headers will be read\n * and a mschmd_header structure will be returned, with a full list of\n * files.\n *\n * In the case of an error occuring, NULL is returned and the error code\n * is available from last_error().\n *\n * The filename pointer should be considered \"in use\" until close() is\n * called on the CHM helpfile.\n *\n * @param self a self-referential pointer to the mschm_decompressor\n * instance being called\n * @param filename the filename of the CHM helpfile. This is passed\n * directly to mspack_system::open().\n * @return a pointer to a mschmd_header structure, or NULL on failure\n * @see close()\n */\n struct mschmd_header *(*open)(struct mschm_decompressor *self,\n const char *filename);\n\n /**\n * Closes a previously opened CHM helpfile.\n *\n * This closes a CHM helpfile, frees the mschmd_header and all\n * mschmd_file structures associated with it (if any). This works on\n * both helpfiles opened with open() and helpfiles opened with\n * fast_open().\n *\n * The CHM header pointer is now invalid and cannot be used again. All\n * mschmd_file pointers referencing that CHM are also now invalid, and\n * cannot be used again.\n *\n * @param self a self-referential pointer to the mschm_decompressor\n * instance being called\n * @param chm the CHM helpfile to close\n * @see open(), fast_open()\n */\n void (*close)(struct mschm_decompressor *self,\n struct mschmd_header *chm);\n\n /**\n * Extracts a file from a CHM helpfile.\n *\n * This extracts a file from a CHM helpfile and writes it to the given\n * filename. The filename of the file, mscabd_file::filename, is not\n * used by extract(), but can be used by the caller as a guide for\n * constructing an appropriate filename.\n *\n * This method works both with files found in the mschmd_header::files\n * and mschmd_header::sysfiles list and mschmd_file structures generated\n * on the fly by fast_find().\n *\n * @param self a self-referential pointer to the mschm_decompressor\n * instance being called\n * @param file the file to be decompressed\n * @param filename the filename of the file being written to\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*extract)(struct mschm_decompressor *self,\n struct mschmd_file *file,\n const char *filename);\n\n /**\n * Returns the error code set by the most recently called method.\n *\n * This is useful for open() and fast_open(), which do not return an\n * error code directly.\n *\n * @param self a self-referential pointer to the mschm_decompressor\n * instance being called\n * @return the most recent error code\n * @see open(), extract()\n */\n int (*last_error)(struct mschm_decompressor *self);\n\n /**\n * Opens a CHM helpfile quickly.\n *\n * If the file opened is a valid CHM helpfile, only essential headers\n * will be read. A mschmd_header structure will be still be returned, as\n * with open(), but the mschmd_header::files field will be NULL. No\n * files details will be automatically read. The fast_find() method\n * must be used to obtain file details.\n *\n * In the case of an error occuring, NULL is returned and the error code\n * is available from last_error().\n *\n * The filename pointer should be considered \"in use\" until close() is\n * called on the CHM helpfile.\n *\n * @param self a self-referential pointer to the mschm_decompressor\n * instance being called\n * @param filename the filename of the CHM helpfile. This is passed\n * directly to mspack_system::open().\n * @return a pointer to a mschmd_header structure, or NULL on failure\n * @see open(), close(), fast_find(), extract()\n */\n struct mschmd_header *(*fast_open)(struct mschm_decompressor *self,\n const char *filename);\n\n /**\n * Finds file details quickly.\n *\n * Instead of reading all CHM helpfile headers and building a list of\n * files, fast_open() and fast_find() are intended for finding file\n * details only when they are needed. The CHM file format includes an\n * on-disk file index to allow this.\n *\n * Given a case-sensitive filename, fast_find() will search the on-disk\n * index for that file.\n *\n * If the file was found, the caller-provided mschmd_file structure will\n * be filled out like so:\n * - section: the correct value for the found file\n * - offset: the correct value for the found file\n * - length: the correct value for the found file\n * - all other structure elements: NULL or 0\n *\n * If the file was not found, MSPACK_ERR_OK will still be returned as the\n * result, but the caller-provided structure will be filled out like so:\n * - section: NULL\n * - offset: 0\n * - length: 0\n * - all other structure elements: NULL or 0\n *\n * This method is intended to be used in conjunction with CHM helpfiles\n * opened with fast_open(), but it also works with helpfiles opened\n * using the regular open().\n *\n * @param self a self-referential pointer to the mschm_decompressor\n * instance being called\n * @param chm the CHM helpfile to search for the file\n * @param filename the filename of the file to search for\n * @param f_ptr a pointer to a caller-provded mschmd_file structure\n * @param f_size sizeof(struct mschmd_file)\n * @return an error code, or MSPACK_ERR_OK if successful\n * @see open(), close(), fast_find(), extract()\n */\n int (*fast_find)(struct mschm_decompressor *self,\n struct mschmd_header *chm,\n const char *filename,\n struct mschmd_file *f_ptr,\n int f_size);\n};\n\n/* --- support for .LIT (EBook) file format -------------------------------- */\n\n/** TODO */\nstruct mslit_compressor {\n int dummy; \n};\n\n/** TODO */\nstruct mslit_decompressor {\n int dummy; \n};\n\n\n/* --- support for .HLP (MS Help) file format ------------------------------ */\n\n/** TODO */\nstruct mshlp_compressor {\n int dummy; \n};\n\n/** TODO */\nstruct mshlp_decompressor {\n int dummy; \n};\n\n\n/* --- support for SZDD file format ---------------------------------------- */\n\n/** msszdd_compressor::set_param() parameter: the missing character */\n#define MSSZDDC_PARAM_MISSINGCHAR (0)\n\n/** msszddd_header::format value - a regular SZDD file */\n#define MSSZDD_FMT_NORMAL (0)\n\n/** msszddd_header::format value - a special QBasic SZDD file */\n#define MSSZDD_FMT_QBASIC (1)\n\n/**\n * A structure which represents an SZDD compressed file.\n *\n * All fields are READ ONLY.\n */\nstruct msszddd_header {\n /** The file format; either #MSSZDD_FMT_NORMAL or #MSSZDD_FMT_QBASIC */\n int format;\n\n /** The amount of data in the SZDD file once uncompressed. */\n off_t length;\n\n /**\n * The last character in the filename, traditionally replaced with an\n * underscore to show the file is compressed. The null character is used\n * to show that this character has not been stored (e.g. because the\n * filename is not known). Generally, only characters that may appear in\n * an MS-DOS filename (except \".\") are valid.\n */\n char missing_char;\n};\n\n/**\n * A compressor for the SZDD file format.\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_szdd_compressor(), mspack_destroy_szdd_compressor()\n */\nstruct msszdd_compressor {\n /**\n * Reads an input file and creates a compressed output file in the\n * SZDD compressed file format. The SZDD compression format is quick\n * but gives poor compression. It is possible for the compressed output\n * file to be larger than the input file.\n *\n * Conventionally, SZDD compressed files have the final character in\n * their filename replaced with an underscore, to show they are\n * compressed. The missing character is stored in the compressed file\n * itself. This is due to the restricted filename conventions of MS-DOS,\n * most operating systems, such as UNIX, simply append another file\n * extension to the existing filename. As mspack does not deal with\n * filenames, this is left up to you. If you wish to set the missing\n * character stored in the file header, use set_param() with the\n * #MSSZDDC_PARAM_MISSINGCHAR parameter.\n *\n * \"Stream\" compression (where the length of the input data is not\n * known) is not possible. The length of the input data is stored in the\n * header of the SZDD file and must therefore be known before any data\n * is compressed. Due to technical limitations of the file format, the\n * maximum size of uncompressed file that will be accepted is 2147483647\n * bytes.\n *\n * @param self a self-referential pointer to the msszdd_compressor\n * instance being called\n * @param input the name of the file to compressed. This is passed\n * passed directly to mspack_system::open()\n * @param output the name of the file to write compressed data to.\n * This is passed directly to mspack_system::open().\n * @param length the length of the uncompressed file, or -1 to indicate\n * that this should be determined automatically by using\n * mspack_system::seek() on the input file.\n * @return an error code, or MSPACK_ERR_OK if successful\n * @see set_param()\n */\n int (*compress)(struct msszdd_compressor *self,\n const char *input,\n const char *output,\n off_t length);\n\n /**\n * Sets an SZDD compression engine parameter.\n *\n * The following parameters are defined:\n\n * - #MSSZDDC_PARAM_CHARACTER: the \"missing character\", the last character\n * in the uncompressed file's filename, which is traditionally replaced\n * with an underscore to show the file is compressed. Traditionally,\n * this can only be a character that is a valid part of an MS-DOS,\n * filename, but libmspack permits any character between 0x00 and 0xFF\n * to be stored. 0x00 is the default, and it represents \"no character\n * stored\".\n *\n * @param self a self-referential pointer to the msszdd_compressor\n * instance being called\n * @param param the parameter to set\n * @param value the value to set the parameter to\n * @return MSPACK_ERR_OK if all is OK, or MSPACK_ERR_ARGS if there\n * is a problem with either parameter or value.\n * @see compress()\n */\n int (*set_param)(struct msszdd_compressor *self,\n int param,\n unsigned int value);\n\n /**\n * Returns the error code set by the most recently called method.\n *\n * @param self a self-referential pointer to the msszdd_compressor\n * instance being called\n * @return the most recent error code\n * @see compress()\n */\n int (*last_error)(struct mschm_decompressor *self);\n};\n\n/**\n * A decompressor for SZDD compressed files.\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_szdd_decompressor(), mspack_destroy_szdd_decompressor()\n */\nstruct msszdd_decompressor {\n /**\n * Opens a SZDD file and reads the header.\n *\n * If the file opened is a valid SZDD file, all headers will be read and\n * a msszddd_header structure will be returned.\n *\n * In the case of an error occuring, NULL is returned and the error code\n * is available from last_error().\n *\n * The filename pointer should be considered \"in use\" until close() is\n * called on the SZDD file.\n *\n * @param self a self-referential pointer to the msszdd_decompressor\n * instance being called\n * @param filename the filename of the SZDD compressed file. This is\n * passed directly to mspack_system::open().\n * @return a pointer to a msszddd_header structure, or NULL on failure\n * @see close()\n */\n struct msszddd_header *(*open)(struct msszdd_decompressor *self,\n const char *filename);\n\n /**\n * Closes a previously opened SZDD file.\n *\n * This closes a SZDD file and frees the msszddd_header associated with\n * it.\n *\n * The SZDD header pointer is now invalid and cannot be used again.\n *\n * @param self a self-referential pointer to the msszdd_decompressor\n * instance being called\n * @param szdd the SZDD file to close\n * @see open()\n */\n void (*close)(struct msszdd_decompressor *self,\n struct msszddd_header *szdd);\n\n /**\n * Extracts the compressed data from a SZDD file.\n *\n * This decompresses the compressed SZDD data stream and writes it to\n * an output file.\n *\n * @param self a self-referential pointer to the msszdd_decompressor\n * instance being called\n * @param szdd the SZDD file to extract data from\n * @param filename the filename to write the decompressed data to. This\n * is passed directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*extract)(struct msszdd_decompressor *self,\n struct msszddd_header *szdd,\n const char *filename);\n\n /**\n * Decompresses an SZDD file to an output file in one step.\n *\n * This opens an SZDD file as input, reads the header, then decompresses\n * the compressed data immediately to an output file, finally closing\n * both the input and output file. It is more convenient to use than\n * open() then extract() then close(), if you do not need to know the\n * SZDD output size or missing character.\n *\n * @param self a self-referential pointer to the msszdd_decompressor\n * instance being called\n * @param input the filename of the input SZDD file. This is passed\n * directly to mspack_system::open().\n * @param output the filename to write the decompressed data to. This\n * is passed directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*decompress)(struct msszdd_decompressor *self,\n const char *input,\n const char *output);\n\n /**\n * Returns the error code set by the most recently called method.\n *\n * This is useful for open() which does not return an\n * error code directly.\n *\n * @param self a self-referential pointer to the msszdd_decompressor\n * instance being called\n * @return the most recent error code\n * @see open(), extract(), decompress()\n */\n int (*last_error)(struct msszdd_decompressor *self);\n};\n\n/* --- support for KWAJ file format ---------------------------------------- */\n\n/** mskwaj_compressor::set_param() parameter: compression type */\n#define MSKWAJC_PARAM_COMP_TYPE (0)\n\n/** mskwaj_compressor::set_param() parameter: include the length of the\n * uncompressed file in the header?\n */\n#define MSKWAJC_PARAM_INCLUDE_LENGTH (1)\n\n/** KWAJ compression type: no compression. */\n#define MSKWAJ_COMP_NONE (0)\n/** KWAJ compression type: no compression, 0xFF XOR \"encryption\". */\n#define MSKWAJ_COMP_XOR (1)\n/** KWAJ compression type: LZSS (same method as SZDD) */\n#define MSKWAJ_COMP_SZDD (2)\n/** KWAJ compression type: LZ+Huffman compression */\n#define MSKWAJ_COMP_LZH (3)\n/** KWAJ compression type: MSZIP */\n#define MSKWAJ_COMP_MSZIP (4)\n\n/** KWAJ optional header flag: decompressed file length is included */\n#define MSKWAJ_HDR_HASLENGTH (0x01)\n\n/** KWAJ optional header flag: unknown 2-byte structure is included */\n#define MSKWAJ_HDR_HASUNKNOWN1 (0x02)\n\n/** KWAJ optional header flag: unknown multi-sized structure is included */\n#define MSKWAJ_HDR_HASUNKNOWN2 (0x04)\n\n/** KWAJ optional header flag: file name (no extension) is included */\n#define MSKWAJ_HDR_HASFILENAME (0x08)\n\n/** KWAJ optional header flag: file extension is included */\n#define MSKWAJ_HDR_HASFILEEXT (0x10)\n\n/** KWAJ optional header flag: extra text is included */\n#define MSKWAJ_HDR_HASEXTRATEXT (0x20)\n\n/**\n * A structure which represents an KWAJ compressed file.\n *\n * All fields are READ ONLY.\n */\nstruct mskwajd_header {\n /** The compression type; should be one of #MSKWAJ_COMP_NONE,\n * #MSKWAJ_COMP_XOR, #MSKWAJ_COMP_SZDD or #MSKWAJ_COMP_LZH\n */\n unsigned short comp_type;\n\n /** The offset in the file where the compressed data stream begins */\n off_t data_offset;\n\n /** Flags indicating which optional headers were included. */\n int headers;\n\n /** The amount of uncompressed data in the file, or 0 if not present. */\n off_t length;\n\n /** output filename, or NULL if not present */\n char *filename;\n\n /** extra uncompressed data (usually text) in the header.\n * This data can contain nulls so use extra_length to get the size.\n */\n char *extra;\n\n /** length of extra uncompressed data in the header */\n unsigned short extra_length;\n};\n\n/**\n * A compressor for the KWAJ file format.\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_kwaj_compressor(), mspack_destroy_kwaj_compressor()\n */\nstruct mskwaj_compressor {\n /**\n * Reads an input file and creates a compressed output file in the\n * KWAJ compressed file format. The KWAJ compression format is quick\n * but gives poor compression. It is possible for the compressed output\n * file to be larger than the input file.\n *\n * @param self a self-referential pointer to the mskwaj_compressor\n * instance being called\n * @param input the name of the file to compressed. This is passed\n * passed directly to mspack_system::open()\n * @param output the name of the file to write compressed data to.\n * This is passed directly to mspack_system::open().\n * @param length the length of the uncompressed file, or -1 to indicate\n * that this should be determined automatically by using\n * mspack_system::seek() on the input file.\n * @return an error code, or MSPACK_ERR_OK if successful\n * @see set_param()\n */\n int (*compress)(struct mskwaj_compressor *self,\n const char *input,\n const char *output,\n off_t length);\n\n /**\n * Sets an KWAJ compression engine parameter.\n *\n * The following parameters are defined:\n *\n * - #MSKWAJC_PARAM_COMP_TYPE: the compression method to use. Must\n * be one of #MSKWAJC_COMP_NONE, #MSKWAJC_COMP_XOR, #MSKWAJ_COMP_SZDD\n * or #MSKWAJ_COMP_LZH. The default is #MSKWAJ_COMP_LZH.\n *\n * - #MSKWAJC_PARAM_INCLUDE_LENGTH: a boolean; should the compressed\n * output file should include the uncompressed length of the input\n * file in the header? This adds 4 bytes to the size of the output\n * file. A value of zero says \"no\", non-zero says \"yes\". The default\n * is \"no\".\n *\n * @param self a self-referential pointer to the mskwaj_compressor\n * instance being called\n * @param param the parameter to set\n * @param value the value to set the parameter to\n * @return MSPACK_ERR_OK if all is OK, or MSPACK_ERR_ARGS if there\n * is a problem with either parameter or value.\n * @see generate()\n */\n int (*set_param)(struct mskwaj_compressor *self,\n int param,\n unsigned int value);\n\n\n /**\n * Sets the original filename of the file before compression,\n * which will be stored in the header of the output file.\n *\n * The filename should be a null-terminated string, it must be an\n * MS-DOS \"8.3\" type filename (up to 8 bytes for the filename, then\n * optionally a \".\" and up to 3 bytes for a filename extension).\n *\n * If NULL is passed as the filename, no filename is included in the\n * header. This is the default.\n *\n * @param self a self-referential pointer to the mskwaj_compressor\n * instance being called\n * @param filename the original filename to use\n * @return MSPACK_ERR_OK if all is OK, or MSPACK_ERR_ARGS if the\n * filename is too long\n */\n int (*set_filename)(struct mskwaj_compressor *self,\n const char *filename);\n\n /**\n * Sets arbitrary data that will be stored in the header of the\n * output file, uncompressed. It can be up to roughly 64 kilobytes,\n * as the overall size of the header must not exceed 65535 bytes.\n * The data can contain null bytes if desired.\n *\n * If NULL is passed as the data pointer, or zero is passed as the\n * length, no extra data is included in the header. This is the\n * default.\n *\n * @param self a self-referential pointer to the mskwaj_compressor\n * instance being called\n * @param data a pointer to the data to be stored in the header\n * @param bytes the length of the data in bytes\n * @return MSPACK_ERR_OK if all is OK, or MSPACK_ERR_ARGS extra data\n * is too long\n */\n int (*set_extra_data)(struct mskwaj_compressor *self,\n void *data,\n size_t bytes);\n\n /**\n * Returns the error code set by the most recently called method.\n *\n * @param self a self-referential pointer to the mskwaj_compressor\n * instance being called\n * @return the most recent error code\n * @see compress()\n */\n int (*last_error)(struct mschm_decompressor *self);\n};\n\n/**\n * A decompressor for KWAJ compressed files.\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_kwaj_decompressor(), mspack_destroy_kwaj_decompressor()\n */\nstruct mskwaj_decompressor {\n /**\n * Opens a KWAJ file and reads the header.\n *\n * If the file opened is a valid KWAJ file, all headers will be read and\n * a mskwajd_header structure will be returned.\n *\n * In the case of an error occuring, NULL is returned and the error code\n * is available from last_error().\n *\n * The filename pointer should be considered \"in use\" until close() is\n * called on the KWAJ file.\n *\n * @param self a self-referential pointer to the mskwaj_decompressor\n * instance being called\n * @param filename the filename of the KWAJ compressed file. This is\n * passed directly to mspack_system::open().\n * @return a pointer to a mskwajd_header structure, or NULL on failure\n * @see close()\n */\n struct mskwajd_header *(*open)(struct mskwaj_decompressor *self,\n const char *filename);\n\n /**\n * Closes a previously opened KWAJ file.\n *\n * This closes a KWAJ file and frees the mskwajd_header associated\n * with it. The KWAJ header pointer is now invalid and cannot be\n * used again.\n *\n * @param self a self-referential pointer to the mskwaj_decompressor\n * instance being called\n * @param kwaj the KWAJ file to close\n * @see open()\n */\n void (*close)(struct mskwaj_decompressor *self,\n struct mskwajd_header *kwaj);\n\n /**\n * Extracts the compressed data from a KWAJ file.\n *\n * This decompresses the compressed KWAJ data stream and writes it to\n * an output file.\n *\n * @param self a self-referential pointer to the mskwaj_decompressor\n * instance being called\n * @param kwaj the KWAJ file to extract data from\n * @param filename the filename to write the decompressed data to. This\n * is passed directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*extract)(struct mskwaj_decompressor *self,\n struct mskwajd_header *kwaj,\n const char *filename);\n\n /**\n * Decompresses an KWAJ file to an output file in one step.\n *\n * This opens an KWAJ file as input, reads the header, then decompresses\n * the compressed data immediately to an output file, finally closing\n * both the input and output file. It is more convenient to use than\n * open() then extract() then close(), if you do not need to know the\n * KWAJ output size or output filename.\n *\n * @param self a self-referential pointer to the mskwaj_decompressor\n * instance being called\n * @param input the filename of the input KWAJ file. This is passed\n * directly to mspack_system::open().\n * @param output the filename to write the decompressed data to. This\n * is passed directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*decompress)(struct mskwaj_decompressor *self,\n const char *input,\n const char *output);\n\n /**\n * Returns the error code set by the most recently called method.\n *\n * This is useful for open() which does not return an\n * error code directly.\n *\n * @param self a self-referential pointer to the mskwaj_decompressor\n * instance being called\n * @return the most recent error code\n * @see open(), search()\n */\n int (*last_error)(struct mskwaj_decompressor *self);\n};\n\n/* --- support for .LZX (Offline Address Book) file format ----------------- */\n\n/**\n * A compressor for the Offline Address Book (OAB) format.\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_oab_compressor(), mspack_destroy_oab_compressor()\n */\nstruct msoab_compressor {\n /**\n * Compress a full OAB file.\n *\n * The input file will be read and the compressed contents written to the\n * output file.\n *\n * @param self a self-referential pointer to the msoab_decompressor\n * instance being called\n * @param input the filename of the input file. This is passed\n * directly to mspack_system::open().\n * @param output the filename of the output file. This is passed\n * directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*compress) (struct msoab_compressor *self,\n const char *input,\n const char *output);\n\n /**\n * Generate a compressed incremental OAB patch file.\n *\n * The two uncompressed files \"input\" and \"base\" will be read, and an\n * incremental patch to generate \"input\" from \"base\" will be written to\n * the output file.\n *\n * @param self a self-referential pointer to the msoab_compressor\n * instance being called\n * @param input the filename of the input file containing the new\n * version of its contents. This is passed directly\n * to mspack_system::open().\n * @param base the filename of the original base file containing\n * the old version of its contents, against which the\n * incremental patch shall generated. This is passed\n * directly to mspack_system::open().\n * @param output the filename of the output file. This is passed\n * directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*compress_incremental) (struct msoab_compressor *self,\n const char *input,\n const char *base,\n const char *output);\n};\n\n/**\n * A decompressor for .LZX (Offline Address Book) files\n *\n * All fields are READ ONLY.\n *\n * @see mspack_create_oab_decompressor(), mspack_destroy_oab_decompressor()\n */\nstruct msoab_decompressor {\n /**\n * Decompresses a full Offline Address Book file.\n *\n * If the input file is a valid compressed Offline Address Book file, \n * it will be read and the decompressed contents will be written to\n * the output file.\n *\n * @param self a self-referential pointer to the msoab_decompressor\n * instance being called\n * @param input the filename of the input file. This is passed\n * directly to mspack_system::open().\n * @param output the filename of the output file. This is passed\n * directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*decompress) (struct msoab_decompressor *self,\n const char *input,\n const char *output);\n\n /**\n * Decompresses an Offline Address Book with an incremental patch file.\n *\n * This requires both a full UNCOMPRESSED Offline Address Book file to\n * act as the \"base\", and a compressed incremental patch file as input.\n * If the input file is valid, it will be decompressed with reference to\n * the base file, and the decompressed contents will be written to the\n * output file.\n *\n * There is no way to tell what the right base file is for the given\n * incremental patch, but if you get it wrong, this will usually result\n * in incorrect data being decompressed, which will then fail a checksum\n * test.\n *\n * @param self a self-referential pointer to the msoab_decompressor\n * instance being called\n * @param input the filename of the input file. This is passed\n * directly to mspack_system::open().\n * @param base the filename of the base file to which the\n * incremental patch shall be applied. This is passed\n * directly to mspack_system::open().\n * @param output the filename of the output file. This is passed\n * directly to mspack_system::open().\n * @return an error code, or MSPACK_ERR_OK if successful\n */\n int (*decompress_incremental) (struct msoab_decompressor *self,\n const char *input,\n const char *base,\n const char *output);\n};\n\n#ifdef __cplusplus\n}\n#endif\n\n#endif\n" }, { "path": "third_party/mspack/readbits.h", "content": "/* This file is part of libmspack.\n * (C) 2003-2010 Stuart Caie.\n *\n * libmspack is free software; you can redistribute it and/or modify it under\n * the terms of the GNU Lesser General Public License (LGPL) version 2.1\n *\n * For further details, see the file COPYING.LIB distributed with libmspack\n */\n\n#ifndef MSPACK_READBITS_H\n#define MSPACK_READBITS_H 1\n\n/* this header defines macros that read data streams by\n * the individual bits\n *\n * INIT_BITS initialises bitstream state in state structure\n * STORE_BITS stores bitstream state in state structure\n * RESTORE_BITS restores bitstream state from state structure\n * ENSURE_BITS(n) ensure there are at least N bits in the bit buffer\n * READ_BITS(var,n) takes N bits from the buffer and puts them in var\n * PEEK_BITS(n) extracts without removing N bits from the bit buffer\n * REMOVE_BITS(n) removes N bits from the bit buffer\n *\n * READ_BITS simply calls ENSURE_BITS, PEEK_BITS and REMOVE_BITS,\n * which means it's limited to reading the number of bits you can\n * ensure at any one time. It also fails if asked to read zero bits.\n * If you need to read zero bits, or more bits than can be ensured in\n * one go, use READ_MANY_BITS instead.\n *\n * These macros have variable names baked into them, so to use them\n * you have to define some macros:\n * - BITS_TYPE: the type name of your state structure\n * - BITS_VAR: the variable that points to your state structure\n * - define BITS_ORDER_MSB if bits are read from the MSB, or\n * define BITS_ORDER_LSB if bits are read from the LSB\n * - READ_BYTES: some code that reads more data into the bit buffer,\n * it should use READ_IF_NEEDED (calls read_input if the byte buffer\n * is empty), then INJECT_BITS(data,n) to put data from the byte\n * buffer into the bit buffer.\n *\n * You also need to define some variables and structure members:\n * - unsigned char *i_ptr; // current position in the byte buffer\n * - unsigned char *i_end; // end of the byte buffer\n * - unsigned int bit_buffer; // the bit buffer itself\n * - unsigned int bits_left; // number of bits remaining\n *\n * If you use read_input() and READ_IF_NEEDED, they also expect these\n * structure members:\n * - struct mspack_system *sys; // to access sys->read()\n * - unsigned int error; // to record/return read errors\n * - unsigned char input_end; // to mark reaching the EOF\n * - unsigned char *inbuf; // the input byte buffer\n * - unsigned int inbuf_size; // the size of the input byte buffer\n *\n * Your READ_BYTES implementation should read data from *i_ptr and\n * put them in the bit buffer. READ_IF_NEEDED will call read_input()\n * if i_ptr reaches i_end, and will fill up inbuf and set i_ptr to\n * the start of inbuf and i_end to the end of inbuf.\n *\n * If you're reading in MSB order, the routines work by using the area\n * beyond the MSB and the LSB of the bit buffer as a free source of\n * zeroes when shifting. This avoids having to mask any bits. So we\n * have to know the bit width of the bit buffer variable. We use\n * and CHAR_BIT to find the size of the bit buffer in bits.\n *\n * If you are reading in LSB order, bits need to be masked. Normally\n * this is done by computing the mask: N bits are masked by the value\n * (1<\n#endif\n#ifndef CHAR_BIT\n# define CHAR_BIT (8)\n#endif\n#define BITBUF_WIDTH (sizeof(bit_buffer) * CHAR_BIT)\n\n#define INIT_BITS do { \\\n BITS_VAR->i_ptr = &BITS_VAR->inbuf[0]; \\\n BITS_VAR->i_end = &BITS_VAR->inbuf[0]; \\\n BITS_VAR->bit_buffer = 0; \\\n BITS_VAR->bits_left = 0; \\\n BITS_VAR->input_end = 0; \\\n} while (0)\n\n#define STORE_BITS do { \\\n BITS_VAR->i_ptr = i_ptr; \\\n BITS_VAR->i_end = i_end; \\\n BITS_VAR->bit_buffer = bit_buffer; \\\n BITS_VAR->bits_left = bits_left; \\\n} while (0)\n\n#define RESTORE_BITS do { \\\n i_ptr = BITS_VAR->i_ptr; \\\n i_end = BITS_VAR->i_end; \\\n bit_buffer = BITS_VAR->bit_buffer; \\\n bits_left = BITS_VAR->bits_left; \\\n} while (0)\n\n#define ENSURE_BITS(nbits) do { \\\n while (bits_left < (nbits)) READ_BYTES; \\\n} while (0)\n\n#define READ_BITS(val, nbits) do { \\\n ENSURE_BITS(nbits); \\\n (val) = PEEK_BITS(nbits); \\\n REMOVE_BITS(nbits); \\\n} while (0)\n\n#define READ_MANY_BITS(val, bits) do { \\\n unsigned char needed = (bits), bitrun; \\\n (val) = 0; \\\n while (needed > 0) { \\\n if (bits_left <= (BITBUF_WIDTH - 16)) READ_BYTES; \\\n bitrun = (bits_left < needed) ? bits_left : needed; \\\n (val) = ((val) << bitrun) | PEEK_BITS(bitrun); \\\n REMOVE_BITS(bitrun); \\\n needed -= bitrun; \\\n } \\\n} while (0)\n\n#ifdef BITS_ORDER_MSB\n# define PEEK_BITS(nbits) (bit_buffer >> (BITBUF_WIDTH - (nbits)))\n# define REMOVE_BITS(nbits) ((bit_buffer <<= (nbits)), (bits_left -= (nbits)))\n# define INJECT_BITS(bitdata,nbits) ((bit_buffer |= \\\n (bitdata) << (BITBUF_WIDTH - (nbits) - bits_left)), (bits_left += (nbits)))\n#else /* BITS_ORDER_LSB */\n# define PEEK_BITS(nbits) (bit_buffer & ((1 << (nbits))-1))\n# define REMOVE_BITS(nbits) ((bit_buffer >>= (nbits)), (bits_left -= (nbits)))\n# define INJECT_BITS(bitdata,nbits) ((bit_buffer |= \\\n (bitdata) << bits_left), (bits_left += (nbits)))\n#endif\n\n#ifdef BITS_LSB_TABLE\n/* lsb_bit_mask[n] = (1 << n) - 1 */\nstatic const unsigned short lsb_bit_mask[17] = {\n 0x0000, 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,\n 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff\n};\n# define PEEK_BITS_T(nbits) (bit_buffer & lsb_bit_mask[(nbits)])\n# define READ_BITS_T(val, nbits) do { \\\n ENSURE_BITS(nbits); \\\n (val) = PEEK_BITS_T(nbits); \\\n REMOVE_BITS(nbits); \\\n} while (0)\n#endif\n\n#ifndef BITS_NO_READ_INPUT\n# define READ_IF_NEEDED do { \\\n if (i_ptr >= i_end) { \\\n if (read_input(BITS_VAR)) \\\n return BITS_VAR->error; \\\n i_ptr = BITS_VAR->i_ptr; \\\n i_end = BITS_VAR->i_end; \\\n } \\\n} while (0)\n\nstatic int read_input(BITS_TYPE *p) {\n int read = p->sys->read(p->input, &p->inbuf[0], (int)p->inbuf_size);\n if (read < 0) return p->error = MSPACK_ERR_READ;\n\n /* we might overrun the input stream by asking for bits we don't use,\n * so fake 2 more bytes at the end of input */\n if (read == 0) {\n if (p->input_end) {\n D((\"out of input bytes\"))\n return p->error = MSPACK_ERR_READ;\n }\n else {\n read = 2;\n p->inbuf[0] = p->inbuf[1] = 0;\n p->input_end = 1;\n }\n }\n\n /* update i_ptr and i_end */\n p->i_ptr = &p->inbuf[0];\n p->i_end = &p->inbuf[read];\n return MSPACK_ERR_OK;\n}\n#endif\n#endif\n" }, { "path": "third_party/mspack/readhuff.h", "content": "/* This file is part of libmspack.\n * (C) 2003-2014 Stuart Caie.\n *\n * libmspack is free software; you can redistribute it and/or modify it under\n * the terms of the GNU Lesser General Public License (LGPL) version 2.1\n *\n * For further details, see the file COPYING.LIB distributed with libmspack\n */\n\n#ifndef MSPACK_READHUFF_H\n#define MSPACK_READHUFF_H 1\n\n/* This implements a fast Huffman tree decoding system. */\n\n#if !(defined(BITS_ORDER_MSB) || defined(BITS_ORDER_LSB))\n# error \"readhuff.h is used in conjunction with readbits.h, include that first\"\n#endif\n#if !(defined(TABLEBITS) && defined(MAXSYMBOLS))\n# error \"define TABLEBITS(tbl) and MAXSYMBOLS(tbl) before using readhuff.h\"\n#endif\n#if !(defined(HUFF_TABLE) && defined(HUFF_LEN))\n# error \"define HUFF_TABLE(tbl) and HUFF_LEN(tbl) before using readhuff.h\"\n#endif\n#ifndef HUFF_ERROR\n# error \"define HUFF_ERROR before using readhuff.h\"\n#endif\n#ifndef HUFF_MAXBITS\n# define HUFF_MAXBITS 16\n#endif\n\n/* Decodes the next huffman symbol from the input bitstream into var.\n * Do not use this macro on a table unless build_decode_table() succeeded.\n */\n#define READ_HUFFSYM(tbl, var) do { \\\n ENSURE_BITS(HUFF_MAXBITS); \\\n sym = HUFF_TABLE(tbl, PEEK_BITS(TABLEBITS(tbl))); \\\n if (sym >= MAXSYMBOLS(tbl)) HUFF_TRAVERSE(tbl); \\\n (var) = sym; \\\n i = HUFF_LEN(tbl, sym); \\\n REMOVE_BITS(i); \\\n} while (0)\n\n#ifdef BITS_ORDER_LSB\n# define HUFF_TRAVERSE(tbl) do { \\\n i = TABLEBITS(tbl) - 1; \\\n do { \\\n if (i++ > HUFF_MAXBITS) HUFF_ERROR; \\\n sym = HUFF_TABLE(tbl, \\\n (sym << 1) | ((bit_buffer >> i) & 1)); \\\n } while (sym >= MAXSYMBOLS(tbl)); \\\n} while (0)\n#else\n#define HUFF_TRAVERSE(tbl) do { \\\n i = 1 << (BITBUF_WIDTH - TABLEBITS(tbl)); \\\n do { \\\n if ((i >>= 1) == 0) HUFF_ERROR; \\\n sym = HUFF_TABLE(tbl, \\\n (sym << 1) | ((bit_buffer & i) ? 1 : 0)); \\\n } while (sym >= MAXSYMBOLS(tbl)); \\\n} while (0)\n#endif\n\n/* make_decode_table(nsyms, nbits, length[], table[])\n *\n * This function was originally coded by David Tritscher.\n * It builds a fast huffman decoding table from\n * a canonical huffman code lengths table.\n *\n * nsyms = total number of symbols in this huffman tree.\n * nbits = any symbols with a code length of nbits or less can be decoded\n * in one lookup of the table.\n * length = A table to get code lengths from [0 to nsyms-1]\n * table = The table to fill up with decoded symbols and pointers.\n * Should be ((1<> 1; /* don't do 0 length codes */\n\n /* fill entries for codes short enough for a direct mapping */\n for (bit_num = 1; bit_num <= nbits; bit_num++) {\n for (sym = 0; sym < nsyms; sym++) {\n if (length[sym] != bit_num) continue;\n#ifdef BITS_ORDER_MSB\n leaf = pos;\n#else\n /* reverse the significant bits */\n fill = length[sym]; reverse = pos >> (nbits - fill); leaf = 0;\n do {leaf <<= 1; leaf |= reverse & 1; reverse >>= 1;} while (--fill);\n#endif\n\n if((pos += bit_mask) > table_mask) return 1; /* table overrun */\n\n /* fill all possible lookups of this symbol with the symbol itself */\n#ifdef BITS_ORDER_MSB\n for (fill = bit_mask; fill-- > 0;) table[leaf++] = sym;\n#else\n fill = bit_mask; next_symbol = 1 << bit_num;\n do { table[leaf] = sym; leaf += next_symbol; } while (--fill);\n#endif\n }\n bit_mask >>= 1;\n }\n\n /* exit with success if table is now complete */\n if (pos == table_mask) return 0;\n\n /* mark all remaining table entries as unused */\n for (sym = pos; sym < table_mask; sym++) {\n#ifdef BITS_ORDER_MSB\n table[sym] = 0xFFFF;\n#else\n reverse = sym; leaf = 0; fill = nbits;\n do { leaf <<= 1; leaf |= reverse & 1; reverse >>= 1; } while (--fill);\n table[leaf] = 0xFFFF;\n#endif\n }\n\n /* next_symbol = base of allocation for long codes */\n next_symbol = ((table_mask >> 1) < nsyms) ? nsyms : (table_mask >> 1);\n\n /* give ourselves room for codes to grow by up to 16 more bits.\n * codes now start at bit nbits+16 and end at (nbits+16-codelength) */\n pos <<= 16;\n table_mask <<= 16;\n bit_mask = 1 << 15;\n\n for (bit_num = nbits+1; bit_num <= HUFF_MAXBITS; bit_num++) {\n for (sym = 0; sym < nsyms; sym++) {\n if (length[sym] != bit_num) continue;\n if (pos >= table_mask) return 1; /* table overflow */\n\n#ifdef BITS_ORDER_MSB\n leaf = pos >> 16;\n#else\n /* leaf = the first nbits of the code, reversed */\n reverse = pos >> 16; leaf = 0; fill = nbits;\n do {leaf <<= 1; leaf |= reverse & 1; reverse >>= 1;} while (--fill);\n#endif\n for (fill = 0; fill < (bit_num - nbits); fill++) {\n /* if this path hasn't been taken yet, 'allocate' two entries */\n if (table[leaf] == 0xFFFF) {\n table[(next_symbol << 1) ] = 0xFFFF;\n table[(next_symbol << 1) + 1 ] = 0xFFFF;\n table[leaf] = next_symbol++;\n }\n\n /* follow the path and select either left or right for next bit */\n leaf = table[leaf] << 1;\n if ((pos >> (15-fill)) & 1) leaf++;\n }\n table[leaf] = sym;\n pos += bit_mask;\n }\n bit_mask >>= 1;\n }\n\n /* full table? */\n return (pos == table_mask) ? 0 : 1;\n}\n#endif\n" }, { "path": "third_party/mspack/system.c", "content": "/* This file is part of libmspack.\n * (C) 2003-2004 Stuart Caie.\n *\n * libmspack is free software; you can redistribute it and/or modify it under\n * the terms of the GNU Lesser General Public License (LGPL) version 2.1\n *\n * For further details, see the file COPYING.LIB distributed with libmspack\n */\n\n#ifdef HAVE_CONFIG_H\n# include \n#endif\n\n#include \n\n#if !LARGEFILE_SUPPORT\nconst char *largefile_msg = \"library not compiled to support large files.\";\n#endif\n\n\nint mspack_version(int entity) {\n switch (entity) {\n /* CHM decoder version 1 -> 2 changes:\n * - added mschmd_sec_mscompressed::spaninfo\n * - added mschmd_header::first_pmgl\n * - added mschmd_header::last_pmgl\n * - added mschmd_header::chunk_cache;\n */\n case MSPACK_VER_MSCHMD:\n /* CAB decoder version 1 -> 2 changes:\n * - added MSCABD_PARAM_SALVAGE\n */\n case MSPACK_VER_MSCABD:\n return 2;\n case MSPACK_VER_LIBRARY:\n case MSPACK_VER_SYSTEM:\n case MSPACK_VER_MSSZDDD:\n case MSPACK_VER_MSKWAJD:\n case MSPACK_VER_MSOABD:\n return 1;\n case MSPACK_VER_MSCABC:\n case MSPACK_VER_MSCHMC:\n case MSPACK_VER_MSLITD:\n case MSPACK_VER_MSLITC:\n case MSPACK_VER_MSHLPD:\n case MSPACK_VER_MSHLPC:\n case MSPACK_VER_MSSZDDC:\n case MSPACK_VER_MSKWAJC:\n case MSPACK_VER_MSOABC:\n return 0;\n }\n return -1;\n}\n\nint mspack_sys_selftest_internal(int offt_size) {\n return (sizeof(off_t) == offt_size) ? MSPACK_ERR_OK : MSPACK_ERR_SEEK;\n}\n\n/* validates a system structure */\nint mspack_valid_system(struct mspack_system *sys) {\n return (sys != NULL) && (sys->open != NULL) && (sys->close != NULL) &&\n (sys->read != NULL) && (sys->write != NULL) && (sys->seek != NULL) &&\n (sys->tell != NULL) && (sys->message != NULL) && (sys->alloc != NULL) &&\n (sys->free != NULL) && (sys->copy != NULL) && (sys->null_ptr == NULL);\n}\n\n/* returns the length of a file opened for reading */\nint mspack_sys_filelen(struct mspack_system *system,\n struct mspack_file *file, off_t *length)\n{\n off_t current;\n\n if (!system || !file || !length) return MSPACK_ERR_OPEN;\n\n /* get current offset */\n current = system->tell(file);\n\n /* seek to end of file */\n if (system->seek(file, (off_t) 0, MSPACK_SYS_SEEK_END)) {\n return MSPACK_ERR_SEEK;\n }\n\n /* get offset of end of file */\n *length = system->tell(file);\n\n /* seek back to original offset */\n if (system->seek(file, current, MSPACK_SYS_SEEK_START)) {\n return MSPACK_ERR_SEEK;\n }\n\n return MSPACK_ERR_OK;\n}\n\n\n\n/* definition of mspack_default_system -- if the library is compiled with\n * MSPACK_NO_DEFAULT_SYSTEM, no default system will be provided. Otherwise,\n * an appropriate default system (e.g. the standard C library, or some native\n * API calls)\n */\n\n#ifdef MSPACK_NO_DEFAULT_SYSTEM\nstruct mspack_system *mspack_default_system = NULL;\n#else\n\n/* implementation of mspack_default_system for standard C library */\n\n#include \n#include \n#include \n#include \n\nstruct mspack_file_p {\n FILE *fh;\n const char *name;\n};\n\nstatic struct mspack_file *msp_open(struct mspack_system *self,\n const char *filename, int mode)\n{\n struct mspack_file_p *fh;\n const char *fmode;\n\n switch (mode) {\n case MSPACK_SYS_OPEN_READ: fmode = \"rb\"; break;\n case MSPACK_SYS_OPEN_WRITE: fmode = \"wb\"; break;\n case MSPACK_SYS_OPEN_UPDATE: fmode = \"r+b\"; break;\n case MSPACK_SYS_OPEN_APPEND: fmode = \"ab\"; break;\n default: return NULL;\n }\n\n if ((fh = (struct mspack_file_p *) malloc(sizeof(struct mspack_file_p)))) {\n fh->name = filename;\n if ((fh->fh = fopen(filename, fmode))) return (struct mspack_file *) fh;\n free(fh);\n }\n return NULL;\n}\n\nstatic void msp_close(struct mspack_file *file) {\n struct mspack_file_p *self = (struct mspack_file_p *) file;\n if (self) {\n fclose(self->fh);\n free(self);\n }\n}\n\nstatic int msp_read(struct mspack_file *file, void *buffer, int bytes) {\n struct mspack_file_p *self = (struct mspack_file_p *) file;\n if (self && buffer && bytes >= 0) {\n size_t count = fread(buffer, 1, (size_t) bytes, self->fh);\n if (!ferror(self->fh)) return (int) count;\n }\n return -1;\n}\n\nstatic int msp_write(struct mspack_file *file, void *buffer, int bytes) {\n struct mspack_file_p *self = (struct mspack_file_p *) file;\n if (self && buffer && bytes >= 0) {\n size_t count = fwrite(buffer, 1, (size_t) bytes, self->fh);\n if (!ferror(self->fh)) return (int) count;\n }\n return -1;\n}\n\nstatic int msp_seek(struct mspack_file *file, off_t offset, int mode) {\n struct mspack_file_p *self = (struct mspack_file_p *) file;\n if (self) {\n switch (mode) {\n case MSPACK_SYS_SEEK_START: mode = SEEK_SET; break;\n case MSPACK_SYS_SEEK_CUR: mode = SEEK_CUR; break;\n case MSPACK_SYS_SEEK_END: mode = SEEK_END; break;\n default: return -1;\n }\n#if HAVE_FSEEKO\n return fseeko(self->fh, offset, mode);\n#else\n return fseek(self->fh, offset, mode);\n#endif\n }\n return -1;\n}\n\nstatic off_t msp_tell(struct mspack_file *file) {\n struct mspack_file_p *self = (struct mspack_file_p *) file;\n#if HAVE_FSEEKO\n return (self) ? (off_t) ftello(self->fh) : 0;\n#else\n return (self) ? (off_t) ftell(self->fh) : 0;\n#endif\n}\n\nstatic void msp_msg(struct mspack_file *file, const char *format, ...) {\n va_list ap;\n if (file) fprintf(stderr, \"%s: \", ((struct mspack_file_p *) file)->name);\n va_start(ap, format);\n vfprintf(stderr, format, ap);\n va_end(ap);\n fputc((int) '\\n', stderr);\n fflush(stderr);\n}\n\nstatic void *msp_alloc(struct mspack_system *self, size_t bytes) {\n#if DEBUG\n /* make uninitialised data obvious */\n char *buf = malloc(bytes + 8);\n if (buf) memset(buf, 0xDC, bytes);\n *((size_t *)buf) = bytes;\n return &buf[8];\n#else\n return malloc(bytes);\n#endif\n}\n\nstatic void msp_free(void *buffer) {\n#if DEBUG\n char *buf = buffer;\n size_t bytes;\n if (buf) {\n buf -= 8;\n bytes = *((size_t *)buf);\n /* make freed data obvious */\n memset(buf, 0xED, bytes);\n free(buf);\n }\n#else\n free(buffer);\n#endif\n}\n\nstatic void msp_copy(void *src, void *dest, size_t bytes) {\n memcpy(dest, src, bytes);\n}\n\nstatic struct mspack_system msp_system = {\n &msp_open, &msp_close, &msp_read, &msp_write, &msp_seek,\n &msp_tell, &msp_msg, &msp_alloc, &msp_free, &msp_copy, NULL\n};\n\nstruct mspack_system *mspack_default_system = &msp_system;\n\n#endif\n" }, { "path": "third_party/mspack/system.h", "content": "/* This file is part of libmspack.\n * (C) 2003-2018 Stuart Caie.\n *\n * libmspack is free software; you can redistribute it and/or modify it under\n * the terms of the GNU Lesser General Public License (LGPL) version 2.1\n *\n * For further details, see the file COPYING.LIB distributed with libmspack\n */\n\n#ifndef MSPACK_SYSTEM_H\n#define MSPACK_SYSTEM_H 1\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\n/* ensure config.h is read before mspack.h */\n#ifdef HAVE_CONFIG_H\n# include \n#endif\n\n#include \n\n/* assume exists */\n#include \n\n/* fix for problem with GCC 4 and glibc (thanks to Ville Skytta)\n * http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=150429\n */\n#ifdef read\n# undef read\n#endif\n\n/* Old GCCs don't have __func__, but __FUNCTION__:\n * http://gcc.gnu.org/onlinedocs/gcc/Function-Names.html\n */\n#if __STDC_VERSION__ < 199901L\n# if __GNUC__ >= 2\n# define __func__ __FUNCTION__\n# else\n# define __func__ \"\"\n# endif\n#endif\n\n#if DEBUG\n# include \n# define D(x) do { printf(\"%s:%d (%s) \",__FILE__, __LINE__, __func__); \\\n printf x ; fputc('\\n', stdout); fflush(stdout);} while (0);\n#else\n# define D(x)\n#endif\n\n/* CAB supports searching through files over 4GB in size, and the CHM file\n * format actively uses 64-bit offsets. These can only be fully supported\n * if the system the code runs on supports large files. If not, the library\n * will work as normal using only 32-bit arithmetic, but if an offset\n * greater than 2GB is detected, an error message indicating the library\n * can't support the file should be printed.\n */\n#if HAVE_INTTYPES_H\n# include \n#else\n# define PRId64 \"lld\"\n# define PRIu64 \"llu\"\n# define PRId32 \"ld\"\n# define PRIu32 \"lu\"\n#endif\n\n#include \n#if ((defined(_FILE_OFFSET_BITS) && _FILE_OFFSET_BITS >= 64) || \\\n (defined(FILESIZEBITS) && FILESIZEBITS >= 64) || \\\n defined(_LARGEFILE_SOURCE) || defined(_LARGEFILE64_SOURCE) || \\\n SIZEOF_OFF_T >= 8)\n# define LARGEFILE_SUPPORT 1\n# define LD PRId64\n# define LU PRIu64\n#else\nextern const char *largefile_msg;\n# define LD PRId32\n# define LU PRIu32\n#endif\n\n/* endian-neutral reading of little-endian data */\n#define __egi32(a,n) ( ((((unsigned char *) a)[n+3]) << 24) | \\\n ((((unsigned char *) a)[n+2]) << 16) | \\\n ((((unsigned char *) a)[n+1]) << 8) | \\\n ((((unsigned char *) a)[n+0])))\n#define EndGetI64(a) ((((unsigned long long int) __egi32(a,4)) << 32) | \\\n ((unsigned int) __egi32(a,0)))\n#define EndGetI32(a) __egi32(a,0)\n#define EndGetI16(a) ((((a)[1])<<8)|((a)[0]))\n\n/* endian-neutral reading of big-endian data */\n#define EndGetM32(a) (((((unsigned char *) a)[0]) << 24) | \\\n ((((unsigned char *) a)[1]) << 16) | \\\n ((((unsigned char *) a)[2]) << 8) | \\\n ((((unsigned char *) a)[3])))\n#define EndGetM16(a) ((((a)[0])<<8)|((a)[1]))\n\nextern struct mspack_system *mspack_default_system;\n\n/* returns the length of a file opened for reading */\nextern int mspack_sys_filelen(struct mspack_system *system,\n struct mspack_file *file, off_t *length);\n\n/* validates a system structure */\nextern int mspack_valid_system(struct mspack_system *sys);\n\n#ifdef __cplusplus\n}\n#endif\n\n#endif\n" }, { "path": "third_party/mspack.lua", "content": "group(\"third_party\")\nproject(\"mspack\")\n uuid(\"0881692A-75A1-4E7B-87D8-BB9108CEDEA4\")\n kind(\"StaticLib\")\n language(\"C++\")\n links({\n \"xenia-base\",\n })\n defines({\n \"_LIB\",\n \"HAVE_CONFIG_H\",\n })\n removedefines({\n \"_UNICODE\",\n \"UNICODE\",\n })\n includedirs({\n \"mspack\",\n })\n files({\n \"mspack/logging.cc\",\n \"mspack/lzx.h\",\n \"mspack/lzxd.c\",\n \"mspack/mspack.h\",\n \"mspack/readbits.h\",\n \"mspack/readhuff.h\",\n \"mspack/system.c\",\n \"mspack/system.h\",\n })\n\n filter(\"platforms:Windows\")\n defines({\n })\n filter(\"platforms:Linux\")\n defines({\n })\n" }, { "path": "third_party/pe/pe_image.h", "content": "/* Xenia: minor tweaks to bring up to date with winnt.h */\n\n/* Copyright (c) Microsoft Corporation. All rights reserved. */\n/* NT image format (to be used when the Win32 SDK version of WINNT.H cannot) */\n\n#ifndef UNALIGNED\n/* BUGBUG fixme */\n#define UNALIGNED\n#endif\n\n#ifdef _MSC_VER\n#pragma pack (1)\n#endif\n\n// MODIFIED: on Windows a few definitions aren't in WINNT\n#ifdef IMAGE_NT_SIGNATURE\n#define IMAGE_SIZEOF_NT_OPTIONAL_HEADER 224\n#define IMAGE_FILE_MACHINE_POWERPCBE 0x01F2 // IBM PowerPC Big-Endian\n#endif\n\n/* Skip if WINNT.H already included. We check IMAGE_NT_SIGNATURE instead of _WINNT_,\n * because WinCE's version of WINNT.H defines _WINNT_ but it doesn't include everything here\n */\n#ifndef IMAGE_NT_SIGNATURE\n\n#define IMAGE_DOS_SIGNATURE 0x5A4D /* MZ */\n#define IMAGE_OS2_SIGNATURE 0x454E /* NE */\n#define IMAGE_OS2_SIGNATURE_LE 0x454C /* LE */\n#define IMAGE_NT_SIGNATURE 0x00004550 /* PE00 */\n\ntypedef struct _IMAGE_DOS_HEADER { /* DOS .EXE header */\n uint16_t e_magic; /* Magic number */\n uint16_t e_cblp; /* Bytes on last page of file */\n uint16_t e_cp; /* Pages in file */\n uint16_t e_crlc; /* Relocations */\n uint16_t e_cparhdr; /* Size of header in paragraphs */\n uint16_t e_minalloc; /* Minimum extra paragraphs needed */\n uint16_t e_maxalloc; /* Maximum extra paragraphs needed */\n uint16_t e_ss; /* Initial (relative) SS value */\n uint16_t e_sp; /* Initial SP value */\n uint16_t e_csum; /* Checksum */\n uint16_t e_ip; /* Initial IP value */\n uint16_t e_cs; /* Initial (relative) CS value */\n uint16_t e_lfarlc; /* File address of relocation table */\n uint16_t e_ovno; /* Overlay number */\n uint16_t e_res[4]; /* Reserved words */\n uint16_t e_oemid; /* OEM identifier (for e_oeminfo) */\n uint16_t e_oeminfo; /* OEM information; e_oemid specific */\n uint16_t e_res2[10]; /* Reserved words */\n int32_t e_lfanew; /* File address of new exe header */\n } IMAGE_DOS_HEADER, *PIMAGE_DOS_HEADER;\n\ntypedef struct _IMAGE_OS2_HEADER { /* OS/2 .EXE header */\n uint16_t ne_magic; /* Magic number */\n int8_t ne_ver; /* Version number */\n int8_t ne_rev; /* Revision number */\n uint16_t ne_enttab; /* Offset of Entry Table */\n uint16_t ne_cbenttab; /* Number of bytes in Entry Table */\n int32_t ne_crc; /* Checksum of whole file */\n uint16_t ne_flags; /* Flag word */\n uint16_t ne_autodata; /* Automatic data segment number */\n uint16_t ne_heap; /* Initial heap allocation */\n uint16_t ne_stack; /* Initial stack allocation */\n int32_t ne_csip; /* Initial CS:IP setting */\n int32_t ne_sssp; /* Initial SS:SP setting */\n uint16_t ne_cseg; /* Count of file segments */\n uint16_t ne_cmod; /* Entries in Module Reference Table */\n uint16_t ne_cbnrestab; /* Size of non-resident name table */\n uint16_t ne_segtab; /* Offset of Segment Table */\n uint16_t ne_rsrctab; /* Offset of Resource Table */\n uint16_t ne_restab; /* Offset of resident name table */\n uint16_t ne_modtab; /* Offset of Module Reference Table */\n uint16_t ne_imptab; /* Offset of Imported Names Table */\n int32_t ne_nrestab; /* Offset of Non-resident Names Table */\n uint16_t ne_cmovent; /* Count of movable entries */\n uint16_t ne_align; /* Segment alignment shift count */\n uint16_t ne_cres; /* Count of resource segments */\n uint8_t ne_exetyp; /* Target Operating system */\n uint8_t ne_flagsothers; /* Other .EXE flags */\n uint16_t ne_pretthunks; /* offset to return thunks */\n uint16_t ne_psegrefbytes; /* offset to segment ref. bytes */\n uint16_t ne_swaparea; /* Minimum code swap area size */\n uint16_t ne_expver; /* Expected Windows version number */\n } IMAGE_OS2_HEADER, *PIMAGE_OS2_HEADER;\n\n/*\n * File header format.\n */\n\ntypedef struct _IMAGE_FILE_HEADER {\n uint16_t Machine;\n uint16_t NumberOfSections;\n uint32_t TimeDateStamp;\n uint32_t PointerToSymbolTable;\n uint32_t NumberOfSymbols;\n uint16_t SizeOfOptionalHeader;\n uint16_t Characteristics;\n} IMAGE_FILE_HEADER, *PIMAGE_FILE_HEADER;\n\n#define IMAGE_SIZEOF_FILE_HEADER 20\n\n#define IMAGE_FILE_RELOCS_STRIPPED 0x0001 /* Relocation info stripped from file. */\n#define IMAGE_FILE_EXECUTABLE_IMAGE 0x0002 /* File is executable (i.e. no unresolved externel references). */\n#define IMAGE_FILE_LINE_NUMS_STRIPPED 0x0004 /* Line nunbers stripped from file. */\n#define IMAGE_FILE_LOCAL_SYMS_STRIPPED 0x0008 /* Local symbols stripped from file. */\n#define IMAGE_FILE_MINIMAL_OBJECT 0x0010 /* Reserved. */\n#define IMAGE_FILE_UPDATE_OBJECT 0x0020 /* Reserved. */\n#define IMAGE_FILE_16BIT_MACHINE 0x0040 /* 16 bit word machine. */\n#define IMAGE_FILE_BYTES_REVERSED_LO 0x0080 /* Bytes of machine word are reversed. */\n#define IMAGE_FILE_32BIT_MACHINE 0x0100 /* 32 bit word machine. */\n#define IMAGE_FILE_DEBUG_STRIPPED 0x0200 /* Debugging info stripped from file in .DBG file */\n#define IMAGE_FILE_PATCH 0x0400 /* Reserved. */\n#define IMAGE_FILE_SYSTEM 0x1000 /* System File. */\n#define IMAGE_FILE_DLL 0x2000 /* File is a DLL. */\n#define IMAGE_FILE_BYTES_REVERSED_HI 0x8000 /* Bytes of machine word are reversed. */\n\n#define IMAGE_FILE_MACHINE_UNKNOWN 0\n#define IMAGE_FILE_MACHINE_I860 0x14d /* Intel 860. */\n#define IMAGE_FILE_MACHINE_I386 0x14c /* Intel 386. */\n#define IMAGE_FILE_MACHINE_R3000 0x162 /* MIPS little-endian, 0540 big-endian */\n#define IMAGE_FILE_MACHINE_R4000 0x166 /* MIPS little-endian */\n#define IMAGE_FILE_MACHINE_R10000 0x0168 // MIPS little-endian\n#define IMAGE_FILE_MACHINE_ALPHA 0x184 /* Alpha_AXP */\n#define IMAGE_FILE_MACHINE_POWERPC 0x01F0 // IBM PowerPC Little-Endian\n#define IMAGE_FILE_MACHINE_POWERPCBE 0x01F2 // IBM PowerPC Big-Endian\n#define IMAGE_FILE_MACHINE_SH3 0x01a2 // SH3 little-endian\n#define IMAGE_FILE_MACHINE_SH3E 0x01a4 // SH3E little-endian\n#define IMAGE_FILE_MACHINE_SH4 0x01a6 // SH4 little-endian\n#define IMAGE_FILE_MACHINE_ARM 0x01c0 // ARM Little-Endian\n#define IMAGE_FILE_MACHINE_THUMB 0x01c2\n#define IMAGE_FILE_MACHINE_IA64 0x0200 // Intel 64\n#define IMAGE_FILE_MACHINE_MIPS16 0x0266 // MIPS\n#define IMAGE_FILE_MACHINE_MIPSFPU 0x0366 // MIPS\n#define IMAGE_FILE_MACHINE_MIPSFPU16 0x0466 // MIPS\n#define IMAGE_FILE_MACHINE_ALPHA64 0x0284 // ALPHA64\n#define IMAGE_FILE_MACHINE_AXP64 IMAGE_FILE_MACHINE_ALPHA64\n#define IMAGE_FILE_MACHINE_CEF 0xC0EF\n\n/*\n * Directory format.\n */\n\ntypedef struct _IMAGE_DATA_DIRECTORY {\n uint32_t VirtualAddress;\n uint32_t Size;\n} IMAGE_DATA_DIRECTORY, *PIMAGE_DATA_DIRECTORY;\n\n#define IMAGE_NUMBEROF_DIRECTORY_ENTRIES 16\n\n/*\n * Optional header format.\n */\n\ntypedef struct _IMAGE_OPTIONAL_HEADER {\n /*\n * Standard fields.\n */\n\n uint16_t Magic;\n uint8_t MajorLinkerVersion;\n uint8_t MinorLinkerVersion;\n uint32_t SizeOfCode;\n uint32_t SizeOfInitializedData;\n uint32_t SizeOfUninitializedData;\n uint32_t AddressOfEntryPoint;\n uint32_t BaseOfCode;\n uint32_t BaseOfData;\n\n /*\n * NT additional fields.\n */\n\n uint32_t ImageBase;\n uint32_t SectionAlignment;\n uint32_t FileAlignment;\n uint16_t MajorOperatingSystemVersion;\n uint16_t MinorOperatingSystemVersion;\n uint16_t MajorImageVersion;\n uint16_t MinorImageVersion;\n uint16_t MajorSubsystemVersion;\n uint16_t MinorSubsystemVersion;\n uint32_t Reserved1;\n uint32_t SizeOfImage;\n uint32_t SizeOfHeaders;\n uint32_t CheckSum;\n uint16_t Subsystem;\n uint16_t DllCharacteristics;\n uint32_t SizeOfStackReserve;\n uint32_t SizeOfStackCommit;\n uint32_t SizeOfHeapReserve;\n uint32_t SizeOfHeapCommit;\n uint32_t LoaderFlags;\n uint32_t NumberOfRvaAndSizes;\n IMAGE_DATA_DIRECTORY DataDirectory[IMAGE_NUMBEROF_DIRECTORY_ENTRIES];\n} IMAGE_OPTIONAL_HEADER, *PIMAGE_OPTIONAL_HEADER;\n\ntypedef IMAGE_OPTIONAL_HEADER IMAGE_OPTIONAL_HEADER32;\n\n#define IMAGE_SIZEOF_STD_OPTIONAL_HEADER 28\n#define IMAGE_SIZEOF_NT_OPTIONAL_HEADER 224\n\n#define IMAGE_NT_OPTIONAL_HDR32_MAGIC 0x10b\n\n#ifdef _WIN64\n#define IMAGE_NT_OPTIONAL_HDR_MAGIC IMAGE_NT_OPTIONAL_HDR64_MAGIC\n#else\n#define IMAGE_NT_OPTIONAL_HDR_MAGIC IMAGE_NT_OPTIONAL_HDR32_MAGIC\n#endif\n\ntypedef struct _IMAGE_NT_HEADERS {\n uint32_t Signature;\n IMAGE_FILE_HEADER FileHeader;\n IMAGE_OPTIONAL_HEADER OptionalHeader;\n} IMAGE_NT_HEADERS, *PIMAGE_NT_HEADERS;\n\ntypedef IMAGE_NT_HEADERS IMAGE_NT_HEADERS32;\n\n#define IMAGE_FIRST_SECTION( ntheader ) ((PIMAGE_SECTION_HEADER) \\\n ((uint8_t*)ntheader + \\\n offsetof( IMAGE_NT_HEADERS, OptionalHeader ) + \\\n ((PIMAGE_NT_HEADERS)(ntheader))->FileHeader.SizeOfOptionalHeader \\\n ))\n\n\n/* Subsystem Values */\n\n#define IMAGE_SUBSYSTEM_UNKNOWN 0 /* Unknown subsystem. */\n#define IMAGE_SUBSYSTEM_NATIVE 1 /* Image doesn't require a subsystem. */\n#define IMAGE_SUBSYSTEM_WINDOWS_GUI 2 /* Image runs in the Windows GUI subsystem. */\n#define IMAGE_SUBSYSTEM_WINDOWS_CUI 3 /* Image runs in the Windows character subsystem. */\n#define IMAGE_SUBSYSTEM_OS2_CUI 5 /* image runs in the OS/2 character subsystem. */\n#define IMAGE_SUBSYSTEM_POSIX_CUI 7 /* image run in the Posix character subsystem. */\n#define IMAGE_SUBSYSTEM_XBOX 14\n\n/* Dll Characteristics */\n\n#define IMAGE_LIBRARY_PROCESS_INIT 1 /* Dll has a process initialization routine. */\n#define IMAGE_LIBRARY_PROCESS_TERM 2 /* Dll has a thread termination routine. */\n#define IMAGE_LIBRARY_THREAD_INIT 4 /* Dll has a thread initialization routine. */\n#define IMAGE_LIBRARY_THREAD_TERM 8 /* Dll has a thread termination routine. */\n\n/*\n * Loader Flags\n */\n\n#define IMAGE_LOADER_FLAGS_BREAK_ON_LOAD 0x00000001\n#define IMAGE_LOADER_FLAGS_DEBUG_ON_LOAD 0x00000002\n\n\n/* Directory Entries */\n\n#define IMAGE_DIRECTORY_ENTRY_EXPORT 0 /* Export Directory */\n#define IMAGE_DIRECTORY_ENTRY_IMPORT 1 /* Import Directory */\n#define IMAGE_DIRECTORY_ENTRY_RESOURCE 2 /* Resource Directory */\n#define IMAGE_DIRECTORY_ENTRY_EXCEPTION 3 /* Exception Directory */\n#define IMAGE_DIRECTORY_ENTRY_SECURITY 4 /* Security Directory */\n#define IMAGE_DIRECTORY_ENTRY_BASERELOC 5 /* Base Relocation Table */\n#define IMAGE_DIRECTORY_ENTRY_DEBUG 6 /* Debug Directory */\n#define IMAGE_DIRECTORY_ENTRY_COPYRIGHT 7 /* Description String */\n#define IMAGE_DIRECTORY_ENTRY_GLOBALPTR 8 /* Machine Value (MIPS GP) */\n#define IMAGE_DIRECTORY_ENTRY_TLS 9 /* TLS Directory */\n#define IMAGE_DIRECTORY_ENTRY_LOAD_CONFIG 10 /* Load Configuration Directory */\n\n#ifdef USES_COMPLUS20\n#define IMAGE_DIRECTORY_ENTRY_COMHEADER 14 /* COM+ data */\n#endif\n\n\n/*\n * Section header format.\n */\n\n#define IMAGE_SIZEOF_SHORT_NAME 8\n\ntypedef struct _IMAGE_SECTION_HEADER {\n uint8_t Name[IMAGE_SIZEOF_SHORT_NAME];\n union {\n uint32_t PhysicalAddress;\n uint32_t VirtualSize;\n } Misc;\n uint32_t VirtualAddress;\n uint32_t SizeOfRawData;\n uint32_t PointerToRawData;\n uint32_t PointerToRelocations;\n uint32_t PointerToLinenumbers;\n uint16_t NumberOfRelocations;\n uint16_t NumberOfLinenumbers;\n uint32_t Characteristics;\n} IMAGE_SECTION_HEADER, *PIMAGE_SECTION_HEADER;\n\n#define IMAGE_SIZEOF_SECTION_HEADER 40\n\n#define IMAGE_SCN_TYPE_REGULAR 0x00000000 /*\n#define IMAGE_SCN_TYPE_DUMMY 0x00000001 // Reserved. */\n#define IMAGE_SCN_TYPE_NO_LOAD 0x00000002 /* Reserved. */\n#define IMAGE_SCN_TYPE_GROUPED 0x00000004 /* Used for 16-bit offset code. */\n#define IMAGE_SCN_TYPE_NO_PAD 0x00000008 /* Reserved. */\n#define IMAGE_SCN_TYPE_COPY 0x00000010 /* Reserved. */\n\n#define IMAGE_SCN_CNT_CODE 0x00000020 /* Section contains code. */\n#define IMAGE_SCN_CNT_INITIALIZED_DATA 0x00000040 /* Section contains initialized data. */\n#define IMAGE_SCN_CNT_UNINITIALIZED_DATA 0x00000080 /* Section contains uninitialized data. */\n\n#define IMAGE_SCN_LNK_OTHER 0x00000100 /* Reserved. */\n#define IMAGE_SCN_LNK_INFO 0x00000200 /* Section contains comments or some other type of information. */\n#define IMAGE_SCN_LNK_OVERLAY 0x00000400 /* Section contains an overlay. */\n#define IMAGE_SCN_LNK_REMOVE 0x00000800 /* Section contents will not become part of image. */\n#define IMAGE_SCN_LNK_COMDAT 0x00001000 /* Section contents comdat. */\n\n#define IMAGE_SCN_ALIGN_1BYTES 0x00100000\n#define IMAGE_SCN_ALIGN_2BYTES 0x00200000\n#define IMAGE_SCN_ALIGN_4BYTES 0x00300000\n#define IMAGE_SCN_ALIGN_8BYTES 0x00400000\n#define IMAGE_SCN_ALIGN_16BYTES 0x00500000 /* Default alignment if no others are specified. */\n#define IMAGE_SCN_ALIGN_32BYTES 0x00600000\n#define IMAGE_SCN_ALIGN_64BYTES 0x00700000\n\n#define IMAGE_SCN_MEM_DISCARDABLE 0x02000000 /* Section can be discarded. */\n#define IMAGE_SCN_MEM_NOT_CACHED 0x04000000 /* Section is not cachable. */\n#define IMAGE_SCN_MEM_NOT_PAGED 0x08000000 /* Section is not pageable. */\n#define IMAGE_SCN_MEM_SHARED 0x10000000 /* Section is shareable. */\n#define IMAGE_SCN_MEM_EXECUTE 0x20000000 /* Section is executable. */\n#define IMAGE_SCN_MEM_READ 0x40000000 /* Section is readable. */\n#define IMAGE_SCN_MEM_WRITE 0x80000000 /* Section is writeable. */\n\n/*\n * Symbol format.\n */\n\ntypedef struct _IMAGE_SYMBOL {\n union {\n uint8_t ShortName[8];\n struct {\n uint32_t Short; /* if 0, use LongName */\n uint32_t Long; /* offset into string table */\n } Name;\n uint8_t* LongName[2];\n } N;\n uint32_t Value;\n int16_t SectionNumber;\n uint16_t Type;\n uint8_t StorageClass;\n uint8_t NumberOfAuxSymbols;\n} IMAGE_SYMBOL;\ntypedef IMAGE_SYMBOL UNALIGNED *PIMAGE_SYMBOL;\n\n#define IMAGE_SIZEOF_SYMBOL 18\n\n/*\n * Section values.\n *\n * Symbols have a section number of the section in which they are\n * defined. Otherwise, section numbers have the following meanings:\n */\n\n#define IMAGE_SYM_UNDEFINED (int16 )0 /* Symbol is undefined or is common. */\n#define IMAGE_SYM_ABSOLUTE (int16 )-1 /* Symbol is an absolute value. */\n#define IMAGE_SYM_DEBUG (int16 )-2 /* Symbol is a special debug item. */\n\n/*\n * Type (fundamental) values.\n */\n\n#define IMAGE_SYM_TYPE_NULL 0 /* no type. */\n#define IMAGE_SYM_TYPE_VOID 1\n#define IMAGE_SYM_TYPE_CHAR 2 /* type character. */\n#define IMAGE_SYM_TYPE_SHORT 3 /* type short integer. */\n#define IMAGE_SYM_TYPE_INT 4\n#define IMAGE_SYM_TYPE_LONG 5\n#define IMAGE_SYM_TYPE_FLOAT 6\n#define IMAGE_SYM_TYPE_DOUBLE 7\n#define IMAGE_SYM_TYPE_STRUCT 8\n#define IMAGE_SYM_TYPE_UNION 9\n#define IMAGE_SYM_TYPE_ENUM 10 /* enumeration. */\n#define IMAGE_SYM_TYPE_MOE 11 /* member of enumeration. */\n#define IMAGE_SYM_TYPE_UCHAR 12\n#define IMAGE_SYM_TYPE_USHORT 13\n#define IMAGE_SYM_TYPE_UINT 14\n#define IMAGE_SYM_TYPE_ULONG 15\n\n/*\n * Type (derived) values.\n */\n\n#define IMAGE_SYM_DTYPE_NULL 0 /* no derived type. */\n#define IMAGE_SYM_DTYPE_POINTER 1 /* pointer. */\n#define IMAGE_SYM_DTYPE_FUNCTION 2 /* function. */\n#define IMAGE_SYM_DTYPE_ARRAY 3 /* array. */\n\n/*\n * Storage classes.\n */\n\n#define IMAGE_SYM_CLASS_END_OF_FUNCTION (uint8_t )-1\n#define IMAGE_SYM_CLASS_NULL 0\n#define IMAGE_SYM_CLASS_AUTOMATIC 1\n#define IMAGE_SYM_CLASS_EXTERNAL 2\n#define IMAGE_SYM_CLASS_STATIC 3\n#define IMAGE_SYM_CLASS_REGISTER 4\n#define IMAGE_SYM_CLASS_EXTERNAL_DEF 5\n#define IMAGE_SYM_CLASS_LABEL 6\n#define IMAGE_SYM_CLASS_UNDEFINED_LABEL 7\n#define IMAGE_SYM_CLASS_MEMBER_OF_STRUCT 8\n#define IMAGE_SYM_CLASS_ARGUMENT 9\n#define IMAGE_SYM_CLASS_STRUCT_TAG 10\n#define IMAGE_SYM_CLASS_MEMBER_OF_UNION 11\n#define IMAGE_SYM_CLASS_UNION_TAG 12\n#define IMAGE_SYM_CLASS_TYPE_DEFINITION 13\n#define IMAGE_SYM_CLASS_UNDEFINED_STATIC 14\n#define IMAGE_SYM_CLASS_ENUM_TAG 15\n#define IMAGE_SYM_CLASS_MEMBER_OF_ENUM 16\n#define IMAGE_SYM_CLASS_REGISTER_PARAM 17\n#define IMAGE_SYM_CLASS_BIT_FIELD 18\n#define IMAGE_SYM_CLASS_BLOCK 100\n#define IMAGE_SYM_CLASS_FUNCTION 101\n#define IMAGE_SYM_CLASS_END_OF_STRUCT 102\n#define IMAGE_SYM_CLASS_FILE 103\n#define IMAGE_SYM_CLASS_SECTION 104\n#define IMAGE_SYM_CLASS_WEAK_EXTERNAL 105\n\n/* type packing constants */\n\n#define N_BTMASK 017\n#define N_TMASK 060\n#define N_TMASK1 0300\n#define N_TMASK2 0360\n#define N_BTSHFT 4\n#define N_TSHIFT 2\n\n/* MACROS */\n\n/* Basic Type of x */\n#define BTYPE(x) ((x) & N_BTMASK)\n\n/* Is x a pointer? */\n#ifndef ISPTR\n#define ISPTR(x) (((x) & N_TMASK) == (IMAGE_SYM_DTYPE_POINTER << N_BTSHFT))\n#endif\n\n/* Is x a function? */\n#ifndef ISFCN\n#define ISFCN(x) (((x) & N_TMASK) == (IMAGE_SYM_DTYPE_FUNCTION << N_BTSHFT))\n#endif\n\n/* Is x an array? */\n\n#ifndef ISARY\n#define ISARY(x) (((x) & N_TMASK) == (IMAGE_SYM_DTYPE_ARRAY << N_BTSHFT))\n#endif\n\n/* Is x a structure, union, or enumeration TAG? */\n#ifndef ISTAG\n#define ISTAG(x) ((x)==IMAGE_SYM_CLASS_STRUCT_TAG || (x)==IMAGE_SYM_CLASS_UNION_TAG || (x)==IMAGE_SYM_CLASS_ENUM_TAG)\n#endif\n\n#ifndef INCREF\n#define INCREF(x) ((((x)&~N_BTMASK)<>N_TSHIFT)&~N_BTMASK)|((x)&N_BTMASK))\n#endif\n\n/*\n * Auxiliary entry format.\n */\n\ntypedef union _IMAGE_AUX_SYMBOL {\n struct {\n uint32_t TagIndex; /* struct, union, or enum tag index */\n union {\n struct {\n uint16_t Linenumber; /* declaration line number */\n uint16_t Size; /* size of struct, union, or enum */\n } LnSz;\n uint32_t TotalSize;\n } Misc;\n union {\n struct { /* if ISFCN, tag, or .bb */\n uint32_t PointerToLinenumber;\n uint32_t PointerToNextFunction;\n } Function;\n struct { /* if ISARY, up to 4 dimen. */\n uint16_t Dimension[4];\n } Array;\n } FcnAry;\n uint16_t TvIndex; /* tv index */\n } Sym;\n struct {\n uint8_t Name[IMAGE_SIZEOF_SYMBOL];\n } File;\n struct {\n uint32_t Length; /* section length */\n uint16_t NumberOfRelocations; /* number of relocation entries */\n uint16_t NumberOfLinenumbers; /* number of line numbers */\n uint32_t CheckSum; /* checksum for communal */\n int16_t Number; /* section number to associate with */\n uint8_t Selection; /* communal selection type */\n } Section;\n} IMAGE_AUX_SYMBOL;\ntypedef IMAGE_AUX_SYMBOL UNALIGNED *PIMAGE_AUX_SYMBOL;\n\n#define IMAGE_SIZEOF_AUX_SYMBOL 18\n\n/*\n * Communal selection types.\n */\n\n#define IMAGE_COMDAT_SELECT_UNKNOWN 0\n#define IMAGE_COMDAT_SELECT_NODUPLICATES 1\n#define IMAGE_COMDAT_SELECT_ANY 2\n#define IMAGE_COMDAT_SELECT_SAME_SIZE 3\n#define IMAGE_COMDAT_SELECT_EXACT_MATCH 4\n#define IMAGE_COMDAT_SELECT_ASSOCIATIVE 5\n\n#define IMAGE_WEAK_EXTERN_SEARCH_UNKNOWN 0\n#define IMAGE_WEAK_EXTERN_SEARCH_NOLIBRARY 1\n#define IMAGE_WEAK_EXTERN_SEARCH_LIBRARY 2\n\n\n/*\n * Relocation format.\n */\n\ntypedef struct _IMAGE_RELOCATION {\n uint32_t VirtualAddress;\n uint32_t SymbolTableIndex;\n uint16_t Type;\n} IMAGE_RELOCATION;\ntypedef IMAGE_RELOCATION UNALIGNED *PIMAGE_RELOCATION;\n\n#define IMAGE_SIZEOF_RELOCATION 10\n\n/*\n * I860 relocation types.\n */\n\n#define IMAGE_REL_I860_ABSOLUTE 0 /* Reference is absolute, no relocation is necessary */\n#define IMAGE_REL_I860_DIR32 06 /* Direct 32-bit reference to the symbols virtual address */\n#define IMAGE_REL_I860_DIR32NB 07\n#define IMAGE_REL_I860_SECTION 012\n#define IMAGE_REL_I860_SECREL 013\n#define IMAGE_REL_I860_PAIR 034\n#define IMAGE_REL_I860_HIGH 036\n#define IMAGE_REL_I860_LOW0 037\n#define IMAGE_REL_I860_LOW1 040\n#define IMAGE_REL_I860_LOW2 041\n#define IMAGE_REL_I860_LOW3 042\n#define IMAGE_REL_I860_LOW4 043\n#define IMAGE_REL_I860_SPLIT0 044\n#define IMAGE_REL_I860_SPLIT1 045\n#define IMAGE_REL_I860_SPLIT2 046\n#define IMAGE_REL_I860_HIGHADJ 047\n#define IMAGE_REL_I860_BRADDR 050\n\n/*\n * I386 relocation types.\n */\n\n#define IMAGE_REL_I386_ABSOLUTE 0 /* Reference is absolute, no relocation is necessary */\n#define IMAGE_REL_I386_DIR16 01 /* Direct 16-bit reference to the symbols virtual address */\n#define IMAGE_REL_I386_REL16 02 /* PC-relative 16-bit reference to the symbols virtual address */\n#define IMAGE_REL_I386_DIR32 06 /* Direct 32-bit reference to the symbols virtual address */\n#define IMAGE_REL_I386_DIR32NB 07 /* Direct 32-bit reference to the symbols virtual address, base not included */\n#define IMAGE_REL_I386_SEG12 011 /* Direct 16-bit reference to the segment-selector bits of a 32-bit virtual address */\n#define IMAGE_REL_I386_SECTION 012\n#define IMAGE_REL_I386_SECREL 013\n#define IMAGE_REL_I386_REL32 024 /* PC-relative 32-bit reference to the symbols virtual address */\n\n/*\n * MIPS relocation types.\n */\n\n#define IMAGE_REL_MIPS_ABSOLUTE 0 /* Reference is absolute, no relocation is necessary */\n#define IMAGE_REL_MIPS_REFHALF 01\n#define IMAGE_REL_MIPS_REFWORD 02\n#define IMAGE_REL_MIPS_JMPADDR 03\n#define IMAGE_REL_MIPS_REFHI 04\n#define IMAGE_REL_MIPS_REFLO 05\n#define IMAGE_REL_MIPS_GPREL 06\n#define IMAGE_REL_MIPS_LITERAL 07\n#define IMAGE_REL_MIPS_SECTION 012\n#define IMAGE_REL_MIPS_SECREL 013\n#define IMAGE_REL_MIPS_REFWORDNB 042\n#define IMAGE_REL_MIPS_PAIR 045\n\n/*\n * Alpha Relocation types.\n */\n\n#define IMAGE_REL_ALPHA_ABSOLUTE 0x0\n#define IMAGE_REL_ALPHA_REFLONG 0x1\n#define IMAGE_REL_ALPHA_REFQUAD 0x2\n#define IMAGE_REL_ALPHA_GPREL32 0x3\n#define IMAGE_REL_ALPHA_LITERAL 0x4\n#define IMAGE_REL_ALPHA_LITUSE 0x5\n#define IMAGE_REL_ALPHA_GPDISP 0x6\n#define IMAGE_REL_ALPHA_BRADDR 0x7\n#define IMAGE_REL_ALPHA_HINT 0x8\n#define IMAGE_REL_ALPHA_INLINE_REFLONG 0x9\n#define IMAGE_REL_ALPHA_REFHI 0xA\n#define IMAGE_REL_ALPHA_REFLO 0xB\n#define IMAGE_REL_ALPHA_PAIR 0xC\n#define IMAGE_REL_ALPHA_MATCH 0xD\n#define IMAGE_REL_ALPHA_SECTION 0xE\n#define IMAGE_REL_ALPHA_SECREL 0xF\n#define IMAGE_REL_ALPHA_REFLONGNB 0x10\n\n/*\n * Based relocation format.\n */\n\ntypedef struct _IMAGE_BASE_RELOCATION {\n uint32_t VirtualAddress;\n uint32_t SizeOfBlock;\n/* uint16_t TypeOffset[1]; */\n} IMAGE_BASE_RELOCATION, *PIMAGE_BASE_RELOCATION;\n\n#define IMAGE_SIZEOF_BASE_RELOCATION 8\n\n/*\n * Based relocation types.\n */\n\n#define IMAGE_REL_BASED_ABSOLUTE 0\n#define IMAGE_REL_BASED_HIGH 1\n#define IMAGE_REL_BASED_LOW 2\n#define IMAGE_REL_BASED_HIGHLOW 3\n#define IMAGE_REL_BASED_HIGHADJ 4\n#define IMAGE_REL_BASED_MIPS_JMPADDR 5\n#define IMAGE_REL_BASED_I860_BRADDR 6\n#define IMAGE_REL_BASED_I860_SPLIT 7\n\n/*\n * Line number format.\n */\n\ntypedef struct _IMAGE_LINENUMBER {\n union {\n uint32_t SymbolTableIndex; /* Symbol table index of function name if Linenumber is 0. */\n uint32_t VirtualAddress; /* Virtual address of line number. */\n } Type;\n uint16_t Linenumber; /* Line number. */\n} IMAGE_LINENUMBER;\ntypedef IMAGE_LINENUMBER UNALIGNED *PIMAGE_LINENUMBER;\n\n#define IMAGE_SIZEOF_LINENUMBER 6\n\n/*\n * Archive format.\n */\n\n#define IMAGE_ARCHIVE_START_SIZE 8\n#define IMAGE_ARCHIVE_START \"!\\n\"\n#define IMAGE_ARCHIVE_END \"`\\n\"\n#define IMAGE_ARCHIVE_PAD \"\\n\"\n#define IMAGE_ARCHIVE_LINKER_MEMBER \"/ \"\n#define IMAGE_ARCHIVE_LONGNAMES_MEMBER \"// \"\n\ntypedef struct _IMAGE_ARCHIVE_MEMBER_HEADER {\n uint8_t Name[16]; /* File member name - `/' terminated. */\n uint8_t Date[12]; /* File member date - decimal. */\n uint8_t UserID[6]; /* File member user id - decimal. */\n uint8_t GroupID[6]; /* File member group id - decimal. */\n uint8_t Mode[8]; /* File member mode - octal. */\n uint8_t Size[10]; /* File member size - decimal. */\n uint8_t EndHeader[2]; /* String to end header. */\n} IMAGE_ARCHIVE_MEMBER_HEADER, *PIMAGE_ARCHIVE_MEMBER_HEADER;\n\n#define IMAGE_SIZEOF_ARCHIVE_MEMBER_HDR 60\n\n/*\n * DLL support.\n */\n\n/*\n * Export Format\n */\n\ntypedef struct _IMAGE_EXPORT_DIRECTORY {\n uint32_t Characteristics;\n uint32_t TimeDateStamp;\n uint16_t MajorVersion;\n uint16_t MinorVersion;\n uint32_t Name;\n uint32_t Base;\n uint32_t NumberOfFunctions;\n uint32_t NumberOfNames;\n uint32_t **AddressOfFunctions;\n uint32_t **AddressOfNames;\n uint16_t **AddressOfNameOrdinals;\n} IMAGE_EXPORT_DIRECTORY, *PIMAGE_EXPORT_DIRECTORY;\n\n/*\n * Import Format\n */\n\ntypedef struct _IMAGE_IMPORT_BY_NAME {\n uint16_t Hint;\n uint8_t Name[1];\n} IMAGE_IMPORT_BY_NAME, *PIMAGE_IMPORT_BY_NAME;\n\ntypedef struct _IMAGE_THUNK_DATA {\n union {\n uint32_t *Function;\n uint32_t Ordinal;\n PIMAGE_IMPORT_BY_NAME AddressOfData;\n } u1;\n} IMAGE_THUNK_DATA, *PIMAGE_THUNK_DATA;\n\n#define IMAGE_ORDINAL_FLAG 0x80000000\n#define IMAGE_SNAP_BY_ORDINAL(Ordinal) ((Ordinal & IMAGE_ORDINAL_FLAG) != 0)\n#define IMAGE_ORDINAL(Ordinal) (Ordinal & 0xffff)\n\ntypedef struct _IMAGE_IMPORT_DESCRIPTOR {\n uint32_t Characteristics;\n uint32_t TimeDateStamp;\n uint32_t ForwarderChain;\n uint32_t Name;\n PIMAGE_THUNK_DATA FirstThunk;\n} IMAGE_IMPORT_DESCRIPTOR, *PIMAGE_IMPORT_DESCRIPTOR;\n\n/*\n * Thread Local Storage\n */\n\ntypedef void (*PIMAGE_TLS_CALLBACK) (\n void* DllHandle,\n uint32_t Reason,\n void* Reserved\n );\n\ntypedef struct _IMAGE_TLS_DIRECTORY {\n uint32_t StartAddressOfRawData;\n uint32_t EndAddressOfRawData;\n uint32_t *AddressOfIndex;\n PIMAGE_TLS_CALLBACK *AddressOfCallBacks;\n uint32_t SizeOfZeroFill;\n uint32_t Characteristics;\n} IMAGE_TLS_DIRECTORY, *PIMAGE_TLS_DIRECTORY;\n\n\n/*\n * Resource Format.\n */\n\n/*\n * Resource directory consists of two counts, following by a variable length\n * array of directory entries. The first count is the number of entries at\n * beginning of the array that have actual names associated with each entry.\n * The entries are in ascending order, case insensitive strings. The second\n * count is the number of entries that immediately follow the named entries.\n * This second count identifies the number of entries that have 31-bit integer\n * Ids as their name. These entries are also sorted in ascending order.\n *\n * This structure allows fast lookup by either name or number, but for any\n * given resource entry only one form of lookup is supported, not both.\n * This is consistant with the syntax of the .RC file and the .RES file.\n */\n\ntypedef struct _IMAGE_RESOURCE_DIRECTORY {\n uint32_t Characteristics;\n uint32_t TimeDateStamp;\n uint16_t MajorVersion;\n uint16_t MinorVersion;\n uint16_t NumberOfNamedEntries;\n uint16_t NumberOfIdEntries;\n/* IMAGE_RESOURCE_DIRECTORY_ENTRY DirectoryEntries[]; */\n} IMAGE_RESOURCE_DIRECTORY, *PIMAGE_RESOURCE_DIRECTORY;\n\n#define IMAGE_RESOURCE_NAME_IS_STRING 0x80000000\n#define IMAGE_RESOURCE_DATA_IS_DIRECTORY 0x80000000\n\n/*\n * Each directory contains the 32-bit Name of the entry and an offset,\n * relative to the beginning of the resource directory of the data associated\n * with this directory entry. If the name of the entry is an actual text\n * string instead of an integer Id, then the high order bit of the name field\n * is set to one and the low order 31-bits are an offset, relative to the\n * beginning of the resource directory of the string, which is of type\n * IMAGE_RESOURCE_DIRECTORY_STRING. Otherwise the high bit is clear and the\n * low-order 31-bits are the integer Id that identify this resource directory\n * entry. If the directory entry is yet another resource directory (i.e. a\n * subdirectory), then the high order bit of the offset field will be\n * set to indicate this. Otherwise the high bit is clear and the offset\n * field points to a resource data entry.\n */\n\ntypedef struct _IMAGE_RESOURCE_DIRECTORY_ENTRY {\n uint32_t Name;\n uint32_t OffsetToData;\n} IMAGE_RESOURCE_DIRECTORY_ENTRY, *PIMAGE_RESOURCE_DIRECTORY_ENTRY;\n\n/*\n * For resource directory entries that have actual string names, the Name\n * field of the directory entry points to an object of the following type.\n * All of these string objects are stored together after the last resource\n * directory entry and before the first resource data object. This minimizes\n * the impact of these variable length objects on the alignment of the fixed\n * size directory entry objects.\n */\n\ntypedef struct _IMAGE_RESOURCE_DIRECTORY_STRING {\n uint16_t Length;\n char NameString[ 1 ];\n} IMAGE_RESOURCE_DIRECTORY_STRING, *PIMAGE_RESOURCE_DIRECTORY_STRING;\n\n\ntypedef struct _IMAGE_RESOURCE_DIR_STRING_U {\n uint16_t Length;\n wchar_t NameString[ 1 ];\n} IMAGE_RESOURCE_DIR_STRING_U, *PIMAGE_RESOURCE_DIR_STRING_U;\n\n/*\n * Each resource data entry describes a leaf node in the resource directory\n * tree. It contains an offset, relative to the beginning of the resource\n * directory of the data for the resource, a size field that gives the number\n * of bytes of data at that offset, a CodePage that should be used when\n * decoding code point values within the resource data. Typically for new\n * applications the code page would be the unicode code page.\n */\n\ntypedef struct _IMAGE_RESOURCE_DATA_ENTRY {\n uint32_t OffsetToData;\n uint32_t Size;\n uint32_t CodePage;\n uint32_t Reserved;\n} IMAGE_RESOURCE_DATA_ENTRY, *PIMAGE_RESOURCE_DATA_ENTRY;\n\n/*\n * Load Configuration Directory Entry\n */\n\ntypedef struct _IMAGE_LOAD_CONFIG_DIRECTORY {\n uint32_t Characteristics;\n uint32_t TimeDateStamp;\n uint16_t MajorVersion;\n uint16_t MinorVersion;\n uint32_t GlobalFlagsClear;\n uint32_t GlobalFlagsSet;\n uint32_t CriticalSectionDefaultTimeout;\n uint32_t DeCommitFreeBlockThreshold;\n uint32_t DeCommitTotalFreeThreshold;\n uint32_t Reserved[ 8 ];\n} IMAGE_LOAD_CONFIG_DIRECTORY, *PIMAGE_LOAD_CONFIG_DIRECTORY;\n\n\n/*\n * Function table entry format for MIPS/ALPHA images. Function table is\n * pointed to by the IMAGE_DIRECTORY_ENTRY_EXCEPTION directory entry.\n * This definition duplicates ones in ntmips.h and ntalpha.h for use\n * by portable image file mungers.\n */\n\ntypedef struct _IMAGE_RUNTIME_FUNCTION_ENTRY {\n uint32_t BeginAddress;\n uint32_t EndAddress;\n void* ExceptionHandler;\n void* HandlerData;\n uint32_t PrologEndAddress;\n} IMAGE_RUNTIME_FUNCTION_ENTRY, *PIMAGE_RUNTIME_FUNCTION_ENTRY;\n\n/*\n * Debug Format\n */\n\ntypedef struct _IMAGE_DEBUG_DIRECTORY {\n uint32_t Characteristics;\n uint32_t TimeDateStamp;\n uint16_t MajorVersion;\n uint16_t MinorVersion;\n uint32_t Type;\n uint32_t SizeOfData;\n uint32_t AddressOfRawData;\n uint32_t PointerToRawData;\n} IMAGE_DEBUG_DIRECTORY, *PIMAGE_DEBUG_DIRECTORY;\n\n#define IMAGE_DEBUG_TYPE_UNKNOWN 0\n#define IMAGE_DEBUG_TYPE_COFF 1\n#define IMAGE_DEBUG_TYPE_CODEVIEW 2\n#define IMAGE_DEBUG_TYPE_FPO 3\n#define IMAGE_DEBUG_TYPE_MISC 4\n#define IMAGE_DEBUG_TYPE_EXCEPTION 5\n#define IMAGE_DEBUG_TYPE_FIXUP 6\n#define IMAGE_DEBUG_TYPE_RESERVED6 7\n#define IMAGE_DEBUG_TYPE_RESERVED7 8\n\ntypedef struct _IMAGE_COFF_SYMBOLS_HEADER {\n uint32_t NumberOfSymbols;\n uint32_t LvaToFirstSymbol;\n uint32_t NumberOfLinenumbers;\n uint32_t LvaToFirstLinenumber;\n uint32_t RvaToFirstByteOfCode;\n uint32_t RvaToLastByteOfCode;\n uint32_t RvaToFirstByteOfData;\n uint32_t RvaToLastByteOfData;\n} IMAGE_COFF_SYMBOLS_HEADER, *PIMAGE_COFF_SYMBOLS_HEADER;\n\n#define FRAME_FPO 0\n#define FRAME_TRAP 1\n#define FRAME_TSS 2\n\n#ifdef _MSC_VER\n#pragma warning(disable:4214)\n#endif\ntypedef struct _FPO_DATA {\n uint32_t ulOffStart; /* offset 1st byte of function code */\n uint32_t cbProcSize; /* # bytes in function */\n uint32_t cdwLocals; /* # bytes in locals/4 */\n uint16_t cdwParams; /* # bytes in params/4 */\n uint16_t cbProlog : 8; /* # bytes in prolog */\n uint16_t cbRegs : 3; /* # regs saved */\n uint16_t fHasSEH : 1; /* TRUE if SEH in func */\n uint16_t fUseBP : 1; /* TRUE if EBP has been allocated */\n uint16_t reserved : 1; /* reserved for future use */\n uint16_t cbFrame : 2; /* frame type */\n} FPO_DATA, *PFPO_DATA;\n#define SIZEOF_RFPO_DATA 16\n#ifdef _MSC_VER\n#pragma warning(default:4214)\n#endif\n\n#define IMAGE_DEBUG_MISC_EXENAME 1\n\ntypedef struct _IMAGE_DEBUG_MISC {\n uint32_t DataType; /* type of misc data, see defines */\n uint32_t Length; /* total length of record, rounded to four */\n /* byte multiple. */\n uint8_t Unicode; /* TRUE if data is unicode string */\n uint8_t Reserved[ 3 ];\n uint8_t Data[ 1 ]; /* Actual data */\n} IMAGE_DEBUG_MISC, *PIMAGE_DEBUG_MISC;\n\n\n/*\n * Debugging information can be stripped from an image file and placed\n * in a separate .DBG file, whose file name part is the same as the\n * image file name part (e.g. symbols for CMD.EXE could be stripped\n * and placed in CMD.DBG). This is indicated by the IMAGE_FILE_DEBUG_STRIPPED\n * flag in the Characteristics field of the file header. The beginning of\n * the .DBG file contains the following structure which captures certain\n * information from the image file. This allows a debug to proceed even if\n * the original image file is not accessable. This header is followed by\n * zero of more IMAGE_SECTION_HEADER structures, followed by zero or more\n * IMAGE_DEBUG_DIRECTORY structures. The latter structures and those in\n * the image file contain file offsets relative to the beginning of the\n * .DBG file.\n *\n * If symbols have been stripped from an image, the IMAGE_DEBUG_MISC structure\n * is left in the image file, but not mapped. This allows a debugger to\n * compute the name of the .DBG file, from the name of the image in the\n * IMAGE_DEBUG_MISC structure.\n */\n\ntypedef struct _IMAGE_SEPARATE_DEBUG_HEADER {\n uint16_t Signature;\n uint16_t Flags;\n uint16_t Machine;\n uint16_t Characteristics;\n uint32_t TimeDateStamp;\n uint32_t CheckSum;\n uint32_t ImageBase;\n uint32_t SizeOfImage;\n uint32_t NumberOfSections;\n uint32_t ExportedNamesSize;\n uint32_t DebugDirectorySize;\n uint32_t Reserved[ 3 ];\n} IMAGE_SEPARATE_DEBUG_HEADER, *PIMAGE_SEPARATE_DEBUG_HEADER;\n\n#define IMAGE_SEPARATE_DEBUG_SIGNATURE 0x4944\n\n#endif /* IMAGE_NT_SIGNATURE */\n\n\n#ifdef USES_COMPLUS20\n/*\n * COM+ 2.0 header structure.\n */\ntypedef struct IMAGE_COR20_HEADER\n{\n /* Header versioning */\n uint32_t cb;\n uint16_t MajorRuntimeVersion;\n uint16_t MinorRuntimeVersion;\n\n /* Symbol table and startup information */\n IMAGE_DATA_DIRECTORY MetaData;\n uint32_t Flags;\n uint32_t EntryPointToken;\n\n /* Binding information */\n IMAGE_DATA_DIRECTORY Resources;\n IMAGE_DATA_DIRECTORY StrongNameSignature;\n\n /* Regular fixup and binding information */\n IMAGE_DATA_DIRECTORY CodeManagerTable;\n IMAGE_DATA_DIRECTORY VTableFixups;\n IMAGE_DATA_DIRECTORY ExportAddressTableJumps;\n\n /* Managed Native Code */\n IMAGE_DATA_DIRECTORY EEInfoTable;\n IMAGE_DATA_DIRECTORY HelperTable;\n IMAGE_DATA_DIRECTORY DynamicInfo;\n IMAGE_DATA_DIRECTORY DelayLoadInfo;\n IMAGE_DATA_DIRECTORY ModuleImage;\n IMAGE_DATA_DIRECTORY ExternalFixups;\n IMAGE_DATA_DIRECTORY RidMap;\n IMAGE_DATA_DIRECTORY DebugMap;\n\n /*@Todo: this is obsolete, being replaced by pdata in the PE header.*/\n IMAGE_DATA_DIRECTORY IPMap;\n\n} IMAGE_COR20_HEADER;\n\n#endif /* USES_COMPLUS20 */\n\n#ifdef _MSC_VER\n#pragma pack ()\n#endif\n\n/*\n * End Image Format\n */\n" }, { "path": "third_party/renderdoc/renderdoc_app.h", "content": "/******************************************************************************\n * The MIT License (MIT)\n * \n * Copyright (c) 2015-2016 Baldur Karlsson\n * \n * Permission is hereby granted, free of charge, to any person obtaining a copy\n * of this software and associated documentation files (the \"Software\"), to deal\n * in the Software without restriction, including without limitation the rights\n * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell\n * copies of the Software, and to permit persons to whom the Software is\n * furnished to do so, subject to the following conditions:\n * \n * The above copyright notice and this permission notice shall be included in\n * all copies or substantial portions of the Software.\n * \n * THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\n * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\n * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN\n * THE SOFTWARE.\n ******************************************************************************/\n\n#pragma once\n\n#if !defined(RENDERDOC_NO_STDINT)\n#include \n#endif\n\n#if defined(WIN32)\n\t#define RENDERDOC_CC __cdecl\n#elif defined(__linux__)\n\t#define RENDERDOC_CC\n#else\n\t#error \"Unknown platform\"\n#endif\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\t\n//////////////////////////////////////////////////////////////////////////////////////////////////\n// RenderDoc capture options\n// \n\ntypedef enum\n{\n\t// Allow the application to enable vsync\n\t//\n\t// Default - enabled\n\t//\n\t// 1 - The application can enable or disable vsync at will\n\t// 0 - vsync is force disabled\n\teRENDERDOC_Option_AllowVSync = 0,\n\t\n\t// Allow the application to enable fullscreen\n\t//\n\t// Default - enabled\n\t//\n\t// 1 - The application can enable or disable fullscreen at will\n\t// 0 - fullscreen is force disabled\n\teRENDERDOC_Option_AllowFullscreen = 1,\n\n\t// Record API debugging events and messages \n\t//\n\t// Default - disabled\n\t//\n\t// 1 - Enable built-in API debugging features and records the results into\n\t// the capture logfile, which is matched up with events on replay\n\t// 0 - no API debugging is forcibly enabled\n\teRENDERDOC_Option_DebugDeviceMode = 2,\n\t\n\t// Capture CPU callstacks for API events \n\t//\n\t// Default - disabled\n\t//\n\t// 1 - Enables capturing of callstacks\n\t// 0 - no callstacks are captured\n\teRENDERDOC_Option_CaptureCallstacks = 3,\n\t\n\t// When capturing CPU callstacks, only capture them from drawcalls.\n\t// This option does nothing without the above option being enabled\n\t//\n\t// Default - disabled\n\t//\n\t// 1 - Only captures callstacks for drawcall type API events.\n\t// Ignored if CaptureCallstacks is disabled\n\t// 0 - Callstacks, if enabled, are captured for every event.\n\teRENDERDOC_Option_CaptureCallstacksOnlyDraws = 4,\n\n\t// Specify a delay in seconds to wait for a debugger to attach, after\n\t// creating or injecting into a process, before continuing to allow it to run.\n\t//\n\t// 0 indicates no delay, and the process will run immediately after injection\n\t//\n\t// Default - 0 seconds\n\t//\n\teRENDERDOC_Option_DelayForDebugger = 5,\n\t\n\t// Verify any writes to mapped buffers, by checking the memory after the\n\t// bounds of the returned pointer to detect any modification.\n\t//\n\t// Default - disabled\n\t//\n\t// 1 - Verify any writes to mapped buffers\n\t// 0 - No verification is performed, and overwriting bounds may cause\n\t// crashes or corruption in RenderDoc\n\teRENDERDOC_Option_VerifyMapWrites = 6,\n\t\n\t// Hooks any system API calls that create child processes, and injects\n\t// RenderDoc into them recursively with the same options.\n\t//\n\t// Default - disabled\n\t//\n\t// 1 - Hooks into spawned child processes\n\t// 0 - Child processes are not hooked by RenderDoc\n\teRENDERDOC_Option_HookIntoChildren = 7,\n\n\t// By default RenderDoc only includes resources in the final logfile necessary\n\t// for that frame, this allows you to override that behaviour.\n\t//\n\t// Default - disabled\n\t//\n\t// 1 - all live resources at the time of capture are included in the log\n\t// and available for inspection\n\t// 0 - only the resources referenced by the captured frame are included\n\teRENDERDOC_Option_RefAllResources = 8,\n\n\t// By default RenderDoc skips saving initial states for resources where the\n\t// previous contents don't appear to be used, assuming that writes before\n\t// reads indicate previous contents aren't used.\n\t//\n\t// Default - disabled\n\t//\n\t// 1 - initial contents at the start of each captured frame are saved, even if\n\t// they are later overwritten or cleared before being used.\n\t// 0 - unless a read is detected, initial contents will not be saved and will\n\t// appear as black or empty data.\n\teRENDERDOC_Option_SaveAllInitials = 9,\n\n\t// In APIs that allow for the recording of command lists to be replayed later,\n\t// RenderDoc may choose to not capture command lists before a frame capture is\n\t// triggered, to reduce overheads. This means any command lists recorded once\n\t// and replayed many times will not be available and may cause a failure to\n\t// capture.\n\t//\n\t// Note this is only true for APIs where multithreading is difficult or\n\t// discouraged. Newer APIs like Vulkan and D3D12 will ignore this option\n\t// and always capture all command lists since the API is heavily oriented\n\t// around it and the overheads have been reduced by API design.\n\t//\n\t// 1 - All command lists are captured from the start of the application\n\t// 0 - Command lists are only captured if their recording begins during\n\t// the period when a frame capture is in progress.\n\teRENDERDOC_Option_CaptureAllCmdLists = 10,\n\n\t// Mute API debugging output when the debug device mode option is enabled \n\t//\n\t// Default - enabled\n\t//\n\t// 1 - Mute any API debug messages from being displayed or passed through\n\t// 0 - API debugging is displayed as normal\n\teRENDERDOC_Option_DebugOutputMute = 11,\n\n} RENDERDOC_CaptureOption;\n\n// Sets an option that controls how RenderDoc behaves on capture.\n//\n// Returns 1 if the option and value are valid\n// Returns 0 if either is invalid and the option is unchanged\ntypedef int (RENDERDOC_CC *pRENDERDOC_SetCaptureOptionU32)(RENDERDOC_CaptureOption opt, uint32_t val);\ntypedef int (RENDERDOC_CC *pRENDERDOC_SetCaptureOptionF32)(RENDERDOC_CaptureOption opt, float val);\n\n// Gets the current value of an option as a uint32_t\n//\n// If the option is invalid, 0xffffffff is returned\ntypedef uint32_t (RENDERDOC_CC *pRENDERDOC_GetCaptureOptionU32)(RENDERDOC_CaptureOption opt);\n\n// Gets the current value of an option as a float\n//\n// If the option is invalid, -FLT_MAX is returned\ntypedef float (RENDERDOC_CC *pRENDERDOC_GetCaptureOptionF32)(RENDERDOC_CaptureOption opt);\n\ntypedef enum\n{\n\t// '0' - '9' matches ASCII values\n\teRENDERDOC_Key_0 = 0x30,\n\teRENDERDOC_Key_1 = 0x31,\n\teRENDERDOC_Key_2 = 0x32,\n\teRENDERDOC_Key_3 = 0x33,\n\teRENDERDOC_Key_4 = 0x34,\n\teRENDERDOC_Key_5 = 0x35,\n\teRENDERDOC_Key_6 = 0x36,\n\teRENDERDOC_Key_7 = 0x37,\n\teRENDERDOC_Key_8 = 0x38,\n\teRENDERDOC_Key_9 = 0x39,\n\n\t// 'A' - 'Z' matches ASCII values\n\teRENDERDOC_Key_A = 0x41,\n\teRENDERDOC_Key_B = 0x42,\n\teRENDERDOC_Key_C = 0x43,\n\teRENDERDOC_Key_D = 0x44,\n\teRENDERDOC_Key_E = 0x45,\n\teRENDERDOC_Key_F = 0x46,\n\teRENDERDOC_Key_G = 0x47,\n\teRENDERDOC_Key_H = 0x48,\n\teRENDERDOC_Key_I = 0x49,\n\teRENDERDOC_Key_J = 0x4A,\n\teRENDERDOC_Key_K = 0x4B,\n\teRENDERDOC_Key_L = 0x4C,\n\teRENDERDOC_Key_M = 0x4D,\n\teRENDERDOC_Key_N = 0x4E,\n\teRENDERDOC_Key_O = 0x4F,\n\teRENDERDOC_Key_P = 0x50,\n\teRENDERDOC_Key_Q = 0x51,\n\teRENDERDOC_Key_R = 0x52,\n\teRENDERDOC_Key_S = 0x53,\n\teRENDERDOC_Key_T = 0x54,\n\teRENDERDOC_Key_U = 0x55,\n\teRENDERDOC_Key_V = 0x56,\n\teRENDERDOC_Key_W = 0x57,\n\teRENDERDOC_Key_X = 0x58,\n\teRENDERDOC_Key_Y = 0x59,\n\teRENDERDOC_Key_Z = 0x5A,\n\n\t// leave the rest of the ASCII range free\n\t// in case we want to use it later\n\teRENDERDOC_Key_NonPrintable = 0x100,\n\n\teRENDERDOC_Key_Divide,\n\teRENDERDOC_Key_Multiply,\n\teRENDERDOC_Key_Subtract,\n\teRENDERDOC_Key_Plus,\n\n\teRENDERDOC_Key_F1,\n\teRENDERDOC_Key_F2,\n\teRENDERDOC_Key_F3,\n\teRENDERDOC_Key_F4,\n\teRENDERDOC_Key_F5,\n\teRENDERDOC_Key_F6,\n\teRENDERDOC_Key_F7,\n\teRENDERDOC_Key_F8,\n\teRENDERDOC_Key_F9,\n\teRENDERDOC_Key_F10,\n\teRENDERDOC_Key_F11,\n\teRENDERDOC_Key_F12,\n\n\teRENDERDOC_Key_Home,\n\teRENDERDOC_Key_End,\n\teRENDERDOC_Key_Insert,\n\teRENDERDOC_Key_Delete,\n\teRENDERDOC_Key_PageUp,\n\teRENDERDOC_Key_PageDn,\n\n\teRENDERDOC_Key_Backspace,\n\teRENDERDOC_Key_Tab,\n\teRENDERDOC_Key_PrtScrn,\n\teRENDERDOC_Key_Pause,\n\n\teRENDERDOC_Key_Max,\n} RENDERDOC_InputButton;\n\n// Sets which key or keys can be used to toggle focus between multiple windows\n//\n// If keys is NULL or num is 0, toggle keys will be disabled\ntypedef void (RENDERDOC_CC *pRENDERDOC_SetFocusToggleKeys)(RENDERDOC_InputButton *keys, int num);\n\n// Sets which key or keys can be used to capture the next frame\n//\n// If keys is NULL or num is 0, captures keys will be disabled\ntypedef void (RENDERDOC_CC *pRENDERDOC_SetCaptureKeys)(RENDERDOC_InputButton *keys, int num);\n\ntypedef enum\n{\n\t// This single bit controls whether the overlay is enabled or disabled globally\n\teRENDERDOC_Overlay_Enabled = 0x1,\n\n\t// Show the average framerate over several seconds as well as min/max\n\teRENDERDOC_Overlay_FrameRate = 0x2,\n\n\t// Show the current frame number\n\teRENDERDOC_Overlay_FrameNumber = 0x4,\n\n\t// Show a list of recent captures, and how many captures have been made\n\teRENDERDOC_Overlay_CaptureList = 0x8,\n\n\t// Default values for the overlay mask\n\teRENDERDOC_Overlay_Default =\n\t (eRENDERDOC_Overlay_Enabled|\n\t eRENDERDOC_Overlay_FrameRate|\n\t eRENDERDOC_Overlay_FrameNumber|\n\t eRENDERDOC_Overlay_CaptureList),\n\n\t// Enable all bits\n\teRENDERDOC_Overlay_All = ~0U,\n\n\t// Disable all bits\n\teRENDERDOC_Overlay_None = 0,\n} RENDERDOC_OverlayBits;\n\n// returns the overlay bits that have been set\ntypedef uint32_t (RENDERDOC_CC *pRENDERDOC_GetOverlayBits)();\n// sets the overlay bits with an and & or mask\ntypedef void (RENDERDOC_CC *pRENDERDOC_MaskOverlayBits)(uint32_t And, uint32_t Or);\n\n// this function will attempt to shut down RenderDoc.\n//\n// Note: that this will only work correctly if done immediately after\n// the dll is loaded, before any API work happens. RenderDoc will remove its\n// injected hooks and shut down. Behaviour is undefined if this is called\n// after any API functions have been called.\ntypedef void (RENDERDOC_CC *pRENDERDOC_Shutdown)();\n\n// This function will unload RenderDoc's crash handler.\n//\n// If you use your own crash handler and don't want RenderDoc's handler to\n// intercede, you can call this function to unload it and any unhandled\n// exceptions will pass to the next handler.\ntypedef void (RENDERDOC_CC *pRENDERDOC_UnloadCrashHandler)();\n\n// Sets the logfile path template\n//\n// logfile is a UTF-8 string that gives a template for how captures will be named\n// and where they will be saved.\n//\n// Any extension is stripped off the path, and captures are saved in the directory\n// specified, and named with the filename and the frame number appended. If the\n// directory does not exist it will be created, including any parent directories.\n//\n// If pathtemplate is NULL, the template will remain unchanged\n//\n// Example:\n//\n// SetLogFilePathTemplate(\"my_captures/example\");\n//\n// Capture #1 -> my_captures/example_frame123.rdc\n// Capture #2 -> my_captures/example_frame456.rdc\ntypedef void (RENDERDOC_CC *pRENDERDOC_SetLogFilePathTemplate)(const char *pathtemplate);\n\n// returns the current logfile template, see SetLogFileTemplate above, as a UTF-8 string\ntypedef const char* (RENDERDOC_CC *pRENDERDOC_GetLogFilePathTemplate)();\n\n// returns the number of captures that have been made\ntypedef uint32_t (RENDERDOC_CC *pRENDERDOC_GetNumCaptures)();\n\n// This function returns the details of a capture, by index. New captures are added\n// to the end of the list.\n//\n// logfile will be filled with the absolute path to the capture file, as a UTF-8 string\n// pathlength will be written with the length in bytes of the logfile string\n// timestamp will be written with the time of the capture, in seconds since the Unix epoch\n//\n// Any of the parameters can be NULL and they'll be skipped.\n//\n// The function will return 1 if the capture index is valid, or 0 if the index is invalid\n// If the index is invalid, the values will be unchanged\n//\n// Note: when captures are deleted in the UI they will remain in this list, so the\n// logfile path may not exist anymore.\ntypedef uint32_t (RENDERDOC_CC *pRENDERDOC_GetCapture)(uint32_t idx, char *logfile, uint32_t *pathlength, uint64_t *timestamp);\n\n// capture the next frame on whichever window and API is currently considered active\ntypedef void (RENDERDOC_CC *pRENDERDOC_TriggerCapture)();\n\n// returns 1 if the RenderDoc UI is connected to this application, 0 otherwise\ntypedef uint32_t (RENDERDOC_CC *pRENDERDOC_IsRemoteAccessConnected)();\n\n// This function will launch the Replay UI associated with the RenderDoc library injected\n// into the running application.\n//\n// if connectRemoteAccess is 1, the Replay UI will be launched with a command line parameter\n// to connect to this application\n// cmdline is the rest of the command line, as a UTF-8 string. E.g. a captures to open\n// if cmdline is NULL, the command line will be empty.\n//\n// returns the PID of the replay UI if successful, 0 if not successful.\ntypedef uint32_t (RENDERDOC_CC *pRENDERDOC_LaunchReplayUI)(uint32_t connectRemoteAccess, const char *cmdline);\n\n// RenderDoc can return a higher version than requested if it's backwards compatible,\n// this function returns the actual version returned. If a parameter is NULL, it will be\n// ignored and the others will be filled out.\ntypedef void (RENDERDOC_CC *pRENDERDOC_GetAPIVersion)(int *major, int *minor, int *patch);\n\n//////////////////////////////////////////////////////////////////////////\n// Capturing functions\n//\n\n// A device pointer is a pointer to the API's root handle.\n//\n// This would be an ID3D11Device, HGLRC/GLXContext, ID3D12Device, etc\ntypedef void* RENDERDOC_DevicePointer;\n\n// A window handle is the OS's native window handle\n//\n// This would be an HWND, GLXDrawable, etc\ntypedef void* RENDERDOC_WindowHandle;\n\n// This sets the RenderDoc in-app overlay in the API/window pair as 'active' and it will\n// respond to keypresses. Neither parameter can be NULL\ntypedef void (RENDERDOC_CC *pRENDERDOC_SetActiveWindow)(RENDERDOC_DevicePointer device, RENDERDOC_WindowHandle wndHandle);\n\n// When choosing either a device pointer or a window handle to capture, you can pass NULL.\n// Passing NULL specifies a 'wildcard' match against anything. This allows you to specify\n// any API rendering to a specific window, or a specific API instance rendering to any window,\n// or in the simplest case of one window and one API, you can just pass NULL for both.\n//\n// In either case, if there are two or more possible matching (device,window) pairs it\n// is undefined which one will be captured.\n//\n// Note: for headless rendering you can pass NULL for the window handle and either specify\n// a device pointer or leave it NULL as above.\n\n// Immediately starts capturing API calls on the specified device pointer and window handle.\n//\n// If there is no matching thing to capture (e.g. no supported API has been initialised),\n// this will do nothing.\n//\n// The results are undefined (including crashes) if two captures are started overlapping,\n// even on separate devices and/oror windows.\ntypedef void (RENDERDOC_CC *pRENDERDOC_StartFrameCapture)(RENDERDOC_DevicePointer device, RENDERDOC_WindowHandle wndHandle);\n\n// Returns whether or not a frame capture is currently ongoing anywhere.\n//\n// This will return 1 if a capture is ongoing, and 0 if there is no capture running\ntypedef uint32_t (RENDERDOC_CC *pRENDERDOC_IsFrameCapturing)();\n\n// Ends capturing immediately.\n//\n// This will return 1 if the capture succeeded, and 0 if there was an error capturing.\ntypedef uint32_t (RENDERDOC_CC *pRENDERDOC_EndFrameCapture)(RENDERDOC_DevicePointer device, RENDERDOC_WindowHandle wndHandle);\n\n//////////////////////////////////////////////////////////////////////////////////////////////////\n// RenderDoc API versions\n// \n\n// RenderDoc uses semantic versioning (http://semver.org/).\n// \n// MAJOR version is incremented when incompatible API changes happen.\n// MINOR version is incremented when functionality is added in a backwards-compatible manner.\n// PATCH version is incremented when backwards-compatible bug fixes happen.\n//\n// Note that this means the API returned can be higher than the one you might have requested.\n// e.g. if you are running against a newer RenderDoc that supports 1.0.1, it will be returned\n// instead of 1.0.0. You can check this with the GetAPIVersion entry point\ntypedef enum\n{\n\teRENDERDOC_API_Version_1_0_0 = 10000, // RENDERDOC_API_1_0_0 = 1 00 00\n\teRENDERDOC_API_Version_1_0_1 = 10001, // RENDERDOC_API_1_0_1 = 1 00 01\n} RENDERDOC_Version;\n\n// API version changelog:\n//\n// 1.0.0 - initial release\n// 1.0.1 - Bugfix: IsFrameCapturing() was returning false for captures that were triggered\n// by keypress or TriggerCapture, instead of Start/EndFrameCapture.\n\n// eRENDERDOC_API_Version_1_0_1\ntypedef struct\n{\n\tpRENDERDOC_GetAPIVersion GetAPIVersion;\n\n\tpRENDERDOC_SetCaptureOptionU32 SetCaptureOptionU32;\n\tpRENDERDOC_SetCaptureOptionF32 SetCaptureOptionF32;\n\n\tpRENDERDOC_GetCaptureOptionU32 GetCaptureOptionU32;\n\tpRENDERDOC_GetCaptureOptionF32 GetCaptureOptionF32;\n\n\tpRENDERDOC_SetFocusToggleKeys SetFocusToggleKeys;\n\tpRENDERDOC_SetCaptureKeys SetCaptureKeys;\n\n\tpRENDERDOC_GetOverlayBits GetOverlayBits;\n\tpRENDERDOC_MaskOverlayBits MaskOverlayBits;\n\n\tpRENDERDOC_Shutdown Shutdown;\n\tpRENDERDOC_UnloadCrashHandler UnloadCrashHandler;\n\n\tpRENDERDOC_SetLogFilePathTemplate SetLogFilePathTemplate;\n\tpRENDERDOC_GetLogFilePathTemplate GetLogFilePathTemplate;\n\n\tpRENDERDOC_GetNumCaptures GetNumCaptures;\n\tpRENDERDOC_GetCapture GetCapture;\n\n\tpRENDERDOC_TriggerCapture TriggerCapture;\n\n\tpRENDERDOC_IsRemoteAccessConnected IsRemoteAccessConnected;\n\tpRENDERDOC_LaunchReplayUI LaunchReplayUI;\n\n\tpRENDERDOC_SetActiveWindow SetActiveWindow;\n\n\tpRENDERDOC_StartFrameCapture StartFrameCapture;\n\tpRENDERDOC_IsFrameCapturing IsFrameCapturing;\n\tpRENDERDOC_EndFrameCapture EndFrameCapture;\n} RENDERDOC_API_1_0_1;\n\ntypedef RENDERDOC_API_1_0_1 RENDERDOC_API_1_0_0;\n\n//////////////////////////////////////////////////////////////////////////////////////////////////\n// RenderDoc API entry point\n// \n// This entry point can be obtained via GetProcAddress/dlsym if RenderDoc is available.\n// \n// The name is the same as the typedef - \"RENDERDOC_GetAPI\"\n//\n// This function is not thread safe, and should not be called on multiple threads at once.\n// Ideally, call this once as early as possible in your application's startup, before doing\n// any API work, since some configuration functionality etc has to be done also before\n// initialising any APIs.\n//\n// Parameters:\n// version is a single value from the RENDERDOC_Version above.\n//\n// outAPIPointers will be filled out with a pointer to the corresponding struct of function\n// pointers.\n//\n// Returns:\n// 1 - if the outAPIPointers has been filled with a pointer to the API struct requested\n// 0 - if the requested version is not supported or the arguments are invalid.\n//\ntypedef int (RENDERDOC_CC *pRENDERDOC_GetAPI)(RENDERDOC_Version version, void **outAPIPointers);\n\n#ifdef __cplusplus\n} // extern \"C\"\n#endif\n" }, { "path": "third_party/snappy.lua", "content": "group(\"third_party\")\nproject(\"snappy\")\n uuid(\"bb143d61-3fd4-44c2-8b7e-04cc538ba2c7\")\n kind(\"StaticLib\")\n language(\"C++\")\n\n defines({\n \"_LIB\",\n })\n files({\n \"snappy/snappy-internal.h\",\n \"snappy/snappy-sinksource.cc\",\n \"snappy/snappy-sinksource.h\",\n \"snappy/snappy-stubs-internal.cc\",\n \"snappy/snappy-stubs-internal.h\",\n \"snappy/snappy-stubs-public.h\",\n \"snappy/snappy.cc\",\n \"snappy/snappy.h\",\n })\n\n filter(\"platforms:Windows\")\n warnings(\"Off\") -- Too many warnings.\n" }, { "path": "third_party/stb/stb_image.h", "content": "/* stb_image - v2.08 - public domain image loader - http://nothings.org/stb_image.h\n no warranty implied; use at your own risk\n\n Do this:\n #define STB_IMAGE_IMPLEMENTATION\n before you include this file in *one* C or C++ file to create the implementation.\n\n // i.e. it should look like this:\n #include ...\n #include ...\n #include ...\n #define STB_IMAGE_IMPLEMENTATION\n #include \"stb_image.h\"\n\n You can #define STBI_ASSERT(x) before the #include to avoid using assert.h.\n And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free\n\n\n QUICK NOTES:\n Primarily of interest to game developers and other people who can\n avoid problematic images and only need the trivial interface\n\n JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib)\n PNG 1/2/4/8-bit-per-channel (16 bpc not supported)\n\n TGA (not sure what subset, if a subset)\n BMP non-1bpp, non-RLE\n PSD (composited view only, no extra channels, 8/16 bit-per-channel)\n\n GIF (*comp always reports as 4-channel)\n HDR (radiance rgbE format)\n PIC (Softimage PIC)\n PNM (PPM and PGM binary only)\n\n Animated GIF still needs a proper API, but here's one way to do it:\n http://gist.github.com/urraka/685d9a6340b26b830d49\n\n - decode from memory or through FILE (define STBI_NO_STDIO to remove code)\n - decode from arbitrary I/O callbacks\n - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON)\n\n Full documentation under \"DOCUMENTATION\" below.\n\n\n Revision 2.00 release notes:\n\n - Progressive JPEG is now supported.\n\n - PPM and PGM binary formats are now supported, thanks to Ken Miller.\n\n - x86 platforms now make use of SSE2 SIMD instructions for\n JPEG decoding, and ARM platforms can use NEON SIMD if requested.\n This work was done by Fabian \"ryg\" Giesen. SSE2 is used by\n default, but NEON must be enabled explicitly; see docs.\n\n With other JPEG optimizations included in this version, we see\n 2x speedup on a JPEG on an x86 machine, and a 1.5x speedup\n on a JPEG on an ARM machine, relative to previous versions of this\n library. The same results will not obtain for all JPGs and for all\n x86/ARM machines. (Note that progressive JPEGs are significantly\n slower to decode than regular JPEGs.) This doesn't mean that this\n is the fastest JPEG decoder in the land; rather, it brings it\n closer to parity with standard libraries. If you want the fastest\n decode, look elsewhere. (See \"Philosophy\" section of docs below.)\n\n See final bullet items below for more info on SIMD.\n\n - Added STBI_MALLOC, STBI_REALLOC, and STBI_FREE macros for replacing\n the memory allocator. Unlike other STBI libraries, these macros don't\n support a context parameter, so if you need to pass a context in to\n the allocator, you'll have to store it in a global or a thread-local\n variable.\n\n - Split existing STBI_NO_HDR flag into two flags, STBI_NO_HDR and\n STBI_NO_LINEAR.\n STBI_NO_HDR: suppress implementation of .hdr reader format\n STBI_NO_LINEAR: suppress high-dynamic-range light-linear float API\n\n - You can suppress implementation of any of the decoders to reduce\n your code footprint by #defining one or more of the following\n symbols before creating the implementation.\n\n STBI_NO_JPEG\n STBI_NO_PNG\n STBI_NO_BMP\n STBI_NO_PSD\n STBI_NO_TGA\n STBI_NO_GIF\n STBI_NO_HDR\n STBI_NO_PIC\n STBI_NO_PNM (.ppm and .pgm)\n\n - You can request *only* certain decoders and suppress all other ones\n (this will be more forward-compatible, as addition of new decoders\n doesn't require you to disable them explicitly):\n\n STBI_ONLY_JPEG\n STBI_ONLY_PNG\n STBI_ONLY_BMP\n STBI_ONLY_PSD\n STBI_ONLY_TGA\n STBI_ONLY_GIF\n STBI_ONLY_HDR\n STBI_ONLY_PIC\n STBI_ONLY_PNM (.ppm and .pgm)\n\n Note that you can define multiples of these, and you will get all\n of them (\"only x\" and \"only y\" is interpreted to mean \"only x&y\").\n\n - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still\n want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB\n\n - Compilation of all SIMD code can be suppressed with\n #define STBI_NO_SIMD\n It should not be necessary to disable SIMD unless you have issues\n compiling (e.g. using an x86 compiler which doesn't support SSE\n intrinsics or that doesn't support the method used to detect\n SSE2 support at run-time), and even those can be reported as\n bugs so I can refine the built-in compile-time checking to be\n smarter.\n\n - The old STBI_SIMD system which allowed installing a user-defined\n IDCT etc. has been removed. If you need this, don't upgrade. My\n assumption is that almost nobody was doing this, and those who\n were will find the built-in SIMD more satisfactory anyway.\n\n - RGB values computed for JPEG images are slightly different from\n previous versions of stb_image. (This is due to using less\n integer precision in SIMD.) The C code has been adjusted so\n that the same RGB values will be computed regardless of whether\n SIMD support is available, so your app should always produce\n consistent results. But these results are slightly different from\n previous versions. (Specifically, about 3% of available YCbCr values\n will compute different RGB results from pre-1.49 versions by +-1;\n most of the deviating values are one smaller in the G channel.)\n\n - If you must produce consistent results with previous versions of\n stb_image, #define STBI_JPEG_OLD and you will get the same results\n you used to; however, you will not get the SIMD speedups for\n the YCbCr-to-RGB conversion step (although you should still see\n significant JPEG speedup from the other changes).\n\n Please note that STBI_JPEG_OLD is a temporary feature; it will be\n removed in future versions of the library. It is only intended for\n near-term back-compatibility use.\n\n\n Latest revision history:\n 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA\n 2.07 (2015-09-13) partial animated GIF support\n limited 16-bit PSD support\n minor bugs, code cleanup, and compiler warnings\n 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value\n 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning\n 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit\n 2.03 (2015-04-12) additional corruption checking\n stbi_set_flip_vertically_on_load\n fix NEON support; fix mingw support\n 2.02 (2015-01-19) fix incorrect assert, fix warning\n 2.01 (2015-01-17) fix various warnings\n 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG\n 2.00 (2014-12-25) optimize JPEG, including x86 SSE2 & ARM NEON SIMD\n progressive JPEG\n PGM/PPM support\n STBI_MALLOC,STBI_REALLOC,STBI_FREE\n STBI_NO_*, STBI_ONLY_*\n GIF bugfix\n 1.48 (2014-12-14) fix incorrectly-named assert()\n 1.47 (2014-12-14) 1/2/4-bit PNG support (both grayscale and paletted)\n optimize PNG\n fix bug in interlaced PNG with user-specified channel count\n\n See end of file for full revision history.\n\n\n ============================ Contributors =========================\n\n Image formats Bug fixes & warning fixes\n Sean Barrett (jpeg, png, bmp) Marc LeBlanc\n Nicolas Schulz (hdr, psd) Christpher Lloyd\n Jonathan Dummer (tga) Dave Moore\n Jean-Marc Lienher (gif) Won Chun\n Tom Seddon (pic) the Horde3D community\n Thatcher Ulrich (psd) Janez Zemva\n Ken Miller (pgm, ppm) Jonathan Blow\n urraka@github (animated gif) Laurent Gomila\n Aruelien Pocheville\n Ryamond Barbiero\n David Woo\n Extensions, features Martin Golini\n Jetro Lauha (stbi_info) Roy Eltham\n Martin \"SpartanJ\" Golini (stbi_info) Luke Graham\n James \"moose2000\" Brown (iPhone PNG) Thomas Ruf\n Ben \"Disch\" Wenger (io callbacks) John Bartholomew\n Omar Cornut (1/2/4-bit PNG) Ken Hamada\n Nicolas Guillemot (vertical flip) Cort Stratton\n Richard Mitton (16-bit PSD) Blazej Dariusz Roszkowski\n Thibault Reuille\n Paul Du Bois\n Guillaume George\n Jerry Jansson\n Hayaki Saito\n Johan Duparc\n Ronny Chevalier\n Optimizations & bugfixes Michal Cichon\n Fabian \"ryg\" Giesen Tero Hanninen\n Arseny Kapoulkine Sergio Gonzalez\n Cass Everitt\n Engin Manap\n If your name should be here but Martins Mozeiko\n isn't, let Sean know. Joseph Thomson\n Phil Jordan\n Nathan Reed\n Michaelangel007@github\n Nick Verigakis\n\nLICENSE\n\nThis software is in the public domain. Where that dedication is not\nrecognized, you are granted a perpetual, irrevocable license to copy,\ndistribute, and modify this file as you see fit.\n\n*/\n\n#ifndef STBI_INCLUDE_STB_IMAGE_H\n#define STBI_INCLUDE_STB_IMAGE_H\n\n// DOCUMENTATION\n//\n// Limitations:\n// - no 16-bit-per-channel PNG\n// - no 12-bit-per-channel JPEG\n// - no JPEGs with arithmetic coding\n// - no 1-bit BMP\n// - GIF always returns *comp=4\n//\n// Basic usage (see HDR discussion below for HDR usage):\n// int x,y,n;\n// unsigned char *data = stbi_load(filename, &x, &y, &n, 0);\n// // ... process data if not NULL ...\n// // ... x = width, y = height, n = # 8-bit components per pixel ...\n// // ... replace '0' with '1'..'4' to force that many components per pixel\n// // ... but 'n' will always be the number that it would have been if you said 0\n// stbi_image_free(data)\n//\n// Standard parameters:\n// int *x -- outputs image width in pixels\n// int *y -- outputs image height in pixels\n// int *comp -- outputs # of image components in image file\n// int req_comp -- if non-zero, # of image components requested in result\n//\n// The return value from an image loader is an 'unsigned char *' which points\n// to the pixel data, or NULL on an allocation failure or if the image is\n// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels,\n// with each pixel consisting of N interleaved 8-bit components; the first\n// pixel pointed to is top-left-most in the image. There is no padding between\n// image scanlines or between pixels, regardless of format. The number of\n// components N is 'req_comp' if req_comp is non-zero, or *comp otherwise.\n// If req_comp is non-zero, *comp has the number of components that _would_\n// have been output otherwise. E.g. if you set req_comp to 4, you will always\n// get RGBA output, but you can check *comp to see if it's trivially opaque\n// because e.g. there were only 3 channels in the source image.\n//\n// An output image with N components has the following components interleaved\n// in this order in each pixel:\n//\n// N=#comp components\n// 1 grey\n// 2 grey, alpha\n// 3 red, green, blue\n// 4 red, green, blue, alpha\n//\n// If image loading fails for any reason, the return value will be NULL,\n// and *x, *y, *comp will be unchanged. The function stbi_failure_reason()\n// can be queried for an extremely brief, end-user unfriendly explanation\n// of why the load failed. Define STBI_NO_FAILURE_STRINGS to avoid\n// compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly\n// more user-friendly ones.\n//\n// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized.\n//\n// ===========================================================================\n//\n// Philosophy\n//\n// stb libraries are designed with the following priorities:\n//\n// 1. easy to use\n// 2. easy to maintain\n// 3. good performance\n//\n// Sometimes I let \"good performance\" creep up in priority over \"easy to maintain\",\n// and for best performance I may provide less-easy-to-use APIs that give higher\n// performance, in addition to the easy to use ones. Nevertheless, it's important\n// to keep in mind that from the standpoint of you, a client of this library,\n// all you care about is #1 and #3, and stb libraries do not emphasize #3 above all.\n//\n// Some secondary priorities arise directly from the first two, some of which\n// make more explicit reasons why performance can't be emphasized.\n//\n// - Portable (\"ease of use\")\n// - Small footprint (\"easy to maintain\")\n// - No dependencies (\"ease of use\")\n//\n// ===========================================================================\n//\n// I/O callbacks\n//\n// I/O callbacks allow you to read from arbitrary sources, like packaged\n// files or some other source. Data read from callbacks are processed\n// through a small internal buffer (currently 128 bytes) to try to reduce\n// overhead.\n//\n// The three functions you must define are \"read\" (reads some bytes of data),\n// \"skip\" (skips some bytes of data), \"eof\" (reports if the stream is at the end).\n//\n// ===========================================================================\n//\n// SIMD support\n//\n// The JPEG decoder will try to automatically use SIMD kernels on x86 when\n// supported by the compiler. For ARM Neon support, you must explicitly\n// request it.\n//\n// (The old do-it-yourself SIMD API is no longer supported in the current\n// code.)\n//\n// On x86, SSE2 will automatically be used when available based on a run-time\n// test; if not, the generic C versions are used as a fall-back. On ARM targets,\n// the typical path is to have separate builds for NEON and non-NEON devices\n// (at least this is true for iOS and Android). Therefore, the NEON support is\n// toggled by a build flag: define STBI_NEON to get NEON loops.\n//\n// The output of the JPEG decoder is slightly different from versions where\n// SIMD support was introduced (that is, for versions before 1.49). The\n// difference is only +-1 in the 8-bit RGB channels, and only on a small\n// fraction of pixels. You can force the pre-1.49 behavior by defining\n// STBI_JPEG_OLD, but this will disable some of the SIMD decoding path\n// and hence cost some performance.\n//\n// If for some reason you do not want to use any of SIMD code, or if\n// you have issues compiling it, you can disable it entirely by\n// defining STBI_NO_SIMD.\n//\n// ===========================================================================\n//\n// HDR image support (disable by defining STBI_NO_HDR)\n//\n// stb_image now supports loading HDR images in general, and currently\n// the Radiance .HDR file format, although the support is provided\n// generically. You can still load any file through the existing interface;\n// if you attempt to load an HDR file, it will be automatically remapped to\n// LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1;\n// both of these constants can be reconfigured through this interface:\n//\n// stbi_hdr_to_ldr_gamma(2.2f);\n// stbi_hdr_to_ldr_scale(1.0f);\n//\n// (note, do not use _inverse_ constants; stbi_image will invert them\n// appropriately).\n//\n// Additionally, there is a new, parallel interface for loading files as\n// (linear) floats to preserve the full dynamic range:\n//\n// float *data = stbi_loadf(filename, &x, &y, &n, 0);\n//\n// If you load LDR images through this interface, those images will\n// be promoted to floating point values, run through the inverse of\n// constants corresponding to the above:\n//\n// stbi_ldr_to_hdr_scale(1.0f);\n// stbi_ldr_to_hdr_gamma(2.2f);\n//\n// Finally, given a filename (or an open file or memory block--see header\n// file for details) containing image data, you can query for the \"most\n// appropriate\" interface to use (that is, whether the image is HDR or\n// not), using:\n//\n// stbi_is_hdr(char *filename);\n//\n// ===========================================================================\n//\n// iPhone PNG support:\n//\n// By default we convert iphone-formatted PNGs back to RGB, even though\n// they are internally encoded differently. You can disable this conversion\n// by by calling stbi_convert_iphone_png_to_rgb(0), in which case\n// you will always just get the native iphone \"format\" through (which\n// is BGR stored in RGB).\n//\n// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per\n// pixel to remove any premultiplied alpha *only* if the image file explicitly\n// says there's premultiplied data (currently only happens in iPhone images,\n// and only if iPhone convert-to-rgb processing is on).\n//\n\n\n#ifndef STBI_NO_STDIO\n#include \n#endif // STBI_NO_STDIO\n\n#define STBI_VERSION 1\n\nenum\n{\n STBI_default = 0, // only used for req_comp\n\n STBI_grey = 1,\n STBI_grey_alpha = 2,\n STBI_rgb = 3,\n STBI_rgb_alpha = 4\n};\n\ntypedef unsigned char stbi_uc;\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\n#ifdef STB_IMAGE_STATIC\n#define STBIDEF static\n#else\n#define STBIDEF extern\n#endif\n\n//////////////////////////////////////////////////////////////////////////////\n//\n// PRIMARY API - works on images of any type\n//\n\n//\n// load image by filename, open file, or memory buffer\n//\n\ntypedef struct\n{\n int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read\n void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative\n int (*eof) (void *user); // returns nonzero if we are at end of file/data\n} stbi_io_callbacks;\n\nSTBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *comp, int req_comp);\nSTBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *comp, int req_comp);\nSTBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *comp, int req_comp);\n\n#ifndef STBI_NO_STDIO\nSTBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp);\n// for stbi_load_from_file, file pointer is left pointing immediately after image\n#endif\n\n#ifndef STBI_NO_LINEAR\n STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *comp, int req_comp);\n STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp);\n STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp);\n\n #ifndef STBI_NO_STDIO\n STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *comp, int req_comp);\n #endif\n#endif\n\n#ifndef STBI_NO_HDR\n STBIDEF void stbi_hdr_to_ldr_gamma(float gamma);\n STBIDEF void stbi_hdr_to_ldr_scale(float scale);\n#endif\n\n#ifndef STBI_NO_LINEAR\n STBIDEF void stbi_ldr_to_hdr_gamma(float gamma);\n STBIDEF void stbi_ldr_to_hdr_scale(float scale);\n#endif // STBI_NO_HDR\n\n// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR\nSTBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user);\nSTBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len);\n#ifndef STBI_NO_STDIO\nSTBIDEF int stbi_is_hdr (char const *filename);\nSTBIDEF int stbi_is_hdr_from_file(FILE *f);\n#endif // STBI_NO_STDIO\n\n\n// get a VERY brief reason for failure\n// NOT THREADSAFE\nSTBIDEF const char *stbi_failure_reason (void);\n\n// free the loaded image -- this is just free()\nSTBIDEF void stbi_image_free (void *retval_from_stbi_load);\n\n// get image dimensions & components without fully decoding\nSTBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp);\nSTBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp);\n\n#ifndef STBI_NO_STDIO\nSTBIDEF int stbi_info (char const *filename, int *x, int *y, int *comp);\nSTBIDEF int stbi_info_from_file (FILE *f, int *x, int *y, int *comp);\n\n#endif\n\n\n\n// for image formats that explicitly notate that they have premultiplied alpha,\n// we just return the colors as stored in the file. set this flag to force\n// unpremultiplication. results are undefined if the unpremultiply overflow.\nSTBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply);\n\n// indicate whether we should process iphone images back to canonical format,\n// or just pass them through \"as-is\"\nSTBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert);\n\n// flip the image vertically, so the first pixel in the output array is the bottom left\nSTBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip);\n\n// ZLIB client - used by PNG, available for other purposes\n\nSTBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen);\nSTBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header);\nSTBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen);\nSTBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);\n\nSTBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen);\nSTBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);\n\n\n#ifdef __cplusplus\n}\n#endif\n\n//\n//\n//// end header file /////////////////////////////////////////////////////\n#endif // STBI_INCLUDE_STB_IMAGE_H\n\n#ifdef STB_IMAGE_IMPLEMENTATION\n\n#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \\\n || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \\\n || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \\\n || defined(STBI_ONLY_ZLIB)\n #ifndef STBI_ONLY_JPEG\n #define STBI_NO_JPEG\n #endif\n #ifndef STBI_ONLY_PNG\n #define STBI_NO_PNG\n #endif\n #ifndef STBI_ONLY_BMP\n #define STBI_NO_BMP\n #endif\n #ifndef STBI_ONLY_PSD\n #define STBI_NO_PSD\n #endif\n #ifndef STBI_ONLY_TGA\n #define STBI_NO_TGA\n #endif\n #ifndef STBI_ONLY_GIF\n #define STBI_NO_GIF\n #endif\n #ifndef STBI_ONLY_HDR\n #define STBI_NO_HDR\n #endif\n #ifndef STBI_ONLY_PIC\n #define STBI_NO_PIC\n #endif\n #ifndef STBI_ONLY_PNM\n #define STBI_NO_PNM\n #endif\n#endif\n\n#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB)\n#define STBI_NO_ZLIB\n#endif\n\n\n#include \n#include // ptrdiff_t on osx\n#include \n#include \n\n#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR)\n#include // ldexp\n#endif\n\n#ifndef STBI_NO_STDIO\n#include \n#endif\n\n#ifndef STBI_ASSERT\n#include \n#define STBI_ASSERT(x) assert(x)\n#endif\n\n\n#ifndef _MSC_VER\n #ifdef __cplusplus\n #define stbi_inline inline\n #else\n #define stbi_inline\n #endif\n#else\n #define stbi_inline __forceinline\n#endif\n\n\n#ifdef _MSC_VER\ntypedef unsigned short stbi__uint16;\ntypedef signed short stbi__int16;\ntypedef unsigned int stbi__uint32;\ntypedef signed int stbi__int32;\n#else\n#include \ntypedef uint16_t stbi__uint16;\ntypedef int16_t stbi__int16;\ntypedef uint32_t stbi__uint32;\ntypedef int32_t stbi__int32;\n#endif\n\n// should produce compiler error if size is wrong\ntypedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1];\n\n#ifdef _MSC_VER\n#define STBI_NOTUSED(v) (void)(v)\n#else\n#define STBI_NOTUSED(v) (void)sizeof(v)\n#endif\n\n#ifdef _MSC_VER\n#define STBI_HAS_LROTL\n#endif\n\n#ifdef STBI_HAS_LROTL\n #define stbi_lrot(x,y) _lrotl(x,y)\n#else\n #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (32 - (y))))\n#endif\n\n#if defined(STBI_MALLOC) && defined(STBI_FREE) && defined(STBI_REALLOC)\n// ok\n#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC)\n// ok\n#else\n#error \"Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC.\"\n#endif\n\n#ifndef STBI_MALLOC\n#define STBI_MALLOC(sz) malloc(sz)\n#define STBI_REALLOC(p,sz) realloc(p,sz)\n#define STBI_FREE(p) free(p)\n#endif\n\n// x86/x64 detection\n#if defined(__x86_64__) || defined(_M_X64)\n#define STBI__X64_TARGET\n#elif defined(__i386) || defined(_M_IX86)\n#define STBI__X86_TARGET\n#endif\n\n#if defined(__GNUC__) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET)) && !defined(__SSE2__) && !defined(STBI_NO_SIMD)\n// NOTE: not clear do we actually need this for the 64-bit path?\n// gcc doesn't support sse2 intrinsics unless you compile with -msse2,\n// (but compiling with -msse2 allows the compiler to use SSE2 everywhere;\n// this is just broken and gcc are jerks for not fixing it properly\n// http://www.virtualdub.org/blog/pivot/entry.php?id=363 )\n#define STBI_NO_SIMD\n#endif\n\n#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD)\n// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET\n//\n// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the\n// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant.\n// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not\n// simultaneously enabling \"-mstackrealign\".\n//\n// See https://github.com/nothings/stb/issues/81 for more information.\n//\n// So default to no SSE2 on 32-bit MinGW. If you've read this far and added\n// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2.\n#define STBI_NO_SIMD\n#endif\n\n#if !defined(STBI_NO_SIMD) && defined(STBI__X86_TARGET)\n#define STBI_SSE2\n#include \n\n#ifdef _MSC_VER\n\n#if _MSC_VER >= 1400 // not VC6\n#include // __cpuid\nstatic int stbi__cpuid3(void)\n{\n int info[4];\n __cpuid(info,1);\n return info[3];\n}\n#else\nstatic int stbi__cpuid3(void)\n{\n int res;\n __asm {\n mov eax,1\n cpuid\n mov res,edx\n }\n return res;\n}\n#endif\n\n#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name\n\nstatic int stbi__sse2_available()\n{\n int info3 = stbi__cpuid3();\n return ((info3 >> 26) & 1) != 0;\n}\n#else // assume GCC-style if not VC++\n#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))\n\nstatic int stbi__sse2_available()\n{\n#if defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__) >= 408 // GCC 4.8 or later\n // GCC 4.8+ has a nice way to do this\n return __builtin_cpu_supports(\"sse2\");\n#else\n // portable way to do this, preferably without using GCC inline ASM?\n // just bail for now.\n return 0;\n#endif\n}\n#endif\n#endif\n\n// ARM NEON\n#if defined(STBI_NO_SIMD) && defined(STBI_NEON)\n#undef STBI_NEON\n#endif\n\n#ifdef STBI_NEON\n#include \n// assume GCC or Clang on ARM targets\n#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))\n#endif\n\n#ifndef STBI_SIMD_ALIGN\n#define STBI_SIMD_ALIGN(type, name) type name\n#endif\n\n///////////////////////////////////////////////\n//\n// stbi__context struct and start_xxx functions\n\n// stbi__context structure is our basic context used by all images, so it\n// contains all the IO context, plus some basic image information\ntypedef struct\n{\n stbi__uint32 img_x, img_y;\n int img_n, img_out_n;\n\n stbi_io_callbacks io;\n void *io_user_data;\n\n int read_from_callbacks;\n int buflen;\n stbi_uc buffer_start[128];\n\n stbi_uc *img_buffer, *img_buffer_end;\n stbi_uc *img_buffer_original, *img_buffer_original_end;\n} stbi__context;\n\n\nstatic void stbi__refill_buffer(stbi__context *s);\n\n// initialize a memory-decode context\nstatic void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len)\n{\n s->io.read = NULL;\n s->read_from_callbacks = 0;\n s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer;\n s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len;\n}\n\n// initialize a callback-based context\nstatic void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user)\n{\n s->io = *c;\n s->io_user_data = user;\n s->buflen = sizeof(s->buffer_start);\n s->read_from_callbacks = 1;\n s->img_buffer_original = s->buffer_start;\n stbi__refill_buffer(s);\n s->img_buffer_original_end = s->img_buffer_end;\n}\n\n#ifndef STBI_NO_STDIO\n\nstatic int stbi__stdio_read(void *user, char *data, int size)\n{\n return (int) fread(data,1,size,(FILE*) user);\n}\n\nstatic void stbi__stdio_skip(void *user, int n)\n{\n fseek((FILE*) user, n, SEEK_CUR);\n}\n\nstatic int stbi__stdio_eof(void *user)\n{\n return feof((FILE*) user);\n}\n\nstatic stbi_io_callbacks stbi__stdio_callbacks =\n{\n stbi__stdio_read,\n stbi__stdio_skip,\n stbi__stdio_eof,\n};\n\nstatic void stbi__start_file(stbi__context *s, FILE *f)\n{\n stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f);\n}\n\n//static void stop_file(stbi__context *s) { }\n\n#endif // !STBI_NO_STDIO\n\nstatic void stbi__rewind(stbi__context *s)\n{\n // conceptually rewind SHOULD rewind to the beginning of the stream,\n // but we just rewind to the beginning of the initial buffer, because\n // we only use it after doing 'test', which only ever looks at at most 92 bytes\n s->img_buffer = s->img_buffer_original;\n s->img_buffer_end = s->img_buffer_original_end;\n}\n\n#ifndef STBI_NO_JPEG\nstatic int stbi__jpeg_test(stbi__context *s);\nstatic stbi_uc *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n#ifndef STBI_NO_PNG\nstatic int stbi__png_test(stbi__context *s);\nstatic stbi_uc *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__png_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n#ifndef STBI_NO_BMP\nstatic int stbi__bmp_test(stbi__context *s);\nstatic stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n#ifndef STBI_NO_TGA\nstatic int stbi__tga_test(stbi__context *s);\nstatic stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n#ifndef STBI_NO_PSD\nstatic int stbi__psd_test(stbi__context *s);\nstatic stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n#ifndef STBI_NO_HDR\nstatic int stbi__hdr_test(stbi__context *s);\nstatic float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n#ifndef STBI_NO_PIC\nstatic int stbi__pic_test(stbi__context *s);\nstatic stbi_uc *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n#ifndef STBI_NO_GIF\nstatic int stbi__gif_test(stbi__context *s);\nstatic stbi_uc *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n#ifndef STBI_NO_PNM\nstatic int stbi__pnm_test(stbi__context *s);\nstatic stbi_uc *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp);\nstatic int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp);\n#endif\n\n// this is not threadsafe\nstatic const char *stbi__g_failure_reason;\n\nSTBIDEF const char *stbi_failure_reason(void)\n{\n return stbi__g_failure_reason;\n}\n\nstatic int stbi__err(const char *str)\n{\n stbi__g_failure_reason = str;\n return 0;\n}\n\nstatic void *stbi__malloc(size_t size)\n{\n return STBI_MALLOC(size);\n}\n\n// stbi__err - error\n// stbi__errpf - error returning pointer to float\n// stbi__errpuc - error returning pointer to unsigned char\n\n#ifdef STBI_NO_FAILURE_STRINGS\n #define stbi__err(x,y) 0\n#elif defined(STBI_FAILURE_USERMSG)\n #define stbi__err(x,y) stbi__err(y)\n#else\n #define stbi__err(x,y) stbi__err(x)\n#endif\n\n#define stbi__errpf(x,y) ((float *)(size_t) (stbi__err(x,y)?NULL:NULL))\n#define stbi__errpuc(x,y) ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL))\n\nSTBIDEF void stbi_image_free(void *retval_from_stbi_load)\n{\n STBI_FREE(retval_from_stbi_load);\n}\n\n#ifndef STBI_NO_LINEAR\nstatic float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp);\n#endif\n\n#ifndef STBI_NO_HDR\nstatic stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp);\n#endif\n\nstatic int stbi__vertically_flip_on_load = 0;\n\nSTBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip)\n{\n stbi__vertically_flip_on_load = flag_true_if_should_flip;\n}\n\nstatic unsigned char *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n #ifndef STBI_NO_JPEG\n if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp);\n #endif\n #ifndef STBI_NO_PNG\n if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp);\n #endif\n #ifndef STBI_NO_BMP\n if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp);\n #endif\n #ifndef STBI_NO_GIF\n if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp);\n #endif\n #ifndef STBI_NO_PSD\n if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp);\n #endif\n #ifndef STBI_NO_PIC\n if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp);\n #endif\n #ifndef STBI_NO_PNM\n if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp);\n #endif\n\n #ifndef STBI_NO_HDR\n if (stbi__hdr_test(s)) {\n float *hdr = stbi__hdr_load(s, x,y,comp,req_comp);\n return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp);\n }\n #endif\n\n #ifndef STBI_NO_TGA\n // test tga last because it's a crappy test!\n if (stbi__tga_test(s))\n return stbi__tga_load(s,x,y,comp,req_comp);\n #endif\n\n return stbi__errpuc(\"unknown image type\", \"Image not of any known type, or corrupt\");\n}\n\nstatic unsigned char *stbi__load_flip(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n unsigned char *result = stbi__load_main(s, x, y, comp, req_comp);\n\n if (stbi__vertically_flip_on_load && result != NULL) {\n int w = *x, h = *y;\n int depth = req_comp ? req_comp : *comp;\n int row,col,z;\n stbi_uc temp;\n\n // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once\n for (row = 0; row < (h>>1); row++) {\n for (col = 0; col < w; col++) {\n for (z = 0; z < depth; z++) {\n temp = result[(row * w + col) * depth + z];\n result[(row * w + col) * depth + z] = result[((h - row - 1) * w + col) * depth + z];\n result[((h - row - 1) * w + col) * depth + z] = temp;\n }\n }\n }\n }\n\n return result;\n}\n\n#ifndef STBI_NO_HDR\nstatic void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp)\n{\n if (stbi__vertically_flip_on_load && result != NULL) {\n int w = *x, h = *y;\n int depth = req_comp ? req_comp : *comp;\n int row,col,z;\n float temp;\n\n // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once\n for (row = 0; row < (h>>1); row++) {\n for (col = 0; col < w; col++) {\n for (z = 0; z < depth; z++) {\n temp = result[(row * w + col) * depth + z];\n result[(row * w + col) * depth + z] = result[((h - row - 1) * w + col) * depth + z];\n result[((h - row - 1) * w + col) * depth + z] = temp;\n }\n }\n }\n }\n}\n#endif\n\n#ifndef STBI_NO_STDIO\n\nstatic FILE *stbi__fopen(char const *filename, char const *mode)\n{\n FILE *f;\n#if defined(_MSC_VER) && _MSC_VER >= 1400\n if (0 != fopen_s(&f, filename, mode))\n f=0;\n#else\n f = fopen(filename, mode);\n#endif\n return f;\n}\n\n\nSTBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp)\n{\n FILE *f = stbi__fopen(filename, \"rb\");\n unsigned char *result;\n if (!f) return stbi__errpuc(\"can't fopen\", \"Unable to open file\");\n result = stbi_load_from_file(f,x,y,comp,req_comp);\n fclose(f);\n return result;\n}\n\nSTBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)\n{\n unsigned char *result;\n stbi__context s;\n stbi__start_file(&s,f);\n result = stbi__load_flip(&s,x,y,comp,req_comp);\n if (result) {\n // need to 'unget' all the characters in the IO buffer\n fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);\n }\n return result;\n}\n#endif //!STBI_NO_STDIO\n\nSTBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)\n{\n stbi__context s;\n stbi__start_mem(&s,buffer,len);\n return stbi__load_flip(&s,x,y,comp,req_comp);\n}\n\nSTBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)\n{\n stbi__context s;\n stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);\n return stbi__load_flip(&s,x,y,comp,req_comp);\n}\n\n#ifndef STBI_NO_LINEAR\nstatic float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n unsigned char *data;\n #ifndef STBI_NO_HDR\n if (stbi__hdr_test(s)) {\n float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp);\n if (hdr_data)\n stbi__float_postprocess(hdr_data,x,y,comp,req_comp);\n return hdr_data;\n }\n #endif\n data = stbi__load_flip(s, x, y, comp, req_comp);\n if (data)\n return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp);\n return stbi__errpf(\"unknown image type\", \"Image not of any known type, or corrupt\");\n}\n\nSTBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)\n{\n stbi__context s;\n stbi__start_mem(&s,buffer,len);\n return stbi__loadf_main(&s,x,y,comp,req_comp);\n}\n\nSTBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp)\n{\n stbi__context s;\n stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);\n return stbi__loadf_main(&s,x,y,comp,req_comp);\n}\n\n#ifndef STBI_NO_STDIO\nSTBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp)\n{\n float *result;\n FILE *f = stbi__fopen(filename, \"rb\");\n if (!f) return stbi__errpf(\"can't fopen\", \"Unable to open file\");\n result = stbi_loadf_from_file(f,x,y,comp,req_comp);\n fclose(f);\n return result;\n}\n\nSTBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp)\n{\n stbi__context s;\n stbi__start_file(&s,f);\n return stbi__loadf_main(&s,x,y,comp,req_comp);\n}\n#endif // !STBI_NO_STDIO\n\n#endif // !STBI_NO_LINEAR\n\n// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is\n// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always\n// reports false!\n\nSTBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len)\n{\n #ifndef STBI_NO_HDR\n stbi__context s;\n stbi__start_mem(&s,buffer,len);\n return stbi__hdr_test(&s);\n #else\n STBI_NOTUSED(buffer);\n STBI_NOTUSED(len);\n return 0;\n #endif\n}\n\n#ifndef STBI_NO_STDIO\nSTBIDEF int stbi_is_hdr (char const *filename)\n{\n FILE *f = stbi__fopen(filename, \"rb\");\n int result=0;\n if (f) {\n result = stbi_is_hdr_from_file(f);\n fclose(f);\n }\n return result;\n}\n\nSTBIDEF int stbi_is_hdr_from_file(FILE *f)\n{\n #ifndef STBI_NO_HDR\n stbi__context s;\n stbi__start_file(&s,f);\n return stbi__hdr_test(&s);\n #else\n STBI_NOTUSED(f);\n return 0;\n #endif\n}\n#endif // !STBI_NO_STDIO\n\nSTBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user)\n{\n #ifndef STBI_NO_HDR\n stbi__context s;\n stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user);\n return stbi__hdr_test(&s);\n #else\n STBI_NOTUSED(clbk);\n STBI_NOTUSED(user);\n return 0;\n #endif\n}\n\nstatic float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f;\nstatic float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f;\n\n#ifndef STBI_NO_LINEAR\nSTBIDEF void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; }\nSTBIDEF void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; }\n#endif\n\nSTBIDEF void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; }\nSTBIDEF void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; }\n\n\n//////////////////////////////////////////////////////////////////////////////\n//\n// Common code used by all image loaders\n//\n\nenum\n{\n STBI__SCAN_load=0,\n STBI__SCAN_type,\n STBI__SCAN_header\n};\n\nstatic void stbi__refill_buffer(stbi__context *s)\n{\n int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen);\n if (n == 0) {\n // at end of file, treat same as if from memory, but need to handle case\n // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file\n s->read_from_callbacks = 0;\n s->img_buffer = s->buffer_start;\n s->img_buffer_end = s->buffer_start+1;\n *s->img_buffer = 0;\n } else {\n s->img_buffer = s->buffer_start;\n s->img_buffer_end = s->buffer_start + n;\n }\n}\n\nstbi_inline static stbi_uc stbi__get8(stbi__context *s)\n{\n if (s->img_buffer < s->img_buffer_end)\n return *s->img_buffer++;\n if (s->read_from_callbacks) {\n stbi__refill_buffer(s);\n return *s->img_buffer++;\n }\n return 0;\n}\n\nstbi_inline static int stbi__at_eof(stbi__context *s)\n{\n if (s->io.read) {\n if (!(s->io.eof)(s->io_user_data)) return 0;\n // if feof() is true, check if buffer = end\n // special case: we've only got the special 0 character at the end\n if (s->read_from_callbacks == 0) return 1;\n }\n\n return s->img_buffer >= s->img_buffer_end;\n}\n\nstatic void stbi__skip(stbi__context *s, int n)\n{\n if (n < 0) {\n s->img_buffer = s->img_buffer_end;\n return;\n }\n if (s->io.read) {\n int blen = (int) (s->img_buffer_end - s->img_buffer);\n if (blen < n) {\n s->img_buffer = s->img_buffer_end;\n (s->io.skip)(s->io_user_data, n - blen);\n return;\n }\n }\n s->img_buffer += n;\n}\n\nstatic int stbi__getn(stbi__context *s, stbi_uc *buffer, int n)\n{\n if (s->io.read) {\n int blen = (int) (s->img_buffer_end - s->img_buffer);\n if (blen < n) {\n int res, count;\n\n memcpy(buffer, s->img_buffer, blen);\n\n count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen);\n res = (count == (n-blen));\n s->img_buffer = s->img_buffer_end;\n return res;\n }\n }\n\n if (s->img_buffer+n <= s->img_buffer_end) {\n memcpy(buffer, s->img_buffer, n);\n s->img_buffer += n;\n return 1;\n } else\n return 0;\n}\n\nstatic int stbi__get16be(stbi__context *s)\n{\n int z = stbi__get8(s);\n return (z << 8) + stbi__get8(s);\n}\n\nstatic stbi__uint32 stbi__get32be(stbi__context *s)\n{\n stbi__uint32 z = stbi__get16be(s);\n return (z << 16) + stbi__get16be(s);\n}\n\n#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF)\n// nothing\n#else\nstatic int stbi__get16le(stbi__context *s)\n{\n int z = stbi__get8(s);\n return z + (stbi__get8(s) << 8);\n}\n#endif\n\n#ifndef STBI_NO_BMP\nstatic stbi__uint32 stbi__get32le(stbi__context *s)\n{\n stbi__uint32 z = stbi__get16le(s);\n return z + (stbi__get16le(s) << 16);\n}\n#endif\n\n#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings\n\n\n//////////////////////////////////////////////////////////////////////////////\n//\n// generic converter from built-in img_n to req_comp\n// individual types do this automatically as much as possible (e.g. jpeg\n// does all cases internally since it needs to colorspace convert anyway,\n// and it never has alpha, so very few cases ). png can automatically\n// interleave an alpha=255 channel, but falls back to this for other cases\n//\n// assume data buffer is malloced, so malloc a new one and free that one\n// only failure mode is malloc failing\n\nstatic stbi_uc stbi__compute_y(int r, int g, int b)\n{\n return (stbi_uc) (((r*77) + (g*150) + (29*b)) >> 8);\n}\n\nstatic unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y)\n{\n int i,j;\n unsigned char *good;\n\n if (req_comp == img_n) return data;\n STBI_ASSERT(req_comp >= 1 && req_comp <= 4);\n\n good = (unsigned char *) stbi__malloc(req_comp * x * y);\n if (good == NULL) {\n STBI_FREE(data);\n return stbi__errpuc(\"outofmem\", \"Out of memory\");\n }\n\n for (j=0; j < (int) y; ++j) {\n unsigned char *src = data + j * x * img_n ;\n unsigned char *dest = good + j * x * req_comp;\n\n #define COMBO(a,b) ((a)*8+(b))\n #define CASE(a,b) case COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)\n // convert source image with img_n components to one with req_comp components;\n // avoid switch per pixel, so use switch per scanline and massive macros\n switch (COMBO(img_n, req_comp)) {\n CASE(1,2) dest[0]=src[0], dest[1]=255; break;\n CASE(1,3) dest[0]=dest[1]=dest[2]=src[0]; break;\n CASE(1,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=255; break;\n CASE(2,1) dest[0]=src[0]; break;\n CASE(2,3) dest[0]=dest[1]=dest[2]=src[0]; break;\n CASE(2,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=src[1]; break;\n CASE(3,4) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2],dest[3]=255; break;\n CASE(3,1) dest[0]=stbi__compute_y(src[0],src[1],src[2]); break;\n CASE(3,2) dest[0]=stbi__compute_y(src[0],src[1],src[2]), dest[1] = 255; break;\n CASE(4,1) dest[0]=stbi__compute_y(src[0],src[1],src[2]); break;\n CASE(4,2) dest[0]=stbi__compute_y(src[0],src[1],src[2]), dest[1] = src[3]; break;\n CASE(4,3) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2]; break;\n default: STBI_ASSERT(0);\n }\n #undef CASE\n }\n\n STBI_FREE(data);\n return good;\n}\n\n#ifndef STBI_NO_LINEAR\nstatic float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp)\n{\n int i,k,n;\n float *output = (float *) stbi__malloc(x * y * comp * sizeof(float));\n if (output == NULL) { STBI_FREE(data); return stbi__errpf(\"outofmem\", \"Out of memory\"); }\n // compute number of non-alpha components\n if (comp & 1) n = comp; else n = comp-1;\n for (i=0; i < x*y; ++i) {\n for (k=0; k < n; ++k) {\n output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale);\n }\n if (k < comp) output[i*comp + k] = data[i*comp+k]/255.0f;\n }\n STBI_FREE(data);\n return output;\n}\n#endif\n\n#ifndef STBI_NO_HDR\n#define stbi__float2int(x) ((int) (x))\nstatic stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp)\n{\n int i,k,n;\n stbi_uc *output = (stbi_uc *) stbi__malloc(x * y * comp);\n if (output == NULL) { STBI_FREE(data); return stbi__errpuc(\"outofmem\", \"Out of memory\"); }\n // compute number of non-alpha components\n if (comp & 1) n = comp; else n = comp-1;\n for (i=0; i < x*y; ++i) {\n for (k=0; k < n; ++k) {\n float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f;\n if (z < 0) z = 0;\n if (z > 255) z = 255;\n output[i*comp + k] = (stbi_uc) stbi__float2int(z);\n }\n if (k < comp) {\n float z = data[i*comp+k] * 255 + 0.5f;\n if (z < 0) z = 0;\n if (z > 255) z = 255;\n output[i*comp + k] = (stbi_uc) stbi__float2int(z);\n }\n }\n STBI_FREE(data);\n return output;\n}\n#endif\n\n//////////////////////////////////////////////////////////////////////////////\n//\n// \"baseline\" JPEG/JFIF decoder\n//\n// simple implementation\n// - doesn't support delayed output of y-dimension\n// - simple interface (only one output format: 8-bit interleaved RGB)\n// - doesn't try to recover corrupt jpegs\n// - doesn't allow partial loading, loading multiple at once\n// - still fast on x86 (copying globals into locals doesn't help x86)\n// - allocates lots of intermediate memory (full size of all components)\n// - non-interleaved case requires this anyway\n// - allows good upsampling (see next)\n// high-quality\n// - upsampled channels are bilinearly interpolated, even across blocks\n// - quality integer IDCT derived from IJG's 'slow'\n// performance\n// - fast huffman; reasonable integer IDCT\n// - some SIMD kernels for common paths on targets with SSE2/NEON\n// - uses a lot of intermediate memory, could cache poorly\n\n#ifndef STBI_NO_JPEG\n\n// huffman decoding acceleration\n#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache\n\ntypedef struct\n{\n stbi_uc fast[1 << FAST_BITS];\n // weirdly, repacking this into AoS is a 10% speed loss, instead of a win\n stbi__uint16 code[256];\n stbi_uc values[256];\n stbi_uc size[257];\n unsigned int maxcode[18];\n int delta[17]; // old 'firstsymbol' - old 'firstcode'\n} stbi__huffman;\n\ntypedef struct\n{\n stbi__context *s;\n stbi__huffman huff_dc[4];\n stbi__huffman huff_ac[4];\n stbi_uc dequant[4][64];\n stbi__int16 fast_ac[4][1 << FAST_BITS];\n\n// sizes for components, interleaved MCUs\n int img_h_max, img_v_max;\n int img_mcu_x, img_mcu_y;\n int img_mcu_w, img_mcu_h;\n\n// definition of jpeg image component\n struct\n {\n int id;\n int h,v;\n int tq;\n int hd,ha;\n int dc_pred;\n\n int x,y,w2,h2;\n stbi_uc *data;\n void *raw_data, *raw_coeff;\n stbi_uc *linebuf;\n short *coeff; // progressive only\n int coeff_w, coeff_h; // number of 8x8 coefficient blocks\n } img_comp[4];\n\n stbi__uint32 code_buffer; // jpeg entropy-coded buffer\n int code_bits; // number of valid bits\n unsigned char marker; // marker seen while filling entropy buffer\n int nomore; // flag if we saw a marker so must stop\n\n int progressive;\n int spec_start;\n int spec_end;\n int succ_high;\n int succ_low;\n int eob_run;\n\n int scan_n, order[4];\n int restart_interval, todo;\n\n// kernels\n void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]);\n void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step);\n stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs);\n} stbi__jpeg;\n\nstatic int stbi__build_huffman(stbi__huffman *h, int *count)\n{\n int i,j,k=0,code;\n // build size list for each symbol (from JPEG spec)\n for (i=0; i < 16; ++i)\n for (j=0; j < count[i]; ++j)\n h->size[k++] = (stbi_uc) (i+1);\n h->size[k] = 0;\n\n // compute actual symbols (from jpeg spec)\n code = 0;\n k = 0;\n for(j=1; j <= 16; ++j) {\n // compute delta to add to code to compute symbol id\n h->delta[j] = k - code;\n if (h->size[k] == j) {\n while (h->size[k] == j)\n h->code[k++] = (stbi__uint16) (code++);\n if (code-1 >= (1 << j)) return stbi__err(\"bad code lengths\",\"Corrupt JPEG\");\n }\n // compute largest code + 1 for this size, preshifted as needed later\n h->maxcode[j] = code << (16-j);\n code <<= 1;\n }\n h->maxcode[j] = 0xffffffff;\n\n // build non-spec acceleration table; 255 is flag for not-accelerated\n memset(h->fast, 255, 1 << FAST_BITS);\n for (i=0; i < k; ++i) {\n int s = h->size[i];\n if (s <= FAST_BITS) {\n int c = h->code[i] << (FAST_BITS-s);\n int m = 1 << (FAST_BITS-s);\n for (j=0; j < m; ++j) {\n h->fast[c+j] = (stbi_uc) i;\n }\n }\n }\n return 1;\n}\n\n// build a table that decodes both magnitude and value of small ACs in\n// one go.\nstatic void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h)\n{\n int i;\n for (i=0; i < (1 << FAST_BITS); ++i) {\n stbi_uc fast = h->fast[i];\n fast_ac[i] = 0;\n if (fast < 255) {\n int rs = h->values[fast];\n int run = (rs >> 4) & 15;\n int magbits = rs & 15;\n int len = h->size[fast];\n\n if (magbits && len + magbits <= FAST_BITS) {\n // magnitude code followed by receive_extend code\n int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits);\n int m = 1 << (magbits - 1);\n if (k < m) k += (-1 << magbits) + 1;\n // if the result is small enough, we can fit it in fast_ac table\n if (k >= -128 && k <= 127)\n fast_ac[i] = (stbi__int16) ((k << 8) + (run << 4) + (len + magbits));\n }\n }\n }\n}\n\nstatic void stbi__grow_buffer_unsafe(stbi__jpeg *j)\n{\n do {\n int b = j->nomore ? 0 : stbi__get8(j->s);\n if (b == 0xff) {\n int c = stbi__get8(j->s);\n if (c != 0) {\n j->marker = (unsigned char) c;\n j->nomore = 1;\n return;\n }\n }\n j->code_buffer |= b << (24 - j->code_bits);\n j->code_bits += 8;\n } while (j->code_bits <= 24);\n}\n\n// (1 << n) - 1\nstatic stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535};\n\n// decode a jpeg huffman value from the bitstream\nstbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h)\n{\n unsigned int temp;\n int c,k;\n\n if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\n\n // look at the top FAST_BITS and determine what symbol ID it is,\n // if the code is <= FAST_BITS\n c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);\n k = h->fast[c];\n if (k < 255) {\n int s = h->size[k];\n if (s > j->code_bits)\n return -1;\n j->code_buffer <<= s;\n j->code_bits -= s;\n return h->values[k];\n }\n\n // naive test is to shift the code_buffer down so k bits are\n // valid, then test against maxcode. To speed this up, we've\n // preshifted maxcode left so that it has (16-k) 0s at the\n // end; in other words, regardless of the number of bits, it\n // wants to be compared against something shifted to have 16;\n // that way we don't need to shift inside the loop.\n temp = j->code_buffer >> 16;\n for (k=FAST_BITS+1 ; ; ++k)\n if (temp < h->maxcode[k])\n break;\n if (k == 17) {\n // error! code not found\n j->code_bits -= 16;\n return -1;\n }\n\n if (k > j->code_bits)\n return -1;\n\n // convert the huffman code to the symbol id\n c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k];\n STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]);\n\n // convert the id to a symbol\n j->code_bits -= k;\n j->code_buffer <<= k;\n return h->values[c];\n}\n\n// bias[n] = (-1<code_bits < n) stbi__grow_buffer_unsafe(j);\n\n sgn = (stbi__int32)j->code_buffer >> 31; // sign bit is always in MSB\n k = stbi_lrot(j->code_buffer, n);\n STBI_ASSERT(n >= 0 && n < (int) (sizeof(stbi__bmask)/sizeof(*stbi__bmask)));\n j->code_buffer = k & ~stbi__bmask[n];\n k &= stbi__bmask[n];\n j->code_bits -= n;\n return k + (stbi__jbias[n] & ~sgn);\n}\n\n// get some unsigned bits\nstbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n)\n{\n unsigned int k;\n if (j->code_bits < n) stbi__grow_buffer_unsafe(j);\n k = stbi_lrot(j->code_buffer, n);\n j->code_buffer = k & ~stbi__bmask[n];\n k &= stbi__bmask[n];\n j->code_bits -= n;\n return k;\n}\n\nstbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j)\n{\n unsigned int k;\n if (j->code_bits < 1) stbi__grow_buffer_unsafe(j);\n k = j->code_buffer;\n j->code_buffer <<= 1;\n --j->code_bits;\n return k & 0x80000000;\n}\n\n// given a value that's at position X in the zigzag stream,\n// where does it appear in the 8x8 matrix coded as row-major?\nstatic stbi_uc stbi__jpeg_dezigzag[64+15] =\n{\n 0, 1, 8, 16, 9, 2, 3, 10,\n 17, 24, 32, 25, 18, 11, 4, 5,\n 12, 19, 26, 33, 40, 48, 41, 34,\n 27, 20, 13, 6, 7, 14, 21, 28,\n 35, 42, 49, 56, 57, 50, 43, 36,\n 29, 22, 15, 23, 30, 37, 44, 51,\n 58, 59, 52, 45, 38, 31, 39, 46,\n 53, 60, 61, 54, 47, 55, 62, 63,\n // let corrupt input sample past end\n 63, 63, 63, 63, 63, 63, 63, 63,\n 63, 63, 63, 63, 63, 63, 63\n};\n\n// decode one 64-entry block--\nstatic int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi_uc *dequant)\n{\n int diff,dc,k;\n int t;\n\n if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\n t = stbi__jpeg_huff_decode(j, hdc);\n if (t < 0) return stbi__err(\"bad huffman code\",\"Corrupt JPEG\");\n\n // 0 all the ac values now so we can do it 32-bits at a time\n memset(data,0,64*sizeof(data[0]));\n\n diff = t ? stbi__extend_receive(j, t) : 0;\n dc = j->img_comp[b].dc_pred + diff;\n j->img_comp[b].dc_pred = dc;\n data[0] = (short) (dc * dequant[0]);\n\n // decode AC components, see JPEG spec\n k = 1;\n do {\n unsigned int zig;\n int c,r,s;\n if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\n c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);\n r = fac[c];\n if (r) { // fast-AC path\n k += (r >> 4) & 15; // run\n s = r & 15; // combined length\n j->code_buffer <<= s;\n j->code_bits -= s;\n // decode into unzigzag'd location\n zig = stbi__jpeg_dezigzag[k++];\n data[zig] = (short) ((r >> 8) * dequant[zig]);\n } else {\n int rs = stbi__jpeg_huff_decode(j, hac);\n if (rs < 0) return stbi__err(\"bad huffman code\",\"Corrupt JPEG\");\n s = rs & 15;\n r = rs >> 4;\n if (s == 0) {\n if (rs != 0xf0) break; // end block\n k += 16;\n } else {\n k += r;\n // decode into unzigzag'd location\n zig = stbi__jpeg_dezigzag[k++];\n data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]);\n }\n }\n } while (k < 64);\n return 1;\n}\n\nstatic int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b)\n{\n int diff,dc;\n int t;\n if (j->spec_end != 0) return stbi__err(\"can't merge dc and ac\", \"Corrupt JPEG\");\n\n if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\n\n if (j->succ_high == 0) {\n // first scan for DC coefficient, must be first\n memset(data,0,64*sizeof(data[0])); // 0 all the ac values now\n t = stbi__jpeg_huff_decode(j, hdc);\n diff = t ? stbi__extend_receive(j, t) : 0;\n\n dc = j->img_comp[b].dc_pred + diff;\n j->img_comp[b].dc_pred = dc;\n data[0] = (short) (dc << j->succ_low);\n } else {\n // refinement scan for DC coefficient\n if (stbi__jpeg_get_bit(j))\n data[0] += (short) (1 << j->succ_low);\n }\n return 1;\n}\n\n// @OPTIMIZE: store non-zigzagged during the decode passes,\n// and only de-zigzag when dequantizing\nstatic int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac)\n{\n int k;\n if (j->spec_start == 0) return stbi__err(\"can't merge dc and ac\", \"Corrupt JPEG\");\n\n if (j->succ_high == 0) {\n int shift = j->succ_low;\n\n if (j->eob_run) {\n --j->eob_run;\n return 1;\n }\n\n k = j->spec_start;\n do {\n unsigned int zig;\n int c,r,s;\n if (j->code_bits < 16) stbi__grow_buffer_unsafe(j);\n c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1);\n r = fac[c];\n if (r) { // fast-AC path\n k += (r >> 4) & 15; // run\n s = r & 15; // combined length\n j->code_buffer <<= s;\n j->code_bits -= s;\n zig = stbi__jpeg_dezigzag[k++];\n data[zig] = (short) ((r >> 8) << shift);\n } else {\n int rs = stbi__jpeg_huff_decode(j, hac);\n if (rs < 0) return stbi__err(\"bad huffman code\",\"Corrupt JPEG\");\n s = rs & 15;\n r = rs >> 4;\n if (s == 0) {\n if (r < 15) {\n j->eob_run = (1 << r);\n if (r)\n j->eob_run += stbi__jpeg_get_bits(j, r);\n --j->eob_run;\n break;\n }\n k += 16;\n } else {\n k += r;\n zig = stbi__jpeg_dezigzag[k++];\n data[zig] = (short) (stbi__extend_receive(j,s) << shift);\n }\n }\n } while (k <= j->spec_end);\n } else {\n // refinement scan for these AC coefficients\n\n short bit = (short) (1 << j->succ_low);\n\n if (j->eob_run) {\n --j->eob_run;\n for (k = j->spec_start; k <= j->spec_end; ++k) {\n short *p = &data[stbi__jpeg_dezigzag[k]];\n if (*p != 0)\n if (stbi__jpeg_get_bit(j))\n if ((*p & bit)==0) {\n if (*p > 0)\n *p += bit;\n else\n *p -= bit;\n }\n }\n } else {\n k = j->spec_start;\n do {\n int r,s;\n int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh\n if (rs < 0) return stbi__err(\"bad huffman code\",\"Corrupt JPEG\");\n s = rs & 15;\n r = rs >> 4;\n if (s == 0) {\n if (r < 15) {\n j->eob_run = (1 << r) - 1;\n if (r)\n j->eob_run += stbi__jpeg_get_bits(j, r);\n r = 64; // force end of block\n } else {\n // r=15 s=0 should write 16 0s, so we just do\n // a run of 15 0s and then write s (which is 0),\n // so we don't have to do anything special here\n }\n } else {\n if (s != 1) return stbi__err(\"bad huffman code\", \"Corrupt JPEG\");\n // sign bit\n if (stbi__jpeg_get_bit(j))\n s = bit;\n else\n s = -bit;\n }\n\n // advance by r\n while (k <= j->spec_end) {\n short *p = &data[stbi__jpeg_dezigzag[k++]];\n if (*p != 0) {\n if (stbi__jpeg_get_bit(j))\n if ((*p & bit)==0) {\n if (*p > 0)\n *p += bit;\n else\n *p -= bit;\n }\n } else {\n if (r == 0) {\n *p = (short) s;\n break;\n }\n --r;\n }\n }\n } while (k <= j->spec_end);\n }\n }\n return 1;\n}\n\n// take a -128..127 value and stbi__clamp it and convert to 0..255\nstbi_inline static stbi_uc stbi__clamp(int x)\n{\n // trick to use a single test to catch both cases\n if ((unsigned int) x > 255) {\n if (x < 0) return 0;\n if (x > 255) return 255;\n }\n return (stbi_uc) x;\n}\n\n#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5)))\n#define stbi__fsh(x) ((x) << 12)\n\n// derived from jidctint -- DCT_ISLOW\n#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \\\n int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \\\n p2 = s2; \\\n p3 = s6; \\\n p1 = (p2+p3) * stbi__f2f(0.5411961f); \\\n t2 = p1 + p3*stbi__f2f(-1.847759065f); \\\n t3 = p1 + p2*stbi__f2f( 0.765366865f); \\\n p2 = s0; \\\n p3 = s4; \\\n t0 = stbi__fsh(p2+p3); \\\n t1 = stbi__fsh(p2-p3); \\\n x0 = t0+t3; \\\n x3 = t0-t3; \\\n x1 = t1+t2; \\\n x2 = t1-t2; \\\n t0 = s7; \\\n t1 = s5; \\\n t2 = s3; \\\n t3 = s1; \\\n p3 = t0+t2; \\\n p4 = t1+t3; \\\n p1 = t0+t3; \\\n p2 = t1+t2; \\\n p5 = (p3+p4)*stbi__f2f( 1.175875602f); \\\n t0 = t0*stbi__f2f( 0.298631336f); \\\n t1 = t1*stbi__f2f( 2.053119869f); \\\n t2 = t2*stbi__f2f( 3.072711026f); \\\n t3 = t3*stbi__f2f( 1.501321110f); \\\n p1 = p5 + p1*stbi__f2f(-0.899976223f); \\\n p2 = p5 + p2*stbi__f2f(-2.562915447f); \\\n p3 = p3*stbi__f2f(-1.961570560f); \\\n p4 = p4*stbi__f2f(-0.390180644f); \\\n t3 += p1+p4; \\\n t2 += p2+p3; \\\n t1 += p2+p4; \\\n t0 += p1+p3;\n\nstatic void stbi__idct_block(stbi_uc *out, int out_stride, short data[64])\n{\n int i,val[64],*v=val;\n stbi_uc *o;\n short *d = data;\n\n // columns\n for (i=0; i < 8; ++i,++d, ++v) {\n // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing\n if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0\n && d[40]==0 && d[48]==0 && d[56]==0) {\n // no shortcut 0 seconds\n // (1|2|3|4|5|6|7)==0 0 seconds\n // all separate -0.047 seconds\n // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds\n int dcterm = d[0] << 2;\n v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm;\n } else {\n STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56])\n // constants scaled things up by 1<<12; let's bring them back\n // down, but keep 2 extra bits of precision\n x0 += 512; x1 += 512; x2 += 512; x3 += 512;\n v[ 0] = (x0+t3) >> 10;\n v[56] = (x0-t3) >> 10;\n v[ 8] = (x1+t2) >> 10;\n v[48] = (x1-t2) >> 10;\n v[16] = (x2+t1) >> 10;\n v[40] = (x2-t1) >> 10;\n v[24] = (x3+t0) >> 10;\n v[32] = (x3-t0) >> 10;\n }\n }\n\n for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) {\n // no fast case since the first 1D IDCT spread components out\n STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7])\n // constants scaled things up by 1<<12, plus we had 1<<2 from first\n // loop, plus horizontal and vertical each scale by sqrt(8) so together\n // we've got an extra 1<<3, so 1<<17 total we need to remove.\n // so we want to round that, which means adding 0.5 * 1<<17,\n // aka 65536. Also, we'll end up with -128 to 127 that we want\n // to encode as 0..255 by adding 128, so we'll add that before the shift\n x0 += 65536 + (128<<17);\n x1 += 65536 + (128<<17);\n x2 += 65536 + (128<<17);\n x3 += 65536 + (128<<17);\n // tried computing the shifts into temps, or'ing the temps to see\n // if any were out of range, but that was slower\n o[0] = stbi__clamp((x0+t3) >> 17);\n o[7] = stbi__clamp((x0-t3) >> 17);\n o[1] = stbi__clamp((x1+t2) >> 17);\n o[6] = stbi__clamp((x1-t2) >> 17);\n o[2] = stbi__clamp((x2+t1) >> 17);\n o[5] = stbi__clamp((x2-t1) >> 17);\n o[3] = stbi__clamp((x3+t0) >> 17);\n o[4] = stbi__clamp((x3-t0) >> 17);\n }\n}\n\n#ifdef STBI_SSE2\n// sse2 integer IDCT. not the fastest possible implementation but it\n// produces bit-identical results to the generic C version so it's\n// fully \"transparent\".\nstatic void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])\n{\n // This is constructed to match our regular (generic) integer IDCT exactly.\n __m128i row0, row1, row2, row3, row4, row5, row6, row7;\n __m128i tmp;\n\n // dot product constant: even elems=x, odd elems=y\n #define dct_const(x,y) _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y))\n\n // out(0) = c0[even]*x + c0[odd]*y (c0, x, y 16-bit, out 32-bit)\n // out(1) = c1[even]*x + c1[odd]*y\n #define dct_rot(out0,out1, x,y,c0,c1) \\\n __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \\\n __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \\\n __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \\\n __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \\\n __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \\\n __m128i out1##_h = _mm_madd_epi16(c0##hi, c1)\n\n // out = in << 12 (in 16-bit, out 32-bit)\n #define dct_widen(out, in) \\\n __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \\\n __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4)\n\n // wide add\n #define dct_wadd(out, a, b) \\\n __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \\\n __m128i out##_h = _mm_add_epi32(a##_h, b##_h)\n\n // wide sub\n #define dct_wsub(out, a, b) \\\n __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \\\n __m128i out##_h = _mm_sub_epi32(a##_h, b##_h)\n\n // butterfly a/b, add bias, then shift by \"s\" and pack\n #define dct_bfly32o(out0, out1, a,b,bias,s) \\\n { \\\n __m128i abiased_l = _mm_add_epi32(a##_l, bias); \\\n __m128i abiased_h = _mm_add_epi32(a##_h, bias); \\\n dct_wadd(sum, abiased, b); \\\n dct_wsub(dif, abiased, b); \\\n out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \\\n out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \\\n }\n\n // 8-bit interleave step (for transposes)\n #define dct_interleave8(a, b) \\\n tmp = a; \\\n a = _mm_unpacklo_epi8(a, b); \\\n b = _mm_unpackhi_epi8(tmp, b)\n\n // 16-bit interleave step (for transposes)\n #define dct_interleave16(a, b) \\\n tmp = a; \\\n a = _mm_unpacklo_epi16(a, b); \\\n b = _mm_unpackhi_epi16(tmp, b)\n\n #define dct_pass(bias,shift) \\\n { \\\n /* even part */ \\\n dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \\\n __m128i sum04 = _mm_add_epi16(row0, row4); \\\n __m128i dif04 = _mm_sub_epi16(row0, row4); \\\n dct_widen(t0e, sum04); \\\n dct_widen(t1e, dif04); \\\n dct_wadd(x0, t0e, t3e); \\\n dct_wsub(x3, t0e, t3e); \\\n dct_wadd(x1, t1e, t2e); \\\n dct_wsub(x2, t1e, t2e); \\\n /* odd part */ \\\n dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \\\n dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \\\n __m128i sum17 = _mm_add_epi16(row1, row7); \\\n __m128i sum35 = _mm_add_epi16(row3, row5); \\\n dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \\\n dct_wadd(x4, y0o, y4o); \\\n dct_wadd(x5, y1o, y5o); \\\n dct_wadd(x6, y2o, y5o); \\\n dct_wadd(x7, y3o, y4o); \\\n dct_bfly32o(row0,row7, x0,x7,bias,shift); \\\n dct_bfly32o(row1,row6, x1,x6,bias,shift); \\\n dct_bfly32o(row2,row5, x2,x5,bias,shift); \\\n dct_bfly32o(row3,row4, x3,x4,bias,shift); \\\n }\n\n __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f));\n __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f));\n __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f));\n __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f));\n __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f));\n __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f));\n __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f));\n __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f));\n\n // rounding biases in column/row passes, see stbi__idct_block for explanation.\n __m128i bias_0 = _mm_set1_epi32(512);\n __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17));\n\n // load\n row0 = _mm_load_si128((const __m128i *) (data + 0*8));\n row1 = _mm_load_si128((const __m128i *) (data + 1*8));\n row2 = _mm_load_si128((const __m128i *) (data + 2*8));\n row3 = _mm_load_si128((const __m128i *) (data + 3*8));\n row4 = _mm_load_si128((const __m128i *) (data + 4*8));\n row5 = _mm_load_si128((const __m128i *) (data + 5*8));\n row6 = _mm_load_si128((const __m128i *) (data + 6*8));\n row7 = _mm_load_si128((const __m128i *) (data + 7*8));\n\n // column pass\n dct_pass(bias_0, 10);\n\n {\n // 16bit 8x8 transpose pass 1\n dct_interleave16(row0, row4);\n dct_interleave16(row1, row5);\n dct_interleave16(row2, row6);\n dct_interleave16(row3, row7);\n\n // transpose pass 2\n dct_interleave16(row0, row2);\n dct_interleave16(row1, row3);\n dct_interleave16(row4, row6);\n dct_interleave16(row5, row7);\n\n // transpose pass 3\n dct_interleave16(row0, row1);\n dct_interleave16(row2, row3);\n dct_interleave16(row4, row5);\n dct_interleave16(row6, row7);\n }\n\n // row pass\n dct_pass(bias_1, 17);\n\n {\n // pack\n __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7\n __m128i p1 = _mm_packus_epi16(row2, row3);\n __m128i p2 = _mm_packus_epi16(row4, row5);\n __m128i p3 = _mm_packus_epi16(row6, row7);\n\n // 8bit 8x8 transpose pass 1\n dct_interleave8(p0, p2); // a0e0a1e1...\n dct_interleave8(p1, p3); // c0g0c1g1...\n\n // transpose pass 2\n dct_interleave8(p0, p1); // a0c0e0g0...\n dct_interleave8(p2, p3); // b0d0f0h0...\n\n // transpose pass 3\n dct_interleave8(p0, p2); // a0b0c0d0...\n dct_interleave8(p1, p3); // a4b4c4d4...\n\n // store\n _mm_storel_epi64((__m128i *) out, p0); out += out_stride;\n _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride;\n _mm_storel_epi64((__m128i *) out, p2); out += out_stride;\n _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride;\n _mm_storel_epi64((__m128i *) out, p1); out += out_stride;\n _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride;\n _mm_storel_epi64((__m128i *) out, p3); out += out_stride;\n _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e));\n }\n\n#undef dct_const\n#undef dct_rot\n#undef dct_widen\n#undef dct_wadd\n#undef dct_wsub\n#undef dct_bfly32o\n#undef dct_interleave8\n#undef dct_interleave16\n#undef dct_pass\n}\n\n#endif // STBI_SSE2\n\n#ifdef STBI_NEON\n\n// NEON integer IDCT. should produce bit-identical\n// results to the generic C version.\nstatic void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64])\n{\n int16x8_t row0, row1, row2, row3, row4, row5, row6, row7;\n\n int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f));\n int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f));\n int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f));\n int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f));\n int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f));\n int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f));\n int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f));\n int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f));\n int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f));\n int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f));\n int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f));\n int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f));\n\n#define dct_long_mul(out, inq, coeff) \\\n int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \\\n int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff)\n\n#define dct_long_mac(out, acc, inq, coeff) \\\n int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \\\n int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff)\n\n#define dct_widen(out, inq) \\\n int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \\\n int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12)\n\n// wide add\n#define dct_wadd(out, a, b) \\\n int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \\\n int32x4_t out##_h = vaddq_s32(a##_h, b##_h)\n\n// wide sub\n#define dct_wsub(out, a, b) \\\n int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \\\n int32x4_t out##_h = vsubq_s32(a##_h, b##_h)\n\n// butterfly a/b, then shift using \"shiftop\" by \"s\" and pack\n#define dct_bfly32o(out0,out1, a,b,shiftop,s) \\\n { \\\n dct_wadd(sum, a, b); \\\n dct_wsub(dif, a, b); \\\n out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \\\n out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \\\n }\n\n#define dct_pass(shiftop, shift) \\\n { \\\n /* even part */ \\\n int16x8_t sum26 = vaddq_s16(row2, row6); \\\n dct_long_mul(p1e, sum26, rot0_0); \\\n dct_long_mac(t2e, p1e, row6, rot0_1); \\\n dct_long_mac(t3e, p1e, row2, rot0_2); \\\n int16x8_t sum04 = vaddq_s16(row0, row4); \\\n int16x8_t dif04 = vsubq_s16(row0, row4); \\\n dct_widen(t0e, sum04); \\\n dct_widen(t1e, dif04); \\\n dct_wadd(x0, t0e, t3e); \\\n dct_wsub(x3, t0e, t3e); \\\n dct_wadd(x1, t1e, t2e); \\\n dct_wsub(x2, t1e, t2e); \\\n /* odd part */ \\\n int16x8_t sum15 = vaddq_s16(row1, row5); \\\n int16x8_t sum17 = vaddq_s16(row1, row7); \\\n int16x8_t sum35 = vaddq_s16(row3, row5); \\\n int16x8_t sum37 = vaddq_s16(row3, row7); \\\n int16x8_t sumodd = vaddq_s16(sum17, sum35); \\\n dct_long_mul(p5o, sumodd, rot1_0); \\\n dct_long_mac(p1o, p5o, sum17, rot1_1); \\\n dct_long_mac(p2o, p5o, sum35, rot1_2); \\\n dct_long_mul(p3o, sum37, rot2_0); \\\n dct_long_mul(p4o, sum15, rot2_1); \\\n dct_wadd(sump13o, p1o, p3o); \\\n dct_wadd(sump24o, p2o, p4o); \\\n dct_wadd(sump23o, p2o, p3o); \\\n dct_wadd(sump14o, p1o, p4o); \\\n dct_long_mac(x4, sump13o, row7, rot3_0); \\\n dct_long_mac(x5, sump24o, row5, rot3_1); \\\n dct_long_mac(x6, sump23o, row3, rot3_2); \\\n dct_long_mac(x7, sump14o, row1, rot3_3); \\\n dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \\\n dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \\\n dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \\\n dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \\\n }\n\n // load\n row0 = vld1q_s16(data + 0*8);\n row1 = vld1q_s16(data + 1*8);\n row2 = vld1q_s16(data + 2*8);\n row3 = vld1q_s16(data + 3*8);\n row4 = vld1q_s16(data + 4*8);\n row5 = vld1q_s16(data + 5*8);\n row6 = vld1q_s16(data + 6*8);\n row7 = vld1q_s16(data + 7*8);\n\n // add DC bias\n row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0));\n\n // column pass\n dct_pass(vrshrn_n_s32, 10);\n\n // 16bit 8x8 transpose\n {\n// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively.\n// whether compilers actually get this is another story, sadly.\n#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; }\n#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); }\n#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); }\n\n // pass 1\n dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6\n dct_trn16(row2, row3);\n dct_trn16(row4, row5);\n dct_trn16(row6, row7);\n\n // pass 2\n dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4\n dct_trn32(row1, row3);\n dct_trn32(row4, row6);\n dct_trn32(row5, row7);\n\n // pass 3\n dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0\n dct_trn64(row1, row5);\n dct_trn64(row2, row6);\n dct_trn64(row3, row7);\n\n#undef dct_trn16\n#undef dct_trn32\n#undef dct_trn64\n }\n\n // row pass\n // vrshrn_n_s32 only supports shifts up to 16, we need\n // 17. so do a non-rounding shift of 16 first then follow\n // up with a rounding shift by 1.\n dct_pass(vshrn_n_s32, 16);\n\n {\n // pack and round\n uint8x8_t p0 = vqrshrun_n_s16(row0, 1);\n uint8x8_t p1 = vqrshrun_n_s16(row1, 1);\n uint8x8_t p2 = vqrshrun_n_s16(row2, 1);\n uint8x8_t p3 = vqrshrun_n_s16(row3, 1);\n uint8x8_t p4 = vqrshrun_n_s16(row4, 1);\n uint8x8_t p5 = vqrshrun_n_s16(row5, 1);\n uint8x8_t p6 = vqrshrun_n_s16(row6, 1);\n uint8x8_t p7 = vqrshrun_n_s16(row7, 1);\n\n // again, these can translate into one instruction, but often don't.\n#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; }\n#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); }\n#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); }\n\n // sadly can't use interleaved stores here since we only write\n // 8 bytes to each scan line!\n\n // 8x8 8-bit transpose pass 1\n dct_trn8_8(p0, p1);\n dct_trn8_8(p2, p3);\n dct_trn8_8(p4, p5);\n dct_trn8_8(p6, p7);\n\n // pass 2\n dct_trn8_16(p0, p2);\n dct_trn8_16(p1, p3);\n dct_trn8_16(p4, p6);\n dct_trn8_16(p5, p7);\n\n // pass 3\n dct_trn8_32(p0, p4);\n dct_trn8_32(p1, p5);\n dct_trn8_32(p2, p6);\n dct_trn8_32(p3, p7);\n\n // store\n vst1_u8(out, p0); out += out_stride;\n vst1_u8(out, p1); out += out_stride;\n vst1_u8(out, p2); out += out_stride;\n vst1_u8(out, p3); out += out_stride;\n vst1_u8(out, p4); out += out_stride;\n vst1_u8(out, p5); out += out_stride;\n vst1_u8(out, p6); out += out_stride;\n vst1_u8(out, p7);\n\n#undef dct_trn8_8\n#undef dct_trn8_16\n#undef dct_trn8_32\n }\n\n#undef dct_long_mul\n#undef dct_long_mac\n#undef dct_widen\n#undef dct_wadd\n#undef dct_wsub\n#undef dct_bfly32o\n#undef dct_pass\n}\n\n#endif // STBI_NEON\n\n#define STBI__MARKER_none 0xff\n// if there's a pending marker from the entropy stream, return that\n// otherwise, fetch from the stream and get a marker. if there's no\n// marker, return 0xff, which is never a valid marker value\nstatic stbi_uc stbi__get_marker(stbi__jpeg *j)\n{\n stbi_uc x;\n if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; }\n x = stbi__get8(j->s);\n if (x != 0xff) return STBI__MARKER_none;\n while (x == 0xff)\n x = stbi__get8(j->s);\n return x;\n}\n\n// in each scan, we'll have scan_n components, and the order\n// of the components is specified by order[]\n#define STBI__RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7)\n\n// after a restart interval, stbi__jpeg_reset the entropy decoder and\n// the dc prediction\nstatic void stbi__jpeg_reset(stbi__jpeg *j)\n{\n j->code_bits = 0;\n j->code_buffer = 0;\n j->nomore = 0;\n j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = 0;\n j->marker = STBI__MARKER_none;\n j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff;\n j->eob_run = 0;\n // no more than 1<<31 MCUs if no restart_interal? that's plenty safe,\n // since we don't even allow 1<<30 pixels\n}\n\nstatic int stbi__parse_entropy_coded_data(stbi__jpeg *z)\n{\n stbi__jpeg_reset(z);\n if (!z->progressive) {\n if (z->scan_n == 1) {\n int i,j;\n STBI_SIMD_ALIGN(short, data[64]);\n int n = z->order[0];\n // non-interleaved data, we just need to process one block at a time,\n // in trivial scanline order\n // number of blocks to do just depends on how many actual \"pixels\" this\n // component has, independent of interleaved MCU blocking and such\n int w = (z->img_comp[n].x+7) >> 3;\n int h = (z->img_comp[n].y+7) >> 3;\n for (j=0; j < h; ++j) {\n for (i=0; i < w; ++i) {\n int ha = z->img_comp[n].ha;\n if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;\n z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);\n // every data block is an MCU, so countdown the restart interval\n if (--z->todo <= 0) {\n if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);\n // if it's NOT a restart, then just bail, so we get corrupt data\n // rather than no data\n if (!STBI__RESTART(z->marker)) return 1;\n stbi__jpeg_reset(z);\n }\n }\n }\n return 1;\n } else { // interleaved\n int i,j,k,x,y;\n STBI_SIMD_ALIGN(short, data[64]);\n for (j=0; j < z->img_mcu_y; ++j) {\n for (i=0; i < z->img_mcu_x; ++i) {\n // scan an interleaved mcu... process scan_n components in order\n for (k=0; k < z->scan_n; ++k) {\n int n = z->order[k];\n // scan out an mcu's worth of this component; that's just determined\n // by the basic H and V specified for the component\n for (y=0; y < z->img_comp[n].v; ++y) {\n for (x=0; x < z->img_comp[n].h; ++x) {\n int x2 = (i*z->img_comp[n].h + x)*8;\n int y2 = (j*z->img_comp[n].v + y)*8;\n int ha = z->img_comp[n].ha;\n if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0;\n z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data);\n }\n }\n }\n // after all interleaved components, that's an interleaved MCU,\n // so now count down the restart interval\n if (--z->todo <= 0) {\n if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);\n if (!STBI__RESTART(z->marker)) return 1;\n stbi__jpeg_reset(z);\n }\n }\n }\n return 1;\n }\n } else {\n if (z->scan_n == 1) {\n int i,j;\n int n = z->order[0];\n // non-interleaved data, we just need to process one block at a time,\n // in trivial scanline order\n // number of blocks to do just depends on how many actual \"pixels\" this\n // component has, independent of interleaved MCU blocking and such\n int w = (z->img_comp[n].x+7) >> 3;\n int h = (z->img_comp[n].y+7) >> 3;\n for (j=0; j < h; ++j) {\n for (i=0; i < w; ++i) {\n short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);\n if (z->spec_start == 0) {\n if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))\n return 0;\n } else {\n int ha = z->img_comp[n].ha;\n if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha]))\n return 0;\n }\n // every data block is an MCU, so countdown the restart interval\n if (--z->todo <= 0) {\n if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);\n if (!STBI__RESTART(z->marker)) return 1;\n stbi__jpeg_reset(z);\n }\n }\n }\n return 1;\n } else { // interleaved\n int i,j,k,x,y;\n for (j=0; j < z->img_mcu_y; ++j) {\n for (i=0; i < z->img_mcu_x; ++i) {\n // scan an interleaved mcu... process scan_n components in order\n for (k=0; k < z->scan_n; ++k) {\n int n = z->order[k];\n // scan out an mcu's worth of this component; that's just determined\n // by the basic H and V specified for the component\n for (y=0; y < z->img_comp[n].v; ++y) {\n for (x=0; x < z->img_comp[n].h; ++x) {\n int x2 = (i*z->img_comp[n].h + x);\n int y2 = (j*z->img_comp[n].v + y);\n short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w);\n if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n))\n return 0;\n }\n }\n }\n // after all interleaved components, that's an interleaved MCU,\n // so now count down the restart interval\n if (--z->todo <= 0) {\n if (z->code_bits < 24) stbi__grow_buffer_unsafe(z);\n if (!STBI__RESTART(z->marker)) return 1;\n stbi__jpeg_reset(z);\n }\n }\n }\n return 1;\n }\n }\n}\n\nstatic void stbi__jpeg_dequantize(short *data, stbi_uc *dequant)\n{\n int i;\n for (i=0; i < 64; ++i)\n data[i] *= dequant[i];\n}\n\nstatic void stbi__jpeg_finish(stbi__jpeg *z)\n{\n if (z->progressive) {\n // dequantize and idct the data\n int i,j,n;\n for (n=0; n < z->s->img_n; ++n) {\n int w = (z->img_comp[n].x+7) >> 3;\n int h = (z->img_comp[n].y+7) >> 3;\n for (j=0; j < h; ++j) {\n for (i=0; i < w; ++i) {\n short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w);\n stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]);\n z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data);\n }\n }\n }\n }\n}\n\nstatic int stbi__process_marker(stbi__jpeg *z, int m)\n{\n int L;\n switch (m) {\n case STBI__MARKER_none: // no marker found\n return stbi__err(\"expected marker\",\"Corrupt JPEG\");\n\n case 0xDD: // DRI - specify restart interval\n if (stbi__get16be(z->s) != 4) return stbi__err(\"bad DRI len\",\"Corrupt JPEG\");\n z->restart_interval = stbi__get16be(z->s);\n return 1;\n\n case 0xDB: // DQT - define quantization table\n L = stbi__get16be(z->s)-2;\n while (L > 0) {\n int q = stbi__get8(z->s);\n int p = q >> 4;\n int t = q & 15,i;\n if (p != 0) return stbi__err(\"bad DQT type\",\"Corrupt JPEG\");\n if (t > 3) return stbi__err(\"bad DQT table\",\"Corrupt JPEG\");\n for (i=0; i < 64; ++i)\n z->dequant[t][stbi__jpeg_dezigzag[i]] = stbi__get8(z->s);\n L -= 65;\n }\n return L==0;\n\n case 0xC4: // DHT - define huffman table\n L = stbi__get16be(z->s)-2;\n while (L > 0) {\n stbi_uc *v;\n int sizes[16],i,n=0;\n int q = stbi__get8(z->s);\n int tc = q >> 4;\n int th = q & 15;\n if (tc > 1 || th > 3) return stbi__err(\"bad DHT header\",\"Corrupt JPEG\");\n for (i=0; i < 16; ++i) {\n sizes[i] = stbi__get8(z->s);\n n += sizes[i];\n }\n L -= 17;\n if (tc == 0) {\n if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0;\n v = z->huff_dc[th].values;\n } else {\n if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0;\n v = z->huff_ac[th].values;\n }\n for (i=0; i < n; ++i)\n v[i] = stbi__get8(z->s);\n if (tc != 0)\n stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th);\n L -= n;\n }\n return L==0;\n }\n // check for comment block or APP blocks\n if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) {\n stbi__skip(z->s, stbi__get16be(z->s)-2);\n return 1;\n }\n return 0;\n}\n\n// after we see SOS\nstatic int stbi__process_scan_header(stbi__jpeg *z)\n{\n int i;\n int Ls = stbi__get16be(z->s);\n z->scan_n = stbi__get8(z->s);\n if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err(\"bad SOS component count\",\"Corrupt JPEG\");\n if (Ls != 6+2*z->scan_n) return stbi__err(\"bad SOS len\",\"Corrupt JPEG\");\n for (i=0; i < z->scan_n; ++i) {\n int id = stbi__get8(z->s), which;\n int q = stbi__get8(z->s);\n for (which = 0; which < z->s->img_n; ++which)\n if (z->img_comp[which].id == id)\n break;\n if (which == z->s->img_n) return 0; // no match\n z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return stbi__err(\"bad DC huff\",\"Corrupt JPEG\");\n z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return stbi__err(\"bad AC huff\",\"Corrupt JPEG\");\n z->order[i] = which;\n }\n\n {\n int aa;\n z->spec_start = stbi__get8(z->s);\n z->spec_end = stbi__get8(z->s); // should be 63, but might be 0\n aa = stbi__get8(z->s);\n z->succ_high = (aa >> 4);\n z->succ_low = (aa & 15);\n if (z->progressive) {\n if (z->spec_start > 63 || z->spec_end > 63 || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13)\n return stbi__err(\"bad SOS\", \"Corrupt JPEG\");\n } else {\n if (z->spec_start != 0) return stbi__err(\"bad SOS\",\"Corrupt JPEG\");\n if (z->succ_high != 0 || z->succ_low != 0) return stbi__err(\"bad SOS\",\"Corrupt JPEG\");\n z->spec_end = 63;\n }\n }\n\n return 1;\n}\n\nstatic int stbi__process_frame_header(stbi__jpeg *z, int scan)\n{\n stbi__context *s = z->s;\n int Lf,p,i,q, h_max=1,v_max=1,c;\n Lf = stbi__get16be(s); if (Lf < 11) return stbi__err(\"bad SOF len\",\"Corrupt JPEG\"); // JPEG\n p = stbi__get8(s); if (p != 8) return stbi__err(\"only 8-bit\",\"JPEG format not supported: 8-bit only\"); // JPEG baseline\n s->img_y = stbi__get16be(s); if (s->img_y == 0) return stbi__err(\"no header height\", \"JPEG format not supported: delayed height\"); // Legal, but we don't handle it--but neither does IJG\n s->img_x = stbi__get16be(s); if (s->img_x == 0) return stbi__err(\"0 width\",\"Corrupt JPEG\"); // JPEG requires\n c = stbi__get8(s);\n if (c != 3 && c != 1) return stbi__err(\"bad component count\",\"Corrupt JPEG\"); // JFIF requires\n s->img_n = c;\n for (i=0; i < c; ++i) {\n z->img_comp[i].data = NULL;\n z->img_comp[i].linebuf = NULL;\n }\n\n if (Lf != 8+3*s->img_n) return stbi__err(\"bad SOF len\",\"Corrupt JPEG\");\n\n for (i=0; i < s->img_n; ++i) {\n z->img_comp[i].id = stbi__get8(s);\n if (z->img_comp[i].id != i+1) // JFIF requires\n if (z->img_comp[i].id != i) // some version of jpegtran outputs non-JFIF-compliant files!\n return stbi__err(\"bad component ID\",\"Corrupt JPEG\");\n q = stbi__get8(s);\n z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err(\"bad H\",\"Corrupt JPEG\");\n z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err(\"bad V\",\"Corrupt JPEG\");\n z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err(\"bad TQ\",\"Corrupt JPEG\");\n }\n\n if (scan != STBI__SCAN_load) return 1;\n\n if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err(\"too large\", \"Image too large to decode\");\n\n for (i=0; i < s->img_n; ++i) {\n if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h;\n if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v;\n }\n\n // compute interleaved mcu info\n z->img_h_max = h_max;\n z->img_v_max = v_max;\n z->img_mcu_w = h_max * 8;\n z->img_mcu_h = v_max * 8;\n z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w;\n z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h;\n\n for (i=0; i < s->img_n; ++i) {\n // number of effective pixels (e.g. for non-interleaved MCU)\n z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max;\n z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max;\n // to simplify generation, we'll allocate enough memory to decode\n // the bogus oversized data from using interleaved MCUs and their\n // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't\n // discard the extra data until colorspace conversion\n z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8;\n z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8;\n z->img_comp[i].raw_data = stbi__malloc(z->img_comp[i].w2 * z->img_comp[i].h2+15);\n\n if (z->img_comp[i].raw_data == NULL) {\n for(--i; i >= 0; --i) {\n STBI_FREE(z->img_comp[i].raw_data);\n z->img_comp[i].raw_data = NULL;\n }\n return stbi__err(\"outofmem\", \"Out of memory\");\n }\n // align blocks for idct using mmx/sse\n z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15);\n z->img_comp[i].linebuf = NULL;\n if (z->progressive) {\n z->img_comp[i].coeff_w = (z->img_comp[i].w2 + 7) >> 3;\n z->img_comp[i].coeff_h = (z->img_comp[i].h2 + 7) >> 3;\n z->img_comp[i].raw_coeff = STBI_MALLOC(z->img_comp[i].coeff_w * z->img_comp[i].coeff_h * 64 * sizeof(short) + 15);\n z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15);\n } else {\n z->img_comp[i].coeff = 0;\n z->img_comp[i].raw_coeff = 0;\n }\n }\n\n return 1;\n}\n\n// use comparisons since in some cases we handle more than one case (e.g. SOF)\n#define stbi__DNL(x) ((x) == 0xdc)\n#define stbi__SOI(x) ((x) == 0xd8)\n#define stbi__EOI(x) ((x) == 0xd9)\n#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2)\n#define stbi__SOS(x) ((x) == 0xda)\n\n#define stbi__SOF_progressive(x) ((x) == 0xc2)\n\nstatic int stbi__decode_jpeg_header(stbi__jpeg *z, int scan)\n{\n int m;\n z->marker = STBI__MARKER_none; // initialize cached marker to empty\n m = stbi__get_marker(z);\n if (!stbi__SOI(m)) return stbi__err(\"no SOI\",\"Corrupt JPEG\");\n if (scan == STBI__SCAN_type) return 1;\n m = stbi__get_marker(z);\n while (!stbi__SOF(m)) {\n if (!stbi__process_marker(z,m)) return 0;\n m = stbi__get_marker(z);\n while (m == STBI__MARKER_none) {\n // some files have extra padding after their blocks, so ok, we'll scan\n if (stbi__at_eof(z->s)) return stbi__err(\"no SOF\", \"Corrupt JPEG\");\n m = stbi__get_marker(z);\n }\n }\n z->progressive = stbi__SOF_progressive(m);\n if (!stbi__process_frame_header(z, scan)) return 0;\n return 1;\n}\n\n// decode image to YCbCr format\nstatic int stbi__decode_jpeg_image(stbi__jpeg *j)\n{\n int m;\n for (m = 0; m < 4; m++) {\n j->img_comp[m].raw_data = NULL;\n j->img_comp[m].raw_coeff = NULL;\n }\n j->restart_interval = 0;\n if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0;\n m = stbi__get_marker(j);\n while (!stbi__EOI(m)) {\n if (stbi__SOS(m)) {\n if (!stbi__process_scan_header(j)) return 0;\n if (!stbi__parse_entropy_coded_data(j)) return 0;\n if (j->marker == STBI__MARKER_none ) {\n // handle 0s at the end of image data from IP Kamera 9060\n while (!stbi__at_eof(j->s)) {\n int x = stbi__get8(j->s);\n if (x == 255) {\n j->marker = stbi__get8(j->s);\n break;\n } else if (x != 0) {\n return stbi__err(\"junk before marker\", \"Corrupt JPEG\");\n }\n }\n // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0\n }\n } else {\n if (!stbi__process_marker(j, m)) return 0;\n }\n m = stbi__get_marker(j);\n }\n if (j->progressive)\n stbi__jpeg_finish(j);\n return 1;\n}\n\n// static jfif-centered resampling (across block boundaries)\n\ntypedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1,\n int w, int hs);\n\n#define stbi__div4(x) ((stbi_uc) ((x) >> 2))\n\nstatic stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\n{\n STBI_NOTUSED(out);\n STBI_NOTUSED(in_far);\n STBI_NOTUSED(w);\n STBI_NOTUSED(hs);\n return in_near;\n}\n\nstatic stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\n{\n // need to generate two samples vertically for every one in input\n int i;\n STBI_NOTUSED(hs);\n for (i=0; i < w; ++i)\n out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2);\n return out;\n}\n\nstatic stbi_uc* stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\n{\n // need to generate two samples horizontally for every one in input\n int i;\n stbi_uc *input = in_near;\n\n if (w == 1) {\n // if only one sample, can't do any interpolation\n out[0] = out[1] = input[0];\n return out;\n }\n\n out[0] = input[0];\n out[1] = stbi__div4(input[0]*3 + input[1] + 2);\n for (i=1; i < w-1; ++i) {\n int n = 3*input[i]+2;\n out[i*2+0] = stbi__div4(n+input[i-1]);\n out[i*2+1] = stbi__div4(n+input[i+1]);\n }\n out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2);\n out[i*2+1] = input[w-1];\n\n STBI_NOTUSED(in_far);\n STBI_NOTUSED(hs);\n\n return out;\n}\n\n#define stbi__div16(x) ((stbi_uc) ((x) >> 4))\n\nstatic stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\n{\n // need to generate 2x2 samples for every one in input\n int i,t0,t1;\n if (w == 1) {\n out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);\n return out;\n }\n\n t1 = 3*in_near[0] + in_far[0];\n out[0] = stbi__div4(t1+2);\n for (i=1; i < w; ++i) {\n t0 = t1;\n t1 = 3*in_near[i]+in_far[i];\n out[i*2-1] = stbi__div16(3*t0 + t1 + 8);\n out[i*2 ] = stbi__div16(3*t1 + t0 + 8);\n }\n out[w*2-1] = stbi__div4(t1+2);\n\n STBI_NOTUSED(hs);\n\n return out;\n}\n\n#if defined(STBI_SSE2) || defined(STBI_NEON)\nstatic stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\n{\n // need to generate 2x2 samples for every one in input\n int i=0,t0,t1;\n\n if (w == 1) {\n out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2);\n return out;\n }\n\n t1 = 3*in_near[0] + in_far[0];\n // process groups of 8 pixels for as long as we can.\n // note we can't handle the last pixel in a row in this loop\n // because we need to handle the filter boundary conditions.\n for (; i < ((w-1) & ~7); i += 8) {\n#if defined(STBI_SSE2)\n // load and perform the vertical filtering pass\n // this uses 3*x + y = 4*x + (y - x)\n __m128i zero = _mm_setzero_si128();\n __m128i farb = _mm_loadl_epi64((__m128i *) (in_far + i));\n __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i));\n __m128i farw = _mm_unpacklo_epi8(farb, zero);\n __m128i nearw = _mm_unpacklo_epi8(nearb, zero);\n __m128i diff = _mm_sub_epi16(farw, nearw);\n __m128i nears = _mm_slli_epi16(nearw, 2);\n __m128i curr = _mm_add_epi16(nears, diff); // current row\n\n // horizontal filter works the same based on shifted vers of current\n // row. \"prev\" is current row shifted right by 1 pixel; we need to\n // insert the previous pixel value (from t1).\n // \"next\" is current row shifted left by 1 pixel, with first pixel\n // of next block of 8 pixels added in.\n __m128i prv0 = _mm_slli_si128(curr, 2);\n __m128i nxt0 = _mm_srli_si128(curr, 2);\n __m128i prev = _mm_insert_epi16(prv0, t1, 0);\n __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7);\n\n // horizontal filter, polyphase implementation since it's convenient:\n // even pixels = 3*cur + prev = cur*4 + (prev - cur)\n // odd pixels = 3*cur + next = cur*4 + (next - cur)\n // note the shared term.\n __m128i bias = _mm_set1_epi16(8);\n __m128i curs = _mm_slli_epi16(curr, 2);\n __m128i prvd = _mm_sub_epi16(prev, curr);\n __m128i nxtd = _mm_sub_epi16(next, curr);\n __m128i curb = _mm_add_epi16(curs, bias);\n __m128i even = _mm_add_epi16(prvd, curb);\n __m128i odd = _mm_add_epi16(nxtd, curb);\n\n // interleave even and odd pixels, then undo scaling.\n __m128i int0 = _mm_unpacklo_epi16(even, odd);\n __m128i int1 = _mm_unpackhi_epi16(even, odd);\n __m128i de0 = _mm_srli_epi16(int0, 4);\n __m128i de1 = _mm_srli_epi16(int1, 4);\n\n // pack and write output\n __m128i outv = _mm_packus_epi16(de0, de1);\n _mm_storeu_si128((__m128i *) (out + i*2), outv);\n#elif defined(STBI_NEON)\n // load and perform the vertical filtering pass\n // this uses 3*x + y = 4*x + (y - x)\n uint8x8_t farb = vld1_u8(in_far + i);\n uint8x8_t nearb = vld1_u8(in_near + i);\n int16x8_t diff = vreinterpretq_s16_u16(vsubl_u8(farb, nearb));\n int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2));\n int16x8_t curr = vaddq_s16(nears, diff); // current row\n\n // horizontal filter works the same based on shifted vers of current\n // row. \"prev\" is current row shifted right by 1 pixel; we need to\n // insert the previous pixel value (from t1).\n // \"next\" is current row shifted left by 1 pixel, with first pixel\n // of next block of 8 pixels added in.\n int16x8_t prv0 = vextq_s16(curr, curr, 7);\n int16x8_t nxt0 = vextq_s16(curr, curr, 1);\n int16x8_t prev = vsetq_lane_s16(t1, prv0, 0);\n int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7);\n\n // horizontal filter, polyphase implementation since it's convenient:\n // even pixels = 3*cur + prev = cur*4 + (prev - cur)\n // odd pixels = 3*cur + next = cur*4 + (next - cur)\n // note the shared term.\n int16x8_t curs = vshlq_n_s16(curr, 2);\n int16x8_t prvd = vsubq_s16(prev, curr);\n int16x8_t nxtd = vsubq_s16(next, curr);\n int16x8_t even = vaddq_s16(curs, prvd);\n int16x8_t odd = vaddq_s16(curs, nxtd);\n\n // undo scaling and round, then store with even/odd phases interleaved\n uint8x8x2_t o;\n o.val[0] = vqrshrun_n_s16(even, 4);\n o.val[1] = vqrshrun_n_s16(odd, 4);\n vst2_u8(out + i*2, o);\n#endif\n\n // \"previous\" value for next iter\n t1 = 3*in_near[i+7] + in_far[i+7];\n }\n\n t0 = t1;\n t1 = 3*in_near[i] + in_far[i];\n out[i*2] = stbi__div16(3*t1 + t0 + 8);\n\n for (++i; i < w; ++i) {\n t0 = t1;\n t1 = 3*in_near[i]+in_far[i];\n out[i*2-1] = stbi__div16(3*t0 + t1 + 8);\n out[i*2 ] = stbi__div16(3*t1 + t0 + 8);\n }\n out[w*2-1] = stbi__div4(t1+2);\n\n STBI_NOTUSED(hs);\n\n return out;\n}\n#endif\n\nstatic stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs)\n{\n // resample with nearest-neighbor\n int i,j;\n STBI_NOTUSED(in_far);\n for (i=0; i < w; ++i)\n for (j=0; j < hs; ++j)\n out[i*hs+j] = in_near[i];\n return out;\n}\n\n#ifdef STBI_JPEG_OLD\n// this is the same YCbCr-to-RGB calculation that stb_image has used\n// historically before the algorithm changes in 1.49\n#define float2fixed(x) ((int) ((x) * 65536 + 0.5))\nstatic void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step)\n{\n int i;\n for (i=0; i < count; ++i) {\n int y_fixed = (y[i] << 16) + 32768; // rounding\n int r,g,b;\n int cr = pcr[i] - 128;\n int cb = pcb[i] - 128;\n r = y_fixed + cr*float2fixed(1.40200f);\n g = y_fixed - cr*float2fixed(0.71414f) - cb*float2fixed(0.34414f);\n b = y_fixed + cb*float2fixed(1.77200f);\n r >>= 16;\n g >>= 16;\n b >>= 16;\n if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }\n if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }\n if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }\n out[0] = (stbi_uc)r;\n out[1] = (stbi_uc)g;\n out[2] = (stbi_uc)b;\n out[3] = 255;\n out += step;\n }\n}\n#else\n// this is a reduced-precision calculation of YCbCr-to-RGB introduced\n// to make sure the code produces the same results in both SIMD and scalar\n#define float2fixed(x) (((int) ((x) * 4096.0f + 0.5f)) << 8)\nstatic void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step)\n{\n int i;\n for (i=0; i < count; ++i) {\n int y_fixed = (y[i] << 20) + (1<<19); // rounding\n int r,g,b;\n int cr = pcr[i] - 128;\n int cb = pcb[i] - 128;\n r = y_fixed + cr* float2fixed(1.40200f);\n g = y_fixed + (cr*-float2fixed(0.71414f)) + ((cb*-float2fixed(0.34414f)) & 0xffff0000);\n b = y_fixed + cb* float2fixed(1.77200f);\n r >>= 20;\n g >>= 20;\n b >>= 20;\n if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }\n if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }\n if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }\n out[0] = (stbi_uc)r;\n out[1] = (stbi_uc)g;\n out[2] = (stbi_uc)b;\n out[3] = 255;\n out += step;\n }\n}\n#endif\n\n#if defined(STBI_SSE2) || defined(STBI_NEON)\nstatic void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step)\n{\n int i = 0;\n\n#ifdef STBI_SSE2\n // step == 3 is pretty ugly on the final interleave, and i'm not convinced\n // it's useful in practice (you wouldn't use it for textures, for example).\n // so just accelerate step == 4 case.\n if (step == 4) {\n // this is a fairly straightforward implementation and not super-optimized.\n __m128i signflip = _mm_set1_epi8(-0x80);\n __m128i cr_const0 = _mm_set1_epi16( (short) ( 1.40200f*4096.0f+0.5f));\n __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f));\n __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f));\n __m128i cb_const1 = _mm_set1_epi16( (short) ( 1.77200f*4096.0f+0.5f));\n __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128);\n __m128i xw = _mm_set1_epi16(255); // alpha channel\n\n for (; i+7 < count; i += 8) {\n // load\n __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i));\n __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i));\n __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i));\n __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128\n __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128\n\n // unpack to short (and left-shift cr, cb by 8)\n __m128i yw = _mm_unpacklo_epi8(y_bias, y_bytes);\n __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased);\n __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased);\n\n // color transform\n __m128i yws = _mm_srli_epi16(yw, 4);\n __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw);\n __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw);\n __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1);\n __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1);\n __m128i rws = _mm_add_epi16(cr0, yws);\n __m128i gwt = _mm_add_epi16(cb0, yws);\n __m128i bws = _mm_add_epi16(yws, cb1);\n __m128i gws = _mm_add_epi16(gwt, cr1);\n\n // descale\n __m128i rw = _mm_srai_epi16(rws, 4);\n __m128i bw = _mm_srai_epi16(bws, 4);\n __m128i gw = _mm_srai_epi16(gws, 4);\n\n // back to byte, set up for transpose\n __m128i brb = _mm_packus_epi16(rw, bw);\n __m128i gxb = _mm_packus_epi16(gw, xw);\n\n // transpose to interleave channels\n __m128i t0 = _mm_unpacklo_epi8(brb, gxb);\n __m128i t1 = _mm_unpackhi_epi8(brb, gxb);\n __m128i o0 = _mm_unpacklo_epi16(t0, t1);\n __m128i o1 = _mm_unpackhi_epi16(t0, t1);\n\n // store\n _mm_storeu_si128((__m128i *) (out + 0), o0);\n _mm_storeu_si128((__m128i *) (out + 16), o1);\n out += 32;\n }\n }\n#endif\n\n#ifdef STBI_NEON\n // in this version, step=3 support would be easy to add. but is there demand?\n if (step == 4) {\n // this is a fairly straightforward implementation and not super-optimized.\n uint8x8_t signflip = vdup_n_u8(0x80);\n int16x8_t cr_const0 = vdupq_n_s16( (short) ( 1.40200f*4096.0f+0.5f));\n int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f));\n int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f));\n int16x8_t cb_const1 = vdupq_n_s16( (short) ( 1.77200f*4096.0f+0.5f));\n\n for (; i+7 < count; i += 8) {\n // load\n uint8x8_t y_bytes = vld1_u8(y + i);\n uint8x8_t cr_bytes = vld1_u8(pcr + i);\n uint8x8_t cb_bytes = vld1_u8(pcb + i);\n int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip));\n int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip));\n\n // expand to s16\n int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4));\n int16x8_t crw = vshll_n_s8(cr_biased, 7);\n int16x8_t cbw = vshll_n_s8(cb_biased, 7);\n\n // color transform\n int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0);\n int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0);\n int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1);\n int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1);\n int16x8_t rws = vaddq_s16(yws, cr0);\n int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1);\n int16x8_t bws = vaddq_s16(yws, cb1);\n\n // undo scaling, round, convert to byte\n uint8x8x4_t o;\n o.val[0] = vqrshrun_n_s16(rws, 4);\n o.val[1] = vqrshrun_n_s16(gws, 4);\n o.val[2] = vqrshrun_n_s16(bws, 4);\n o.val[3] = vdup_n_u8(255);\n\n // store, interleaving r/g/b/a\n vst4_u8(out, o);\n out += 8*4;\n }\n }\n#endif\n\n for (; i < count; ++i) {\n int y_fixed = (y[i] << 20) + (1<<19); // rounding\n int r,g,b;\n int cr = pcr[i] - 128;\n int cb = pcb[i] - 128;\n r = y_fixed + cr* float2fixed(1.40200f);\n g = y_fixed + cr*-float2fixed(0.71414f) + ((cb*-float2fixed(0.34414f)) & 0xffff0000);\n b = y_fixed + cb* float2fixed(1.77200f);\n r >>= 20;\n g >>= 20;\n b >>= 20;\n if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; }\n if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; }\n if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; }\n out[0] = (stbi_uc)r;\n out[1] = (stbi_uc)g;\n out[2] = (stbi_uc)b;\n out[3] = 255;\n out += step;\n }\n}\n#endif\n\n// set up the kernels\nstatic void stbi__setup_jpeg(stbi__jpeg *j)\n{\n j->idct_block_kernel = stbi__idct_block;\n j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row;\n j->resample_row_hv_2_kernel = stbi__resample_row_hv_2;\n\n#ifdef STBI_SSE2\n if (stbi__sse2_available()) {\n j->idct_block_kernel = stbi__idct_simd;\n #ifndef STBI_JPEG_OLD\n j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;\n #endif\n j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;\n }\n#endif\n\n#ifdef STBI_NEON\n j->idct_block_kernel = stbi__idct_simd;\n #ifndef STBI_JPEG_OLD\n j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd;\n #endif\n j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd;\n#endif\n}\n\n// clean up the temporary component buffers\nstatic void stbi__cleanup_jpeg(stbi__jpeg *j)\n{\n int i;\n for (i=0; i < j->s->img_n; ++i) {\n if (j->img_comp[i].raw_data) {\n STBI_FREE(j->img_comp[i].raw_data);\n j->img_comp[i].raw_data = NULL;\n j->img_comp[i].data = NULL;\n }\n if (j->img_comp[i].raw_coeff) {\n STBI_FREE(j->img_comp[i].raw_coeff);\n j->img_comp[i].raw_coeff = 0;\n j->img_comp[i].coeff = 0;\n }\n if (j->img_comp[i].linebuf) {\n STBI_FREE(j->img_comp[i].linebuf);\n j->img_comp[i].linebuf = NULL;\n }\n }\n}\n\ntypedef struct\n{\n resample_row_func resample;\n stbi_uc *line0,*line1;\n int hs,vs; // expansion factor in each axis\n int w_lores; // horizontal pixels pre-expansion\n int ystep; // how far through vertical expansion we are\n int ypos; // which pre-expansion row we're on\n} stbi__resample;\n\nstatic stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp)\n{\n int n, decode_n;\n z->s->img_n = 0; // make stbi__cleanup_jpeg safe\n\n // validate req_comp\n if (req_comp < 0 || req_comp > 4) return stbi__errpuc(\"bad req_comp\", \"Internal error\");\n\n // load a jpeg image from whichever source, but leave in YCbCr format\n if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; }\n\n // determine actual number of components to generate\n n = req_comp ? req_comp : z->s->img_n;\n\n if (z->s->img_n == 3 && n < 3)\n decode_n = 1;\n else\n decode_n = z->s->img_n;\n\n // resample and color-convert\n {\n int k;\n unsigned int i,j;\n stbi_uc *output;\n stbi_uc *coutput[4];\n\n stbi__resample res_comp[4];\n\n for (k=0; k < decode_n; ++k) {\n stbi__resample *r = &res_comp[k];\n\n // allocate line buffer big enough for upsampling off the edges\n // with upsample factor of 4\n z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3);\n if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc(\"outofmem\", \"Out of memory\"); }\n\n r->hs = z->img_h_max / z->img_comp[k].h;\n r->vs = z->img_v_max / z->img_comp[k].v;\n r->ystep = r->vs >> 1;\n r->w_lores = (z->s->img_x + r->hs-1) / r->hs;\n r->ypos = 0;\n r->line0 = r->line1 = z->img_comp[k].data;\n\n if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1;\n else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2;\n else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2;\n else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel;\n else r->resample = stbi__resample_row_generic;\n }\n\n // can't error after this so, this is safe\n output = (stbi_uc *) stbi__malloc(n * z->s->img_x * z->s->img_y + 1);\n if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc(\"outofmem\", \"Out of memory\"); }\n\n // now go ahead and resample\n for (j=0; j < z->s->img_y; ++j) {\n stbi_uc *out = output + n * z->s->img_x * j;\n for (k=0; k < decode_n; ++k) {\n stbi__resample *r = &res_comp[k];\n int y_bot = r->ystep >= (r->vs >> 1);\n coutput[k] = r->resample(z->img_comp[k].linebuf,\n y_bot ? r->line1 : r->line0,\n y_bot ? r->line0 : r->line1,\n r->w_lores, r->hs);\n if (++r->ystep >= r->vs) {\n r->ystep = 0;\n r->line0 = r->line1;\n if (++r->ypos < z->img_comp[k].y)\n r->line1 += z->img_comp[k].w2;\n }\n }\n if (n >= 3) {\n stbi_uc *y = coutput[0];\n if (z->s->img_n == 3) {\n z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);\n } else\n for (i=0; i < z->s->img_x; ++i) {\n out[0] = out[1] = out[2] = y[i];\n out[3] = 255; // not used if n==3\n out += n;\n }\n } else {\n stbi_uc *y = coutput[0];\n if (n == 1)\n for (i=0; i < z->s->img_x; ++i) out[i] = y[i];\n else\n for (i=0; i < z->s->img_x; ++i) *out++ = y[i], *out++ = 255;\n }\n }\n stbi__cleanup_jpeg(z);\n *out_x = z->s->img_x;\n *out_y = z->s->img_y;\n if (comp) *comp = z->s->img_n; // report original components, not output\n return output;\n }\n}\n\nstatic unsigned char *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n stbi__jpeg j;\n j.s = s;\n stbi__setup_jpeg(&j);\n return load_jpeg_image(&j, x,y,comp,req_comp);\n}\n\nstatic int stbi__jpeg_test(stbi__context *s)\n{\n int r;\n stbi__jpeg j;\n j.s = s;\n stbi__setup_jpeg(&j);\n r = stbi__decode_jpeg_header(&j, STBI__SCAN_type);\n stbi__rewind(s);\n return r;\n}\n\nstatic int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp)\n{\n if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) {\n stbi__rewind( j->s );\n return 0;\n }\n if (x) *x = j->s->img_x;\n if (y) *y = j->s->img_y;\n if (comp) *comp = j->s->img_n;\n return 1;\n}\n\nstatic int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp)\n{\n stbi__jpeg j;\n j.s = s;\n return stbi__jpeg_info_raw(&j, x, y, comp);\n}\n#endif\n\n// public domain zlib decode v0.2 Sean Barrett 2006-11-18\n// simple implementation\n// - all input must be provided in an upfront buffer\n// - all output is written to a single output buffer (can malloc/realloc)\n// performance\n// - fast huffman\n\n#ifndef STBI_NO_ZLIB\n\n// fast-way is faster to check than jpeg huffman, but slow way is slower\n#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables\n#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1)\n\n// zlib-style huffman encoding\n// (jpegs packs from left, zlib from right, so can't share code)\ntypedef struct\n{\n stbi__uint16 fast[1 << STBI__ZFAST_BITS];\n stbi__uint16 firstcode[16];\n int maxcode[17];\n stbi__uint16 firstsymbol[16];\n stbi_uc size[288];\n stbi__uint16 value[288];\n} stbi__zhuffman;\n\nstbi_inline static int stbi__bitreverse16(int n)\n{\n n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1);\n n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2);\n n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4);\n n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8);\n return n;\n}\n\nstbi_inline static int stbi__bit_reverse(int v, int bits)\n{\n STBI_ASSERT(bits <= 16);\n // to bit reverse n bits, reverse 16 and shift\n // e.g. 11 bits, bit reverse and shift away 5\n return stbi__bitreverse16(v) >> (16-bits);\n}\n\nstatic int stbi__zbuild_huffman(stbi__zhuffman *z, stbi_uc *sizelist, int num)\n{\n int i,k=0;\n int code, next_code[16], sizes[17];\n\n // DEFLATE spec for generating codes\n memset(sizes, 0, sizeof(sizes));\n memset(z->fast, 0, sizeof(z->fast));\n for (i=0; i < num; ++i)\n ++sizes[sizelist[i]];\n sizes[0] = 0;\n for (i=1; i < 16; ++i)\n if (sizes[i] > (1 << i))\n return stbi__err(\"bad sizes\", \"Corrupt PNG\");\n code = 0;\n for (i=1; i < 16; ++i) {\n next_code[i] = code;\n z->firstcode[i] = (stbi__uint16) code;\n z->firstsymbol[i] = (stbi__uint16) k;\n code = (code + sizes[i]);\n if (sizes[i])\n if (code-1 >= (1 << i)) return stbi__err(\"bad codelengths\",\"Corrupt PNG\");\n z->maxcode[i] = code << (16-i); // preshift for inner loop\n code <<= 1;\n k += sizes[i];\n }\n z->maxcode[16] = 0x10000; // sentinel\n for (i=0; i < num; ++i) {\n int s = sizelist[i];\n if (s) {\n int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s];\n stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i);\n z->size [c] = (stbi_uc ) s;\n z->value[c] = (stbi__uint16) i;\n if (s <= STBI__ZFAST_BITS) {\n int j = stbi__bit_reverse(next_code[s],s);\n while (j < (1 << STBI__ZFAST_BITS)) {\n z->fast[j] = fastv;\n j += (1 << s);\n }\n }\n ++next_code[s];\n }\n }\n return 1;\n}\n\n// zlib-from-memory implementation for PNG reading\n// because PNG allows splitting the zlib stream arbitrarily,\n// and it's annoying structurally to have PNG call ZLIB call PNG,\n// we require PNG read all the IDATs and combine them into a single\n// memory buffer\n\ntypedef struct\n{\n stbi_uc *zbuffer, *zbuffer_end;\n int num_bits;\n stbi__uint32 code_buffer;\n\n char *zout;\n char *zout_start;\n char *zout_end;\n int z_expandable;\n\n stbi__zhuffman z_length, z_distance;\n} stbi__zbuf;\n\nstbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z)\n{\n if (z->zbuffer >= z->zbuffer_end) return 0;\n return *z->zbuffer++;\n}\n\nstatic void stbi__fill_bits(stbi__zbuf *z)\n{\n do {\n STBI_ASSERT(z->code_buffer < (1U << z->num_bits));\n z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits;\n z->num_bits += 8;\n } while (z->num_bits <= 24);\n}\n\nstbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n)\n{\n unsigned int k;\n if (z->num_bits < n) stbi__fill_bits(z);\n k = z->code_buffer & ((1 << n) - 1);\n z->code_buffer >>= n;\n z->num_bits -= n;\n return k;\n}\n\nstatic int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z)\n{\n int b,s,k;\n // not resolved by fast table, so compute it the slow way\n // use jpeg approach, which requires MSbits at top\n k = stbi__bit_reverse(a->code_buffer, 16);\n for (s=STBI__ZFAST_BITS+1; ; ++s)\n if (k < z->maxcode[s])\n break;\n if (s == 16) return -1; // invalid code!\n // code size is s, so:\n b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s];\n STBI_ASSERT(z->size[b] == s);\n a->code_buffer >>= s;\n a->num_bits -= s;\n return z->value[b];\n}\n\nstbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z)\n{\n int b,s;\n if (a->num_bits < 16) stbi__fill_bits(a);\n b = z->fast[a->code_buffer & STBI__ZFAST_MASK];\n if (b) {\n s = b >> 9;\n a->code_buffer >>= s;\n a->num_bits -= s;\n return b & 511;\n }\n return stbi__zhuffman_decode_slowpath(a, z);\n}\n\nstatic int stbi__zexpand(stbi__zbuf *z, char *zout, int n) // need to make room for n bytes\n{\n char *q;\n int cur, limit;\n z->zout = zout;\n if (!z->z_expandable) return stbi__err(\"output buffer limit\",\"Corrupt PNG\");\n cur = (int) (z->zout - z->zout_start);\n limit = (int) (z->zout_end - z->zout_start);\n while (cur + n > limit)\n limit *= 2;\n q = (char *) STBI_REALLOC(z->zout_start, limit);\n if (q == NULL) return stbi__err(\"outofmem\", \"Out of memory\");\n z->zout_start = q;\n z->zout = q + cur;\n z->zout_end = q + limit;\n return 1;\n}\n\nstatic int stbi__zlength_base[31] = {\n 3,4,5,6,7,8,9,10,11,13,\n 15,17,19,23,27,31,35,43,51,59,\n 67,83,99,115,131,163,195,227,258,0,0 };\n\nstatic int stbi__zlength_extra[31]=\n{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 };\n\nstatic int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,\n257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0};\n\nstatic int stbi__zdist_extra[32] =\n{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};\n\nstatic int stbi__parse_huffman_block(stbi__zbuf *a)\n{\n char *zout = a->zout;\n for(;;) {\n int z = stbi__zhuffman_decode(a, &a->z_length);\n if (z < 256) {\n if (z < 0) return stbi__err(\"bad huffman code\",\"Corrupt PNG\"); // error in huffman codes\n if (zout >= a->zout_end) {\n if (!stbi__zexpand(a, zout, 1)) return 0;\n zout = a->zout;\n }\n *zout++ = (char) z;\n } else {\n stbi_uc *p;\n int len,dist;\n if (z == 256) {\n a->zout = zout;\n return 1;\n }\n z -= 257;\n len = stbi__zlength_base[z];\n if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]);\n z = stbi__zhuffman_decode(a, &a->z_distance);\n if (z < 0) return stbi__err(\"bad huffman code\",\"Corrupt PNG\");\n dist = stbi__zdist_base[z];\n if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]);\n if (zout - a->zout_start < dist) return stbi__err(\"bad dist\",\"Corrupt PNG\");\n if (zout + len > a->zout_end) {\n if (!stbi__zexpand(a, zout, len)) return 0;\n zout = a->zout;\n }\n p = (stbi_uc *) (zout - dist);\n if (dist == 1) { // run of one byte; common in images.\n stbi_uc v = *p;\n if (len) { do *zout++ = v; while (--len); }\n } else {\n if (len) { do *zout++ = *p++; while (--len); }\n }\n }\n }\n}\n\nstatic int stbi__compute_huffman_codes(stbi__zbuf *a)\n{\n static stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 };\n stbi__zhuffman z_codelength;\n stbi_uc lencodes[286+32+137];//padding for maximum single op\n stbi_uc codelength_sizes[19];\n int i,n;\n\n int hlit = stbi__zreceive(a,5) + 257;\n int hdist = stbi__zreceive(a,5) + 1;\n int hclen = stbi__zreceive(a,4) + 4;\n\n memset(codelength_sizes, 0, sizeof(codelength_sizes));\n for (i=0; i < hclen; ++i) {\n int s = stbi__zreceive(a,3);\n codelength_sizes[length_dezigzag[i]] = (stbi_uc) s;\n }\n if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0;\n\n n = 0;\n while (n < hlit + hdist) {\n int c = stbi__zhuffman_decode(a, &z_codelength);\n if (c < 0 || c >= 19) return stbi__err(\"bad codelengths\", \"Corrupt PNG\");\n if (c < 16)\n lencodes[n++] = (stbi_uc) c;\n else if (c == 16) {\n c = stbi__zreceive(a,2)+3;\n memset(lencodes+n, lencodes[n-1], c);\n n += c;\n } else if (c == 17) {\n c = stbi__zreceive(a,3)+3;\n memset(lencodes+n, 0, c);\n n += c;\n } else {\n STBI_ASSERT(c == 18);\n c = stbi__zreceive(a,7)+11;\n memset(lencodes+n, 0, c);\n n += c;\n }\n }\n if (n != hlit+hdist) return stbi__err(\"bad codelengths\",\"Corrupt PNG\");\n if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0;\n if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0;\n return 1;\n}\n\nstatic int stbi__parse_uncomperssed_block(stbi__zbuf *a)\n{\n stbi_uc header[4];\n int len,nlen,k;\n if (a->num_bits & 7)\n stbi__zreceive(a, a->num_bits & 7); // discard\n // drain the bit-packed data into header\n k = 0;\n while (a->num_bits > 0) {\n header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check\n a->code_buffer >>= 8;\n a->num_bits -= 8;\n }\n STBI_ASSERT(a->num_bits == 0);\n // now fill header the normal way\n while (k < 4)\n header[k++] = stbi__zget8(a);\n len = header[1] * 256 + header[0];\n nlen = header[3] * 256 + header[2];\n if (nlen != (len ^ 0xffff)) return stbi__err(\"zlib corrupt\",\"Corrupt PNG\");\n if (a->zbuffer + len > a->zbuffer_end) return stbi__err(\"read past buffer\",\"Corrupt PNG\");\n if (a->zout + len > a->zout_end)\n if (!stbi__zexpand(a, a->zout, len)) return 0;\n memcpy(a->zout, a->zbuffer, len);\n a->zbuffer += len;\n a->zout += len;\n return 1;\n}\n\nstatic int stbi__parse_zlib_header(stbi__zbuf *a)\n{\n int cmf = stbi__zget8(a);\n int cm = cmf & 15;\n /* int cinfo = cmf >> 4; */\n int flg = stbi__zget8(a);\n if ((cmf*256+flg) % 31 != 0) return stbi__err(\"bad zlib header\",\"Corrupt PNG\"); // zlib spec\n if (flg & 32) return stbi__err(\"no preset dict\",\"Corrupt PNG\"); // preset dictionary not allowed in png\n if (cm != 8) return stbi__err(\"bad compression\",\"Corrupt PNG\"); // DEFLATE required for png\n // window = 1 << (8 + cinfo)... but who cares, we fully buffer output\n return 1;\n}\n\n// @TODO: should statically initialize these for optimal thread safety\nstatic stbi_uc stbi__zdefault_length[288], stbi__zdefault_distance[32];\nstatic void stbi__init_zdefaults(void)\n{\n int i; // use <= to match clearly with spec\n for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8;\n for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9;\n for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7;\n for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8;\n\n for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5;\n}\n\nstatic int stbi__parse_zlib(stbi__zbuf *a, int parse_header)\n{\n int final, type;\n if (parse_header)\n if (!stbi__parse_zlib_header(a)) return 0;\n a->num_bits = 0;\n a->code_buffer = 0;\n do {\n final = stbi__zreceive(a,1);\n type = stbi__zreceive(a,2);\n if (type == 0) {\n if (!stbi__parse_uncomperssed_block(a)) return 0;\n } else if (type == 3) {\n return 0;\n } else {\n if (type == 1) {\n // use fixed code lengths\n if (!stbi__zdefault_distance[31]) stbi__init_zdefaults();\n if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , 288)) return 0;\n if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0;\n } else {\n if (!stbi__compute_huffman_codes(a)) return 0;\n }\n if (!stbi__parse_huffman_block(a)) return 0;\n }\n } while (!final);\n return 1;\n}\n\nstatic int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header)\n{\n a->zout_start = obuf;\n a->zout = obuf;\n a->zout_end = obuf + olen;\n a->z_expandable = exp;\n\n return stbi__parse_zlib(a, parse_header);\n}\n\nSTBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen)\n{\n stbi__zbuf a;\n char *p = (char *) stbi__malloc(initial_size);\n if (p == NULL) return NULL;\n a.zbuffer = (stbi_uc *) buffer;\n a.zbuffer_end = (stbi_uc *) buffer + len;\n if (stbi__do_zlib(&a, p, initial_size, 1, 1)) {\n if (outlen) *outlen = (int) (a.zout - a.zout_start);\n return a.zout_start;\n } else {\n STBI_FREE(a.zout_start);\n return NULL;\n }\n}\n\nSTBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen)\n{\n return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen);\n}\n\nSTBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header)\n{\n stbi__zbuf a;\n char *p = (char *) stbi__malloc(initial_size);\n if (p == NULL) return NULL;\n a.zbuffer = (stbi_uc *) buffer;\n a.zbuffer_end = (stbi_uc *) buffer + len;\n if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) {\n if (outlen) *outlen = (int) (a.zout - a.zout_start);\n return a.zout_start;\n } else {\n STBI_FREE(a.zout_start);\n return NULL;\n }\n}\n\nSTBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen)\n{\n stbi__zbuf a;\n a.zbuffer = (stbi_uc *) ibuffer;\n a.zbuffer_end = (stbi_uc *) ibuffer + ilen;\n if (stbi__do_zlib(&a, obuffer, olen, 0, 1))\n return (int) (a.zout - a.zout_start);\n else\n return -1;\n}\n\nSTBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen)\n{\n stbi__zbuf a;\n char *p = (char *) stbi__malloc(16384);\n if (p == NULL) return NULL;\n a.zbuffer = (stbi_uc *) buffer;\n a.zbuffer_end = (stbi_uc *) buffer+len;\n if (stbi__do_zlib(&a, p, 16384, 1, 0)) {\n if (outlen) *outlen = (int) (a.zout - a.zout_start);\n return a.zout_start;\n } else {\n STBI_FREE(a.zout_start);\n return NULL;\n }\n}\n\nSTBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen)\n{\n stbi__zbuf a;\n a.zbuffer = (stbi_uc *) ibuffer;\n a.zbuffer_end = (stbi_uc *) ibuffer + ilen;\n if (stbi__do_zlib(&a, obuffer, olen, 0, 0))\n return (int) (a.zout - a.zout_start);\n else\n return -1;\n}\n#endif\n\n// public domain \"baseline\" PNG decoder v0.10 Sean Barrett 2006-11-18\n// simple implementation\n// - only 8-bit samples\n// - no CRC checking\n// - allocates lots of intermediate memory\n// - avoids problem of streaming data between subsystems\n// - avoids explicit window management\n// performance\n// - uses stb_zlib, a PD zlib implementation with fast huffman decoding\n\n#ifndef STBI_NO_PNG\ntypedef struct\n{\n stbi__uint32 length;\n stbi__uint32 type;\n} stbi__pngchunk;\n\nstatic stbi__pngchunk stbi__get_chunk_header(stbi__context *s)\n{\n stbi__pngchunk c;\n c.length = stbi__get32be(s);\n c.type = stbi__get32be(s);\n return c;\n}\n\nstatic int stbi__check_png_header(stbi__context *s)\n{\n static stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 };\n int i;\n for (i=0; i < 8; ++i)\n if (stbi__get8(s) != png_sig[i]) return stbi__err(\"bad png sig\",\"Not a PNG\");\n return 1;\n}\n\ntypedef struct\n{\n stbi__context *s;\n stbi_uc *idata, *expanded, *out;\n} stbi__png;\n\n\nenum {\n STBI__F_none=0,\n STBI__F_sub=1,\n STBI__F_up=2,\n STBI__F_avg=3,\n STBI__F_paeth=4,\n // synthetic filters used for first scanline to avoid needing a dummy row of 0s\n STBI__F_avg_first,\n STBI__F_paeth_first\n};\n\nstatic stbi_uc first_row_filter[5] =\n{\n STBI__F_none,\n STBI__F_sub,\n STBI__F_none,\n STBI__F_avg_first,\n STBI__F_paeth_first\n};\n\nstatic int stbi__paeth(int a, int b, int c)\n{\n int p = a + b - c;\n int pa = abs(p-a);\n int pb = abs(p-b);\n int pc = abs(p-c);\n if (pa <= pb && pa <= pc) return a;\n if (pb <= pc) return b;\n return c;\n}\n\nstatic stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 };\n\n// create the png data from post-deflated data\nstatic int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color)\n{\n stbi__context *s = a->s;\n stbi__uint32 i,j,stride = x*out_n;\n stbi__uint32 img_len, img_width_bytes;\n int k;\n int img_n = s->img_n; // copy it into a local for later\n\n STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1);\n a->out = (stbi_uc *) stbi__malloc(x * y * out_n); // extra bytes to write off the end into\n if (!a->out) return stbi__err(\"outofmem\", \"Out of memory\");\n\n img_width_bytes = (((img_n * x * depth) + 7) >> 3);\n img_len = (img_width_bytes + 1) * y;\n if (s->img_x == x && s->img_y == y) {\n if (raw_len != img_len) return stbi__err(\"not enough pixels\",\"Corrupt PNG\");\n } else { // interlaced:\n if (raw_len < img_len) return stbi__err(\"not enough pixels\",\"Corrupt PNG\");\n }\n\n for (j=0; j < y; ++j) {\n stbi_uc *cur = a->out + stride*j;\n stbi_uc *prior = cur - stride;\n int filter = *raw++;\n int filter_bytes = img_n;\n int width = x;\n if (filter > 4)\n return stbi__err(\"invalid filter\",\"Corrupt PNG\");\n\n if (depth < 8) {\n STBI_ASSERT(img_width_bytes <= x);\n cur += x*out_n - img_width_bytes; // store output to the rightmost img_len bytes, so we can decode in place\n filter_bytes = 1;\n width = img_width_bytes;\n }\n\n // if first row, use special filter that doesn't sample previous row\n if (j == 0) filter = first_row_filter[filter];\n\n // handle first byte explicitly\n for (k=0; k < filter_bytes; ++k) {\n switch (filter) {\n case STBI__F_none : cur[k] = raw[k]; break;\n case STBI__F_sub : cur[k] = raw[k]; break;\n case STBI__F_up : cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break;\n case STBI__F_avg : cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); break;\n case STBI__F_paeth : cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(0,prior[k],0)); break;\n case STBI__F_avg_first : cur[k] = raw[k]; break;\n case STBI__F_paeth_first: cur[k] = raw[k]; break;\n }\n }\n\n if (depth == 8) {\n if (img_n != out_n)\n cur[img_n] = 255; // first pixel\n raw += img_n;\n cur += out_n;\n prior += out_n;\n } else {\n raw += 1;\n cur += 1;\n prior += 1;\n }\n\n // this is a little gross, so that we don't switch per-pixel or per-component\n if (depth < 8 || img_n == out_n) {\n int nk = (width - 1)*img_n;\n #define CASE(f) \\\n case f: \\\n for (k=0; k < nk; ++k)\n switch (filter) {\n // \"none\" filter turns into a memcpy here; make that explicit.\n case STBI__F_none: memcpy(cur, raw, nk); break;\n CASE(STBI__F_sub) cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); break;\n CASE(STBI__F_up) cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break;\n CASE(STBI__F_avg) cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); break;\n CASE(STBI__F_paeth) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); break;\n CASE(STBI__F_avg_first) cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); break;\n CASE(STBI__F_paeth_first) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); break;\n }\n #undef CASE\n raw += nk;\n } else {\n STBI_ASSERT(img_n+1 == out_n);\n #define CASE(f) \\\n case f: \\\n for (i=x-1; i >= 1; --i, cur[img_n]=255,raw+=img_n,cur+=out_n,prior+=out_n) \\\n for (k=0; k < img_n; ++k)\n switch (filter) {\n CASE(STBI__F_none) cur[k] = raw[k]; break;\n CASE(STBI__F_sub) cur[k] = STBI__BYTECAST(raw[k] + cur[k-out_n]); break;\n CASE(STBI__F_up) cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break;\n CASE(STBI__F_avg) cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-out_n])>>1)); break;\n CASE(STBI__F_paeth) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-out_n],prior[k],prior[k-out_n])); break;\n CASE(STBI__F_avg_first) cur[k] = STBI__BYTECAST(raw[k] + (cur[k-out_n] >> 1)); break;\n CASE(STBI__F_paeth_first) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-out_n],0,0)); break;\n }\n #undef CASE\n }\n }\n\n // we make a separate pass to expand bits to pixels; for performance,\n // this could run two scanlines behind the above code, so it won't\n // intefere with filtering but will still be in the cache.\n if (depth < 8) {\n for (j=0; j < y; ++j) {\n stbi_uc *cur = a->out + stride*j;\n stbi_uc *in = a->out + stride*j + x*out_n - img_width_bytes;\n // unpack 1/2/4-bit into a 8-bit buffer. allows us to keep the common 8-bit path optimal at minimal cost for 1/2/4-bit\n // png guarante byte alignment, if width is not multiple of 8/4/2 we'll decode dummy trailing data that will be skipped in the later loop\n stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range\n\n // note that the final byte might overshoot and write more data than desired.\n // we can allocate enough data that this never writes out of memory, but it\n // could also overwrite the next scanline. can it overwrite non-empty data\n // on the next scanline? yes, consider 1-pixel-wide scanlines with 1-bit-per-pixel.\n // so we need to explicitly clamp the final ones\n\n if (depth == 4) {\n for (k=x*img_n; k >= 2; k-=2, ++in) {\n *cur++ = scale * ((*in >> 4) );\n *cur++ = scale * ((*in ) & 0x0f);\n }\n if (k > 0) *cur++ = scale * ((*in >> 4) );\n } else if (depth == 2) {\n for (k=x*img_n; k >= 4; k-=4, ++in) {\n *cur++ = scale * ((*in >> 6) );\n *cur++ = scale * ((*in >> 4) & 0x03);\n *cur++ = scale * ((*in >> 2) & 0x03);\n *cur++ = scale * ((*in ) & 0x03);\n }\n if (k > 0) *cur++ = scale * ((*in >> 6) );\n if (k > 1) *cur++ = scale * ((*in >> 4) & 0x03);\n if (k > 2) *cur++ = scale * ((*in >> 2) & 0x03);\n } else if (depth == 1) {\n for (k=x*img_n; k >= 8; k-=8, ++in) {\n *cur++ = scale * ((*in >> 7) );\n *cur++ = scale * ((*in >> 6) & 0x01);\n *cur++ = scale * ((*in >> 5) & 0x01);\n *cur++ = scale * ((*in >> 4) & 0x01);\n *cur++ = scale * ((*in >> 3) & 0x01);\n *cur++ = scale * ((*in >> 2) & 0x01);\n *cur++ = scale * ((*in >> 1) & 0x01);\n *cur++ = scale * ((*in ) & 0x01);\n }\n if (k > 0) *cur++ = scale * ((*in >> 7) );\n if (k > 1) *cur++ = scale * ((*in >> 6) & 0x01);\n if (k > 2) *cur++ = scale * ((*in >> 5) & 0x01);\n if (k > 3) *cur++ = scale * ((*in >> 4) & 0x01);\n if (k > 4) *cur++ = scale * ((*in >> 3) & 0x01);\n if (k > 5) *cur++ = scale * ((*in >> 2) & 0x01);\n if (k > 6) *cur++ = scale * ((*in >> 1) & 0x01);\n }\n if (img_n != out_n) {\n int q;\n // insert alpha = 255\n cur = a->out + stride*j;\n if (img_n == 1) {\n for (q=x-1; q >= 0; --q) {\n cur[q*2+1] = 255;\n cur[q*2+0] = cur[q];\n }\n } else {\n STBI_ASSERT(img_n == 3);\n for (q=x-1; q >= 0; --q) {\n cur[q*4+3] = 255;\n cur[q*4+2] = cur[q*3+2];\n cur[q*4+1] = cur[q*3+1];\n cur[q*4+0] = cur[q*3+0];\n }\n }\n }\n }\n }\n\n return 1;\n}\n\nstatic int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced)\n{\n stbi_uc *final;\n int p;\n if (!interlaced)\n return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color);\n\n // de-interlacing\n final = (stbi_uc *) stbi__malloc(a->s->img_x * a->s->img_y * out_n);\n for (p=0; p < 7; ++p) {\n int xorig[] = { 0,4,0,2,0,1,0 };\n int yorig[] = { 0,0,4,0,2,0,1 };\n int xspc[] = { 8,8,4,4,2,2,1 };\n int yspc[] = { 8,8,8,4,4,2,2 };\n int i,j,x,y;\n // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1\n x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p];\n y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p];\n if (x && y) {\n stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y;\n if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) {\n STBI_FREE(final);\n return 0;\n }\n for (j=0; j < y; ++j) {\n for (i=0; i < x; ++i) {\n int out_y = j*yspc[p]+yorig[p];\n int out_x = i*xspc[p]+xorig[p];\n memcpy(final + out_y*a->s->img_x*out_n + out_x*out_n,\n a->out + (j*x+i)*out_n, out_n);\n }\n }\n STBI_FREE(a->out);\n image_data += img_len;\n image_data_len -= img_len;\n }\n }\n a->out = final;\n\n return 1;\n}\n\nstatic int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n)\n{\n stbi__context *s = z->s;\n stbi__uint32 i, pixel_count = s->img_x * s->img_y;\n stbi_uc *p = z->out;\n\n // compute color-based transparency, assuming we've\n // already got 255 as the alpha value in the output\n STBI_ASSERT(out_n == 2 || out_n == 4);\n\n if (out_n == 2) {\n for (i=0; i < pixel_count; ++i) {\n p[1] = (p[0] == tc[0] ? 0 : 255);\n p += 2;\n }\n } else {\n for (i=0; i < pixel_count; ++i) {\n if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])\n p[3] = 0;\n p += 4;\n }\n }\n return 1;\n}\n\nstatic int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n)\n{\n stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y;\n stbi_uc *p, *temp_out, *orig = a->out;\n\n p = (stbi_uc *) stbi__malloc(pixel_count * pal_img_n);\n if (p == NULL) return stbi__err(\"outofmem\", \"Out of memory\");\n\n // between here and free(out) below, exitting would leak\n temp_out = p;\n\n if (pal_img_n == 3) {\n for (i=0; i < pixel_count; ++i) {\n int n = orig[i]*4;\n p[0] = palette[n ];\n p[1] = palette[n+1];\n p[2] = palette[n+2];\n p += 3;\n }\n } else {\n for (i=0; i < pixel_count; ++i) {\n int n = orig[i]*4;\n p[0] = palette[n ];\n p[1] = palette[n+1];\n p[2] = palette[n+2];\n p[3] = palette[n+3];\n p += 4;\n }\n }\n STBI_FREE(a->out);\n a->out = temp_out;\n\n STBI_NOTUSED(len);\n\n return 1;\n}\n\nstatic int stbi__unpremultiply_on_load = 0;\nstatic int stbi__de_iphone_flag = 0;\n\nSTBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply)\n{\n stbi__unpremultiply_on_load = flag_true_if_should_unpremultiply;\n}\n\nSTBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert)\n{\n stbi__de_iphone_flag = flag_true_if_should_convert;\n}\n\nstatic void stbi__de_iphone(stbi__png *z)\n{\n stbi__context *s = z->s;\n stbi__uint32 i, pixel_count = s->img_x * s->img_y;\n stbi_uc *p = z->out;\n\n if (s->img_out_n == 3) { // convert bgr to rgb\n for (i=0; i < pixel_count; ++i) {\n stbi_uc t = p[0];\n p[0] = p[2];\n p[2] = t;\n p += 3;\n }\n } else {\n STBI_ASSERT(s->img_out_n == 4);\n if (stbi__unpremultiply_on_load) {\n // convert bgr to rgb and unpremultiply\n for (i=0; i < pixel_count; ++i) {\n stbi_uc a = p[3];\n stbi_uc t = p[0];\n if (a) {\n p[0] = p[2] * 255 / a;\n p[1] = p[1] * 255 / a;\n p[2] = t * 255 / a;\n } else {\n p[0] = p[2];\n p[2] = t;\n }\n p += 4;\n }\n } else {\n // convert bgr to rgb\n for (i=0; i < pixel_count; ++i) {\n stbi_uc t = p[0];\n p[0] = p[2];\n p[2] = t;\n p += 4;\n }\n }\n }\n}\n\n#define STBI__PNG_TYPE(a,b,c,d) (((a) << 24) + ((b) << 16) + ((c) << 8) + (d))\n\nstatic int stbi__parse_png_file(stbi__png *z, int scan, int req_comp)\n{\n stbi_uc palette[1024], pal_img_n=0;\n stbi_uc has_trans=0, tc[3];\n stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0;\n int first=1,k,interlace=0, color=0, depth=0, is_iphone=0;\n stbi__context *s = z->s;\n\n z->expanded = NULL;\n z->idata = NULL;\n z->out = NULL;\n\n if (!stbi__check_png_header(s)) return 0;\n\n if (scan == STBI__SCAN_type) return 1;\n\n for (;;) {\n stbi__pngchunk c = stbi__get_chunk_header(s);\n switch (c.type) {\n case STBI__PNG_TYPE('C','g','B','I'):\n is_iphone = 1;\n stbi__skip(s, c.length);\n break;\n case STBI__PNG_TYPE('I','H','D','R'): {\n int comp,filter;\n if (!first) return stbi__err(\"multiple IHDR\",\"Corrupt PNG\");\n first = 0;\n if (c.length != 13) return stbi__err(\"bad IHDR len\",\"Corrupt PNG\");\n s->img_x = stbi__get32be(s); if (s->img_x > (1 << 24)) return stbi__err(\"too large\",\"Very large image (corrupt?)\");\n s->img_y = stbi__get32be(s); if (s->img_y > (1 << 24)) return stbi__err(\"too large\",\"Very large image (corrupt?)\");\n depth = stbi__get8(s); if (depth != 1 && depth != 2 && depth != 4 && depth != 8) return stbi__err(\"1/2/4/8-bit only\",\"PNG not supported: 1/2/4/8-bit only\");\n color = stbi__get8(s); if (color > 6) return stbi__err(\"bad ctype\",\"Corrupt PNG\");\n if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err(\"bad ctype\",\"Corrupt PNG\");\n comp = stbi__get8(s); if (comp) return stbi__err(\"bad comp method\",\"Corrupt PNG\");\n filter= stbi__get8(s); if (filter) return stbi__err(\"bad filter method\",\"Corrupt PNG\");\n interlace = stbi__get8(s); if (interlace>1) return stbi__err(\"bad interlace method\",\"Corrupt PNG\");\n if (!s->img_x || !s->img_y) return stbi__err(\"0-pixel image\",\"Corrupt PNG\");\n if (!pal_img_n) {\n s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0);\n if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err(\"too large\", \"Image too large to decode\");\n if (scan == STBI__SCAN_header) return 1;\n } else {\n // if paletted, then pal_n is our final components, and\n // img_n is # components to decompress/filter.\n s->img_n = 1;\n if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err(\"too large\",\"Corrupt PNG\");\n // if SCAN_header, have to scan to see if we have a tRNS\n }\n break;\n }\n\n case STBI__PNG_TYPE('P','L','T','E'): {\n if (first) return stbi__err(\"first not IHDR\", \"Corrupt PNG\");\n if (c.length > 256*3) return stbi__err(\"invalid PLTE\",\"Corrupt PNG\");\n pal_len = c.length / 3;\n if (pal_len * 3 != c.length) return stbi__err(\"invalid PLTE\",\"Corrupt PNG\");\n for (i=0; i < pal_len; ++i) {\n palette[i*4+0] = stbi__get8(s);\n palette[i*4+1] = stbi__get8(s);\n palette[i*4+2] = stbi__get8(s);\n palette[i*4+3] = 255;\n }\n break;\n }\n\n case STBI__PNG_TYPE('t','R','N','S'): {\n if (first) return stbi__err(\"first not IHDR\", \"Corrupt PNG\");\n if (z->idata) return stbi__err(\"tRNS after IDAT\",\"Corrupt PNG\");\n if (pal_img_n) {\n if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; }\n if (pal_len == 0) return stbi__err(\"tRNS before PLTE\",\"Corrupt PNG\");\n if (c.length > pal_len) return stbi__err(\"bad tRNS len\",\"Corrupt PNG\");\n pal_img_n = 4;\n for (i=0; i < c.length; ++i)\n palette[i*4+3] = stbi__get8(s);\n } else {\n if (!(s->img_n & 1)) return stbi__err(\"tRNS with alpha\",\"Corrupt PNG\");\n if (c.length != (stbi__uint32) s->img_n*2) return stbi__err(\"bad tRNS len\",\"Corrupt PNG\");\n has_trans = 1;\n for (k=0; k < s->img_n; ++k)\n tc[k] = (stbi_uc) (stbi__get16be(s) & 255) * stbi__depth_scale_table[depth]; // non 8-bit images will be larger\n }\n break;\n }\n\n case STBI__PNG_TYPE('I','D','A','T'): {\n if (first) return stbi__err(\"first not IHDR\", \"Corrupt PNG\");\n if (pal_img_n && !pal_len) return stbi__err(\"no PLTE\",\"Corrupt PNG\");\n if (scan == STBI__SCAN_header) { s->img_n = pal_img_n; return 1; }\n if ((int)(ioff + c.length) < (int)ioff) return 0;\n if (ioff + c.length > idata_limit) {\n stbi_uc *p;\n if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096;\n while (ioff + c.length > idata_limit)\n idata_limit *= 2;\n p = (stbi_uc *) STBI_REALLOC(z->idata, idata_limit); if (p == NULL) return stbi__err(\"outofmem\", \"Out of memory\");\n z->idata = p;\n }\n if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err(\"outofdata\",\"Corrupt PNG\");\n ioff += c.length;\n break;\n }\n\n case STBI__PNG_TYPE('I','E','N','D'): {\n stbi__uint32 raw_len, bpl;\n if (first) return stbi__err(\"first not IHDR\", \"Corrupt PNG\");\n if (scan != STBI__SCAN_load) return 1;\n if (z->idata == NULL) return stbi__err(\"no IDAT\",\"Corrupt PNG\");\n // initial guess for decoded data size to avoid unnecessary reallocs\n bpl = (s->img_x * depth + 7) / 8; // bytes per line, per component\n raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */;\n z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone);\n if (z->expanded == NULL) return 0; // zlib should set error\n STBI_FREE(z->idata); z->idata = NULL;\n if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans)\n s->img_out_n = s->img_n+1;\n else\n s->img_out_n = s->img_n;\n if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, depth, color, interlace)) return 0;\n if (has_trans)\n if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0;\n if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2)\n stbi__de_iphone(z);\n if (pal_img_n) {\n // pal_img_n == 3 or 4\n s->img_n = pal_img_n; // record the actual colors we had\n s->img_out_n = pal_img_n;\n if (req_comp >= 3) s->img_out_n = req_comp;\n if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n))\n return 0;\n }\n STBI_FREE(z->expanded); z->expanded = NULL;\n return 1;\n }\n\n default:\n // if critical, fail\n if (first) return stbi__err(\"first not IHDR\", \"Corrupt PNG\");\n if ((c.type & (1 << 29)) == 0) {\n #ifndef STBI_NO_FAILURE_STRINGS\n // not threadsafe\n static char invalid_chunk[] = \"XXXX PNG chunk not known\";\n invalid_chunk[0] = STBI__BYTECAST(c.type >> 24);\n invalid_chunk[1] = STBI__BYTECAST(c.type >> 16);\n invalid_chunk[2] = STBI__BYTECAST(c.type >> 8);\n invalid_chunk[3] = STBI__BYTECAST(c.type >> 0);\n #endif\n return stbi__err(invalid_chunk, \"PNG not supported: unknown PNG chunk type\");\n }\n stbi__skip(s, c.length);\n break;\n }\n // end of PNG chunk, read and skip CRC\n stbi__get32be(s);\n }\n}\n\nstatic unsigned char *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp)\n{\n unsigned char *result=NULL;\n if (req_comp < 0 || req_comp > 4) return stbi__errpuc(\"bad req_comp\", \"Internal error\");\n if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) {\n result = p->out;\n p->out = NULL;\n if (req_comp && req_comp != p->s->img_out_n) {\n result = stbi__convert_format(result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y);\n p->s->img_out_n = req_comp;\n if (result == NULL) return result;\n }\n *x = p->s->img_x;\n *y = p->s->img_y;\n if (n) *n = p->s->img_out_n;\n }\n STBI_FREE(p->out); p->out = NULL;\n STBI_FREE(p->expanded); p->expanded = NULL;\n STBI_FREE(p->idata); p->idata = NULL;\n\n return result;\n}\n\nstatic unsigned char *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n stbi__png p;\n p.s = s;\n return stbi__do_png(&p, x,y,comp,req_comp);\n}\n\nstatic int stbi__png_test(stbi__context *s)\n{\n int r;\n r = stbi__check_png_header(s);\n stbi__rewind(s);\n return r;\n}\n\nstatic int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp)\n{\n if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) {\n stbi__rewind( p->s );\n return 0;\n }\n if (x) *x = p->s->img_x;\n if (y) *y = p->s->img_y;\n if (comp) *comp = p->s->img_n;\n return 1;\n}\n\nstatic int stbi__png_info(stbi__context *s, int *x, int *y, int *comp)\n{\n stbi__png p;\n p.s = s;\n return stbi__png_info_raw(&p, x, y, comp);\n}\n#endif\n\n// Microsoft/Windows BMP image\n\n#ifndef STBI_NO_BMP\nstatic int stbi__bmp_test_raw(stbi__context *s)\n{\n int r;\n int sz;\n if (stbi__get8(s) != 'B') return 0;\n if (stbi__get8(s) != 'M') return 0;\n stbi__get32le(s); // discard filesize\n stbi__get16le(s); // discard reserved\n stbi__get16le(s); // discard reserved\n stbi__get32le(s); // discard data offset\n sz = stbi__get32le(s);\n r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124);\n return r;\n}\n\nstatic int stbi__bmp_test(stbi__context *s)\n{\n int r = stbi__bmp_test_raw(s);\n stbi__rewind(s);\n return r;\n}\n\n\n// returns 0..31 for the highest set bit\nstatic int stbi__high_bit(unsigned int z)\n{\n int n=0;\n if (z == 0) return -1;\n if (z >= 0x10000) n += 16, z >>= 16;\n if (z >= 0x00100) n += 8, z >>= 8;\n if (z >= 0x00010) n += 4, z >>= 4;\n if (z >= 0x00004) n += 2, z >>= 2;\n if (z >= 0x00002) n += 1, z >>= 1;\n return n;\n}\n\nstatic int stbi__bitcount(unsigned int a)\n{\n a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2\n a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4\n a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits\n a = (a + (a >> 8)); // max 16 per 8 bits\n a = (a + (a >> 16)); // max 32 per 8 bits\n return a & 0xff;\n}\n\nstatic int stbi__shiftsigned(int v, int shift, int bits)\n{\n int result;\n int z=0;\n\n if (shift < 0) v <<= -shift;\n else v >>= shift;\n result = v;\n\n z = bits;\n while (z < 8) {\n result += v >> z;\n z += bits;\n }\n return result;\n}\n\nstatic stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n stbi_uc *out;\n unsigned int mr=0,mg=0,mb=0,ma=0, all_a=255;\n stbi_uc pal[256][4];\n int psize=0,i,j,compress=0,width;\n int bpp, flip_vertically, pad, target, offset, hsz;\n if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc(\"not BMP\", \"Corrupt BMP\");\n stbi__get32le(s); // discard filesize\n stbi__get16le(s); // discard reserved\n stbi__get16le(s); // discard reserved\n offset = stbi__get32le(s);\n hsz = stbi__get32le(s);\n if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc(\"unknown BMP\", \"BMP type not supported: unknown\");\n if (hsz == 12) {\n s->img_x = stbi__get16le(s);\n s->img_y = stbi__get16le(s);\n } else {\n s->img_x = stbi__get32le(s);\n s->img_y = stbi__get32le(s);\n }\n if (stbi__get16le(s) != 1) return stbi__errpuc(\"bad BMP\", \"bad BMP\");\n bpp = stbi__get16le(s);\n if (bpp == 1) return stbi__errpuc(\"monochrome\", \"BMP type not supported: 1-bit\");\n flip_vertically = ((int) s->img_y) > 0;\n s->img_y = abs((int) s->img_y);\n if (hsz == 12) {\n if (bpp < 24)\n psize = (offset - 14 - 24) / 3;\n } else {\n compress = stbi__get32le(s);\n if (compress == 1 || compress == 2) return stbi__errpuc(\"BMP RLE\", \"BMP type not supported: RLE\");\n stbi__get32le(s); // discard sizeof\n stbi__get32le(s); // discard hres\n stbi__get32le(s); // discard vres\n stbi__get32le(s); // discard colorsused\n stbi__get32le(s); // discard max important\n if (hsz == 40 || hsz == 56) {\n if (hsz == 56) {\n stbi__get32le(s);\n stbi__get32le(s);\n stbi__get32le(s);\n stbi__get32le(s);\n }\n if (bpp == 16 || bpp == 32) {\n mr = mg = mb = 0;\n if (compress == 0) {\n if (bpp == 32) {\n mr = 0xffu << 16;\n mg = 0xffu << 8;\n mb = 0xffu << 0;\n ma = 0xffu << 24;\n all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0\n } else {\n mr = 31u << 10;\n mg = 31u << 5;\n mb = 31u << 0;\n }\n } else if (compress == 3) {\n mr = stbi__get32le(s);\n mg = stbi__get32le(s);\n mb = stbi__get32le(s);\n // not documented, but generated by photoshop and handled by mspaint\n if (mr == mg && mg == mb) {\n // ?!?!?\n return stbi__errpuc(\"bad BMP\", \"bad BMP\");\n }\n } else\n return stbi__errpuc(\"bad BMP\", \"bad BMP\");\n }\n } else {\n STBI_ASSERT(hsz == 108 || hsz == 124);\n mr = stbi__get32le(s);\n mg = stbi__get32le(s);\n mb = stbi__get32le(s);\n ma = stbi__get32le(s);\n stbi__get32le(s); // discard color space\n for (i=0; i < 12; ++i)\n stbi__get32le(s); // discard color space parameters\n if (hsz == 124) {\n stbi__get32le(s); // discard rendering intent\n stbi__get32le(s); // discard offset of profile data\n stbi__get32le(s); // discard size of profile data\n stbi__get32le(s); // discard reserved\n }\n }\n if (bpp < 16)\n psize = (offset - 14 - hsz) >> 2;\n }\n s->img_n = ma ? 4 : 3;\n if (req_comp && req_comp >= 3) // we can directly decode 3 or 4\n target = req_comp;\n else\n target = s->img_n; // if they want monochrome, we'll post-convert\n out = (stbi_uc *) stbi__malloc(target * s->img_x * s->img_y);\n if (!out) return stbi__errpuc(\"outofmem\", \"Out of memory\");\n if (bpp < 16) {\n int z=0;\n if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc(\"invalid\", \"Corrupt BMP\"); }\n for (i=0; i < psize; ++i) {\n pal[i][2] = stbi__get8(s);\n pal[i][1] = stbi__get8(s);\n pal[i][0] = stbi__get8(s);\n if (hsz != 12) stbi__get8(s);\n pal[i][3] = 255;\n }\n stbi__skip(s, offset - 14 - hsz - psize * (hsz == 12 ? 3 : 4));\n if (bpp == 4) width = (s->img_x + 1) >> 1;\n else if (bpp == 8) width = s->img_x;\n else { STBI_FREE(out); return stbi__errpuc(\"bad bpp\", \"Corrupt BMP\"); }\n pad = (-width)&3;\n for (j=0; j < (int) s->img_y; ++j) {\n for (i=0; i < (int) s->img_x; i += 2) {\n int v=stbi__get8(s),v2=0;\n if (bpp == 4) {\n v2 = v & 15;\n v >>= 4;\n }\n out[z++] = pal[v][0];\n out[z++] = pal[v][1];\n out[z++] = pal[v][2];\n if (target == 4) out[z++] = 255;\n if (i+1 == (int) s->img_x) break;\n v = (bpp == 8) ? stbi__get8(s) : v2;\n out[z++] = pal[v][0];\n out[z++] = pal[v][1];\n out[z++] = pal[v][2];\n if (target == 4) out[z++] = 255;\n }\n stbi__skip(s, pad);\n }\n } else {\n int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0;\n int z = 0;\n int easy=0;\n stbi__skip(s, offset - 14 - hsz);\n if (bpp == 24) width = 3 * s->img_x;\n else if (bpp == 16) width = 2*s->img_x;\n else /* bpp = 32 and pad = 0 */ width=0;\n pad = (-width) & 3;\n if (bpp == 24) {\n easy = 1;\n } else if (bpp == 32) {\n if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000)\n easy = 2;\n }\n if (!easy) {\n if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc(\"bad masks\", \"Corrupt BMP\"); }\n // right shift amt to put high bit in position #7\n rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr);\n gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg);\n bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb);\n ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma);\n }\n for (j=0; j < (int) s->img_y; ++j) {\n if (easy) {\n for (i=0; i < (int) s->img_x; ++i) {\n unsigned char a;\n out[z+2] = stbi__get8(s);\n out[z+1] = stbi__get8(s);\n out[z+0] = stbi__get8(s);\n z += 3;\n a = (easy == 2 ? stbi__get8(s) : 255);\n all_a |= a;\n if (target == 4) out[z++] = a;\n }\n } else {\n for (i=0; i < (int) s->img_x; ++i) {\n stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s));\n int a;\n out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount));\n out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount));\n out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount));\n a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255);\n all_a |= a;\n if (target == 4) out[z++] = STBI__BYTECAST(a);\n }\n }\n stbi__skip(s, pad);\n }\n }\n\n // if alpha channel is all 0s, replace with all 255s\n if (target == 4 && all_a == 0)\n for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4)\n out[i] = 255;\n\n if (flip_vertically) {\n stbi_uc t;\n for (j=0; j < (int) s->img_y>>1; ++j) {\n stbi_uc *p1 = out + j *s->img_x*target;\n stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target;\n for (i=0; i < (int) s->img_x*target; ++i) {\n t = p1[i], p1[i] = p2[i], p2[i] = t;\n }\n }\n }\n\n if (req_comp && req_comp != target) {\n out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y);\n if (out == NULL) return out; // stbi__convert_format frees input on failure\n }\n\n *x = s->img_x;\n *y = s->img_y;\n if (comp) *comp = s->img_n;\n return out;\n}\n#endif\n\n// Targa Truevision - TGA\n// by Jonathan Dummer\n#ifndef STBI_NO_TGA\nstatic int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp)\n{\n int tga_w, tga_h, tga_comp;\n int sz;\n stbi__get8(s); // discard Offset\n sz = stbi__get8(s); // color type\n if( sz > 1 ) {\n stbi__rewind(s);\n return 0; // only RGB or indexed allowed\n }\n sz = stbi__get8(s); // image type\n // only RGB or grey allowed, +/- RLE\n if ((sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11)) return 0;\n stbi__skip(s,9);\n tga_w = stbi__get16le(s);\n if( tga_w < 1 ) {\n stbi__rewind(s);\n return 0; // test width\n }\n tga_h = stbi__get16le(s);\n if( tga_h < 1 ) {\n stbi__rewind(s);\n return 0; // test height\n }\n sz = stbi__get8(s); // bits per pixel\n // only RGB or RGBA or grey allowed\n if ((sz != 8) && (sz != 16) && (sz != 24) && (sz != 32)) {\n stbi__rewind(s);\n return 0;\n }\n tga_comp = sz;\n if (x) *x = tga_w;\n if (y) *y = tga_h;\n if (comp) *comp = tga_comp / 8;\n return 1; // seems to have passed everything\n}\n\nstatic int stbi__tga_test(stbi__context *s)\n{\n int res;\n int sz;\n stbi__get8(s); // discard Offset\n sz = stbi__get8(s); // color type\n if ( sz > 1 ) return 0; // only RGB or indexed allowed\n sz = stbi__get8(s); // image type\n if ( (sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11) ) return 0; // only RGB or grey allowed, +/- RLE\n stbi__get16be(s); // discard palette start\n stbi__get16be(s); // discard palette length\n stbi__get8(s); // discard bits per palette color entry\n stbi__get16be(s); // discard x origin\n stbi__get16be(s); // discard y origin\n if ( stbi__get16be(s) < 1 ) return 0; // test width\n if ( stbi__get16be(s) < 1 ) return 0; // test height\n sz = stbi__get8(s); // bits per pixel\n if ( (sz != 8) && (sz != 16) && (sz != 24) && (sz != 32) )\n res = 0;\n else\n res = 1;\n stbi__rewind(s);\n return res;\n}\n\nstatic stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n // read in the TGA header stuff\n int tga_offset = stbi__get8(s);\n int tga_indexed = stbi__get8(s);\n int tga_image_type = stbi__get8(s);\n int tga_is_RLE = 0;\n int tga_palette_start = stbi__get16le(s);\n int tga_palette_len = stbi__get16le(s);\n int tga_palette_bits = stbi__get8(s);\n int tga_x_origin = stbi__get16le(s);\n int tga_y_origin = stbi__get16le(s);\n int tga_width = stbi__get16le(s);\n int tga_height = stbi__get16le(s);\n int tga_bits_per_pixel = stbi__get8(s);\n int tga_comp = tga_bits_per_pixel / 8;\n int tga_inverted = stbi__get8(s);\n // image data\n unsigned char *tga_data;\n unsigned char *tga_palette = NULL;\n int i, j;\n unsigned char raw_data[4];\n int RLE_count = 0;\n int RLE_repeating = 0;\n int read_next_pixel = 1;\n\n // do a tiny bit of precessing\n if ( tga_image_type >= 8 )\n {\n tga_image_type -= 8;\n tga_is_RLE = 1;\n }\n /* int tga_alpha_bits = tga_inverted & 15; */\n tga_inverted = 1 - ((tga_inverted >> 5) & 1);\n\n // error check\n if ( //(tga_indexed) ||\n (tga_width < 1) || (tga_height < 1) ||\n (tga_image_type < 1) || (tga_image_type > 3) ||\n ((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16) &&\n (tga_bits_per_pixel != 24) && (tga_bits_per_pixel != 32))\n )\n {\n return NULL; // we don't report this as a bad TGA because we don't even know if it's TGA\n }\n\n // If I'm paletted, then I'll use the number of bits from the palette\n if ( tga_indexed )\n {\n tga_comp = tga_palette_bits / 8;\n }\n\n // tga info\n *x = tga_width;\n *y = tga_height;\n if (comp) *comp = tga_comp;\n\n tga_data = (unsigned char*)stbi__malloc( (size_t)tga_width * tga_height * tga_comp );\n if (!tga_data) return stbi__errpuc(\"outofmem\", \"Out of memory\");\n\n // skip to the data's starting position (offset usually = 0)\n stbi__skip(s, tga_offset );\n\n if ( !tga_indexed && !tga_is_RLE) {\n for (i=0; i < tga_height; ++i) {\n int row = tga_inverted ? tga_height -i - 1 : i;\n stbi_uc *tga_row = tga_data + row*tga_width*tga_comp;\n stbi__getn(s, tga_row, tga_width * tga_comp);\n }\n } else {\n // do I need to load a palette?\n if ( tga_indexed)\n {\n // any data to skip? (offset usually = 0)\n stbi__skip(s, tga_palette_start );\n // load the palette\n tga_palette = (unsigned char*)stbi__malloc( tga_palette_len * tga_palette_bits / 8 );\n if (!tga_palette) {\n STBI_FREE(tga_data);\n return stbi__errpuc(\"outofmem\", \"Out of memory\");\n }\n if (!stbi__getn(s, tga_palette, tga_palette_len * tga_palette_bits / 8 )) {\n STBI_FREE(tga_data);\n STBI_FREE(tga_palette);\n return stbi__errpuc(\"bad palette\", \"Corrupt TGA\");\n }\n }\n // load the data\n for (i=0; i < tga_width * tga_height; ++i)\n {\n // if I'm in RLE mode, do I need to get a RLE stbi__pngchunk?\n if ( tga_is_RLE )\n {\n if ( RLE_count == 0 )\n {\n // yep, get the next byte as a RLE command\n int RLE_cmd = stbi__get8(s);\n RLE_count = 1 + (RLE_cmd & 127);\n RLE_repeating = RLE_cmd >> 7;\n read_next_pixel = 1;\n } else if ( !RLE_repeating )\n {\n read_next_pixel = 1;\n }\n } else\n {\n read_next_pixel = 1;\n }\n // OK, if I need to read a pixel, do it now\n if ( read_next_pixel )\n {\n // load however much data we did have\n if ( tga_indexed )\n {\n // read in 1 byte, then perform the lookup\n int pal_idx = stbi__get8(s);\n if ( pal_idx >= tga_palette_len )\n {\n // invalid index\n pal_idx = 0;\n }\n pal_idx *= tga_bits_per_pixel / 8;\n for (j = 0; j*8 < tga_bits_per_pixel; ++j)\n {\n raw_data[j] = tga_palette[pal_idx+j];\n }\n } else\n {\n // read in the data raw\n for (j = 0; j*8 < tga_bits_per_pixel; ++j)\n {\n raw_data[j] = stbi__get8(s);\n }\n }\n // clear the reading flag for the next pixel\n read_next_pixel = 0;\n } // end of reading a pixel\n\n // copy data\n for (j = 0; j < tga_comp; ++j)\n tga_data[i*tga_comp+j] = raw_data[j];\n\n // in case we're in RLE mode, keep counting down\n --RLE_count;\n }\n // do I need to invert the image?\n if ( tga_inverted )\n {\n for (j = 0; j*2 < tga_height; ++j)\n {\n int index1 = j * tga_width * tga_comp;\n int index2 = (tga_height - 1 - j) * tga_width * tga_comp;\n for (i = tga_width * tga_comp; i > 0; --i)\n {\n unsigned char temp = tga_data[index1];\n tga_data[index1] = tga_data[index2];\n tga_data[index2] = temp;\n ++index1;\n ++index2;\n }\n }\n }\n // clear my palette, if I had one\n if ( tga_palette != NULL )\n {\n STBI_FREE( tga_palette );\n }\n }\n\n // swap RGB\n if (tga_comp >= 3)\n {\n unsigned char* tga_pixel = tga_data;\n for (i=0; i < tga_width * tga_height; ++i)\n {\n unsigned char temp = tga_pixel[0];\n tga_pixel[0] = tga_pixel[2];\n tga_pixel[2] = temp;\n tga_pixel += tga_comp;\n }\n }\n\n // convert to target component count\n if (req_comp && req_comp != tga_comp)\n tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height);\n\n // the things I do to get rid of an error message, and yet keep\n // Microsoft's C compilers happy... [8^(\n tga_palette_start = tga_palette_len = tga_palette_bits =\n tga_x_origin = tga_y_origin = 0;\n // OK, done\n return tga_data;\n}\n#endif\n\n// *************************************************************************************************\n// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB\n\n#ifndef STBI_NO_PSD\nstatic int stbi__psd_test(stbi__context *s)\n{\n int r = (stbi__get32be(s) == 0x38425053);\n stbi__rewind(s);\n return r;\n}\n\nstatic stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n int pixelCount;\n int channelCount, compression;\n int channel, i, count, len;\n int bitdepth;\n int w,h;\n stbi_uc *out;\n\n // Check identifier\n if (stbi__get32be(s) != 0x38425053) // \"8BPS\"\n return stbi__errpuc(\"not PSD\", \"Corrupt PSD image\");\n\n // Check file type version.\n if (stbi__get16be(s) != 1)\n return stbi__errpuc(\"wrong version\", \"Unsupported version of PSD image\");\n\n // Skip 6 reserved bytes.\n stbi__skip(s, 6 );\n\n // Read the number of channels (R, G, B, A, etc).\n channelCount = stbi__get16be(s);\n if (channelCount < 0 || channelCount > 16)\n return stbi__errpuc(\"wrong channel count\", \"Unsupported number of channels in PSD image\");\n\n // Read the rows and columns of the image.\n h = stbi__get32be(s);\n w = stbi__get32be(s);\n\n // Make sure the depth is 8 bits.\n bitdepth = stbi__get16be(s);\n if (bitdepth != 8 && bitdepth != 16)\n return stbi__errpuc(\"unsupported bit depth\", \"PSD bit depth is not 8 or 16 bit\");\n\n // Make sure the color mode is RGB.\n // Valid options are:\n // 0: Bitmap\n // 1: Grayscale\n // 2: Indexed color\n // 3: RGB color\n // 4: CMYK color\n // 7: Multichannel\n // 8: Duotone\n // 9: Lab color\n if (stbi__get16be(s) != 3)\n return stbi__errpuc(\"wrong color format\", \"PSD is not in RGB color format\");\n\n // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.)\n stbi__skip(s,stbi__get32be(s) );\n\n // Skip the image resources. (resolution, pen tool paths, etc)\n stbi__skip(s, stbi__get32be(s) );\n\n // Skip the reserved data.\n stbi__skip(s, stbi__get32be(s) );\n\n // Find out if the data is compressed.\n // Known values:\n // 0: no compression\n // 1: RLE compressed\n compression = stbi__get16be(s);\n if (compression > 1)\n return stbi__errpuc(\"bad compression\", \"PSD has an unknown compression format\");\n\n // Create the destination image.\n out = (stbi_uc *) stbi__malloc(4 * w*h);\n if (!out) return stbi__errpuc(\"outofmem\", \"Out of memory\");\n pixelCount = w*h;\n\n // Initialize the data to zero.\n //memset( out, 0, pixelCount * 4 );\n\n // Finally, the image data.\n if (compression) {\n // RLE as used by .PSD and .TIFF\n // Loop until you get the number of unpacked bytes you are expecting:\n // Read the next source byte into n.\n // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally.\n // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times.\n // Else if n is 128, noop.\n // Endloop\n\n // The RLE-compressed data is preceeded by a 2-byte data count for each row in the data,\n // which we're going to just skip.\n stbi__skip(s, h * channelCount * 2 );\n\n // Read the RLE data by channel.\n for (channel = 0; channel < 4; channel++) {\n stbi_uc *p;\n\n p = out+channel;\n if (channel >= channelCount) {\n // Fill this channel with default data.\n for (i = 0; i < pixelCount; i++, p += 4)\n *p = (channel == 3 ? 255 : 0);\n } else {\n // Read the RLE data.\n count = 0;\n while (count < pixelCount) {\n len = stbi__get8(s);\n if (len == 128) {\n // No-op.\n } else if (len < 128) {\n // Copy next len+1 bytes literally.\n len++;\n count += len;\n while (len) {\n *p = stbi__get8(s);\n p += 4;\n len--;\n }\n } else if (len > 128) {\n stbi_uc val;\n // Next -len+1 bytes in the dest are replicated from next source byte.\n // (Interpret len as a negative 8-bit int.)\n len ^= 0x0FF;\n len += 2;\n val = stbi__get8(s);\n count += len;\n while (len) {\n *p = val;\n p += 4;\n len--;\n }\n }\n }\n }\n }\n\n } else {\n // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...)\n // where each channel consists of an 8-bit value for each pixel in the image.\n\n // Read the data by channel.\n for (channel = 0; channel < 4; channel++) {\n stbi_uc *p;\n\n p = out + channel;\n if (channel >= channelCount) {\n // Fill this channel with default data.\n stbi_uc val = channel == 3 ? 255 : 0;\n for (i = 0; i < pixelCount; i++, p += 4)\n *p = val;\n } else {\n // Read the data.\n if (bitdepth == 16) {\n for (i = 0; i < pixelCount; i++, p += 4)\n *p = (stbi_uc) (stbi__get16be(s) >> 8);\n } else {\n for (i = 0; i < pixelCount; i++, p += 4)\n *p = stbi__get8(s);\n }\n }\n }\n }\n\n if (req_comp && req_comp != 4) {\n out = stbi__convert_format(out, 4, req_comp, w, h);\n if (out == NULL) return out; // stbi__convert_format frees input on failure\n }\n\n if (comp) *comp = 4;\n *y = h;\n *x = w;\n\n return out;\n}\n#endif\n\n// *************************************************************************************************\n// Softimage PIC loader\n// by Tom Seddon\n//\n// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format\n// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/\n\n#ifndef STBI_NO_PIC\nstatic int stbi__pic_is4(stbi__context *s,const char *str)\n{\n int i;\n for (i=0; i<4; ++i)\n if (stbi__get8(s) != (stbi_uc)str[i])\n return 0;\n\n return 1;\n}\n\nstatic int stbi__pic_test_core(stbi__context *s)\n{\n int i;\n\n if (!stbi__pic_is4(s,\"\\x53\\x80\\xF6\\x34\"))\n return 0;\n\n for(i=0;i<84;++i)\n stbi__get8(s);\n\n if (!stbi__pic_is4(s,\"PICT\"))\n return 0;\n\n return 1;\n}\n\ntypedef struct\n{\n stbi_uc size,type,channel;\n} stbi__pic_packet;\n\nstatic stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest)\n{\n int mask=0x80, i;\n\n for (i=0; i<4; ++i, mask>>=1) {\n if (channel & mask) {\n if (stbi__at_eof(s)) return stbi__errpuc(\"bad file\",\"PIC file too short\");\n dest[i]=stbi__get8(s);\n }\n }\n\n return dest;\n}\n\nstatic void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src)\n{\n int mask=0x80,i;\n\n for (i=0;i<4; ++i, mask>>=1)\n if (channel&mask)\n dest[i]=src[i];\n}\n\nstatic stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result)\n{\n int act_comp=0,num_packets=0,y,chained;\n stbi__pic_packet packets[10];\n\n // this will (should...) cater for even some bizarre stuff like having data\n // for the same channel in multiple packets.\n do {\n stbi__pic_packet *packet;\n\n if (num_packets==sizeof(packets)/sizeof(packets[0]))\n return stbi__errpuc(\"bad format\",\"too many packets\");\n\n packet = &packets[num_packets++];\n\n chained = stbi__get8(s);\n packet->size = stbi__get8(s);\n packet->type = stbi__get8(s);\n packet->channel = stbi__get8(s);\n\n act_comp |= packet->channel;\n\n if (stbi__at_eof(s)) return stbi__errpuc(\"bad file\",\"file too short (reading packets)\");\n if (packet->size != 8) return stbi__errpuc(\"bad format\",\"packet isn't 8bpp\");\n } while (chained);\n\n *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel?\n\n for(y=0; ytype) {\n default:\n return stbi__errpuc(\"bad format\",\"packet has bad compression type\");\n\n case 0: {//uncompressed\n int x;\n\n for(x=0;xchannel,dest))\n return 0;\n break;\n }\n\n case 1://Pure RLE\n {\n int left=width, i;\n\n while (left>0) {\n stbi_uc count,value[4];\n\n count=stbi__get8(s);\n if (stbi__at_eof(s)) return stbi__errpuc(\"bad file\",\"file too short (pure read count)\");\n\n if (count > left)\n count = (stbi_uc) left;\n\n if (!stbi__readval(s,packet->channel,value)) return 0;\n\n for(i=0; ichannel,dest,value);\n left -= count;\n }\n }\n break;\n\n case 2: {//Mixed RLE\n int left=width;\n while (left>0) {\n int count = stbi__get8(s), i;\n if (stbi__at_eof(s)) return stbi__errpuc(\"bad file\",\"file too short (mixed read count)\");\n\n if (count >= 128) { // Repeated\n stbi_uc value[4];\n\n if (count==128)\n count = stbi__get16be(s);\n else\n count -= 127;\n if (count > left)\n return stbi__errpuc(\"bad file\",\"scanline overrun\");\n\n if (!stbi__readval(s,packet->channel,value))\n return 0;\n\n for(i=0;ichannel,dest,value);\n } else { // Raw\n ++count;\n if (count>left) return stbi__errpuc(\"bad file\",\"scanline overrun\");\n\n for(i=0;ichannel,dest))\n return 0;\n }\n left-=count;\n }\n break;\n }\n }\n }\n }\n\n return result;\n}\n\nstatic stbi_uc *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp)\n{\n stbi_uc *result;\n int i, x,y;\n\n for (i=0; i<92; ++i)\n stbi__get8(s);\n\n x = stbi__get16be(s);\n y = stbi__get16be(s);\n if (stbi__at_eof(s)) return stbi__errpuc(\"bad file\",\"file too short (pic header)\");\n if ((1 << 28) / x < y) return stbi__errpuc(\"too large\", \"Image too large to decode\");\n\n stbi__get32be(s); //skip `ratio'\n stbi__get16be(s); //skip `fields'\n stbi__get16be(s); //skip `pad'\n\n // intermediate buffer is RGBA\n result = (stbi_uc *) stbi__malloc(x*y*4);\n memset(result, 0xff, x*y*4);\n\n if (!stbi__pic_load_core(s,x,y,comp, result)) {\n STBI_FREE(result);\n result=0;\n }\n *px = x;\n *py = y;\n if (req_comp == 0) req_comp = *comp;\n result=stbi__convert_format(result,4,req_comp,x,y);\n\n return result;\n}\n\nstatic int stbi__pic_test(stbi__context *s)\n{\n int r = stbi__pic_test_core(s);\n stbi__rewind(s);\n return r;\n}\n#endif\n\n// *************************************************************************************************\n// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb\n\n#ifndef STBI_NO_GIF\ntypedef struct\n{\n stbi__int16 prefix;\n stbi_uc first;\n stbi_uc suffix;\n} stbi__gif_lzw;\n\ntypedef struct\n{\n int w,h;\n stbi_uc *out, *old_out; // output buffer (always 4 components)\n int flags, bgindex, ratio, transparent, eflags, delay;\n stbi_uc pal[256][4];\n stbi_uc lpal[256][4];\n stbi__gif_lzw codes[4096];\n stbi_uc *color_table;\n int parse, step;\n int lflags;\n int start_x, start_y;\n int max_x, max_y;\n int cur_x, cur_y;\n int line_size;\n} stbi__gif;\n\nstatic int stbi__gif_test_raw(stbi__context *s)\n{\n int sz;\n if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0;\n sz = stbi__get8(s);\n if (sz != '9' && sz != '7') return 0;\n if (stbi__get8(s) != 'a') return 0;\n return 1;\n}\n\nstatic int stbi__gif_test(stbi__context *s)\n{\n int r = stbi__gif_test_raw(s);\n stbi__rewind(s);\n return r;\n}\n\nstatic void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp)\n{\n int i;\n for (i=0; i < num_entries; ++i) {\n pal[i][2] = stbi__get8(s);\n pal[i][1] = stbi__get8(s);\n pal[i][0] = stbi__get8(s);\n pal[i][3] = transp == i ? 0 : 255;\n }\n}\n\nstatic int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info)\n{\n stbi_uc version;\n if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8')\n return stbi__err(\"not GIF\", \"Corrupt GIF\");\n\n version = stbi__get8(s);\n if (version != '7' && version != '9') return stbi__err(\"not GIF\", \"Corrupt GIF\");\n if (stbi__get8(s) != 'a') return stbi__err(\"not GIF\", \"Corrupt GIF\");\n\n stbi__g_failure_reason = \"\";\n g->w = stbi__get16le(s);\n g->h = stbi__get16le(s);\n g->flags = stbi__get8(s);\n g->bgindex = stbi__get8(s);\n g->ratio = stbi__get8(s);\n g->transparent = -1;\n\n if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments\n\n if (is_info) return 1;\n\n if (g->flags & 0x80)\n stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1);\n\n return 1;\n}\n\nstatic int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp)\n{\n stbi__gif g;\n if (!stbi__gif_header(s, &g, comp, 1)) {\n stbi__rewind( s );\n return 0;\n }\n if (x) *x = g.w;\n if (y) *y = g.h;\n return 1;\n}\n\nstatic void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code)\n{\n stbi_uc *p, *c;\n\n // recurse to decode the prefixes, since the linked-list is backwards,\n // and working backwards through an interleaved image would be nasty\n if (g->codes[code].prefix >= 0)\n stbi__out_gif_code(g, g->codes[code].prefix);\n\n if (g->cur_y >= g->max_y) return;\n\n p = &g->out[g->cur_x + g->cur_y];\n c = &g->color_table[g->codes[code].suffix * 4];\n\n if (c[3] >= 128) {\n p[0] = c[2];\n p[1] = c[1];\n p[2] = c[0];\n p[3] = c[3];\n }\n g->cur_x += 4;\n\n if (g->cur_x >= g->max_x) {\n g->cur_x = g->start_x;\n g->cur_y += g->step;\n\n while (g->cur_y >= g->max_y && g->parse > 0) {\n g->step = (1 << g->parse) * g->line_size;\n g->cur_y = g->start_y + (g->step >> 1);\n --g->parse;\n }\n }\n}\n\nstatic stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g)\n{\n stbi_uc lzw_cs;\n stbi__int32 len, init_code;\n stbi__uint32 first;\n stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear;\n stbi__gif_lzw *p;\n\n lzw_cs = stbi__get8(s);\n if (lzw_cs > 12) return NULL;\n clear = 1 << lzw_cs;\n first = 1;\n codesize = lzw_cs + 1;\n codemask = (1 << codesize) - 1;\n bits = 0;\n valid_bits = 0;\n for (init_code = 0; init_code < clear; init_code++) {\n g->codes[init_code].prefix = -1;\n g->codes[init_code].first = (stbi_uc) init_code;\n g->codes[init_code].suffix = (stbi_uc) init_code;\n }\n\n // support no starting clear code\n avail = clear+2;\n oldcode = -1;\n\n len = 0;\n for(;;) {\n if (valid_bits < codesize) {\n if (len == 0) {\n len = stbi__get8(s); // start new block\n if (len == 0)\n return g->out;\n }\n --len;\n bits |= (stbi__int32) stbi__get8(s) << valid_bits;\n valid_bits += 8;\n } else {\n stbi__int32 code = bits & codemask;\n bits >>= codesize;\n valid_bits -= codesize;\n // @OPTIMIZE: is there some way we can accelerate the non-clear path?\n if (code == clear) { // clear code\n codesize = lzw_cs + 1;\n codemask = (1 << codesize) - 1;\n avail = clear + 2;\n oldcode = -1;\n first = 0;\n } else if (code == clear + 1) { // end of stream code\n stbi__skip(s, len);\n while ((len = stbi__get8(s)) > 0)\n stbi__skip(s,len);\n return g->out;\n } else if (code <= avail) {\n if (first) return stbi__errpuc(\"no clear code\", \"Corrupt GIF\");\n\n if (oldcode >= 0) {\n p = &g->codes[avail++];\n if (avail > 4096) return stbi__errpuc(\"too many codes\", \"Corrupt GIF\");\n p->prefix = (stbi__int16) oldcode;\n p->first = g->codes[oldcode].first;\n p->suffix = (code == avail) ? p->first : g->codes[code].first;\n } else if (code == avail)\n return stbi__errpuc(\"illegal code in raster\", \"Corrupt GIF\");\n\n stbi__out_gif_code(g, (stbi__uint16) code);\n\n if ((avail & codemask) == 0 && avail <= 0x0FFF) {\n codesize++;\n codemask = (1 << codesize) - 1;\n }\n\n oldcode = code;\n } else {\n return stbi__errpuc(\"illegal code in raster\", \"Corrupt GIF\");\n }\n }\n }\n}\n\nstatic void stbi__fill_gif_background(stbi__gif *g, int x0, int y0, int x1, int y1)\n{\n int x, y;\n stbi_uc *c = g->pal[g->bgindex];\n for (y = y0; y < y1; y += 4 * g->w) {\n for (x = x0; x < x1; x += 4) {\n stbi_uc *p = &g->out[y + x];\n p[0] = c[2];\n p[1] = c[1];\n p[2] = c[0];\n p[3] = 0;\n }\n }\n}\n\n// this function is designed to support animated gifs, although stb_image doesn't support it\nstatic stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp)\n{\n int i;\n stbi_uc *prev_out = 0;\n\n if (g->out == 0 && !stbi__gif_header(s, g, comp,0))\n return 0; // stbi__g_failure_reason set by stbi__gif_header\n\n prev_out = g->out;\n g->out = (stbi_uc *) stbi__malloc(4 * g->w * g->h);\n if (g->out == 0) return stbi__errpuc(\"outofmem\", \"Out of memory\");\n\n switch ((g->eflags & 0x1C) >> 2) {\n case 0: // unspecified (also always used on 1st frame)\n stbi__fill_gif_background(g, 0, 0, 4 * g->w, 4 * g->w * g->h);\n break;\n case 1: // do not dispose\n if (prev_out) memcpy(g->out, prev_out, 4 * g->w * g->h);\n g->old_out = prev_out;\n break;\n case 2: // dispose to background\n if (prev_out) memcpy(g->out, prev_out, 4 * g->w * g->h);\n stbi__fill_gif_background(g, g->start_x, g->start_y, g->max_x, g->max_y);\n break;\n case 3: // dispose to previous\n if (g->old_out) {\n for (i = g->start_y; i < g->max_y; i += 4 * g->w)\n memcpy(&g->out[i + g->start_x], &g->old_out[i + g->start_x], g->max_x - g->start_x);\n }\n break;\n }\n\n for (;;) {\n switch (stbi__get8(s)) {\n case 0x2C: /* Image Descriptor */\n {\n int prev_trans = -1;\n stbi__int32 x, y, w, h;\n stbi_uc *o;\n\n x = stbi__get16le(s);\n y = stbi__get16le(s);\n w = stbi__get16le(s);\n h = stbi__get16le(s);\n if (((x + w) > (g->w)) || ((y + h) > (g->h)))\n return stbi__errpuc(\"bad Image Descriptor\", \"Corrupt GIF\");\n\n g->line_size = g->w * 4;\n g->start_x = x * 4;\n g->start_y = y * g->line_size;\n g->max_x = g->start_x + w * 4;\n g->max_y = g->start_y + h * g->line_size;\n g->cur_x = g->start_x;\n g->cur_y = g->start_y;\n\n g->lflags = stbi__get8(s);\n\n if (g->lflags & 0x40) {\n g->step = 8 * g->line_size; // first interlaced spacing\n g->parse = 3;\n } else {\n g->step = g->line_size;\n g->parse = 0;\n }\n\n if (g->lflags & 0x80) {\n stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1);\n g->color_table = (stbi_uc *) g->lpal;\n } else if (g->flags & 0x80) {\n if (g->transparent >= 0 && (g->eflags & 0x01)) {\n prev_trans = g->pal[g->transparent][3];\n g->pal[g->transparent][3] = 0;\n }\n g->color_table = (stbi_uc *) g->pal;\n } else\n return stbi__errpuc(\"missing color table\", \"Corrupt GIF\");\n\n o = stbi__process_gif_raster(s, g);\n if (o == NULL) return NULL;\n\n if (prev_trans != -1)\n g->pal[g->transparent][3] = (stbi_uc) prev_trans;\n\n return o;\n }\n\n case 0x21: // Comment Extension.\n {\n int len;\n if (stbi__get8(s) == 0xF9) { // Graphic Control Extension.\n len = stbi__get8(s);\n if (len == 4) {\n g->eflags = stbi__get8(s);\n g->delay = stbi__get16le(s);\n g->transparent = stbi__get8(s);\n } else {\n stbi__skip(s, len);\n break;\n }\n }\n while ((len = stbi__get8(s)) != 0)\n stbi__skip(s, len);\n break;\n }\n\n case 0x3B: // gif stream termination code\n return (stbi_uc *) s; // using '1' causes warning on some compilers\n\n default:\n return stbi__errpuc(\"unknown code\", \"Corrupt GIF\");\n }\n }\n\n STBI_NOTUSED(req_comp);\n}\n\nstatic stbi_uc *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n stbi_uc *u = 0;\n stbi__gif g;\n memset(&g, 0, sizeof(g));\n\n u = stbi__gif_load_next(s, &g, comp, req_comp);\n if (u == (stbi_uc *) s) u = 0; // end of animated gif marker\n if (u) {\n *x = g.w;\n *y = g.h;\n if (req_comp && req_comp != 4)\n u = stbi__convert_format(u, 4, req_comp, g.w, g.h);\n }\n else if (g.out)\n STBI_FREE(g.out);\n\n return u;\n}\n\nstatic int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp)\n{\n return stbi__gif_info_raw(s,x,y,comp);\n}\n#endif\n\n// *************************************************************************************************\n// Radiance RGBE HDR loader\n// originally by Nicolas Schulz\n#ifndef STBI_NO_HDR\nstatic int stbi__hdr_test_core(stbi__context *s)\n{\n const char *signature = \"#?RADIANCE\\n\";\n int i;\n for (i=0; signature[i]; ++i)\n if (stbi__get8(s) != signature[i])\n return 0;\n return 1;\n}\n\nstatic int stbi__hdr_test(stbi__context* s)\n{\n int r = stbi__hdr_test_core(s);\n stbi__rewind(s);\n return r;\n}\n\n#define STBI__HDR_BUFLEN 1024\nstatic char *stbi__hdr_gettoken(stbi__context *z, char *buffer)\n{\n int len=0;\n char c = '\\0';\n\n c = (char) stbi__get8(z);\n\n while (!stbi__at_eof(z) && c != '\\n') {\n buffer[len++] = c;\n if (len == STBI__HDR_BUFLEN-1) {\n // flush to end of line\n while (!stbi__at_eof(z) && stbi__get8(z) != '\\n')\n ;\n break;\n }\n c = (char) stbi__get8(z);\n }\n\n buffer[len] = 0;\n return buffer;\n}\n\nstatic void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp)\n{\n if ( input[3] != 0 ) {\n float f1;\n // Exponent\n f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8));\n if (req_comp <= 2)\n output[0] = (input[0] + input[1] + input[2]) * f1 / 3;\n else {\n output[0] = input[0] * f1;\n output[1] = input[1] * f1;\n output[2] = input[2] * f1;\n }\n if (req_comp == 2) output[1] = 1;\n if (req_comp == 4) output[3] = 1;\n } else {\n switch (req_comp) {\n case 4: output[3] = 1; /* fallthrough */\n case 3: output[0] = output[1] = output[2] = 0;\n break;\n case 2: output[1] = 1; /* fallthrough */\n case 1: output[0] = 0;\n break;\n }\n }\n}\n\nstatic float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n char buffer[STBI__HDR_BUFLEN];\n char *token;\n int valid = 0;\n int width, height;\n stbi_uc *scanline;\n float *hdr_data;\n int len;\n unsigned char count, value;\n int i, j, k, c1,c2, z;\n\n\n // Check identifier\n if (strcmp(stbi__hdr_gettoken(s,buffer), \"#?RADIANCE\") != 0)\n return stbi__errpf(\"not HDR\", \"Corrupt HDR image\");\n\n // Parse header\n for(;;) {\n token = stbi__hdr_gettoken(s,buffer);\n if (token[0] == 0) break;\n if (strcmp(token, \"FORMAT=32-bit_rle_rgbe\") == 0) valid = 1;\n }\n\n if (!valid) return stbi__errpf(\"unsupported format\", \"Unsupported HDR format\");\n\n // Parse width and height\n // can't use sscanf() if we're not using stdio!\n token = stbi__hdr_gettoken(s,buffer);\n if (strncmp(token, \"-Y \", 3)) return stbi__errpf(\"unsupported data layout\", \"Unsupported HDR format\");\n token += 3;\n height = (int) strtol(token, &token, 10);\n while (*token == ' ') ++token;\n if (strncmp(token, \"+X \", 3)) return stbi__errpf(\"unsupported data layout\", \"Unsupported HDR format\");\n token += 3;\n width = (int) strtol(token, NULL, 10);\n\n *x = width;\n *y = height;\n\n if (comp) *comp = 3;\n if (req_comp == 0) req_comp = 3;\n\n // Read data\n hdr_data = (float *) stbi__malloc(height * width * req_comp * sizeof(float));\n\n // Load image data\n // image data is stored as some number of sca\n if ( width < 8 || width >= 32768) {\n // Read flat data\n for (j=0; j < height; ++j) {\n for (i=0; i < width; ++i) {\n stbi_uc rgbe[4];\n main_decode_loop:\n stbi__getn(s, rgbe, 4);\n stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp);\n }\n }\n } else {\n // Read RLE-encoded data\n scanline = NULL;\n\n for (j = 0; j < height; ++j) {\n c1 = stbi__get8(s);\n c2 = stbi__get8(s);\n len = stbi__get8(s);\n if (c1 != 2 || c2 != 2 || (len & 0x80)) {\n // not run-length encoded, so we have to actually use THIS data as a decoded\n // pixel (note this can't be a valid pixel--one of RGB must be >= 128)\n stbi_uc rgbe[4];\n rgbe[0] = (stbi_uc) c1;\n rgbe[1] = (stbi_uc) c2;\n rgbe[2] = (stbi_uc) len;\n rgbe[3] = (stbi_uc) stbi__get8(s);\n stbi__hdr_convert(hdr_data, rgbe, req_comp);\n i = 1;\n j = 0;\n STBI_FREE(scanline);\n goto main_decode_loop; // yes, this makes no sense\n }\n len <<= 8;\n len |= stbi__get8(s);\n if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf(\"invalid decoded scanline length\", \"corrupt HDR\"); }\n if (scanline == NULL) scanline = (stbi_uc *) stbi__malloc(width * 4);\n\n for (k = 0; k < 4; ++k) {\n i = 0;\n while (i < width) {\n count = stbi__get8(s);\n if (count > 128) {\n // Run\n value = stbi__get8(s);\n count -= 128;\n for (z = 0; z < count; ++z)\n scanline[i++ * 4 + k] = value;\n } else {\n // Dump\n for (z = 0; z < count; ++z)\n scanline[i++ * 4 + k] = stbi__get8(s);\n }\n }\n }\n for (i=0; i < width; ++i)\n stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp);\n }\n STBI_FREE(scanline);\n }\n\n return hdr_data;\n}\n\nstatic int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp)\n{\n char buffer[STBI__HDR_BUFLEN];\n char *token;\n int valid = 0;\n\n if (strcmp(stbi__hdr_gettoken(s,buffer), \"#?RADIANCE\") != 0) {\n stbi__rewind( s );\n return 0;\n }\n\n for(;;) {\n token = stbi__hdr_gettoken(s,buffer);\n if (token[0] == 0) break;\n if (strcmp(token, \"FORMAT=32-bit_rle_rgbe\") == 0) valid = 1;\n }\n\n if (!valid) {\n stbi__rewind( s );\n return 0;\n }\n token = stbi__hdr_gettoken(s,buffer);\n if (strncmp(token, \"-Y \", 3)) {\n stbi__rewind( s );\n return 0;\n }\n token += 3;\n *y = (int) strtol(token, &token, 10);\n while (*token == ' ') ++token;\n if (strncmp(token, \"+X \", 3)) {\n stbi__rewind( s );\n return 0;\n }\n token += 3;\n *x = (int) strtol(token, NULL, 10);\n *comp = 3;\n return 1;\n}\n#endif // STBI_NO_HDR\n\n#ifndef STBI_NO_BMP\nstatic int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp)\n{\n int hsz;\n if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') {\n stbi__rewind( s );\n return 0;\n }\n stbi__skip(s,12);\n hsz = stbi__get32le(s);\n if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) {\n stbi__rewind( s );\n return 0;\n }\n if (hsz == 12) {\n *x = stbi__get16le(s);\n *y = stbi__get16le(s);\n } else {\n *x = stbi__get32le(s);\n *y = stbi__get32le(s);\n }\n if (stbi__get16le(s) != 1) {\n stbi__rewind( s );\n return 0;\n }\n *comp = stbi__get16le(s) / 8;\n return 1;\n}\n#endif\n\n#ifndef STBI_NO_PSD\nstatic int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp)\n{\n int channelCount;\n if (stbi__get32be(s) != 0x38425053) {\n stbi__rewind( s );\n return 0;\n }\n if (stbi__get16be(s) != 1) {\n stbi__rewind( s );\n return 0;\n }\n stbi__skip(s, 6);\n channelCount = stbi__get16be(s);\n if (channelCount < 0 || channelCount > 16) {\n stbi__rewind( s );\n return 0;\n }\n *y = stbi__get32be(s);\n *x = stbi__get32be(s);\n if (stbi__get16be(s) != 8) {\n stbi__rewind( s );\n return 0;\n }\n if (stbi__get16be(s) != 3) {\n stbi__rewind( s );\n return 0;\n }\n *comp = 4;\n return 1;\n}\n#endif\n\n#ifndef STBI_NO_PIC\nstatic int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp)\n{\n int act_comp=0,num_packets=0,chained;\n stbi__pic_packet packets[10];\n\n if (!stbi__pic_is4(s,\"\\x53\\x80\\xF6\\x34\")) {\n stbi__rewind(s);\n return 0;\n }\n\n stbi__skip(s, 88);\n\n *x = stbi__get16be(s);\n *y = stbi__get16be(s);\n if (stbi__at_eof(s)) {\n stbi__rewind( s);\n return 0;\n }\n if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) {\n stbi__rewind( s );\n return 0;\n }\n\n stbi__skip(s, 8);\n\n do {\n stbi__pic_packet *packet;\n\n if (num_packets==sizeof(packets)/sizeof(packets[0]))\n return 0;\n\n packet = &packets[num_packets++];\n chained = stbi__get8(s);\n packet->size = stbi__get8(s);\n packet->type = stbi__get8(s);\n packet->channel = stbi__get8(s);\n act_comp |= packet->channel;\n\n if (stbi__at_eof(s)) {\n stbi__rewind( s );\n return 0;\n }\n if (packet->size != 8) {\n stbi__rewind( s );\n return 0;\n }\n } while (chained);\n\n *comp = (act_comp & 0x10 ? 4 : 3);\n\n return 1;\n}\n#endif\n\n// *************************************************************************************************\n// Portable Gray Map and Portable Pixel Map loader\n// by Ken Miller\n//\n// PGM: http://netpbm.sourceforge.net/doc/pgm.html\n// PPM: http://netpbm.sourceforge.net/doc/ppm.html\n//\n// Known limitations:\n// Does not support comments in the header section\n// Does not support ASCII image data (formats P2 and P3)\n// Does not support 16-bit-per-channel\n\n#ifndef STBI_NO_PNM\n\nstatic int stbi__pnm_test(stbi__context *s)\n{\n char p, t;\n p = (char) stbi__get8(s);\n t = (char) stbi__get8(s);\n if (p != 'P' || (t != '5' && t != '6')) {\n stbi__rewind( s );\n return 0;\n }\n return 1;\n}\n\nstatic stbi_uc *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp)\n{\n stbi_uc *out;\n if (!stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n))\n return 0;\n *x = s->img_x;\n *y = s->img_y;\n *comp = s->img_n;\n\n out = (stbi_uc *) stbi__malloc(s->img_n * s->img_x * s->img_y);\n if (!out) return stbi__errpuc(\"outofmem\", \"Out of memory\");\n stbi__getn(s, out, s->img_n * s->img_x * s->img_y);\n\n if (req_comp && req_comp != s->img_n) {\n out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y);\n if (out == NULL) return out; // stbi__convert_format frees input on failure\n }\n return out;\n}\n\nstatic int stbi__pnm_isspace(char c)\n{\n return c == ' ' || c == '\\t' || c == '\\n' || c == '\\v' || c == '\\f' || c == '\\r';\n}\n\nstatic void stbi__pnm_skip_whitespace(stbi__context *s, char *c)\n{\n while (!stbi__at_eof(s) && stbi__pnm_isspace(*c))\n *c = (char) stbi__get8(s);\n}\n\nstatic int stbi__pnm_isdigit(char c)\n{\n return c >= '0' && c <= '9';\n}\n\nstatic int stbi__pnm_getinteger(stbi__context *s, char *c)\n{\n int value = 0;\n\n while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) {\n value = value*10 + (*c - '0');\n *c = (char) stbi__get8(s);\n }\n\n return value;\n}\n\nstatic int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp)\n{\n int maxv;\n char c, p, t;\n\n stbi__rewind( s );\n\n // Get identifier\n p = (char) stbi__get8(s);\n t = (char) stbi__get8(s);\n if (p != 'P' || (t != '5' && t != '6')) {\n stbi__rewind( s );\n return 0;\n }\n\n *comp = (t == '6') ? 3 : 1; // '5' is 1-component .pgm; '6' is 3-component .ppm\n\n c = (char) stbi__get8(s);\n stbi__pnm_skip_whitespace(s, &c);\n\n *x = stbi__pnm_getinteger(s, &c); // read width\n stbi__pnm_skip_whitespace(s, &c);\n\n *y = stbi__pnm_getinteger(s, &c); // read height\n stbi__pnm_skip_whitespace(s, &c);\n\n maxv = stbi__pnm_getinteger(s, &c); // read max value\n\n if (maxv > 255)\n return stbi__err(\"max value > 255\", \"PPM image not 8-bit\");\n else\n return 1;\n}\n#endif\n\nstatic int stbi__info_main(stbi__context *s, int *x, int *y, int *comp)\n{\n #ifndef STBI_NO_JPEG\n if (stbi__jpeg_info(s, x, y, comp)) return 1;\n #endif\n\n #ifndef STBI_NO_PNG\n if (stbi__png_info(s, x, y, comp)) return 1;\n #endif\n\n #ifndef STBI_NO_GIF\n if (stbi__gif_info(s, x, y, comp)) return 1;\n #endif\n\n #ifndef STBI_NO_BMP\n if (stbi__bmp_info(s, x, y, comp)) return 1;\n #endif\n\n #ifndef STBI_NO_PSD\n if (stbi__psd_info(s, x, y, comp)) return 1;\n #endif\n\n #ifndef STBI_NO_PIC\n if (stbi__pic_info(s, x, y, comp)) return 1;\n #endif\n\n #ifndef STBI_NO_PNM\n if (stbi__pnm_info(s, x, y, comp)) return 1;\n #endif\n\n #ifndef STBI_NO_HDR\n if (stbi__hdr_info(s, x, y, comp)) return 1;\n #endif\n\n // test tga last because it's a crappy test!\n #ifndef STBI_NO_TGA\n if (stbi__tga_info(s, x, y, comp))\n return 1;\n #endif\n return stbi__err(\"unknown image type\", \"Image not of any known type, or corrupt\");\n}\n\n#ifndef STBI_NO_STDIO\nSTBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp)\n{\n FILE *f = stbi__fopen(filename, \"rb\");\n int result;\n if (!f) return stbi__err(\"can't fopen\", \"Unable to open file\");\n result = stbi_info_from_file(f, x, y, comp);\n fclose(f);\n return result;\n}\n\nSTBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp)\n{\n int r;\n stbi__context s;\n long pos = ftell(f);\n stbi__start_file(&s, f);\n r = stbi__info_main(&s,x,y,comp);\n fseek(f,pos,SEEK_SET);\n return r;\n}\n#endif // !STBI_NO_STDIO\n\nSTBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp)\n{\n stbi__context s;\n stbi__start_mem(&s,buffer,len);\n return stbi__info_main(&s,x,y,comp);\n}\n\nSTBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp)\n{\n stbi__context s;\n stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user);\n return stbi__info_main(&s,x,y,comp);\n}\n\n#endif // STB_IMAGE_IMPLEMENTATION\n\n/*\n revision history:\n 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA\n 2.07 (2015-09-13) fix compiler warnings\n partial animated GIF support\n limited 16-bit PSD support\n #ifdef unused functions\n bug with < 92 byte PIC,PNM,HDR,TGA\n 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value\n 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning\n 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit\n 2.03 (2015-04-12) extra corruption checking (mmozeiko)\n stbi_set_flip_vertically_on_load (nguillemot)\n fix NEON support; fix mingw support\n 2.02 (2015-01-19) fix incorrect assert, fix warning\n 2.01 (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2\n 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG\n 2.00 (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg)\n progressive JPEG (stb)\n PGM/PPM support (Ken Miller)\n STBI_MALLOC,STBI_REALLOC,STBI_FREE\n GIF bugfix -- seemingly never worked\n STBI_NO_*, STBI_ONLY_*\n 1.48 (2014-12-14) fix incorrectly-named assert()\n 1.47 (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb)\n optimize PNG (ryg)\n fix bug in interlaced PNG with user-specified channel count (stb)\n 1.46 (2014-08-26)\n fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG\n 1.45 (2014-08-16)\n fix MSVC-ARM internal compiler error by wrapping malloc\n 1.44 (2014-08-07)\n various warning fixes from Ronny Chevalier\n 1.43 (2014-07-15)\n fix MSVC-only compiler problem in code changed in 1.42\n 1.42 (2014-07-09)\n don't define _CRT_SECURE_NO_WARNINGS (affects user code)\n fixes to stbi__cleanup_jpeg path\n added STBI_ASSERT to avoid requiring assert.h\n 1.41 (2014-06-25)\n fix search&replace from 1.36 that messed up comments/error messages\n 1.40 (2014-06-22)\n fix gcc struct-initialization warning\n 1.39 (2014-06-15)\n fix to TGA optimization when req_comp != number of components in TGA;\n fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite)\n add support for BMP version 5 (more ignored fields)\n 1.38 (2014-06-06)\n suppress MSVC warnings on integer casts truncating values\n fix accidental rename of 'skip' field of I/O\n 1.37 (2014-06-04)\n remove duplicate typedef\n 1.36 (2014-06-03)\n convert to header file single-file library\n if de-iphone isn't set, load iphone images color-swapped instead of returning NULL\n 1.35 (2014-05-27)\n various warnings\n fix broken STBI_SIMD path\n fix bug where stbi_load_from_file no longer left file pointer in correct place\n fix broken non-easy path for 32-bit BMP (possibly never used)\n TGA optimization by Arseny Kapoulkine\n 1.34 (unknown)\n use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case\n 1.33 (2011-07-14)\n make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements\n 1.32 (2011-07-13)\n support for \"info\" function for all supported filetypes (SpartanJ)\n 1.31 (2011-06-20)\n a few more leak fixes, bug in PNG handling (SpartanJ)\n 1.30 (2011-06-11)\n added ability to load files via callbacks to accomidate custom input streams (Ben Wenger)\n removed deprecated format-specific test/load functions\n removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway\n error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha)\n fix inefficiency in decoding 32-bit BMP (David Woo)\n 1.29 (2010-08-16)\n various warning fixes from Aurelien Pocheville\n 1.28 (2010-08-01)\n fix bug in GIF palette transparency (SpartanJ)\n 1.27 (2010-08-01)\n cast-to-stbi_uc to fix warnings\n 1.26 (2010-07-24)\n fix bug in file buffering for PNG reported by SpartanJ\n 1.25 (2010-07-17)\n refix trans_data warning (Won Chun)\n 1.24 (2010-07-12)\n perf improvements reading from files on platforms with lock-heavy fgetc()\n minor perf improvements for jpeg\n deprecated type-specific functions so we'll get feedback if they're needed\n attempt to fix trans_data warning (Won Chun)\n 1.23 fixed bug in iPhone support\n 1.22 (2010-07-10)\n removed image *writing* support\n stbi_info support from Jetro Lauha\n GIF support from Jean-Marc Lienher\n iPhone PNG-extensions from James Brown\n warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva)\n 1.21 fix use of 'stbi_uc' in header (reported by jon blow)\n 1.20 added support for Softimage PIC, by Tom Seddon\n 1.19 bug in interlaced PNG corruption check (found by ryg)\n 1.18 (2008-08-02)\n fix a threading bug (local mutable static)\n 1.17 support interlaced PNG\n 1.16 major bugfix - stbi__convert_format converted one too many pixels\n 1.15 initialize some fields for thread safety\n 1.14 fix threadsafe conversion bug\n header-file-only version (#define STBI_HEADER_FILE_ONLY before including)\n 1.13 threadsafe\n 1.12 const qualifiers in the API\n 1.11 Support installable IDCT, colorspace conversion routines\n 1.10 Fixes for 64-bit (don't use \"unsigned long\")\n optimized upsampling by Fabian \"ryg\" Giesen\n 1.09 Fix format-conversion for PSD code (bad global variables!)\n 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz\n 1.07 attempt to fix C++ warning/errors again\n 1.06 attempt to fix C++ warning/errors again\n 1.05 fix TGA loading to return correct *comp and use good luminance calc\n 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free\n 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR\n 1.02 support for (subset of) HDR files, float interface for preferred access to them\n 1.01 fix bug: possible bug in handling right-side up bmps... not sure\n fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all\n 1.00 interface to zlib that skips zlib header\n 0.99 correct handling of alpha in palette\n 0.98 TGA loader by lonesock; dynamically add loaders (untested)\n 0.97 jpeg errors on too large a file; also catch another malloc failure\n 0.96 fix detection of invalid v value - particleman@mollyrocket forum\n 0.95 during header scan, seek to markers in case of padding\n 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same\n 0.93 handle jpegtran output; verbose errors\n 0.92 read 4,8,16,24,32-bit BMP files of several formats\n 0.91 output 24-bit Windows 3.0 BMP files\n 0.90 fix a few more warnings; bump version number to approach 1.0\n 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd\n 0.60 fix compiling as c++\n 0.59 fix warnings: merge Dave Moore's -Wall fixes\n 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian\n 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available\n 0.56 fix bug: zlib uncompressed mode len vs. nlen\n 0.55 fix bug: restart_interval not initialized to 0\n 0.54 allow NULL for 'int *comp'\n 0.53 fix bug in png 3->4; speedup png decoding\n 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments\n 0.51 obey req_comp requests, 1-component jpegs return as 1-component,\n on 'test' only check type, not whether we support this variant\n 0.50 (2006-11-19)\n first released version\n*/\n" }, { "path": "third_party/stb/stb_image_write.h", "content": "/* stb_image_write - v1.00 - public domain - http://nothings.org/stb/stb_image_write.h\n writes out PNG/BMP/TGA images to C stdio - Sean Barrett 2010-2015\n no warranty implied; use at your own risk\n\n Before #including,\n\n #define STB_IMAGE_WRITE_IMPLEMENTATION\n\n in the file that you want to have the implementation.\n\n Will probably not work correctly with strict-aliasing optimizations.\n\nABOUT:\n\n This header file is a library for writing images to C stdio. It could be\n adapted to write to memory or a general streaming interface; let me know.\n\n The PNG output is not optimal; it is 20-50% larger than the file\n written by a decent optimizing implementation. This library is designed\n for source code compactness and simplicity, not optimal image file size\n or run-time performance.\n\nBUILDING:\n\n You can #define STBIW_ASSERT(x) before the #include to avoid using assert.h.\n You can #define STBIW_MALLOC(), STBIW_REALLOC(), and STBIW_FREE() to replace\n malloc,realloc,free.\n You can define STBIW_MEMMOVE() to replace memmove()\n\nUSAGE:\n\n There are four functions, one for each image file format:\n\n int stbi_write_png(char const *filename, int w, int h, int comp, const void *data, int stride_in_bytes);\n int stbi_write_bmp(char const *filename, int w, int h, int comp, const void *data);\n int stbi_write_tga(char const *filename, int w, int h, int comp, const void *data);\n int stbi_write_hdr(char const *filename, int w, int h, int comp, const void *data);\n\n There are also four equivalent functions that use an arbitrary write function. You are\n expected to open/close your file-equivalent before and after calling these:\n\n int stbi_write_png_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data, int stride_in_bytes);\n int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data);\n int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data);\n int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data);\n\n where the callback is:\n void stbi_write_func(void *context, void *data, int size);\n\n You can define STBI_WRITE_NO_STDIO to disable the file variant of these\n functions, so the library will not use stdio.h at all. However, this will\n also disable HDR writing, because it requires stdio for formatted output.\n\n Each function returns 0 on failure and non-0 on success.\n\n The functions create an image file defined by the parameters. The image\n is a rectangle of pixels stored from left-to-right, top-to-bottom.\n Each pixel contains 'comp' channels of data stored interleaved with 8-bits\n per channel, in the following order: 1=Y, 2=YA, 3=RGB, 4=RGBA. (Y is\n monochrome color.) The rectangle is 'w' pixels wide and 'h' pixels tall.\n The *data pointer points to the first byte of the top-left-most pixel.\n For PNG, \"stride_in_bytes\" is the distance in bytes from the first byte of\n a row of pixels to the first byte of the next row of pixels.\n\n PNG creates output files with the same number of components as the input.\n The BMP format expands Y to RGB in the file format and does not\n output alpha.\n\n PNG supports writing rectangles of data even when the bytes storing rows of\n data are not consecutive in memory (e.g. sub-rectangles of a larger image),\n by supplying the stride between the beginning of adjacent rows. The other\n formats do not. (Thus you cannot write a native-format BMP through the BMP\n writer, both because it is in BGR order and because it may have padding\n at the end of the line.)\n\n HDR expects linear float data. Since the format is always 32-bit rgb(e)\n data, alpha (if provided) is discarded, and for monochrome data it is\n replicated across all three channels.\n\n TGA supports RLE or non-RLE compressed data. To use non-RLE-compressed\n data, set the global variable 'stbi_write_tga_with_rle' to 0.\n\nCREDITS:\n\n PNG/BMP/TGA\n Sean Barrett\n HDR\n Baldur Karlsson\n TGA monochrome:\n Jean-Sebastien Guay\n misc enhancements:\n Tim Kelsey\n TGA RLE\n Alan Hickman\n initial file IO callback implementation\n Emmanuel Julien\n bugfixes:\n github:Chribba\n Guillaume Chereau\n github:jry2\n\nLICENSE\n\nThis software is in the public domain. Where that dedication is not\nrecognized, you are granted a perpetual, irrevocable license to copy,\ndistribute, and modify this file as you see fit.\n\n*/\n\n#ifndef INCLUDE_STB_IMAGE_WRITE_H\n#define INCLUDE_STB_IMAGE_WRITE_H\n\n#ifdef __cplusplus\nextern \"C\" {\n#endif\n\n#ifdef STB_IMAGE_WRITE_STATIC\n#define STBIWDEF static\n#else\n#define STBIWDEF extern\nextern int stbi_write_tga_with_rle;\n#endif\n\n#ifndef STBI_WRITE_NO_STDIO\nSTBIWDEF int stbi_write_png(char const *filename, int w, int h, int comp, const void *data, int stride_in_bytes);\nSTBIWDEF int stbi_write_bmp(char const *filename, int w, int h, int comp, const void *data);\nSTBIWDEF int stbi_write_tga(char const *filename, int w, int h, int comp, const void *data);\nSTBIWDEF int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data);\n#endif\n\ntypedef void stbi_write_func(void *context, void *data, int size);\n\nSTBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data, int stride_in_bytes);\nSTBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data);\nSTBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data);\nSTBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data);\n\n#ifdef __cplusplus\n}\n#endif\n\n#endif//INCLUDE_STB_IMAGE_WRITE_H\n\n#ifdef STB_IMAGE_WRITE_IMPLEMENTATION\n\n#ifdef _WIN32\n #define _CRT_SECURE_NO_WARNINGS\n #define _CRT_NONSTDC_NO_DEPRECATE\n#endif\n\n#ifndef STBI_WRITE_NO_STDIO\n#include \n#endif // STBI_WRITE_NO_STDIO\n\n#include \n#include \n#include \n#include \n\n#if defined(STBIW_MALLOC) && defined(STBIW_FREE) && defined(STBIW_REALLOC)\n// ok\n#elif !defined(STBIW_MALLOC) && !defined(STBIW_FREE) && !defined(STBIW_REALLOC)\n// ok\n#else\n#error \"Must define all or none of STBIW_MALLOC, STBIW_FREE, and STBIW_REALLOC.\"\n#endif\n\n#ifndef STBIW_MALLOC\n#define STBIW_MALLOC(sz) malloc(sz)\n#define STBIW_REALLOC(p,sz) realloc(p,sz)\n#define STBIW_FREE(p) free(p)\n#endif\n#ifndef STBIW_MEMMOVE\n#define STBIW_MEMMOVE(a,b,sz) memmove(a,b,sz)\n#endif\n\n\n#ifndef STBIW_ASSERT\n#include \n#define STBIW_ASSERT(x) assert(x)\n#endif\n\ntypedef struct\n{\n stbi_write_func *func;\n void *context;\n} stbi__write_context;\n\n// initialize a callback-based context\nstatic void stbi__start_write_callbacks(stbi__write_context *s, stbi_write_func *c, void *context)\n{\n s->func = c;\n s->context = context;\n}\n\n#ifndef STBI_WRITE_NO_STDIO\n\nstatic void stbi__stdio_write(void *context, void *data, int size)\n{\n fwrite(data,1,size,(FILE*) context);\n}\n\nstatic int stbi__start_write_file(stbi__write_context *s, const char *filename)\n{\n FILE *f = fopen(filename, \"wb\");\n stbi__start_write_callbacks(s, stbi__stdio_write, (void *) f);\n return f != NULL;\n}\n\nstatic void stbi__end_write_file(stbi__write_context *s)\n{\n fclose((FILE *)s->context);\n}\n\n#endif // !STBI_WRITE_NO_STDIO\n\ntypedef unsigned int stbiw_uint32;\ntypedef int stb_image_write_test[sizeof(stbiw_uint32)==4 ? 1 : -1];\n\n#ifdef STB_IMAGE_WRITE_STATIC\nstatic int stbi_write_tga_with_rle = 1;\n#else\nint stbi_write_tga_with_rle = 1;\n#endif\n\nstatic void stbiw__writefv(stbi__write_context *s, const char *fmt, va_list v)\n{\n while (*fmt) {\n switch (*fmt++) {\n case ' ': break;\n case '1': { unsigned char x = (unsigned char) va_arg(v, int);\n s->func(s->context,&x,1);\n break; }\n case '2': { int x = va_arg(v,int);\n unsigned char b[2];\n b[0] = (unsigned char) x;\n b[1] = (unsigned char) (x>>8);\n s->func(s->context,b,2);\n break; }\n case '4': { stbiw_uint32 x = va_arg(v,int);\n unsigned char b[4];\n b[0]=(unsigned char)x;\n b[1]=(unsigned char)(x>>8);\n b[2]=(unsigned char)(x>>16);\n b[3]=(unsigned char)(x>>24);\n s->func(s->context,b,4);\n break; }\n default:\n STBIW_ASSERT(0);\n return;\n }\n }\n}\n\nstatic void stbiw__writef(stbi__write_context *s, const char *fmt, ...)\n{\n va_list v;\n va_start(v, fmt);\n stbiw__writefv(s, fmt, v);\n va_end(v);\n}\n\nstatic void stbiw__write3(stbi__write_context *s, unsigned char a, unsigned char b, unsigned char c)\n{\n unsigned char arr[3];\n arr[0] = a, arr[1] = b, arr[2] = c;\n s->func(s->context, arr, 3);\n}\n\nstatic void stbiw__write_pixel(stbi__write_context *s, int rgb_dir, int comp, int write_alpha, int expand_mono, unsigned char *d)\n{\n unsigned char bg[3] = { 255, 0, 255}, px[3];\n int k;\n\n if (write_alpha < 0)\n s->func(s->context, &d[comp - 1], 1);\n\n switch (comp) {\n case 1:\n s->func(s->context,d,1);\n break;\n case 2:\n if (expand_mono)\n stbiw__write3(s, d[0], d[0], d[0]); // monochrome bmp\n else\n s->func(s->context, d, 1); // monochrome TGA\n break;\n case 4:\n if (!write_alpha) {\n // composite against pink background\n for (k = 0; k < 3; ++k)\n px[k] = bg[k] + ((d[k] - bg[k]) * d[3]) / 255;\n stbiw__write3(s, px[1 - rgb_dir], px[1], px[1 + rgb_dir]);\n break;\n }\n /* FALLTHROUGH */\n case 3:\n stbiw__write3(s, d[1 - rgb_dir], d[1], d[1 + rgb_dir]);\n break;\n }\n if (write_alpha > 0)\n s->func(s->context, &d[comp - 1], 1);\n}\n\nstatic void stbiw__write_pixels(stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, void *data, int write_alpha, int scanline_pad, int expand_mono)\n{\n stbiw_uint32 zero = 0;\n int i,j, j_end;\n\n if (y <= 0)\n return;\n\n if (vdir < 0)\n j_end = -1, j = y-1;\n else\n j_end = y, j = 0;\n\n for (; j != j_end; j += vdir) {\n for (i=0; i < x; ++i) {\n unsigned char *d = (unsigned char *) data + (j*x+i)*comp;\n stbiw__write_pixel(s, rgb_dir, comp, write_alpha, expand_mono, d);\n }\n s->func(s->context, &zero, scanline_pad);\n }\n}\n\nstatic int stbiw__outfile(stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, int expand_mono, void *data, int alpha, int pad, const char *fmt, ...)\n{\n if (y < 0 || x < 0) {\n return 0;\n } else {\n va_list v;\n va_start(v, fmt);\n stbiw__writefv(s, fmt, v);\n va_end(v);\n stbiw__write_pixels(s,rgb_dir,vdir,x,y,comp,data,alpha,pad, expand_mono);\n return 1;\n }\n}\n\nstatic int stbi_write_bmp_core(stbi__write_context *s, int x, int y, int comp, const void *data)\n{\n int pad = (-x*3) & 3;\n return stbiw__outfile(s,-1,-1,x,y,comp,1,(void *) data,0,pad,\n \"11 4 22 4\" \"4 44 22 444444\",\n 'B', 'M', 14+40+(x*3+pad)*y, 0,0, 14+40, // file header\n 40, x,y, 1,24, 0,0,0,0,0,0); // bitmap header\n}\n\nSTBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data)\n{\n stbi__write_context s;\n stbi__start_write_callbacks(&s, func, context);\n return stbi_write_bmp_core(&s, x, y, comp, data);\n}\n\n#ifndef STBI_WRITE_NO_STDIO\nSTBIWDEF int stbi_write_bmp(char const *filename, int x, int y, int comp, const void *data)\n{\n stbi__write_context s;\n if (stbi__start_write_file(&s,filename)) {\n int r = stbi_write_bmp_core(&s, x, y, comp, data);\n stbi__end_write_file(&s);\n return r;\n } else\n return 0;\n}\n#endif //!STBI_WRITE_NO_STDIO\n\nstatic int stbi_write_tga_core(stbi__write_context *s, int x, int y, int comp, void *data)\n{\n int has_alpha = (comp == 2 || comp == 4);\n int colorbytes = has_alpha ? comp-1 : comp;\n int format = colorbytes < 2 ? 3 : 2; // 3 color channels (RGB/RGBA) = 2, 1 color channel (Y/YA) = 3\n\n if (y < 0 || x < 0)\n return 0;\n\n if (!stbi_write_tga_with_rle) {\n return stbiw__outfile(s, -1, -1, x, y, comp, 0, (void *) data, has_alpha, 0,\n \"111 221 2222 11\", 0, 0, format, 0, 0, 0, 0, 0, x, y, (colorbytes + has_alpha) * 8, has_alpha * 8);\n } else {\n int i,j,k;\n\n stbiw__writef(s, \"111 221 2222 11\", 0,0,format+8, 0,0,0, 0,0,x,y, (colorbytes + has_alpha) * 8, has_alpha * 8);\n\n for (j = y - 1; j >= 0; --j) {\n unsigned char *row = (unsigned char *) data + j * x * comp;\n int len;\n\n for (i = 0; i < x; i += len) {\n unsigned char *begin = row + i * comp;\n int diff = 1;\n len = 1;\n\n if (i < x - 1) {\n ++len;\n diff = memcmp(begin, row + (i + 1) * comp, comp);\n if (diff) {\n const unsigned char *prev = begin;\n for (k = i + 2; k < x && len < 128; ++k) {\n if (memcmp(prev, row + k * comp, comp)) {\n prev += comp;\n ++len;\n } else {\n --len;\n break;\n }\n }\n } else {\n for (k = i + 2; k < x && len < 128; ++k) {\n if (!memcmp(begin, row + k * comp, comp)) {\n ++len;\n } else {\n break;\n }\n }\n }\n }\n\n if (diff) {\n unsigned char header = (unsigned char) (len - 1);\n s->func(s->context, &header, 1);\n for (k = 0; k < len; ++k) {\n stbiw__write_pixel(s, -1, comp, has_alpha, 0, begin + k * comp);\n }\n } else {\n unsigned char header = (unsigned char) (len - 129);\n s->func(s->context, &header, 1);\n stbiw__write_pixel(s, -1, comp, has_alpha, 0, begin);\n }\n }\n }\n }\n return 1;\n}\n\nint stbi_write_tga_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data)\n{\n stbi__write_context s;\n stbi__start_write_callbacks(&s, func, context);\n return stbi_write_tga_core(&s, x, y, comp, (void *) data);\n}\n\n#ifndef STBI_WRITE_NO_STDIO\nint stbi_write_tga(char const *filename, int x, int y, int comp, const void *data)\n{\n stbi__write_context s;\n if (stbi__start_write_file(&s,filename)) {\n int r = stbi_write_tga_core(&s, x, y, comp, (void *) data);\n stbi__end_write_file(&s);\n return r;\n } else\n return 0;\n}\n#endif\n\n// *************************************************************************************************\n// Radiance RGBE HDR writer\n// by Baldur Karlsson\n#ifndef STBI_WRITE_NO_STDIO\n\n#define stbiw__max(a, b) ((a) > (b) ? (a) : (b))\n\nvoid stbiw__linear_to_rgbe(unsigned char *rgbe, float *linear)\n{\n int exponent;\n float maxcomp = stbiw__max(linear[0], stbiw__max(linear[1], linear[2]));\n\n if (maxcomp < 1e-32) {\n rgbe[0] = rgbe[1] = rgbe[2] = rgbe[3] = 0;\n } else {\n float normalize = (float) frexp(maxcomp, &exponent) * 256.0f/maxcomp;\n\n rgbe[0] = (unsigned char)(linear[0] * normalize);\n rgbe[1] = (unsigned char)(linear[1] * normalize);\n rgbe[2] = (unsigned char)(linear[2] * normalize);\n rgbe[3] = (unsigned char)(exponent + 128);\n }\n}\n\nvoid stbiw__write_run_data(stbi__write_context *s, int length, unsigned char databyte)\n{\n unsigned char lengthbyte = (unsigned char) (length+128);\n STBIW_ASSERT(length+128 <= 255);\n s->func(s->context, &lengthbyte, 1);\n s->func(s->context, &databyte, 1);\n}\n\nvoid stbiw__write_dump_data(stbi__write_context *s, int length, unsigned char *data)\n{\n unsigned char lengthbyte = (unsigned char )(length & 0xff);\n STBIW_ASSERT(length <= 128); // inconsistent with spec but consistent with official code\n s->func(s->context, &lengthbyte, 1);\n s->func(s->context, data, length);\n}\n\nvoid stbiw__write_hdr_scanline(stbi__write_context *s, int width, int ncomp, unsigned char *scratch, float *scanline)\n{\n unsigned char scanlineheader[4] = { 2, 2, 0, 0 };\n unsigned char rgbe[4];\n float linear[3];\n int x;\n\n scanlineheader[2] = (width&0xff00)>>8;\n scanlineheader[3] = (width&0x00ff);\n\n /* skip RLE for images too small or large */\n if (width < 8 || width >= 32768) {\n for (x=0; x < width; x++) {\n switch (ncomp) {\n case 4: /* fallthrough */\n case 3: linear[2] = scanline[x*ncomp + 2];\n linear[1] = scanline[x*ncomp + 1];\n linear[0] = scanline[x*ncomp + 0];\n break;\n default:\n linear[0] = linear[1] = linear[2] = scanline[x*ncomp + 0];\n break;\n }\n stbiw__linear_to_rgbe(rgbe, linear);\n s->func(s->context, rgbe, 4);\n }\n } else {\n int c,r;\n /* encode into scratch buffer */\n for (x=0; x < width; x++) {\n switch(ncomp) {\n case 4: /* fallthrough */\n case 3: linear[2] = scanline[x*ncomp + 2];\n linear[1] = scanline[x*ncomp + 1];\n linear[0] = scanline[x*ncomp + 0];\n break;\n default:\n linear[0] = linear[1] = linear[2] = scanline[x*ncomp + 0];\n break;\n }\n stbiw__linear_to_rgbe(rgbe, linear);\n scratch[x + width*0] = rgbe[0];\n scratch[x + width*1] = rgbe[1];\n scratch[x + width*2] = rgbe[2];\n scratch[x + width*3] = rgbe[3];\n }\n\n s->func(s->context, scanlineheader, 4);\n\n /* RLE each component separately */\n for (c=0; c < 4; c++) {\n unsigned char *comp = &scratch[width*c];\n\n x = 0;\n while (x < width) {\n // find first run\n r = x;\n while (r+2 < width) {\n if (comp[r] == comp[r+1] && comp[r] == comp[r+2])\n break;\n ++r;\n }\n if (r+2 >= width)\n r = width;\n // dump up to first run\n while (x < r) {\n int len = r-x;\n if (len > 128) len = 128;\n stbiw__write_dump_data(s, len, &comp[x]);\n x += len;\n }\n // if there's a run, output it\n if (r+2 < width) { // same test as what we break out of in search loop, so only true if we break'd\n // find next byte after run\n while (r < width && comp[r] == comp[x])\n ++r;\n // output run up to r\n while (x < r) {\n int len = r-x;\n if (len > 127) len = 127;\n stbiw__write_run_data(s, len, comp[x]);\n x += len;\n }\n }\n }\n }\n }\n}\n\nstatic int stbi_write_hdr_core(stbi__write_context *s, int x, int y, int comp, float *data)\n{\n if (y <= 0 || x <= 0 || data == NULL)\n return 0;\n else {\n // Each component is stored separately. Allocate scratch space for full output scanline.\n unsigned char *scratch = (unsigned char *) STBIW_MALLOC(x*4);\n int i, len;\n char buffer[128];\n char header[] = \"#?RADIANCE\\n# Written by stb_image_write.h\\nFORMAT=32-bit_rle_rgbe\\n\";\n s->func(s->context, header, sizeof(header)-1);\n\n len = sprintf(buffer, \"EXPOSURE= 1.0000000000000\\n\\n-Y %d +X %d\\n\", y, x);\n s->func(s->context, buffer, len);\n\n for(i=0; i < y; i++)\n stbiw__write_hdr_scanline(s, x, comp, scratch, data + comp*i*x);\n STBIW_FREE(scratch);\n return 1;\n }\n}\n\nint stbi_write_hdr_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const float *data)\n{\n stbi__write_context s;\n stbi__start_write_callbacks(&s, func, context);\n return stbi_write_hdr_core(&s, x, y, comp, (float *) data);\n}\n\nint stbi_write_hdr(char const *filename, int x, int y, int comp, const float *data)\n{\n stbi__write_context s;\n if (stbi__start_write_file(&s,filename)) {\n int r = stbi_write_hdr_core(&s, x, y, comp, (float *) data);\n stbi__end_write_file(&s);\n return r;\n } else\n return 0;\n}\n#endif // STBI_WRITE_NO_STDIO\n\n\n//////////////////////////////////////////////////////////////////////////////\n//\n// PNG writer\n//\n\n// stretchy buffer; stbiw__sbpush() == vector<>::push_back() -- stbiw__sbcount() == vector<>::size()\n#define stbiw__sbraw(a) ((int *) (a) - 2)\n#define stbiw__sbm(a) stbiw__sbraw(a)[0]\n#define stbiw__sbn(a) stbiw__sbraw(a)[1]\n\n#define stbiw__sbneedgrow(a,n) ((a)==0 || stbiw__sbn(a)+n >= stbiw__sbm(a))\n#define stbiw__sbmaybegrow(a,n) (stbiw__sbneedgrow(a,(n)) ? stbiw__sbgrow(a,n) : 0)\n#define stbiw__sbgrow(a,n) stbiw__sbgrowf((void **) &(a), (n), sizeof(*(a)))\n\n#define stbiw__sbpush(a, v) (stbiw__sbmaybegrow(a,1), (a)[stbiw__sbn(a)++] = (v))\n#define stbiw__sbcount(a) ((a) ? stbiw__sbn(a) : 0)\n#define stbiw__sbfree(a) ((a) ? STBIW_FREE(stbiw__sbraw(a)),0 : 0)\n\nstatic void *stbiw__sbgrowf(void **arr, int increment, int itemsize)\n{\n int m = *arr ? 2*stbiw__sbm(*arr)+increment : increment+1;\n void *p = STBIW_REALLOC(*arr ? stbiw__sbraw(*arr) : 0, itemsize * m + sizeof(int)*2);\n STBIW_ASSERT(p);\n if (p) {\n if (!*arr) ((int *) p)[1] = 0;\n *arr = (void *) ((int *) p + 2);\n stbiw__sbm(*arr) = m;\n }\n return *arr;\n}\n\nstatic unsigned char *stbiw__zlib_flushf(unsigned char *data, unsigned int *bitbuffer, int *bitcount)\n{\n while (*bitcount >= 8) {\n stbiw__sbpush(data, (unsigned char) *bitbuffer);\n *bitbuffer >>= 8;\n *bitcount -= 8;\n }\n return data;\n}\n\nstatic int stbiw__zlib_bitrev(int code, int codebits)\n{\n int res=0;\n while (codebits--) {\n res = (res << 1) | (code & 1);\n code >>= 1;\n }\n return res;\n}\n\nstatic unsigned int stbiw__zlib_countm(unsigned char *a, unsigned char *b, int limit)\n{\n int i;\n for (i=0; i < limit && i < 258; ++i)\n if (a[i] != b[i]) break;\n return i;\n}\n\nstatic unsigned int stbiw__zhash(unsigned char *data)\n{\n stbiw_uint32 hash = data[0] + (data[1] << 8) + (data[2] << 16);\n hash ^= hash << 3;\n hash += hash >> 5;\n hash ^= hash << 4;\n hash += hash >> 17;\n hash ^= hash << 25;\n hash += hash >> 6;\n return hash;\n}\n\n#define stbiw__zlib_flush() (out = stbiw__zlib_flushf(out, &bitbuf, &bitcount))\n#define stbiw__zlib_add(code,codebits) \\\n (bitbuf |= (code) << bitcount, bitcount += (codebits), stbiw__zlib_flush())\n#define stbiw__zlib_huffa(b,c) stbiw__zlib_add(stbiw__zlib_bitrev(b,c),c)\n// default huffman tables\n#define stbiw__zlib_huff1(n) stbiw__zlib_huffa(0x30 + (n), 8)\n#define stbiw__zlib_huff2(n) stbiw__zlib_huffa(0x190 + (n)-144, 9)\n#define stbiw__zlib_huff3(n) stbiw__zlib_huffa(0 + (n)-256,7)\n#define stbiw__zlib_huff4(n) stbiw__zlib_huffa(0xc0 + (n)-280,8)\n#define stbiw__zlib_huff(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : (n) <= 255 ? stbiw__zlib_huff2(n) : (n) <= 279 ? stbiw__zlib_huff3(n) : stbiw__zlib_huff4(n))\n#define stbiw__zlib_huffb(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : stbiw__zlib_huff2(n))\n\n#define stbiw__ZHASH 16384\n\nunsigned char * stbi_zlib_compress(unsigned char *data, int data_len, int *out_len, int quality)\n{\n static unsigned short lengthc[] = { 3,4,5,6,7,8,9,10,11,13,15,17,19,23,27,31,35,43,51,59,67,83,99,115,131,163,195,227,258, 259 };\n static unsigned char lengtheb[]= { 0,0,0,0,0,0,0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0 };\n static unsigned short distc[] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577, 32768 };\n static unsigned char disteb[] = { 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13 };\n unsigned int bitbuf=0;\n int i,j, bitcount=0;\n unsigned char *out = NULL;\n unsigned char **hash_table[stbiw__ZHASH]; // 64KB on the stack!\n if (quality < 5) quality = 5;\n\n stbiw__sbpush(out, 0x78); // DEFLATE 32K window\n stbiw__sbpush(out, 0x5e); // FLEVEL = 1\n stbiw__zlib_add(1,1); // BFINAL = 1\n stbiw__zlib_add(1,2); // BTYPE = 1 -- fixed huffman\n\n for (i=0; i < stbiw__ZHASH; ++i)\n hash_table[i] = NULL;\n\n i=0;\n while (i < data_len-3) {\n // hash next 3 bytes of data to be compressed\n int h = stbiw__zhash(data+i)&(stbiw__ZHASH-1), best=3;\n unsigned char *bestloc = 0;\n unsigned char **hlist = hash_table[h];\n int n = stbiw__sbcount(hlist);\n for (j=0; j < n; ++j) {\n if (hlist[j]-data > i-32768) { // if entry lies within window\n int d = stbiw__zlib_countm(hlist[j], data+i, data_len-i);\n if (d >= best) best=d,bestloc=hlist[j];\n }\n }\n // when hash table entry is too long, delete half the entries\n if (hash_table[h] && stbiw__sbn(hash_table[h]) == 2*quality) {\n STBIW_MEMMOVE(hash_table[h], hash_table[h]+quality, sizeof(hash_table[h][0])*quality);\n stbiw__sbn(hash_table[h]) = quality;\n }\n stbiw__sbpush(hash_table[h],data+i);\n\n if (bestloc) {\n // \"lazy matching\" - check match at *next* byte, and if it's better, do cur byte as literal\n h = stbiw__zhash(data+i+1)&(stbiw__ZHASH-1);\n hlist = hash_table[h];\n n = stbiw__sbcount(hlist);\n for (j=0; j < n; ++j) {\n if (hlist[j]-data > i-32767) {\n int e = stbiw__zlib_countm(hlist[j], data+i+1, data_len-i-1);\n if (e > best) { // if next match is better, bail on current match\n bestloc = NULL;\n break;\n }\n }\n }\n }\n\n if (bestloc) {\n int d = (int) (data+i - bestloc); // distance back\n STBIW_ASSERT(d <= 32767 && best <= 258);\n for (j=0; best > lengthc[j+1]-1; ++j);\n stbiw__zlib_huff(j+257);\n if (lengtheb[j]) stbiw__zlib_add(best - lengthc[j], lengtheb[j]);\n for (j=0; d > distc[j+1]-1; ++j);\n stbiw__zlib_add(stbiw__zlib_bitrev(j,5),5);\n if (disteb[j]) stbiw__zlib_add(d - distc[j], disteb[j]);\n i += best;\n } else {\n stbiw__zlib_huffb(data[i]);\n ++i;\n }\n }\n // write out final bytes\n for (;i < data_len; ++i)\n stbiw__zlib_huffb(data[i]);\n stbiw__zlib_huff(256); // end of block\n // pad with 0 bits to byte boundary\n while (bitcount)\n stbiw__zlib_add(0,1);\n\n for (i=0; i < stbiw__ZHASH; ++i)\n (void) stbiw__sbfree(hash_table[i]);\n\n {\n // compute adler32 on input\n unsigned int k=0, s1=1, s2=0;\n int blocklen = (int) (data_len % 5552);\n j=0;\n while (j < data_len) {\n for (i=0; i < blocklen; ++i) s1 += data[j+i], s2 += s1;\n s1 %= 65521, s2 %= 65521;\n j += blocklen;\n blocklen = 5552;\n }\n stbiw__sbpush(out, (unsigned char) (s2 >> 8));\n stbiw__sbpush(out, (unsigned char) s2);\n stbiw__sbpush(out, (unsigned char) (s1 >> 8));\n stbiw__sbpush(out, (unsigned char) s1);\n }\n *out_len = stbiw__sbn(out);\n // make returned pointer freeable\n STBIW_MEMMOVE(stbiw__sbraw(out), out, *out_len);\n return (unsigned char *) stbiw__sbraw(out);\n}\n\nunsigned int stbiw__crc32(unsigned char *buffer, int len)\n{\n static unsigned int crc_table[256];\n unsigned int crc = ~0u;\n int i,j;\n if (crc_table[1] == 0)\n for(i=0; i < 256; i++)\n for (crc_table[i]=i, j=0; j < 8; ++j)\n crc_table[i] = (crc_table[i] >> 1) ^ (crc_table[i] & 1 ? 0xedb88320 : 0);\n for (i=0; i < len; ++i)\n crc = (crc >> 8) ^ crc_table[buffer[i] ^ (crc & 0xff)];\n return ~crc;\n}\n\n#define stbiw__wpng4(o,a,b,c,d) ((o)[0]=(unsigned char)(a),(o)[1]=(unsigned char)(b),(o)[2]=(unsigned char)(c),(o)[3]=(unsigned char)(d),(o)+=4)\n#define stbiw__wp32(data,v) stbiw__wpng4(data, (v)>>24,(v)>>16,(v)>>8,(v));\n#define stbiw__wptag(data,s) stbiw__wpng4(data, s[0],s[1],s[2],s[3])\n\nstatic void stbiw__wpcrc(unsigned char **data, int len)\n{\n unsigned int crc = stbiw__crc32(*data - len - 4, len+4);\n stbiw__wp32(*data, crc);\n}\n\nstatic unsigned char stbiw__paeth(int a, int b, int c)\n{\n int p = a + b - c, pa = abs(p-a), pb = abs(p-b), pc = abs(p-c);\n if (pa <= pb && pa <= pc) return (unsigned char) a;\n if (pb <= pc) return (unsigned char) b;\n return (unsigned char) c;\n}\n\nunsigned char *stbi_write_png_to_mem(unsigned char *pixels, int stride_bytes, int x, int y, int n, int *out_len)\n{\n int ctype[5] = { -1, 0, 4, 2, 6 };\n unsigned char sig[8] = { 137,80,78,71,13,10,26,10 };\n unsigned char *out,*o, *filt, *zlib;\n signed char *line_buffer;\n int i,j,k,p,zlen;\n\n if (stride_bytes == 0)\n stride_bytes = x * n;\n\n filt = (unsigned char *) STBIW_MALLOC((x*n+1) * y); if (!filt) return 0;\n line_buffer = (signed char *) STBIW_MALLOC(x * n); if (!line_buffer) { STBIW_FREE(filt); return 0; }\n for (j=0; j < y; ++j) {\n static int mapping[] = { 0,1,2,3,4 };\n static int firstmap[] = { 0,1,0,5,6 };\n int *mymap = j ? mapping : firstmap;\n int best = 0, bestval = 0x7fffffff;\n for (p=0; p < 2; ++p) {\n for (k= p?best:0; k < 5; ++k) {\n int type = mymap[k],est=0;\n unsigned char *z = pixels + stride_bytes*j;\n for (i=0; i < n; ++i)\n switch (type) {\n case 0: line_buffer[i] = z[i]; break;\n case 1: line_buffer[i] = z[i]; break;\n case 2: line_buffer[i] = z[i] - z[i-stride_bytes]; break;\n case 3: line_buffer[i] = z[i] - (z[i-stride_bytes]>>1); break;\n case 4: line_buffer[i] = (signed char) (z[i] - stbiw__paeth(0,z[i-stride_bytes],0)); break;\n case 5: line_buffer[i] = z[i]; break;\n case 6: line_buffer[i] = z[i]; break;\n }\n for (i=n; i < x*n; ++i) {\n switch (type) {\n case 0: line_buffer[i] = z[i]; break;\n case 1: line_buffer[i] = z[i] - z[i-n]; break;\n case 2: line_buffer[i] = z[i] - z[i-stride_bytes]; break;\n case 3: line_buffer[i] = z[i] - ((z[i-n] + z[i-stride_bytes])>>1); break;\n case 4: line_buffer[i] = z[i] - stbiw__paeth(z[i-n], z[i-stride_bytes], z[i-stride_bytes-n]); break;\n case 5: line_buffer[i] = z[i] - (z[i-n]>>1); break;\n case 6: line_buffer[i] = z[i] - stbiw__paeth(z[i-n], 0,0); break;\n }\n }\n if (p) break;\n for (i=0; i < x*n; ++i)\n est += abs((signed char) line_buffer[i]);\n if (est < bestval) { bestval = est; best = k; }\n }\n }\n // when we get here, best contains the filter type, and line_buffer contains the data\n filt[j*(x*n+1)] = (unsigned char) best;\n STBIW_MEMMOVE(filt+j*(x*n+1)+1, line_buffer, x*n);\n }\n STBIW_FREE(line_buffer);\n zlib = stbi_zlib_compress(filt, y*( x*n+1), &zlen, 8); // increase 8 to get smaller but use more memory\n STBIW_FREE(filt);\n if (!zlib) return 0;\n\n // each tag requires 12 bytes of overhead\n out = (unsigned char *) STBIW_MALLOC(8 + 12+13 + 12+zlen + 12);\n if (!out) return 0;\n *out_len = 8 + 12+13 + 12+zlen + 12;\n\n o=out;\n STBIW_MEMMOVE(o,sig,8); o+= 8;\n stbiw__wp32(o, 13); // header length\n stbiw__wptag(o, \"IHDR\");\n stbiw__wp32(o, x);\n stbiw__wp32(o, y);\n *o++ = 8;\n *o++ = (unsigned char) ctype[n];\n *o++ = 0;\n *o++ = 0;\n *o++ = 0;\n stbiw__wpcrc(&o,13);\n\n stbiw__wp32(o, zlen);\n stbiw__wptag(o, \"IDAT\");\n STBIW_MEMMOVE(o, zlib, zlen);\n o += zlen;\n STBIW_FREE(zlib);\n stbiw__wpcrc(&o, zlen);\n\n stbiw__wp32(o,0);\n stbiw__wptag(o, \"IEND\");\n stbiw__wpcrc(&o,0);\n\n STBIW_ASSERT(o == out + *out_len);\n\n return out;\n}\n\n#ifndef STBI_WRITE_NO_STDIO\nSTBIWDEF int stbi_write_png(char const *filename, int x, int y, int comp, const void *data, int stride_bytes)\n{\n FILE *f;\n int len;\n unsigned char *png = stbi_write_png_to_mem((unsigned char *) data, stride_bytes, x, y, comp, &len);\n if (png == NULL) return 0;\n f = fopen(filename, \"wb\");\n if (!f) { STBIW_FREE(png); return 0; }\n fwrite(png, 1, len, f);\n fclose(f);\n STBIW_FREE(png);\n return 1;\n}\n#endif\n\nSTBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int stride_bytes)\n{\n int len;\n unsigned char *png = stbi_write_png_to_mem((unsigned char *) data, stride_bytes, x, y, comp, &len);\n if (png == NULL) return 0;\n func(context, png, len);\n STBIW_FREE(png);\n return 1;\n}\n\n#endif // STB_IMAGE_WRITE_IMPLEMENTATION\n\n/* Revision history\n 1.00 (2015-09-14)\n installable file IO function\n 0.99 (2015-09-13)\n warning fixes; TGA rle support\n 0.98 (2015-04-08)\n added STBIW_MALLOC, STBIW_ASSERT etc\n 0.97 (2015-01-18)\n fixed HDR asserts, rewrote HDR rle logic\n 0.96 (2015-01-17)\n add HDR output\n fix monochrome BMP\n 0.95 (2014-08-17)\n add monochrome TGA output\n 0.94 (2014-05-31)\n rename private functions to avoid conflicts with stb_image.h\n 0.93 (2014-05-27)\n warning fixes\n 0.92 (2010-08-01)\n casts to unsigned char to fix warnings\n 0.91 (2010-07-17)\n first public release\n 0.90 first internal release\n*/\n" }, { "path": "third_party/xxhash.lua", "content": "group(\"third_party\")\nproject(\"xxhash\")\n uuid(\"40d4ce21-5448-4399-9f98-589b7e1c23b1\")\n kind(\"StaticLib\")\n language(\"C\")\n links({\n })\n defines({\n \"_LIB\",\n })\n includedirs({\n \"xxhash\",\n })\n files({\n \"xxhash/xxhash.c\",\n \"xxhash/xxhash.h\",\n })\n" }, { "path": "tools/build/premake", "content": "#!/usr/bin/env python3\n\n# Copyright 2015 Ben Vanik. All Rights Reserved.\n\n\"\"\"Premake trampoline script.\n\"\"\"\n\n__author__ = 'ben.vanik@gmail.com (Ben Vanik)'\n\n\nimport json\nimport os\nimport shutil\nimport subprocess\nimport sys\nimport re\n\n\nself_path = os.path.dirname(os.path.abspath(__file__))\nroot_path = os.path.join(self_path, '..', '..')\npremake_submodule_path = os.path.join(root_path, 'third_party', 'premake-core')\npremake_path = premake_submodule_path\n\n\ndef setup_premake_path_override():\n global premake_path\n premake_path = premake_submodule_path\n if sys.platform == 'linux':\n # On Android, the repository may be cloned to the external storage, which\n # doesn't support executables in it.\n # In this case, premake-core needs to be checked out in the internal\n # storage, which supports executables, with all the permissions as set in\n # its repository.\n # On Termux, the home directory is in the internal storage - use it for\n # executing.\n # If xenia-build doesn't have execute permissions, Xenia is in the external\n # storage now.\n try:\n popen = subprocess.Popen(\n ['uname', '-o'], stdout=subprocess.PIPE, stderr=subprocess.DEVNULL,\n universal_newlines=True)\n if popen.communicate()[0] == 'Android\\n':\n xb_file = os.path.join(root_path, 'xenia-build')\n if (os.path.isfile(xb_file) and not os.access(xb_file, os.X_OK) and\n 'HOME' in os.environ):\n premake_path = os.path.join(\n os.environ['HOME'], '.xenia-build', 'premake-core')\n except Exception:\n pass\n\nsetup_premake_path_override()\n\n\ndef main():\n # First try the freshly-built premake.\n premake5_bin = os.path.join(premake_path, 'bin', 'release', 'premake5')\n if not has_bin(premake5_bin):\n # No fresh build, so fallback to checked in copy (which we may not have).\n premake5_bin = os.path.join(self_path, 'bin', 'premake5')\n if not has_bin(premake5_bin):\n # Still no valid binary, so build it.\n print('premake5 executable not found, attempting build...')\n build_premake()\n premake5_bin = os.path.join(premake_path, 'bin', 'release', 'premake5')\n if not has_bin(premake5_bin):\n # Nope, boned.\n print('ERROR: cannot build premake5 executable.')\n sys.exit(1)\n\n # Ensure the submodule has been checked out.\n if not os.path.exists(os.path.join(premake_path, 'scripts', 'package.lua')):\n print('third_party/premake-core was not present; run xb setup...')\n sys.exit(1)\n return\n\n if sys.platform == 'win32':\n # Append the executable extension on windows.\n premake5_bin = premake5_bin + '.exe'\n\n return_code = shell_call([\n premake5_bin,\n '--scripts=%s' % (premake_path),\n ] + sys.argv[1:],\n throw_on_error=False)\n\n sys.exit(return_code)\n\n\ndef build_premake():\n \"\"\"Builds premake from source.\n \"\"\"\n # Ensure that on Android, premake-core is in the internal storage.\n clone_premake_to_internal_storage()\n cwd = os.getcwd()\n try:\n os.chdir(premake_path)\n if sys.platform == 'darwin':\n subprocess.call([\n 'make',\n '-f', 'Bootstrap.mak',\n 'osx',\n ], shell=False)\n elif sys.platform == 'win32':\n # Grab Visual Studio version and execute shell to set up environment.\n vs_version = import_vs_environment()\n if vs_version is None:\n print('ERROR: Visual Studio not found!')\n sys.exit(1)\n return\n\n subprocess.call([\n 'nmake',\n '-f', 'Bootstrap.mak',\n 'windows',\n ], shell=False)\n else:\n subprocess.call([\n 'make',\n '-f', 'Bootstrap.mak',\n 'linux',\n ], shell=False)\n finally:\n os.chdir(cwd)\n pass\n\n\ndef clone_premake_to_internal_storage():\n \"\"\"Clones premake to the Android internal storage so it can be executed.\n \"\"\"\n # premake_path is initialized to a value different than premake_submodule_path\n # if running from the Android external storage, and may not exist yet.\n if premake_path == premake_submodule_path:\n return\n\n # Ensure the submodule has been checked out.\n if not os.path.exists(\n os.path.join(premake_submodule_path, 'scripts', 'package.lua')):\n print('third_party/premake-core was not present; run xb setup...')\n sys.exit(1)\n return\n\n # Create or refresh premake-core in the internal storage.\n print('Cloning premake5 to the internal storage...')\n shutil.rmtree(premake_path, ignore_errors=True)\n os.makedirs(premake_path)\n shell_call([\n 'git',\n 'clone',\n '--recurse-submodules',\n premake_submodule_path,\n premake_path,\n ])\n\n\ndef has_bin(bin):\n \"\"\"Checks whether the given binary is present.\n \"\"\"\n for path in os.environ[\"PATH\"].split(os.pathsep):\n if sys.platform == 'win32':\n exe_file = os.path.join(path, bin + '.exe')\n if os.path.isfile(exe_file) and os.access(exe_file, os.X_OK):\n return True\n else:\n path = path.strip('\"')\n exe_file = os.path.join(path, bin)\n if os.path.isfile(exe_file) and os.access(exe_file, os.X_OK):\n return True\n return None\n\n\ndef shell_call(command, throw_on_error=True, stdout_path=None, stderr_path=None, shell=False):\n \"\"\"Executes a shell command.\n\n Args:\n command: Command to execute, as a list of parameters.\n throw_on_error: Whether to throw an error or return the status code.\n stdout_path: File path to write stdout output to.\n stderr_path: File path to write stderr output to.\n\n Returns:\n If throw_on_error is False the status code of the call will be returned.\n \"\"\"\n stdout_file = None\n if stdout_path:\n stdout_file = open(stdout_path, 'w')\n stderr_file = None\n if stderr_path:\n stderr_file = open(stderr_path, 'w')\n result = 0\n try:\n if throw_on_error:\n result = 1\n subprocess.check_call(command, shell=shell, stdout=stdout_file, stderr=stderr_file)\n result = 0\n else:\n result = subprocess.call(command, shell=shell, stdout=stdout_file, stderr=stderr_file)\n finally:\n if stdout_file:\n stdout_file.close()\n if stderr_file:\n stderr_file.close()\n return result\n\n\ndef import_vs_environment():\n \"\"\"Finds the installed Visual Studio version and imports\n interesting environment variables into os.environ.\n\n Returns:\n A version such as 2015 or None if no installation is found.\n \"\"\"\n version = 0\n install_path = None\n env_tool_args = None\n\n vswhere = subprocess.check_output('tools/vswhere/vswhere.exe -version \"[15,)\" -latest -format json -utf8', shell=False, universal_newlines=True, encoding=\"utf-8\")\n if vswhere:\n vswhere = json.loads(vswhere)\n if vswhere and len(vswhere) > 0:\n version = int(vswhere[0].get(\"catalog\", {}).get(\"productLineVersion\", 2017))\n install_path = vswhere[0].get(\"installationPath\", None)\n\n if version < 2017:\n if 'VS140COMNTOOLS' in os.environ:\n version = 2015\n vcvars_path = os.environ['VS140COMNTOOLS']\n vcvars_path = os.path.join(tools_path, '..\\\\..\\\\vc\\\\vcvarsall.bat')\n env_tool_args = [vcvars_path, 'x64', '&&', 'set']\n else:\n vsdevcmd_path = os.path.join(install_path, 'Common7\\\\Tools\\\\VsDevCmd.bat')\n env_tool_args = [vsdevcmd_path, '-arch=amd64', '-host_arch=amd64']\n\n if version == 0:\n return None\n\n import_subprocess_environment(env_tool_args)\n os.environ['VSVERSION'] = str(version)\n return version\n\n\ndef import_subprocess_environment(args):\n popen = subprocess.Popen(\n args, shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, universal_newlines=True)\n variables, _ = popen.communicate()\n envvars_to_save = (\n 'devenvdir',\n 'include',\n 'lib',\n 'libpath',\n 'path',\n 'pathext',\n 'systemroot',\n 'temp',\n 'tmp',\n 'windowssdkdir',\n )\n for line in variables.splitlines():\n for envvar in envvars_to_save:\n if re.match(envvar + '=', line.lower()):\n var, setting = line.split('=', 1)\n if envvar == 'path':\n setting = os.path.dirname(sys.executable) + os.pathsep + setting\n os.environ[var.upper()] = setting\n break\n\nif __name__ == '__main__':\n main()\n" }, { "path": "tools/build/premake5.lua", "content": "require(\"vstudio\")\n\ninclude(\"scripts/build_paths.lua\")\ninclude(\"scripts/force_compile_as_c.lua\")\ninclude(\"scripts/force_compile_as_cc.lua\")\ninclude(\"scripts/pkg_config.lua\")\ninclude(\"scripts/platform_files.lua\")\ninclude(\"scripts/single_library.lua\")\ninclude(\"scripts/test_suite.lua\")\n" }, { "path": "tools/build/scripts/build_paths.lua", "content": "build_root = \"build\"\nbuild_bin = build_root .. \"/bin/%{cfg.platform}/%{cfg.buildcfg}\"\nbuild_gen = build_root .. \"/gen/%{cfg.platform}/%{cfg.buildcfg}\"\nbuild_obj = build_root .. \"/obj/%{cfg.platform}/%{cfg.buildcfg}\"\n\nbuild_tools = \"tools/build\"\nbuild_scripts = build_tools .. \"/scripts\"\nbuild_tools_src = build_tools .. \"/src\"\n\nif os.istarget(\"android\") then\n platform_suffix = \"android\"\nelseif os.istarget(\"windows\") then\n platform_suffix = \"win\"\nelse\n platform_suffix = \"posix\"\nend\n" }, { "path": "tools/build/scripts/force_compile_as_c.lua", "content": "if premake.override then\n local forced_c_files = {}\n\n -- Forces all of the given .c and .cc files to be compiled as if they were C.\n function force_compile_as_c(files)\n for _, val in ipairs(files) do\n for _, fname in ipairs(os.matchfiles(val)) do\n table.insert(forced_c_files, path.getabsolute(fname))\n end\n end\n end\n\n -- for gmake\n premake.override(path, \"iscfile\", function(base, fname)\n if table.contains(forced_c_files, fname) then\n return true\n else\n return base(fname)\n end\n end)\n -- for msvc\n premake.override(premake.vstudio.vc2010, \"additionalCompileOptions\", function(base, cfg, condition)\n if cfg.abspath and table.contains(forced_c_files, cfg.abspath) then\n if condition == nil or condition == '' then\n _p(3,'CompileAsC')\n else\n _p(3,'CompileAsC', condition)\n end\n end\n return base(cfg, condition)\n end)\nend\n" }, { "path": "tools/build/scripts/force_compile_as_cc.lua", "content": "if premake.override then\n local forced_cc_files = {}\n\n -- Forces all of the given .c files to be compiled as if they were C++.\n function force_compile_as_cc(files)\n for _, val in ipairs(files) do\n for _, fname in ipairs(os.matchfiles(val)) do\n table.insert(forced_cc_files, path.getabsolute(fname))\n end\n end\n end\n\n -- for gmake\n premake.override(path, \"iscfile\", function(base, fname)\n if table.contains(forced_cc_files, fname) then\n return false\n else\n return base(fname)\n end\n end)\n -- for msvc\n premake.override(premake.vstudio.vc2010, \"additionalCompileOptions\", function(base, cfg, condition)\n if cfg.abspath and table.contains(forced_cc_files, cfg.abspath) then\n if condition == nil or condition == '' then\n _p(3,'CompileAsCpp')\n else\n _p(3,'CompileAsCpp', condition)\n end\n end\n return base(cfg, condition)\n end)\nend\n" }, { "path": "tools/build/scripts/pkg_config.lua", "content": "-- Helper methods to use the system pkg-config utility\n\npkg_config = {}\n\nlocal function pkg_config_call(lib, what)\n local result, code = os.outputof(\"pkg-config --\"..what..\" \"..lib)\n if result then\n return result\n else\n error(\"Failed to run 'pkg-config' for library '\"..lib..\"'. Are the development files installed?\")\n end\nend\n\nfunction pkg_config.cflags(lib)\n if not os.istarget(\"linux\") then\n return\n end\n buildoptions({\n pkg_config_call(lib, \"cflags\"),\n })\nend\n\nfunction pkg_config.lflags(lib)\n if not os.istarget(\"linux\") then\n return\n end\n linkoptions({\n pkg_config_call(lib, \"libs-only-L\"),\n pkg_config_call(lib, \"libs-only-other\"),\n })\n -- We can't just drop the stdout of the `--libs` command in\n -- linkoptions because library order matters\n local output = pkg_config_call(lib, \"libs-only-l\")\n for k, flag in next, string.explode(output, \" \") do\n -- remove \"-l\"\n if flag ~= \"\" then\n links(string.sub(flag, 3))\n end\n end\nend\n\nfunction pkg_config.all(lib)\n pkg_config.cflags(lib)\n pkg_config.lflags(lib)\nend\n" }, { "path": "tools/build/scripts/platform_files.lua", "content": "include(\"build_paths.lua\")\ninclude(\"util.lua\")\n\nlocal function match_platform_files(base_path, base_match)\n files({\n base_path..\"/\"..base_match..\".h\",\n base_path..\"/\"..base_match..\".c\",\n base_path..\"/\"..base_match..\".cc\",\n base_path..\"/\"..base_match..\".cpp\",\n base_path..\"/\"..base_match..\".inc\",\n })\n removefiles({base_path..\"/\"..\"**_main.cc\"})\n removefiles({base_path..\"/\"..\"**_test.cc\"})\n removefiles({base_path..\"/\"..\"**_posix.h\", base_path..\"/\"..\"**_posix.cc\"})\n removefiles({base_path..\"/\"..\"**_linux.h\", base_path..\"/\"..\"**_linux.cc\"})\n removefiles({base_path..\"/\"..\"**_gnulinux.h\",\n base_path..\"/\"..\"**_gnulinux.cc\"})\n removefiles({base_path..\"/\"..\"**_x11.h\", base_path..\"/\"..\"**_x11.cc\"})\n removefiles({base_path..\"/\"..\"**_gtk.h\", base_path..\"/\"..\"**_gtk.cc\"})\n removefiles({base_path..\"/\"..\"**_android.h\", base_path..\"/\"..\"**_android.cc\"})\n removefiles({base_path..\"/\"..\"**_mac.h\", base_path..\"/\"..\"**_mac.cc\"})\n removefiles({base_path..\"/\"..\"**_win.h\", base_path..\"/\"..\"**_win.cc\"})\n filter(\"platforms:Windows\")\n files({\n base_path..\"/\"..base_match..\"_win.h\",\n base_path..\"/\"..base_match..\"_win.cc\",\n })\n filter(\"platforms:Linux or Android-*\")\n files({\n base_path..\"/\"..base_match..\"_posix.h\",\n base_path..\"/\"..base_match..\"_posix.cc\",\n base_path..\"/\"..base_match..\"_linux.h\",\n base_path..\"/\"..base_match..\"_linux.cc\",\n })\n filter(\"platforms:Linux\")\n files({\n base_path..\"/\"..base_match..\"_gnulinux.h\",\n base_path..\"/\"..base_match..\"_gnulinux.cc\",\n base_path..\"/\"..base_match..\"_x11.h\",\n base_path..\"/\"..base_match..\"_x11.cc\",\n base_path..\"/\"..base_match..\"_gtk.h\",\n base_path..\"/\"..base_match..\"_gtk.cc\",\n })\n filter(\"platforms:Android-*\")\n files({\n base_path..\"/\"..base_match..\"_android.h\",\n base_path..\"/\"..base_match..\"_android.cc\",\n })\n filter({})\nend\n\n-- Adds all .h and .cc files in the current path that match the current platform\n-- suffix (_win, etc).\nfunction local_platform_files(base_path)\n match_platform_files(base_path or \".\", \"*\")\nend\n\n-- Adds all .h and .cc files in the current path and all subpaths that match\n-- the current platform suffix (_win, etc).\nfunction recursive_platform_files(base_path)\n match_platform_files(base_path or \".\", \"**\")\nend\n" }, { "path": "tools/build/scripts/single_library.lua", "content": "SINGLE_LIBRARY_PLATFORM_PATTERNS = {\n \"Android-*\",\n};\n\nSINGLE_LIBRARY_FILTER =\n \"platforms:\" .. table.concat(SINGLE_LIBRARY_PLATFORM_PATTERNS, \" or \");\nNOT_SINGLE_LIBRARY_FILTER = table.translate(\n SINGLE_LIBRARY_PLATFORM_PATTERNS,\n function(pattern)\n return \"platforms:not \" .. pattern;\n end);\n\nfunction single_library_windowed_app_kind()\n filter(SINGLE_LIBRARY_FILTER);\n kind(\"StaticLib\");\n wholelib(\"On\");\n filter(NOT_SINGLE_LIBRARY_FILTER);\n kind(\"WindowedApp\");\n filter({});\nend\n" }, { "path": "tools/build/scripts/test_suite.lua", "content": "include(\"build_paths.lua\")\ninclude(\"util.lua\")\n\nnewoption({\n trigger = \"test-suite-mode\",\n description = \"Whether to merge all tests in a test_suite into a single project\",\n value = \"MODE\",\n allowed = {\n { \"individual\", \"One binary per test.\" },\n { \"combined\", \"One binary per test suite (default).\" },\n },\n})\n\nlocal function combined_test_suite(test_suite_name, project_root, base_path, config)\n group(\"tests\")\n project(test_suite_name)\n kind(\"ConsoleApp\")\n language(\"C++\")\n includedirs(merge_arrays(config[\"includedirs\"], {\n project_root..\"/\"..build_tools,\n project_root..\"/\"..build_tools_src,\n project_root..\"/\"..build_tools..\"/third_party/catch/include\",\n }))\n libdirs(merge_arrays(config[\"libdirs\"], {\n project_root..\"/\"..build_bin,\n }))\n links(config[\"links\"])\n if config.filtered_links ~= nil then\n for _, filtered_links in ipairs(config.filtered_links) do\n filter(filtered_links.filter)\n links(filtered_links.links)\n end\n filter({})\n end\n defines({\n \"XE_TEST_SUITE_NAME=\\\"\"..test_suite_name..\"\\\"\",\n })\n files({\n project_root..\"/\"..build_tools_src..\"/test_suite_main.cc\",\n project_root..\"/src/xenia/base/console_app_main_\"..platform_suffix..\".cc\",\n base_path..\"/**_test.cc\",\n })\n filter(\"toolset:msc\")\n -- Edit and Continue in MSVC can cause the __LINE__ macro to produce\n -- invalid values, which breaks the usability of Catch2 output on\n -- failed tests.\n editAndContinue(\"Off\")\nend\n\nlocal function split_test_suite(test_suite_name, project_root, base_path, config)\n local test_paths = os.matchfiles(base_path..\"/**_test.cc\")\n for _, file_path in pairs(test_paths) do\n local test_name = file_path:match(\"(.*).cc\")\n group(\"tests/\"..test_suite_name)\n project(test_suite_name..\"-\"..test_name)\n kind(\"ConsoleApp\")\n language(\"C++\")\n includedirs(merge_arrays(config[\"includedirs\"], {\n project_root..\"/\"..build_tools,\n project_root..\"/\"..build_tools_src,\n project_root..\"/\"..build_tools..\"/third_party/catch/include\",\n }))\n libdirs(merge_arrays(config[\"libdirs\"], {\n project_root..\"/\"..build_bin,\n }))\n links(config[\"links\"])\n if config.filtered_links ~= nil then\n for _, filtered_links in ipairs(config.filtered_links) do\n filter(filtered_links.filter)\n links(filtered_links.links)\n end\n filter({})\n end\n files({\n project_root..\"/\"..build_tools_src..\"/test_suite_main.cc\",\n file_path,\n })\n filter(\"toolset:msc\")\n -- Edit and Continue in MSVC can cause the __LINE__ macro to produce\n -- invalid values, which breaks the usability of Catch2 output on\n -- failed tests.\n editAndContinue(\"Off\")\n end\nend\n\n-- Defines a test suite binary.\n-- Can either be a single binary with all tests or one binary per test based on\n-- the --test-suite-mode= option.\nfunction test_suite(\n test_suite_name, -- Project or group name for the entire suite.\n project_root, -- Project root path (with build_tools/ under it).\n base_path, -- Base source path to search for _test.cc files.\n config) -- Include/lib directories and links for binaries.\n if _OPTIONS[\"test-suite-mode\"] == \"individual\" then\n split_test_suite(test_suite_name, project_root, base_path, config)\n else\n combined_test_suite(test_suite_name, project_root, base_path, config)\n end\nend\n" }, { "path": "tools/build/scripts/util.lua", "content": "-- Prints a table and all of its contents.\nfunction print_r(t)\n local print_r_cache={}\n local function sub_print_r(t, indent)\n if (print_r_cache[tostring(t)]) then\n print(indent..\"*\"..tostring(t))\n else\n print_r_cache[tostring(t)]=true\n if (type(t)==\"table\") then\n for pos,val in pairs(t) do\n if (type(val)==\"table\") then\n print(indent..\"[\"..pos..\"] => \"..tostring(t)..\" {\")\n sub_print_r(val,indent..string.rep(\" \",string.len(pos)+8))\n print(indent..string.rep(\" \",string.len(pos)+6)..\"}\")\n elseif (type(val)==\"string\") then\n print(indent..\"[\"..pos..'] => \"'..val..'\"')\n else\n print(indent..\"[\"..pos..\"] => \"..tostring(val))\n end\n end\n else\n print(indent..tostring(t))\n end\n end\n end\n if (type(t)==\"table\") then\n print(tostring(t)..\" {\")\n sub_print_r(t,\" \")\n print(\"}\")\n else\n sub_print_r(t,\" \")\n end\n print()\nend\n\n-- Merges two tables and returns the resulting table.\nfunction merge_tables(t1, t2)\n local result = {}\n for k,v in pairs(t1 or {}) do result[k] = v end\n for k,v in pairs(t2 or {}) do result[k] = v end\n return result\nend\n\n-- Merges to arrays and returns the resulting array.\nfunction merge_arrays(t1, t2)\n local result = {}\n for k,v in pairs(t1 or {}) do result[#result + 1] = v end\n for k,v in pairs(t2 or {}) do result[#result + 1] = v end\n return result\nend\n" }, { "path": "tools/build/src/test_suite_main.cc", "content": "/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright 2021 Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */\n\n#include \n#include \n#include \n#include \n#include \n\n#include \"xenia/base/console_app_main.h\"\n#include \"xenia/base/cvar.h\"\n\n#define CATCH_CONFIG_RUNNER\n#include \"third_party/catch/single_include/catch2/catch.hpp\"\n\nnamespace xe {\nnamespace test_suite {\n\nint test_suite_main(const std::vector& args) {\n // Catch doesn't expose a way to pass a vector of strings, despite building a\n // vector internally.\n int argc = 0;\n std::vector argv;\n for (const auto& arg : args) {\n argv.push_back(arg.c_str());\n argc++;\n }\n\n // Run Catch.\n return Catch::Session().run(argc, argv.data());\n}\n\n} // namespace test_suite\n} // namespace xe\n\n#ifndef XE_TEST_SUITE_NAME\n#error XE_TEST_SUITE_NAME is undefined!\n#endif\n\nXE_DEFINE_CONSOLE_APP_TRANSPARENT(XE_TEST_SUITE_NAME,\n xe::test_suite::test_suite_main);\n" }, { "path": "tools/diff.py", "content": "#!/usr/bin/env python\n# -*- coding: utf-8 -*-\n\"\"\"python diff.py file1 file2 diff\"\"\"\n\nimport difflib\nimport sys\n\ndiff = difflib.unified_diff(\n open(sys.argv[1], encoding='utf-8').readlines(),\n open(sys.argv[2], encoding='utf-8').readlines())\nwith open(sys.argv[3], 'w', encoding='utf-8') as f:\n f.write(''.join(diff))\n" }, { "path": "tools/generate_map.idc", "content": "#include \n\nstatic is_bad_name(s)\n{\n auto p;\n\n if (s == \"\")\n {\n return 1;\n }\n\n p = substr(s, 0, 4);\n if (p == \"unk_\" ||\n p == \"loc_\" ||\n p == \"sub_\" ||\n p == \"off_\" ||\n p == \"flt_\" ||\n p == \"dbl_\")\n {\n return 1;\n }\n\n p = substr(s, 0, 5);\n if (p == \"byte_\" ||\n p == \"word_\")\n {\n return 1;\n }\n\n p = substr(s, 0, 6);\n if (p == \"dword_\" ||\n p == \"qword_\")\n {\n return 1;\n }\n\n p = substr(s, 0, 7);\n if (p == \"locret_\" ||\n p == \"__imp__\")\n {\n return 1;\n }\n\n p = substr(s, 0, 8);\n if (p == \"xam_xex_\")\n {\n return 1;\n }\n\n p = substr(s, 0, 9);\n if (p == \"xboxkrnl_\")\n {\n return 1;\n }\n\n p = substr(s, 0, 10);\n if (p == \"j_xam_xex_\")\n {\n return 1;\n }\n\n p = substr(s, 0, 15);\n if (p == \"j_xboxkrnl_exe_\")\n {\n return 1;\n }\n\n return 0;\n}\n\nstatic main()\n{\n auto just_code;\n auto seg_start, seg_end, seg_base, seg_name, seg_type;\n auto item_start, item_end, item_flags, item_name;\n auto path, handle;\n\n seg_start = FirstSeg();\n seg_end = SegEnd(seg_start);\n if (seg_start == BADADDR ||\n seg_end == BADADDR)\n {\n return;\n }\n\n path = AskFile(1, \"*.map\", \"Save map\");\n if (path == \"\")\n {\n return;\n }\n\n\n just_code = AskYN(0, \"Just code?\") == 1;\n\n handle = fopen(path, \"wb\");\n if (!handle)\n {\n return;\n }\n\n seg_base = 0;\n fprintf(handle, \" Start Length Name Class\\n\");\n do\n {\n seg_base++;\n seg_name = SegName(seg_start);\n seg_type = GetSegmentAttr(seg_start, SEGATTR_TYPE);\n\n fprintf(handle, \" %04x:%08x %08xH %-23s %s\\n\", seg_base, -1, seg_end - seg_start, seg_name, seg_type == 2 ? \"CODE\" : \"DATA\");\n\n seg_start = NextSeg(seg_start);\n seg_end = SegEnd(seg_start);\n }\n while (seg_start != BADADDR &&\n seg_end != BADADDR);\n\n fprintf(handle, \"\\n\");\n fprintf(handle, \" Address Publics by Value Rva+Base Lib:Object\\n\");\n\n seg_start = FirstSeg();\n seg_end = SegEnd(seg_start);\n seg_base = 0;\n do\n {\n seg_base++;\n\n item_start = seg_start;\n while (item_start < seg_end)\n {\n item_end = item_start + 4;\n if (item_end == BADADDR)\n {\n break;\n }\n\n item_name = GetTrueNameEx(BADADDR, item_start);\n if (is_bad_name(item_name) == 0)\n {\n item_flags = GetFlags(item_start);\n if (just_code == 0 || (item_flags & FF_CODE) == FF_CODE)\n {\n fprintf(handle, \" %04x:%08x %-29s %08x %s\\n\", seg_base, item_start - seg_start, item_name, item_start, \"\");\n }\n }\n item_start = item_start + 4;\n }\n\n seg_start = NextSeg(seg_start);\n seg_end = SegEnd(seg_start);\n }\n while (seg_start != BADADDR &&\n seg_end != BADADDR);\n\n fclose(handle);\n}\n" }, { "path": "tools/gpu-trace-diff", "content": "#!/usr/bin/env python3 \n\n# Copyright 2015 Ben Vanik. All Rights Reserved.\n\n\"\"\"GPU trace runner and comparison tool.\n\"\"\"\n\n__author__ = 'ben.vanik@gmail.com (Ben Vanik)'\n\n\nimport argparse\nimport hashlib\nimport math\nimport operator\nimport os\nimport shutil\nimport subprocess\nimport sys\n\nself_path = os.path.dirname(os.path.abspath(__file__))\n\n\ntry:\n from PIL import Image, ImageChops\nexcept:\n print 'ERROR: could not find ImageChops - install with:'\n print ' python -m pip install Pillow'\n sys.exit(1)\n\n\ndef main():\n # Add self to the root search path.\n sys.path.insert(0, os.path.abspath(os.path.dirname(__file__)))\n\n # Check python version.\n if not sys.version_info[:2] == (2, 7):\n # TODO(benvanik): allow things to work with 3, but warn on clang-format.\n print('ERROR: Python 2.7 must be installed and on PATH')\n sys.exit(1)\n return\n\n # Setup main argument parser and common arguments.\n parser = argparse.ArgumentParser(prog='gpu-trace-diff',\n description='Run and diff GPU traces.')\n parser.add_argument(\n '-x', '--executable',\n default='build/bin/Windows/Debug/xenia-gpu-vulkan-trace-dump.exe')\n parser.add_argument('-t', '--trace_file', action='append')\n parser.add_argument('-p', '--trace_path')\n parser.add_argument('-o', '--output_path', default='')\n parser.add_argument('-r', '--reference_path', default='')\n parser.add_argument('-u', '--update_reference_files', action='store_true')\n parser.add_argument('-n', '--generate_missing_reference_files',\n action='store_true')\n args = vars(parser.parse_args(sys.argv[1:]))\n\n exe_path = args['executable']\n if not os.path.exists(exe_path):\n print 'ERROR: executable %s not found, ensure it is built' % (exe_path)\n sys.exit(1)\n return\n\n trace_files = args['trace_file'] or []\n if args['trace_path']:\n for child_path in os.listdir(args['trace_path']):\n if (os.path.splitext(child_path)[1] == '.xenia_gpu_trace'):\n trace_files.append(os.path.join(args['trace_path'], child_path))\n\n # If the user passed no args, die nicely.\n if not trace_files:\n parser.print_help()\n sys.exit(1)\n return\n\n output_path = args['output_path'].replace('/', os.pathsep)\n if not os.path.exists(output_path):\n os.makedirs(output_path)\n\n html_path = os.path.join(output_path, 'results.html')\n if os.path.exists(html_path):\n os.remove(html_path)\n\n reference_path = args['reference_path'].replace('/', os.pathsep)\n if not os.path.exists(reference_path):\n print('Reference path %s not found; forcing to update mode')\n os.makedirs(reference_path)\n args['update_reference_files'] = True\n\n html_file = None\n if not args['update_reference_files']:\n html_file = open(html_path, 'w')\n html_file.write(\"\"\"\n\nGPU Trace Comparison Results\n\n\n\n\n\n\n\n\"\"\")\n html_rel_path = os.path.relpath(os.getcwd(), output_path)\n html_file.write('' % html_rel_path)\n\n diff_count = 0\n for trace_file in trace_files:\n trace_file = trace_file.replace('/', os.pathsep)\n base_path = os.path.dirname(trace_file)\n file_name = os.path.basename(trace_file)\n reference_file_path = os.path.join(reference_path, file_name + '.png')\n output_file_path = os.path.join(output_path, file_name + '.png')\n diff_file_path = os.path.join(output_path, file_name + '.diff.png')\n\n if (args['generate_missing_reference_files'] and\n os.path.exists(reference_file_path)):\n # Only process tracess that are missing reference files.\n continue\n\n print '--------------------------------------------------------------------'\n print ' Trace: %s' % (trace_file)\n print 'Reference: %s' % (reference_file_path)\n\n # Cleanup old files.\n if os.path.exists(output_file_path):\n os.remove(output_file_path)\n if os.path.exists(diff_file_path):\n os.remove(diff_file_path)\n\n # Run the trace dump too to produce a new png.\n run_args = [\n exe_path.replace('/', os.pathsep),\n '--target_trace_file=%s' % (trace_file.replace(os.pathsep, '/')),\n '--trace_dump_path=%s' % (output_path.replace(os.pathsep, '/')),\n ]\n tries_remaining = 3\n while tries_remaining:\n try:\n startupinfo = None\n if os.name == 'nt':\n startupinfo = subprocess.STARTUPINFO()\n startupinfo.dwFlags |= subprocess.STARTF_USESHOWWINDOW\n startupinfo.wShowWindow = 4 # SW_SHOWNOACTIVATE\n proc = subprocess.Popen(\n run_args,\n shell=True,\n startupinfo=startupinfo,\n stdout=subprocess.PIPE,\n stderr=subprocess.PIPE)\n out, err = proc.communicate()\n # print out\n # print err\n except Exception as e:\n print 'ERROR: failed to run trace dump tool'\n print run_args\n print e\n sys.exit(1)\n return\n\n # Ensure the file was generated\n if not os.path.exists(output_file_path):\n print 'New image not generated; died?'\n tries_remaining -= 1\n continue\n else:\n break\n\n if not tries_remaining and not os.path.exists(output_file_path):\n print 'Retries exceeded; aborting'\n sys.exit(1)\n return\n\n # If updating references, move the output into place.\n if args['update_reference_files']:\n print 'Updating reference file...'\n if not os.path.exists(os.path.dirname(reference_file_path)):\n os.makedirs(os.path.dirname(reference_file_path))\n shutil.copy2(output_file_path, reference_file_path)\n continue\n\n # If we didn't have a reference file for this and are in gen mode, just copy\n # and ignore.\n if (not os.path.exists(reference_file_path) and\n args['generate_missing_reference_files']):\n print 'Adding new reference file...'\n if not os.path.exists(os.path.dirname(reference_file_path)):\n os.makedirs(os.path.dirname(reference_file_path))\n shutil.copy2(output_file_path, reference_file_path)\n continue\n\n # Compare files.\n print ' New: %s' % (output_file_path)\n reference_image = Image.open(reference_file_path)\n reference_rgb = reference_image.convert('RGB').tobytes('raw', 'RGB')\n reference_hash = hashlib.sha1(reference_rgb).hexdigest()\n new_image = Image.open(output_file_path)\n new_rgb = new_image.convert('RGB').tobytes('raw', 'RGB')\n new_hash = hashlib.sha1(new_rgb).hexdigest()\n is_diff = reference_hash != new_hash\n if is_diff:\n print 'DIFFERENCE DETECTED! %s != %s' % (reference_hash, new_hash)\n diff_count += 1\n diff_image = ImageChops.difference(reference_image, new_image)\n diff_image.save(diff_file_path)\n else:\n print 'Matches exactly!'\n\n if html_file:\n html_file.write('')\n html_file.write('' % (trace_file))\n html_file.write('')\n html_file.write('')\n if is_diff:\n html_file.write(\n '' % (\n output_file_path))\n html_file.write(\n '' % (\n diff_file_path))\n html_file.write(\n '' % (\n reference_file_path))\n else:\n html_file.write(\n '' % (\n reference_file_path))\n html_file.write('')\n html_file.write('')\n\n print ''\n\n if html_file:\n html_file.write(\"\"\"\n
    OutputDiffReference
    %s
    \n\n\n\"\"\")\n html_file.close()\n\n if not diff_count:\n print 'All files match!'\n sys.exit(0)\n else:\n print '%s files do not match!' % (diff_count)\n sys.exit(1)\n\nif __name__ == '__main__':\n main()\n" }, { "path": "tools/ppc-instructions.xml", "content": "\n\n\n \n \n \n \n \n \n \n add[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n \n \n addc[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n \n \n adde[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n addi [RD], [RA0], [SIMM]\n \n \n \n \n \n \n addic [RD], [RA], [SIMM]\n \n \n \n \n \n \n \n addic. [RD], [RA], [SIMM]\n \n \n \n \n \n addis [RD], [RA0], [SIMM]\n \n \n \n \n \n \n \n \n addme[OE][Rc] [RD], [RA]\n \n \n \n \n \n \n \n \n addze[OE][Rc] [RD], [RA]\n \n \n \n \n \n \n and[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n andc[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n andi. [RA], [RS], [UIMM]\n \n \n \n \n \n \n andis. [RA], [RS], [UIMM]\n \n \n \n \n \n \n b[LK][AA] [ADDR]\n \n \n \n \n \n \n \n \n \n \n \n bc[LK][AA] [BO], [BI], [ADDR]\n \n \n \n \n \n \n \n \n bcctr[LK] [BO], [BI]\n \n \n \n \n \n \n \n \n \n bclr[LK] [BO], [BI]\n \n \n \n \n \n \n cmp [CRFD], [L], [RA], [RB]\n \n \n \n \n \n \n \n cmpi [CRFD], [L], [RA], [SIMM]\n \n \n \n \n \n \n cmpl [CRFD], [L], [RA], [RB]\n \n \n \n \n \n \n cmpli [CRFD], [L], [RA], [UIMM]\n \n \n \n \n \n \n cntlzd [RA], [RS]\n \n \n \n \n \n cntlzw[Rc] [RA], [RS]\n \n \n \n \n \n crand [CRBD], [CRBA], [CRBB]\n \n \n \n \n \n crandc [CRBD], [CRBA], [CRBB]\n \n \n \n \n \n creqv [CRBD], [CRBA], [CRBB]\n \n \n \n \n \n crnand [CRBD], [CRBA], [CRBB]\n \n \n \n \n \n crnor [CRBD], [CRBA], [CRBB]\n \n \n \n \n \n cror [CRBD], [CRBA], [CRBB]\n \n \n \n \n \n crorc [CRBD], [CRBA], [CRBB]\n \n \n \n \n \n crxor [CRBD], [CRBA], [CRBB]\n \n \n \n \n dcbf [RA0], [RB]\n \n \n \n \n dcbi [RA0], [RB]\n \n \n \n \n dcbst [RA0], [RB]\n \n \n \n \n dcbt [RA0], [RB]\n \n \n \n \n dcbtst [RA0], [RB]\n \n \n \n \n dcbz [RA0], [RB]\n \n \n \n \n dcbz128 [RA0], [RB]\n \n \n \n \n \n \n \n divd[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n \n divdu[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n \n divw[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n \n divwu[OE][Rc] [RD], [RA], [RB]\n \n \n eieio\n \n \n \n \n \n \n eqv[Rc] [RA], [RS], [RB]\n \n \n \n \n \n extsb[Rc] [RA], [RS]\n \n \n \n \n \n extsh[Rc] [RA], [RS]\n \n \n \n \n \n extsw[Rc] [RA], [RS]\n \n \n \n \n \n fabs[Rc] [FD], [FB]\n \n \n \n \n \n \n \n fadd[Rc] [FD], [FA], [FB]\n \n \n \n \n \n \n \n fadds[Rc] [FD], [FA], [FB]\n \n \n \n \n \n \n fcfid[Rc] [FD], [FB]\n \n \n \n \n \n \n fcmpo [CRFD], [FA], [FB]\n \n \n \n \n \n \n fcmpu [CRFD], [FA], [FB]\n \n \n \n \n \n \n fctid[Rc] [FD], [FB]\n \n \n \n \n \n \n fctidz[Rc] [FD], [FB]\n \n \n \n \n \n \n fctiw[Rc] [FD], [FB]\n \n \n \n \n \n \n fctiwz[Rc] [FD], [FB]\n \n \n \n \n \n \n \n fdiv[Rc] [FD], [FA], [FB]\n \n \n \n \n \n \n \n fdivs[Rc] [FD], [FA], [FB]\n \n \n \n \n \n \n \n \n fmadd[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n \n \n \n fmadds[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n fmr[Rc] [FD], [FB]\n \n \n \n \n \n \n \n \n fmsub[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n \n \n \n fmsubs[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n \n \n fmul[Rc] [FD], [FA], [FC]\n \n \n \n \n \n \n \n fmuls[Rc] [FD], [FA], [FC]\n \n \n \n \n \n fnabs[Rc] [FD], [FB]\n \n \n \n \n \n fneg[Rc] [FD], [FB]\n \n \n \n \n \n \n \n \n fnmadd[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n \n \n \n fnmadds[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n \n \n \n fnmsub[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n \n \n \n fnmsubs[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n \n fres[Rc] [FD], [FB]\n \n \n \n \n \n \n frsp[Rc] [FD], [FB]\n \n \n \n \n \n \n frsqrte[Rc] [FD], [FB]\n \n \n \n \n \n \n \n fsel[Rc] [FD], [FA], [FC], [FB]\n \n \n \n \n \n \n fsqrt[Rc] [FD], [FB]\n \n \n \n \n \n \n fsqrts[Rc] [FD], [FB]\n \n \n \n \n \n \n \n fsub[Rc] [FD], [FA], [FB]\n \n \n \n \n \n \n \n fsubs[Rc] [FD], [FA], [FB]\n \n \n \n \n icbi [RA], [RB]\n \n \n isync\n \n \n \n \n \n lbz [RD], [d]([RA0])\n \n \n \n \n \n \n lbzu [RD], [d]([RA])\n \n \n \n \n \n \n lbzux [RD], [RA], [RB]\n \n \n \n \n \n lbzx [RD], [RA0], [RB]\n \n \n \n \n \n ld [RD], [ds]([RA0])\n \n \n \n \n \n ldarx [RD], [RA0], [RB]\n \n \n \n \n \n ldbrx [RD], [RA0], [RB]\n \n \n \n \n \n \n ldu [RD], [ds]([RA])\n \n \n \n \n \n \n ldux [RD], [RA], [RB]\n \n \n \n \n \n ldx [RD], [RA0], [RB]\n \n \n \n \n \n lfd [FD], [d]([RA0])\n \n \n \n \n \n \n lfdu [FD], [d]([RA])\n \n \n \n \n \n \n lfdux [FD], [RA], [RB]\n \n \n \n \n \n lfdx [FD], [RA0], [RB]\n \n \n \n \n \n lfs [FD], [d]([RA0])\n \n \n \n \n \n \n lfsu [FD], [d]([RA])\n \n \n \n \n \n \n lfsux [FD], [RA], [RB]\n \n \n \n \n \n lfsx [FD], [RA0], [RB]\n \n \n \n \n \n lha [RD], [d]([RA0])\n \n \n \n \n \n \n lhau [RD], [d]([RA])\n \n \n \n \n \n \n lhaux [RD], [RA], [RB]\n \n \n \n \n \n lhax [RD], [RA0], [RB]\n \n \n \n \n \n lhbrx [RD], [RA0], [RB]\n \n \n \n \n \n lhz [RD], [d]([RA0])\n \n \n \n \n \n \n lhzu [RD], [d]([RA])\n \n \n \n \n \n \n lhzux [RD], [RA], [RB]\n \n \n \n \n \n lhzx [RD], [RA0], [RB]\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n lwa [RD], [ds]([RA0])\n \n \n \n \n \n lwarx [RD], [RA0], [RB]\n \n \n \n \n \n \n lwaux [RD], [RA], [RB]\n \n \n \n \n \n lwax [RD], [RA0], [RB]\n \n \n \n \n \n lwbrx [RD], [RA0], [RB]\n \n \n \n \n \n lwz [RD], [d]([RA0])\n \n \n \n \n \n \n lwzu [RD], [d]([RA])\n \n \n \n \n \n \n lwzux [RD], [RA], [RB]\n \n \n \n \n \n lwzx [RD], [RA0], [RB]\n \n \n \n \n mcrf [CRFD], [CRFS]\n \n \n \n \n \n \n mcrfs [CRFD], [CRFS]\n \n \n \n \n mcrxr [CRFD]\n \n \n \n \n mfcr [RD]\n \n \n \n \n \n mffs[Rc] [RD]\n \n \n \n \n mfmsr [RD]\n \n \n \n \n mfspr [RD], [SPR]\n \n \n \n \n mftb [RD], [TBR]\n \n \n \n \n mtcrf [CRM], [RS]\n \n \n \n \n mtfsb0[Rc] [FPSCRD]\n \n \n \n \n mtfsb1[Rc] [FPSCRD]\n \n \n \n \n \n \n mtfsf[Rc] [FM], [FB]\n \n \n \n \n \n mtfsfi[Rc] [CRFD], [IMM]\n \n \n \n \n mtmsr [RS]\n \n \n \n \n mtmsrd [RS]\n \n \n \n \n mtspr [SPR], [RS]\n \n \n \n \n \n \n mulhd[Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n mulhdu[Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n mulhw[Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n mulhwu[Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n \n mulld[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n mulli [RD], [RA], [SIMM]\n \n \n \n \n \n \n \n mullw[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n nand[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n neg[OE][Rc] [RD], [RA]\n \n \n \n \n \n \n nor[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n or[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n orc[Rc] [RA], [RS], [RB]\n \n \n \n \n \n ori [RA], [RS], [UIMM]\n \n \n \n \n \n oris [RA], [RS], [UIMM]\n \n \n \n \n \n \n \n rldcl[Rc] [RA], [RS], [RB], [MB]\n \n \n \n \n \n \n \n rldcr[Rc] [RA], [RS], [RB], [ME]\n \n \n \n \n \n \n \n rldic[Rc] [RA], [RS], [SH], [MB]\n \n \n \n \n \n \n \n rldicl[Rc] [RA], [RS], [SH], [MB]\n \n \n \n \n \n \n \n rldicr[Rc] [RA], [RS], [SH], [ME]\n \n \n \n \n \n \n \n rldimi[Rc] [RA], [RS], [SH], [MB]\n \n \n \n \n \n \n \n \n rlwimi[Rc] [RA], [RS], [SH], [MB], [ME]\n \n \n \n \n \n \n \n \n rlwinm[Rc] [RA], [RS], [SH], [MB], [ME]\n \n \n \n \n \n \n \n \n rlwnm[Rc] [RA], [RS], [RB], [MB], [ME]\n \n \n \n sc [LEV]\n \n \n \n \n \n \n sld[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n slw[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n \n srad[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n \n sradi[Rc] [RA], [RS], [SH]\n \n \n \n \n \n \n \n sraw[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n \n srawi[Rc] [RA], [RS], [SH]\n \n \n \n \n \n \n srd[Rc] [RA], [RS], [RB]\n \n \n \n \n \n \n srw[Rc] [RA], [RS], [RB]\n \n \n \n \n \n stb [RS], [d]([RA0])\n \n \n \n \n \n \n stbu [RS], [d]([RA])\n \n \n \n \n \n \n stbux [RS], [RA], [RB]\n \n \n \n \n \n stbx [RS], [RA0], [RB]\n \n \n \n \n \n std [RS], [ds]([RA0])\n \n \n \n \n \n stdbrx [RS], [RA0], [RB]\n \n \n \n \n \n \n stdcx. [RS], [RA0], [RB]\n \n \n \n \n \n \n stdu [RS], [ds]([RA])\n \n \n \n \n \n \n stdux [RS], [RA], [RB]\n \n \n \n \n \n stdx [RS], [RA0], [RB]\n \n \n \n \n \n stfd [FS], [d]([RA0])\n \n \n \n \n \n \n stfdu [FS], [d]([RA])\n \n \n \n \n \n \n stfdux [FS], [RA], [RB]\n \n \n \n \n \n stfdx [FS], [RA0], [RB]\n \n \n \n \n \n stfiwx [FS], [RA0], [RB]\n \n \n \n \n \n stfs [FS], [d]([RA0])\n \n \n \n \n \n \n stfsu [FS], [d]([RA])\n \n \n \n \n \n \n stfsux [FS], [RA], [RB]\n \n \n \n \n \n stfsx [FS], [RA], [RB]\n \n \n \n \n \n sth [RS], [d]([RA0])\n \n \n \n \n \n sthbrx [RS], [RA0], [RB]\n \n \n \n \n \n \n sthu [RS], [d]([RA])\n \n \n \n \n \n \n sthux [RS], [RA], [RB]\n \n \n \n \n \n sthx [RS], [RA0], [RB]\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n stw [RS], [d]([RA0])\n \n \n \n \n \n stwbrx [RS], [RA0], [RB]\n \n \n \n \n \n \n stwcx. [RS], [RA0], [RB]\n \n \n \n \n \n \n stwu [RS], [d]([RA])\n \n \n \n \n \n \n stwux [RS], [RA], [RB]\n \n \n \n \n \n stwx [RS], [RA0], [RB]\n \n \n \n \n \n \n \n subf[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n \n subfc[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n \n subfe[OE][Rc] [RD], [RA], [RB]\n \n \n \n \n \n \n subfic [RD], [RA], [SIMM]\n \n \n \n \n \n \n \n \n subfme[OE][Rc] [RD], [RA]\n \n \n \n \n \n \n \n \n subfze[OE][Rc] [RD], [RA]\n \n \n sync\n \n \n \n \n \n td [TO], [RA], [RB]\n \n \n \n \n \n tdi [TO], [RA], [SIMM]\n \n \n \n \n \n tw [TO], [RA], [RB]\n \n \n \n \n \n tw [TO], [RA], [SIMM]\n \n \n \n \n \n \n xor[Rc] [RA], [RS], [RB]\n \n \n \n \n \n xori [RA], [RS], [UIMM]\n \n \n \n \n \n xoris [RA], [RS], [UIMM]\n \n\n\n \n \n \n \n \n lvebx [VD], [RA0], [RB]\n \n \n \n \n \n lvehx [VD], [RA0], [RB]\n \n \n \n \n \n lvewx [VD], [RA0], [RB]\n \n \n \n \n \n lvewx128 [VD], [RA0], [RB]\n \n \n \n \n \n lvsl [VD], [RA0], [RB]\n \n \n \n \n \n lvsl128 [VD], [RA0], [RB]\n \n \n \n \n \n lvsr [VD], [RA0], [RB]\n \n \n \n \n \n lvsr128 [VD], [RA0], [RB]\n \n \n \n \n \n lvx [VD], [RA0], [RB]\n \n \n \n \n \n lvx128 [VD], [RA0], [RB]\n \n \n \n \n \n lvslx [VD], [RA0], [RB]\n \n \n \n \n \n lvxl128 [VD], [RA0], [RB]\n \n \n \n \n \n lvlx [VD], [RA0], [RB]\n \n \n \n \n \n lvlx128 [VD], [RA0], [RB]\n \n \n \n \n \n lvlxl [VD], [RA0], [RB]\n \n \n \n \n \n lvlxl128 [VD], [RA0], [RB]\n \n \n \n \n \n lvrx [VD], [RA0], [RB]\n \n \n \n \n \n lvrx128 [VD], [RA0], [RB]\n \n \n \n \n \n lvrxl [VD], [RA0], [RB]\n \n \n \n \n \n lvrxl128 [VD], [RA0], [RB]\n \n \n \n \n \n \n \n \n \n \n \n \n \n stvebx [VS], [RA0], [RB]\n \n \n \n \n \n stvehx [VS], [RA0], [RB]\n \n \n \n \n \n stvewx [VS], [RA0], [RB]\n \n \n \n \n \n stvewx128 [VS], [RA0], [RB]\n \n \n \n \n \n stvx [VS], [RA0], [RB]\n \n \n \n \n \n stvx128 [VS], [RA0], [RB]\n \n \n \n \n \n stvxl [VS], [RA0], [RB]\n \n \n \n \n \n stvxl128 [VS], [RA0], [RB]\n \n \n \n \n \n stvlx [VS], [RA0], [RB]\n \n \n \n \n \n stvlx128 [VS], [RA0], [RB]\n \n \n \n \n \n stvlxl [VS], [RA0], [RB]\n \n \n \n \n \n stvlxl128 [VS], [RA0], [RB]\n \n \n \n \n \n stvrx [VS], [RA0], [RB]\n \n \n \n \n \n stvrx128 [VS], [RA0], [RB]\n \n \n \n \n \n stvrxl [VS], [RA0], [RB]\n \n \n \n \n \n stvrxl128 [VS], [RA0], [RB]\n \n \n \n \n \n vaddcuw [VD], [VA], [VB]\n \n \n \n \n \n vaddfp [VD], [VA], [VB]\n \n \n \n \n \n vaddfp128 [VD], [VA], [VB]\n \n \n \n \n \n \n vaddsbs [VD], [VA], [VB]\n \n \n \n \n \n \n vaddshs [VD], [VA], [VB]\n \n \n \n \n \n \n vaddsws [VD], [VA], [VB]\n \n \n \n \n \n vaddubm [VD], [VA], [VB]\n \n \n \n \n \n \n vaddubs [VD], [VA], [VB]\n \n \n \n \n \n vadduhm [VD], [VA], [VB]\n \n \n \n \n \n \n vadduhs [VD], [VA], [VB]\n \n \n \n \n \n vadduwm [VD], [VA], [VB]\n \n \n \n \n \n \n vadduws [VD], [VA], [VB]\n \n \n \n \n \n vand [VD], [VA], [VB]\n \n \n \n \n \n vand128 [VD], [VA], [VB]\n \n \n \n \n \n vandc [VD], [VA], [VB]\n \n \n \n \n \n vandc128 [VD], [VA], [VB]\n \n \n \n \n \n vavgsb [VD], [VA], [VB]\n \n \n \n \n \n vavgsh [VD], [VA], [VB]\n \n \n \n \n \n vavgsw [VD], [VA], [VB]\n \n \n \n \n \n vavgub [VD], [VA], [VB]\n \n \n \n \n \n vavguh [VD], [VA], [VB]\n \n \n \n \n \n vavguw [VD], [VA], [VB]\n \n \n \n \n \n vcfsx [VD], [VB], [UIMM]\n \n \n \n \n \n vcfux [VD], [VB], [UIMM]\n \n \n \n \n \n \n vcmpbfp[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpbfp128[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpeqfp[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpeqfp128[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpequb[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpequh[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpequw[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpequw128[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgefp[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgefp128[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgtfp[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgtfp128[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgtsb[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgtsh[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgtsw[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgtub[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgtuh[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vcmpgtuw[Rc] [VD], [VA], [VB]\n \n \n \n \n \n \n vctsxs [VD], [VB], [UIMM]\n \n \n \n \n \n \n vctuxs [VD], [VB], [UIMM]\n \n \n \n \n vexptefp [VD], [VB]\n \n \n \n \n vexptefp128 [VD], [VB]\n \n \n \n \n vlogefp [VD], [VB]\n \n \n \n \n vlogefp128 [VD], [VB]\n \n \n \n \n \n \n vmaddfp [VD], [VA], [VC], [VB]\n \n \n \n \n \n \n vmaddfp128 [VD], [VA], [VB], [VD]\n \n \n \n \n \n vmaxfp [VD], [VA], [VB]\n \n \n \n \n \n vmaxfp128 [VD], [VA], [VB]\n \n \n \n \n \n vmaxsb [VD], [VA], [VB]\n \n \n \n \n \n vmaxsh [VD], [VA], [VB]\n \n \n \n \n \n vmaxsw [VD], [VA], [VB]\n \n \n \n \n \n vmaxub [VD], [VA], [VB]\n \n \n \n \n \n vmaxuh [VD], [VA], [VB]\n \n \n \n \n \n vmaxuw [VD], [VA], [VB]\n \n \n \n \n \n \n \n vmhaddshs [VD], [VA], [VB], [VC]\n \n \n \n \n \n \n \n vmhraddshs [VD], [VA], [VB], [VC]\n \n \n \n \n \n vminfp [VD], [VA], [VB]\n \n \n \n \n \n vminfp128 [VD], [VA], [VB]\n \n \n \n \n \n vminsb [VD], [VA], [VB]\n \n \n \n \n \n vminsh [VD], [VA], [VB]\n \n \n \n \n \n vminsw [VD], [VA], [VB]\n \n \n \n \n \n vminub [VD], [VA], [VB]\n \n \n \n \n \n vminuh [VD], [VA], [VB]\n \n \n \n \n \n vminuw [VD], [VA], [VB]\n \n \n \n \n \n \n vmladduhm [VD], [VA], [VB], [VC]\n \n \n \n \n \n vmrghb [VD], [VA], [VB]\n \n \n \n \n \n vmrghh [VD], [VA], [VB]\n \n \n \n \n \n vmrghw [VD], [VA], [VB]\n \n \n \n \n \n vmrghw128 [VD], [VA], [VB]\n \n \n \n \n \n vmrglb [VD], [VA], [VB]\n \n \n \n \n \n vmrglh [VD], [VA], [VB]\n \n \n \n \n \n vmrglw [VD], [VA], [VB]\n \n \n \n \n \n vmrglw128 [VD], [VA], [VB]\n \n \n \n \n \n \n vmsummbm [VD], [VA], [VB], [VC]\n \n \n \n \n \n \n vmsumshm [VD], [VA], [VB], [VC]\n \n \n \n \n \n \n \n vmsumshs [VD], [VA], [VB], [VC]\n \n \n \n \n \n \n vmsumubm [VD], [VA], [VB], [VC]\n \n \n \n \n \n \n vmsumuhm [VD], [VA], [VB], [VC]\n \n \n \n \n \n \n \n vmsumuhs [VD], [VA], [VB], [VC]\n \n \n \n \n \n vmulesb [VD], [VA], [VB]\n \n \n \n \n \n vmulesh [VD], [VA], [VB]\n \n \n \n \n \n vmuleub [VD], [VA], [VB]\n \n \n \n \n \n vmuleuh [VD], [VA], [VB]\n \n \n \n \n \n vmulosb [VD], [VA], [VB]\n \n \n \n \n \n vmulosh [VD], [VA], [VB]\n \n \n \n \n \n vmuloub [VD], [VA], [VB]\n \n \n \n \n \n vmulouh [VD], [VA], [VB]\n \n \n \n \n \n \n vnmsubfp [VD], [VA], [VC], [VB]\n \n \n \n \n \n \n vnmsubfp128 [VD], [VA], [VD], [VB]\n \n \n \n \n \n vnor [VD], [VA], [VB]\n \n \n \n \n \n vnor128 [VD], [VA], [VB]\n \n \n \n \n \n vor [VD], [VA], [VB]\n \n \n \n \n \n vor128 [VD], [VA], [VB]\n \n \n \n \n \n \n vperm [VD], [VA], [VB], [VC]\n \n \n \n \n \n \n vperm128 [VD], [VA], [VB], [VC]\n \n \n \n \n \n vpkpx [VD], [VA], [VB]\n \n \n \n \n \n \n vpkshss [VD], [VA], [VB]\n \n \n \n \n \n \n vpkshss128 [VD], [VA], [VB]\n \n \n \n \n \n \n vpkshus [VD], [VA], [VB]\n \n \n \n \n \n \n vpkshus128 [VD], [VA], [VB]\n \n \n \n \n \n \n vpkswss [VD], [VA], [VB]\n \n \n \n \n \n \n vpkswss128 [VD], [VA], [VB]\n \n \n \n \n \n \n vpkswus [VD], [VA], [VB]\n \n \n \n \n \n \n vpkswus128 [VD], [VA], [VB]\n \n \n \n \n \n vpkuhum [VD], [VA], [VB]\n \n \n \n \n \n vpkuhum128 [VD], [VA], [VB]\n \n \n \n \n \n \n vpkuhus [VD], [VA], [VB]\n \n \n \n \n \n \n vpkuhus128 [VD], [VA], [VB]\n \n \n \n \n \n vpkuwum [VD], [VA], [VB]\n \n \n \n \n \n vpkuwum128 [VD], [VA], [VB]\n \n \n \n \n \n \n vpkuwus [VD], [VA], [VB]\n \n \n \n \n \n \n vpkuwus128 [VD], [VA], [VB]\n \n \n \n \n vrefp [VD], [VB]\n \n \n \n \n vrefp128 [VD], [VB]\n \n \n \n \n vrfim [VD], [VB]\n \n \n \n \n vrfim128 [VD], [VB]\n \n \n \n \n vrfin [VD], [VB]\n \n \n \n \n vrfin128 [VD], [VB]\n \n \n \n \n vrfip [VD], [VB]\n \n \n \n \n vrfip128 [VD], [VB]\n \n \n \n \n vrfiz [VD], [VB]\n \n \n \n \n vrfiz128 [VD], [VB]\n \n \n \n \n \n vrlb [VD], [VA], [VB]\n \n \n \n \n \n vrlh [VD], [VA], [VB]\n \n \n \n \n \n vrlw [VD], [VA], [VB]\n \n \n \n \n \n vrlw128 [VD], [VA], [VB]\n \n \n \n \n vrsqrtefp [VD], [VB]\n \n \n \n \n vrsqrtefp128 [VD], [VB]\n \n \n \n \n \n \n vsel [VD], [VA], [VB], [VC]\n \n \n \n \n \n \n vsel128 [VD], [VA], [VB], [VD]\n \n \n \n \n \n vsl [VD], [VA], [VB]\n \n \n \n \n \n vslb [VD], [VA], [VB]\n \n \n \n \n \n \n vsldoi [VD], [VA], [VB], [SHB]\n \n \n \n \n \n \n vsldoi128 [VD], [VA], [VB], [SHB]\n \n \n \n \n \n vslh [VD], [VA], [VB]\n \n \n \n \n \n vslo [VD], [VA], [VB]\n \n \n \n \n \n vslo128 [VD], [VA], [VB]\n \n \n \n \n \n vslw [VD], [VA], [VB]\n \n \n \n \n \n vslw128 [VD], [VA], [VB]\n \n \n \n \n \n vspltb [VD], [VB], [UIMM]\n \n \n \n \n \n vsplth [VD], [VB], [UIMM]\n \n \n \n \n vspltisb [VD], [SIMM]\n \n \n \n \n vspltish [VD], [SIMM]\n \n \n \n \n vspltisw [VD], [SIMM]\n \n \n \n \n vspltisw128 [VD], [SIMM]\n \n \n \n \n \n vspltw [VD], [VB], [UIMM]\n \n \n \n \n \n vspltw128 [VD], [VB], [UIMM]\n \n \n \n \n \n vsr [VD], [VA], [VB]\n \n \n \n \n \n vsrab [VD], [VA], [VB]\n \n \n \n \n \n vsrah [VD], [VA], [VB]\n \n \n \n \n \n vsraw [VD], [VA], [VB]\n \n \n \n \n \n vsraw128 [VD], [VA], [VB]\n \n \n \n \n \n vsrb [VD], [VA], [VB]\n \n \n \n \n \n vsrh [VD], [VA], [VB]\n \n \n \n \n \n vsro [VD], [VA], [VB]\n \n \n \n \n \n vsro128 [VD], [VA], [VB]\n \n \n \n \n \n vsrw [VD], [VA], [VB]\n \n \n \n \n \n vsrw128 [VD], [VA], [VB]\n \n \n \n \n \n vsubcuw [VD], [VA], [VB]\n \n \n \n \n \n vsubfp [VD], [VA], [VB]\n \n \n \n \n \n vsubfp128 [VD], [VA], [VB]\n \n \n \n \n \n \n vsubsbs [VD], [VA], [VB]\n \n \n \n \n \n \n vsubshs [VD], [VA], [VB]\n \n \n \n \n \n \n vsubsws [VD], [VA], [VB]\n \n \n \n \n \n vsububm [VD], [VA], [VB]\n \n \n \n \n \n \n vsububs [VD], [VA], [VB]\n \n \n \n \n \n vsubuhm [VD], [VA], [VB]\n \n \n \n \n \n \n vsubuhs [VD], [VA], [VB]\n \n \n \n \n \n vsubuwm [VD], [VA], [VB]\n \n \n \n \n \n \n vsubuws [VD], [VA], [VB]\n \n \n \n \n \n \n vsumsws [VD], [VA], [VB]\n \n \n \n \n \n \n vsum2sws [VD], [VA], [VB]\n \n \n \n \n \n \n vsum4sbs [VD], [VA], [VB]\n \n \n \n \n \n \n vsum4shs [VD], [VA], [VB]\n \n \n \n \n \n \n vsum4ubs [VD], [VA], [VB]\n \n \n \n \n vupkhpx [VD], [VB]\n \n \n \n \n vupkhsb [VD], [VB]\n \n \n \n \n vupkhsb128 [VD], [VB]\n \n \n \n \n vupkhsh [VD], [VB]\n \n \n \n \n vupklpx [VD], [VB]\n \n \n \n \n vupklsb [VD], [VB]\n \n \n \n \n vupklsb128 [VD], [VB]\n \n \n \n \n vupklsh [VD], [VB]\n \n \n \n \n \n vxor [VD], [VA], [VB]\n \n \n \n \n \n vxor128 [VD], [VA], [VB]\n \n\n \n \n \n \n vmulfp128 [VD], [VA], [VB]\n \n \n \n \n \n \n vmaddcfp128 [VD], [VA], [VD], [VB]\n \n \n \n \n \n vmsum3fp128 [VD], [VA], [VB]\n \n \n \n \n \n vmsum4fp128 [VD], [VA], [VB]\n \n \n \n \n \n vpermwi128 [VD], [VB], [UIMM]\n \n \n \n \n \n \n vcfpsxws128 [VD], [VB], [UIMM]\n \n \n \n \n \n \n vcfpuxws128 [VD], [VB], [UIMM]\n \n \n \n \n \n vcsxwfp128 [VD], [VB], [UIMM]\n \n \n \n \n \n vcuxwfp128 [VD], [VB], [UIMM]\n \n \n \n \n \n \n \n \n \n vrlimi128 [VD], [VB], [IMM], [z]\n \n \n \n \n \n \n\n\n" }, { "path": "tools/ppc-table-gen", "content": "#!/usr/bin/env python3 \n\n# Copyright 2015 Ben Vanik & shuffle2. All Rights Reserved.\n\n\"\"\"PPC instruction table generator.\n\nGenerates various headers/sources for looking up and handling PPC instructions.\n\nThis is based on shuffle2's PPC generator:\nhttps://gist.github.com/shuffle2/10015968\n\"\"\"\n\n__author__ = 'ben.vanik@gmail.com (Ben Vanik)'\n\nimport os\nimport sys\nfrom xml.etree.ElementTree import ElementTree, Element, SubElement, tostring, dump\n\n\ncustom_disasm = ['bcx']\n\n\nself_path = os.path.dirname(os.path.abspath(__file__))\n\n\ndef bit_extract(x, leftmost, rightmost):\n return (x >> (32 - 1 - rightmost)) & ((1 << (rightmost - leftmost + 1)) - 1)\n\n\nextended_opcode_bits = {\n 'X': [(21, 30)],\n 'XL': [(21, 30)],\n 'XFX': [(21, 30)],\n 'XFL': [(21, 30)],\n 'VX': [(21, 31)],\n 'VX128': [(22, 25), (27, 27)],\n 'VX128_1': [(21, 27), (30, 31)],\n 'VX128_2': [(22, 22), (27, 27)],\n 'VX128_3': [(21, 27)],\n 'VX128_4': [(21, 23), (26, 27)],\n 'VX128_5': [(27, 27)],\n 'VX128_R': [(22, 24), (27, 27)],\n 'VX128_P': [(21, 22), (26, 27)],\n 'VC': [(22, 31)],\n 'VA': [(26, 31)],\n 'XO': [(22, 30)],\n 'XW': [(25, 30)],\n 'A': [(26, 30)],\n 'DS': [(30, 31)],\n 'MD': [(27, 29)],\n 'MDS': [(27, 30)],\n 'XS': [(21, 29)],\n 'DCBZ': [(6, 10), (21, 30)], # like X\n }\n\n\ndef opcode_primary(insn):\n return bit_extract(insn, 0, 5)\n\n\ndef opcode_extended(insn, form):\n if form in extended_opcode_bits:\n parts = extended_opcode_bits[form]\n value = 0\n shift = 0\n for part in parts:\n shift = max(shift, part[1])\n for part in parts:\n part_value = bit_extract(insn, part[0], part[1])\n value = value | (part_value << (shift - part[1]))\n return value\n else:\n return -1\n\n\nclass Insn:\n pass\n\n\ndef parse_insns(filename):\n root = ElementTree(file = filename)\n insns = []\n # Convert to python types\n for e in root.findall('.//insn'):\n i = Insn()\n i.opcode = int(e.attrib['opcode'], 16)\n i.mnem = e.attrib['mnem']\n i.form = e.attrib['form']\n i.group = e.attrib['group']\n i.desc = e.attrib['desc']\n i.type = 'General'\n if 'sync' in e.attrib and e.attrib['sync'] == 'true':\n i.type = 'Sync'\n i.op_primary = opcode_primary(i.opcode)\n i.op_extended = opcode_extended(i.opcode, i.form)\n i.reads = []\n i.writes = []\n for r in e.findall('.//in'):\n is_conditional = 'conditional' in r.attrib and r.attrib['conditional'] == 'true'\n i.reads.append((r.attrib['field'], is_conditional))\n for w in e.findall('.//out'):\n is_conditional = 'conditional' in w.attrib and w.attrib['conditional'] == 'true'\n i.writes.append((w.attrib['field'], is_conditional))\n i.disasm_str = None\n for d in e.findall('.//disasm'):\n i.disasm_str = d.text\n insns.append(i)\n return insns\n\n\ndef c_mnem(x):\n return x.replace('.', 'x')\n\n\ndef c_group(x):\n return 'k' + x[0].upper() + x[1:]\n\n\ndef c_bool(x):\n return 'true' if x else 'false'\n\n\ndef c_field(x):\n base_name = 'k' + x[0]\n cond_name = 'cond' if x[1] else ''\n return base_name + cond_name\n\n\ndef generate_opcodes(insns):\n l = []\n TAB = ' ' * 2\n def w0(x): l.append(x)\n def w1(x): w0(TAB * 1 + x)\n def w2(x): w0(TAB * 2 + x)\n def w3(x): w0(TAB * 3 + x)\n\n w0('// This code was autogenerated by %s. Do not modify!' % (sys.argv[0]))\n w0('// clang-format off')\n w0('#ifndef XENIA_CPU_PPC_PPC_OPCODE_H_')\n w0('#define XENIA_CPU_PPC_PPC_OPCODE_H_')\n w0('')\n w0('#include ')\n w0('#include ')\n w0('')\n w0('namespace xe {')\n w0('namespace cpu {')\n w0('namespace ppc {')\n w0('')\n\n for i in insns:\n i.mnem = c_mnem(i.mnem)\n insns = sorted(insns, key = lambda i: i.mnem)\n\n w0('// All PPC opcodes in the same order they appear in ppc_opcode_table.h:')\n w0('enum class PPCOpcode : uint32_t {')\n for i in insns:\n w1('%s,' % (i.mnem))\n w1('kInvalid,')\n w0('};')\n\n w0('')\n w0('} // namespace ppc')\n w0('} // namespace cpu')\n w0('} // namespace xe')\n w0('')\n w0('#endif // XENIA_CPU_PPC_PPC_OPCODE_H_')\n w0('')\n\n return '\\n'.join(l)\n\n\ndef generate_table(insns):\n l = []\n TAB = ' ' * 2\n def w0(x): l.append(x)\n def w1(x): w0(TAB * 1 + x)\n def w2(x): w0(TAB * 2 + x)\n def w3(x): w0(TAB * 3 + x)\n\n w0('// This code was autogenerated by %s. Do not modify!' % (sys.argv[0]))\n w0('// clang-format off')\n w0('#include ')\n w0('#include ')\n w0('')\n w0('#include \"xenia/base/assert.h\"')\n w0('#include \"xenia/cpu/ppc/ppc_opcode.h\"')\n w0('#include \"xenia/cpu/ppc/ppc_opcode_info.h\"')\n w0('')\n w0('namespace xe {')\n w0('namespace cpu {')\n w0('namespace ppc {')\n w0('')\n\n for i in insns:\n i.mnem = '\"' + c_mnem(i.mnem) + '\"'\n i.form = c_group(i.form)\n i.group = c_group(i.group)\n i.type = c_group(i.type)\n\n mnem_len = len(max(insns, key = lambda i: len(i.mnem)).mnem)\n form_len = len(max(insns, key = lambda i: len(i.form)).form)\n group_len = len(max(insns, key = lambda i: len(i.group)).group)\n type_len = len(max(insns, key = lambda i: len(i.type)).type)\n\n insns = sorted(insns, key = lambda i: i.mnem)\n\n w0('#define INSTRUCTION(opcode, mnem, form, group, type) \\\\')\n w0(' {PPCOpcodeGroup::group, PPCOpcodeType::type, nullptr}')\n w0('PPCOpcodeInfo ppc_opcode_table[] = {')\n fmt = 'INSTRUCTION(' + ', '.join([\n '0x%08x',\n '%-' + str(mnem_len) + 's',\n '%-' + str(form_len) + 's',\n '%-' + str(group_len) + 's',\n '%-' + str(type_len) + 's',\n ]) + '),'\n for i in insns:\n w1(fmt % (i.opcode, i.mnem, i.form, i.group, i.type))\n w0('};')\n w0('static_assert(sizeof(ppc_opcode_table) / sizeof(PPCOpcodeInfo) == static_cast(PPCOpcode::kInvalid), \"PPC table mismatch - rerun ppc-table-gen\");')\n w0('')\n w0('const PPCOpcodeInfo& GetOpcodeInfo(PPCOpcode opcode) {')\n w1('return ppc_opcode_table[static_cast(opcode)];')\n w0('}')\n w0('void RegisterOpcodeEmitter(PPCOpcode opcode, InstrEmitFn fn) {')\n w1('assert_null(ppc_opcode_table[static_cast(opcode)].emit);')\n w1('ppc_opcode_table[static_cast(opcode)].emit = fn;')\n w0('}')\n\n w0('')\n w0('} // namespace ppc')\n w0('} // namespace cpu')\n w0('} // namespace xe')\n w0('')\n\n return '\\n'.join(l)\n\n\ndef literal_mnem(x):\n x = x.replace('.', '_')\n x = x.replace('\"', '')\n return x\n\ndef generate_token_append(i, token):\n # Rc = . iff Rc=1\n # OE = o iff OE=1\n if token == 'Rc':\n return 'if (d.%s.Rc()) str->Append(\\'.\\');' % (i.o_form)\n elif token == 'OE':\n return 'if (d.%s.OE()) str->Append(\\'o\\');' % (i.o_form)\n elif token == 'LK':\n return 'if (d.%s.LK()) str->Append(\\'l\\');' % (i.o_form)\n elif token == 'AA':\n return 'if (d.%s.AA()) str->Append(\\'a\\');' % (i.o_form)\n elif token in ['RA', 'RA0', 'RB', 'RC', 'RT', 'RS', 'RD']:\n return 'str->AppendFormat(\"r{}\", d.%s.%s());' % (i.o_form, token)\n elif token in ['FA', 'FB', 'FC', 'FT', 'FS', 'FD']:\n return 'str->AppendFormat(\"fr{}\", d.%s.%s());' % (i.o_form, token)\n elif token in ['VA', 'VB', 'VC', 'VT', 'VS', 'VD']:\n return 'str->AppendFormat(\"vr{}\", d.%s.%s());' % (i.o_form, token)\n elif token in ['CRFD', 'CRFS']:\n return 'str->AppendFormat(\"crf{}\", d.%s.%s());' % (i.o_form, token)\n elif token in ['CRBA', 'CRBB', 'CRBD']:\n return 'str->AppendFormat(\"crb{}\", d.%s.%s());' % (i.o_form, token)\n elif token in ['BO', 'BI', 'TO', 'SPR', 'TBR', 'L', 'FM', 'MB', 'ME', 'SH', 'IMM', 'z']:\n return 'str->AppendFormat(\"{}\", d.%s.%s());' % (i.o_form, token)\n elif token == 'UIMM':\n return 'str->AppendFormat(\"0x{:X}\", d.%s.%s());' % (i.o_form, token)\n elif token in ['d', 'ds', 'SIMM']:\n return 'str->AppendFormat(d.%s.%s() < 0 ? \"-0x{:X}\" : \"0x{:X}\", std::abs(d.%s.%s()));' % (i.o_form, token, i.o_form, token)\n elif token == 'ADDR':\n return 'str->AppendFormat(\"0x{:X}\", d.%s.%s());' % (i.o_form, token)\n elif token == 'LEV':\n return 'if (d.%s.%s()) str->AppendFormat(\"{}\", d.%s.%s());' % (i.o_form, token, i.o_form, token)\n return 'str->AppendFormat(\"(UNHANDLED %s)\");' % token\n\n\ndef generate_disasm(insns):\n l = []\n TAB = ' ' * 2\n def w0(x): l.append(x)\n def w1(x): w0(TAB * 1 + x)\n def w2(x): w0(TAB * 2 + x)\n def w3(x): w0(TAB * 3 + x)\n\n w0('// This code was autogenerated by %s. Do not modify!' % (sys.argv[0]))\n w0('// clang-format off')\n w0('#include ')\n w0('#include ')\n w0('')\n w0('#include \"xenia/base/assert.h\"')\n w0('#include \"xenia/cpu/ppc/ppc_decode_data.h\"')\n w0('#include \"xenia/cpu/ppc/ppc_opcode.h\"')\n w0('#include \"xenia/cpu/ppc/ppc_opcode_disasm.h\"')\n w0('#include \"xenia/cpu/ppc/ppc_opcode_info.h\"')\n w0('')\n w0('namespace xe {')\n w0('namespace cpu {')\n w0('namespace ppc {')\n w0('')\n\n for i in insns:\n i.mnem = '\"' + c_mnem(i.mnem) + '\"'\n i.o_form = i.form\n i.form = c_group(i.form)\n i.desc = '\"' + i.desc + '\"'\n i.group = c_group(i.group)\n i.type = c_group(i.type)\n\n mnem_len = len(max(insns, key = lambda i: len(i.mnem)).mnem)\n form_len = len(max(insns, key = lambda i: len(i.form)).form)\n desc_len = len(max(insns, key = lambda i: len(i.desc)).desc)\n group_len = len(max(insns, key = lambda i: len(i.group)).group)\n type_len = len(max(insns, key = lambda i: len(i.type)).type)\n\n insns = sorted(insns, key = lambda i: i.mnem)\n\n # TODO(benvanik): support alts:\n # li [RD], [SIMM]\n for i in insns:\n if not i.disasm_str or literal_mnem(i.mnem) in custom_disasm:\n continue\n w0('void PrintDisasm_%s(const PPCDecodeData& d, StringBuffer* str) {' % (literal_mnem(i.mnem)))\n w1('// ' + i.disasm_str)\n w1('size_t str_start = str->length();')\n current_str = ''\n j = 0\n first_space = False\n while j < len(i.disasm_str):\n c = i.disasm_str[j]\n if c == '[':\n if current_str:\n w1('str->Append(\"%s\");' % (current_str))\n current_str = ''\n token = i.disasm_str[j + 1 : i.disasm_str.index(']', j)]\n j += len(token) + 1\n w1(generate_token_append(i, token))\n else:\n if c == ' ' and not first_space:\n if current_str:\n w1('str->Append(\"%s\");' % (current_str))\n current_str = ''\n w1('PadStringBuffer(str, str_start, kNamePad);')\n first_space = True\n else:\n current_str += c\n j += 1\n if current_str:\n w1('str->Append(\"%s\");' % (current_str))\n if not first_space:\n w1('PadStringBuffer(str, str_start, kNamePad);')\n w0('}')\n\n w0('#define INIT_LIST(...) {__VA_ARGS__}')\n w0('#define INSTRUCTION(opcode, mnem, form, group, type, desc, reads, writes, fn) \\\\')\n w0(' {PPCOpcodeGroup::group, PPCOpcodeFormat::form, opcode, mnem, desc, INIT_LIST reads, INIT_LIST writes, fn}')\n w0('PPCOpcodeDisasmInfo ppc_opcode_disasm_table[] = {')\n fmt = 'INSTRUCTION(' + ', '.join([\n '0x%08x',\n '%-' + str(mnem_len) + 's',\n '%-' + str(form_len) + 's',\n '%-' + str(group_len) + 's',\n '%-' + str(type_len) + 's',\n '%-' + str(desc_len) + 's',\n '(%s)',\n '(%s)',\n '%s',\n ]) + '),'\n for i in insns:\n w1(fmt % (\n i.opcode,\n i.mnem,\n i.form,\n i.group,\n i.type,\n i.desc,\n ','.join(['PPCOpcodeField::' + c_field(r) for r in i.reads]),\n ','.join(['PPCOpcodeField::' + c_field(w) for w in i.writes]),\n ('PrintDisasm_' + literal_mnem(i.mnem)) if i.disasm_str else 'nullptr',\n ))\n w0('};')\n w0('static_assert(sizeof(ppc_opcode_disasm_table) / sizeof(PPCOpcodeDisasmInfo) == static_cast(PPCOpcode::kInvalid), \"PPC table mismatch - rerun ppc-table-gen\");')\n w0('')\n w0('const PPCOpcodeDisasmInfo& GetOpcodeDisasmInfo(PPCOpcode opcode) {')\n w1('return ppc_opcode_disasm_table[static_cast(opcode)];')\n w0('}')\n w0('void RegisterOpcodeDisasm(PPCOpcode opcode, InstrDisasmFn fn) {')\n w1('assert_null(ppc_opcode_disasm_table[static_cast(opcode)].disasm);')\n w1('ppc_opcode_disasm_table[static_cast(opcode)].disasm = fn;')\n w0('}')\n\n w0('')\n w0('} // namespace ppc')\n w0('} // namespace cpu')\n w0('} // namespace xe')\n w0('')\n\n return '\\n'.join(l)\n\n\ndef generate_lookup(insns):\n l = []\n TAB = ' ' * 2\n def w0(x): l.append(x)\n def w1(x): w0(TAB * 1 + x)\n def w2(x): w0(TAB * 2 + x)\n def w3(x): w0(TAB * 3 + x)\n\n for i in insns:\n i.mnem = c_mnem(i.mnem)\n\n w0('// This code was autogenerated by %s. Do not modify!' % (sys.argv[0]))\n w0('// clang-format off')\n w0('#include ')\n w0('#include ')\n w0('')\n w0('#include \"xenia/base/assert.h\"')\n w0('#include \"xenia/cpu/ppc/ppc_opcode.h\"')\n w0('#include \"xenia/cpu/ppc/ppc_opcode_info.h\"')\n w0('')\n w0('namespace xe {')\n w0('namespace cpu {')\n w0('namespace ppc {')\n w0('')\n w0('constexpr uint32_t ExtractBits(uint32_t v, uint32_t a, uint32_t b) {')\n w0(' return (v >> (32 - 1 - b)) & ((1 << (b - a + 1)) - 1);')\n w0('}')\n w0('')\n w0('#define PPC_DECODER_MISS assert_always(); return PPCOpcode::kInvalid')\n w0('#define PPC_DECODER_HIT(name) return PPCOpcode::name;')\n w0('')\n w0('PPCOpcode LookupOpcode(uint32_t code) {')\n w1('switch (ExtractBits(code, 0, 5)) {')\n\n subtables = {}\n for i in sorted(insns, key = lambda i: i.op_primary):\n if i.op_primary not in subtables: subtables[i.op_primary] = []\n subtables[i.op_primary].append(i)\n\n for pri in sorted(subtables.iterkeys()):\n # all the extended encodings (which we care about) end with bit 30. So we want to\n # do the rest of the seach by bitscanning left from bit 30. This is simulated\n # in the C switch-statement by creating leafs for each extended opcode,\n # sorted by bitlength shortest to longest.\n\n if len(subtables[pri]) == 1:\n for i in subtables[pri]:\n # the primary opcode field fully identifies the opcode\n w1('case %i: PPC_DECODER_HIT(%s);' % (i.op_primary, i.mnem))\n continue\n\n extract_groups = {}\n for i in subtables[pri]:\n form_parts = extended_opcode_bits[i.form]\n shift = 0\n for form_part in form_parts:\n shift = max(shift, form_part[1])\n extract_parts = []\n for form_part in form_parts:\n extract_parts.append('(ExtractBits(code, %s, %s) << %s)' % (form_part[0], form_part[1], shift - form_part[1]))\n extract_expression = '|'.join(extract_parts)\n if extract_expression not in extract_groups:\n extract_groups[extract_expression] = (i.form, extract_expression, [])\n extract_groups[extract_expression][2].append(i)\n\n w1('case %i:' % (pri))\n for extract_expression in sorted(extract_groups.iterkeys()):\n (form, extract_expression, group_insns) = extract_groups[extract_expression]\n bit_span_low = 31\n bit_span_high = 0\n form_parts = extended_opcode_bits[form]\n for form_part in form_parts:\n bit_span_low = min(bit_span_low, form_part[0])\n bit_span_high = max(bit_span_high, form_part[1])\n bit_count = bit_span_high - bit_span_low + 1\n w2('switch (%s) {' % (extract_expression))\n for i in sorted(group_insns, key=lambda i: i.op_extended):\n w3('case 0b%s: PPC_DECODER_HIT(%s);' % (\n ('{:0'+str(bit_count)+'b}').format(i.op_extended),\n i.mnem))\n w2('}')\n w2('PPC_DECODER_MISS;')\n\n w1('default: PPC_DECODER_MISS;')\n\n w1('}')\n w0('}')\n w0('')\n w0('} // namespace ppc')\n w0('} // namespace cpu')\n w0('} // namespace xe')\n w0('')\n\n # from this we can see some tables have bits which can be used to determine extended opcoded size:\n # primary opcode 31:\n # 01... = 9 bits (XO form), else 10 bits (X/XFX forms)\n # primary opcode 63:\n # 1.... = 7 bits (A form), else 10 bits (X/XFL forms)\n # primary opcode 4:\n # does not have small bit range to determine size, but you can just use the\n # low 7 bits in order to \"guess\" the opcode. if you assume no invalid\n # encodings are input, only the sequence ...0001000 actually *needs* the upper\n # bits in order to differentiate the opcode (0100001000 = ps_abs, 0010001000 = ps_nabs)\n # otherwise, the low 7bits can be used as the determinant, and a second comparison\n # can be used against the real length of bits to fully match the extended opcode\n #\n # this approach can be generalized for all primary opcodes with extended opcodes of varying lengths:\n # compare bits of smallest length, fall through to comparing larger sizes until found or failure\n # with the optional optimization of discarding further compares for extended opcodes which\n # share top bits with any other extended opcode (at the price of failing to detect invalid opcodes)\n\n return '\\n'.join(l)\n\n\nif __name__ == '__main__':\n ppc_src_path = os.path.join(self_path, '..', 'src', 'xenia', 'cpu', 'ppc')\n insns = parse_insns(os.path.join(self_path, 'ppc-instructions.xml'))\n with open(os.path.join(ppc_src_path, 'ppc_opcode.h'), 'w') as f:\n f.write(generate_opcodes(insns))\n insns = parse_insns(os.path.join(self_path, 'ppc-instructions.xml'))\n with open(os.path.join(ppc_src_path, 'ppc_opcode_table_gen.cc'), 'w') as f:\n f.write(generate_table(insns))\n insns = parse_insns(os.path.join(self_path, 'ppc-instructions.xml'))\n with open(os.path.join(ppc_src_path, 'ppc_opcode_disasm_gen.cc'), 'w') as f:\n f.write(generate_disasm(insns))\n insns = parse_insns(os.path.join(self_path, 'ppc-instructions.xml'))\n with open(os.path.join(ppc_src_path, 'ppc_opcode_lookup_gen.cc'), 'w') as f:\n f.write(generate_lookup(insns))\n" }, { "path": "tools/shader-playground/App.config", "content": "\n\n \n \n\n" }, { "path": "tools/shader-playground/Editor.Designer.cs", "content": "namespace shader_playground {\n partial class Editor {\n /// \n /// Required designer variable.\n /// \n private System.ComponentModel.IContainer components = null;\n\n /// \n /// Clean up any resources being used.\n /// \n /// true if managed resources should be disposed; otherwise, false.\n protected override void Dispose(bool disposing) {\n if (disposing && (components != null)) {\n components.Dispose();\n }\n base.Dispose(disposing);\n }\n\n #region Windows Form Designer generated code\n\n /// \n /// Required method for Designer support - do not modify\n /// the contents of this method with the code editor.\n /// \n private void InitializeComponent() {\n this.wordsTextBox = new System.Windows.Forms.TextBox();\n this.tableLayoutPanel1 = new System.Windows.Forms.TableLayoutPanel();\n this.vertexShaderComboBox = new System.Windows.Forms.ComboBox();\n this.compilerTranslatedTextBox = new System.Windows.Forms.TextBox();\n this.sourceCodeTextBox = new System.Windows.Forms.TextBox();\n this.outputTextBox = new System.Windows.Forms.TextBox();\n this.compilerUcodeTextBox = new System.Windows.Forms.TextBox();\n this.label1 = new System.Windows.Forms.Label();\n this.label2 = new System.Windows.Forms.Label();\n this.label3 = new System.Windows.Forms.Label();\n this.translationComboBox = new System.Windows.Forms.ComboBox();\n this.tableLayoutPanel1.SuspendLayout();\n this.SuspendLayout();\n // \n // wordsTextBox\n // \n this.wordsTextBox.Anchor = ((System.Windows.Forms.AnchorStyles)(((System.Windows.Forms.AnchorStyles.Bottom | System.Windows.Forms.AnchorStyles.Left) \n | System.Windows.Forms.AnchorStyles.Right)));\n this.wordsTextBox.Font = new System.Drawing.Font(\"Consolas\", 8.25F, System.Drawing.FontStyle.Regular, System.Drawing.GraphicsUnit.Point, ((byte)(0)));\n this.wordsTextBox.Location = new System.Drawing.Point(12, 657);\n this.wordsTextBox.Multiline = true;\n this.wordsTextBox.Name = \"wordsTextBox\";\n this.wordsTextBox.ReadOnly = true;\n this.wordsTextBox.Size = new System.Drawing.Size(1631, 137);\n this.wordsTextBox.TabIndex = 11;\n // \n // tableLayoutPanel1\n // \n this.tableLayoutPanel1.Anchor = ((System.Windows.Forms.AnchorStyles)((((System.Windows.Forms.AnchorStyles.Top | System.Windows.Forms.AnchorStyles.Bottom) \n | System.Windows.Forms.AnchorStyles.Left) \n | System.Windows.Forms.AnchorStyles.Right)));\n this.tableLayoutPanel1.AutoSizeMode = System.Windows.Forms.AutoSizeMode.GrowAndShrink;\n this.tableLayoutPanel1.ColumnCount = 4;\n this.tableLayoutPanel1.ColumnStyles.Add(new System.Windows.Forms.ColumnStyle(System.Windows.Forms.SizeType.Percent, 25F));\n this.tableLayoutPanel1.ColumnStyles.Add(new System.Windows.Forms.ColumnStyle(System.Windows.Forms.SizeType.Percent, 25F));\n this.tableLayoutPanel1.ColumnStyles.Add(new System.Windows.Forms.ColumnStyle(System.Windows.Forms.SizeType.Percent, 25F));\n this.tableLayoutPanel1.ColumnStyles.Add(new System.Windows.Forms.ColumnStyle(System.Windows.Forms.SizeType.Percent, 25F));\n this.tableLayoutPanel1.Controls.Add(this.vertexShaderComboBox, 3, 1);\n this.tableLayoutPanel1.Controls.Add(this.compilerTranslatedTextBox, 3, 2);\n this.tableLayoutPanel1.Controls.Add(this.sourceCodeTextBox, 0, 1);\n this.tableLayoutPanel1.Controls.Add(this.outputTextBox, 1, 1);\n this.tableLayoutPanel1.Controls.Add(this.compilerUcodeTextBox, 2, 1);\n this.tableLayoutPanel1.Controls.Add(this.label1, 0, 0);\n this.tableLayoutPanel1.Controls.Add(this.label2, 1, 0);\n this.tableLayoutPanel1.Controls.Add(this.label3, 2, 0);\n this.tableLayoutPanel1.Controls.Add(this.translationComboBox, 3, 0);\n this.tableLayoutPanel1.Location = new System.Drawing.Point(12, 12);\n this.tableLayoutPanel1.Name = \"tableLayoutPanel1\";\n this.tableLayoutPanel1.RowCount = 3;\n this.tableLayoutPanel1.RowStyles.Add(new System.Windows.Forms.RowStyle());\n this.tableLayoutPanel1.RowStyles.Add(new System.Windows.Forms.RowStyle());\n this.tableLayoutPanel1.RowStyles.Add(new System.Windows.Forms.RowStyle(System.Windows.Forms.SizeType.Percent, 100F));\n this.tableLayoutPanel1.RowStyles.Add(new System.Windows.Forms.RowStyle(System.Windows.Forms.SizeType.Absolute, 20F));\n this.tableLayoutPanel1.Size = new System.Drawing.Size(1631, 639);\n this.tableLayoutPanel1.TabIndex = 1;\n // \n // vertexShaderComboBox\n // \n this.vertexShaderComboBox.Anchor = ((System.Windows.Forms.AnchorStyles)(((System.Windows.Forms.AnchorStyles.Top | System.Windows.Forms.AnchorStyles.Left) \n | System.Windows.Forms.AnchorStyles.Right)));\n this.vertexShaderComboBox.DropDownStyle = System.Windows.Forms.ComboBoxStyle.DropDownList;\n this.vertexShaderComboBox.FormattingEnabled = true;\n this.vertexShaderComboBox.Items.AddRange(new object[] {\n \"VS to VS\",\n \"VS to line DS with control point indices\",\n \"VS to line DS with patch index\",\n \"VS to triangle DS with control point indices\",\n \"VS to triangle DS with patch index\",\n \"VS to quad DS with control point indices\",\n \"VS to quad DS with patch index\"});\n this.vertexShaderComboBox.Location = new System.Drawing.Point(1224, 24);\n this.vertexShaderComboBox.Margin = new System.Windows.Forms.Padding(3, 3, 3, 0);\n this.vertexShaderComboBox.Name = \"vertexShaderComboBox\";\n this.vertexShaderComboBox.Size = new System.Drawing.Size(404, 21);\n this.vertexShaderComboBox.TabIndex = 7;\n // \n // compilerTranslatedTextBox\n // \n this.compilerTranslatedTextBox.AcceptsReturn = true;\n this.compilerTranslatedTextBox.AcceptsTab = true;\n this.compilerTranslatedTextBox.Dock = System.Windows.Forms.DockStyle.Fill;\n this.compilerTranslatedTextBox.Font = new System.Drawing.Font(\"Consolas\", 8.25F, System.Drawing.FontStyle.Regular, System.Drawing.GraphicsUnit.Point, ((byte)(0)));\n this.compilerTranslatedTextBox.Location = new System.Drawing.Point(1224, 48);\n this.compilerTranslatedTextBox.Multiline = true;\n this.compilerTranslatedTextBox.Name = \"compilerTranslatedTextBox\";\n this.compilerTranslatedTextBox.ReadOnly = true;\n this.compilerTranslatedTextBox.ScrollBars = System.Windows.Forms.ScrollBars.Vertical;\n this.compilerTranslatedTextBox.Size = new System.Drawing.Size(404, 588);\n this.compilerTranslatedTextBox.TabIndex = 10;\n // \n // sourceCodeTextBox\n // \n this.sourceCodeTextBox.AcceptsReturn = true;\n this.sourceCodeTextBox.AcceptsTab = true;\n this.sourceCodeTextBox.Dock = System.Windows.Forms.DockStyle.Fill;\n this.sourceCodeTextBox.Font = new System.Drawing.Font(\"Consolas\", 8.25F, System.Drawing.FontStyle.Regular, System.Drawing.GraphicsUnit.Point, ((byte)(0)));\n this.sourceCodeTextBox.Location = new System.Drawing.Point(3, 24);\n this.sourceCodeTextBox.Multiline = true;\n this.sourceCodeTextBox.Name = \"sourceCodeTextBox\";\n this.tableLayoutPanel1.SetRowSpan(this.sourceCodeTextBox, 2);\n this.sourceCodeTextBox.ScrollBars = System.Windows.Forms.ScrollBars.Vertical;\n this.sourceCodeTextBox.Size = new System.Drawing.Size(401, 612);\n this.sourceCodeTextBox.TabIndex = 5;\n // \n // outputTextBox\n // \n this.outputTextBox.AcceptsReturn = true;\n this.outputTextBox.AcceptsTab = true;\n this.outputTextBox.Dock = System.Windows.Forms.DockStyle.Fill;\n this.outputTextBox.Font = new System.Drawing.Font(\"Consolas\", 8.25F, System.Drawing.FontStyle.Regular, System.Drawing.GraphicsUnit.Point, ((byte)(0)));\n this.outputTextBox.Location = new System.Drawing.Point(410, 24);\n this.outputTextBox.Multiline = true;\n this.outputTextBox.Name = \"outputTextBox\";\n this.outputTextBox.ReadOnly = true;\n this.tableLayoutPanel1.SetRowSpan(this.outputTextBox, 2);\n this.outputTextBox.ScrollBars = System.Windows.Forms.ScrollBars.Vertical;\n this.outputTextBox.Size = new System.Drawing.Size(401, 612);\n this.outputTextBox.TabIndex = 8;\n // \n // compilerUcodeTextBox\n // \n this.compilerUcodeTextBox.AcceptsReturn = true;\n this.compilerUcodeTextBox.AcceptsTab = true;\n this.compilerUcodeTextBox.Dock = System.Windows.Forms.DockStyle.Fill;\n this.compilerUcodeTextBox.Font = new System.Drawing.Font(\"Consolas\", 8.25F, System.Drawing.FontStyle.Regular, System.Drawing.GraphicsUnit.Point, ((byte)(0)));\n this.compilerUcodeTextBox.Location = new System.Drawing.Point(817, 24);\n this.compilerUcodeTextBox.Multiline = true;\n this.compilerUcodeTextBox.Name = \"compilerUcodeTextBox\";\n this.compilerUcodeTextBox.ReadOnly = true;\n this.tableLayoutPanel1.SetRowSpan(this.compilerUcodeTextBox, 2);\n this.compilerUcodeTextBox.ScrollBars = System.Windows.Forms.ScrollBars.Vertical;\n this.compilerUcodeTextBox.Size = new System.Drawing.Size(401, 612);\n this.compilerUcodeTextBox.TabIndex = 9;\n // \n // label1\n // \n this.label1.AutoSize = true;\n this.label1.Location = new System.Drawing.Point(3, 0);\n this.label1.Name = \"label1\";\n this.label1.Size = new System.Drawing.Size(78, 13);\n this.label1.TabIndex = 2;\n this.label1.Text = \"Input Assembly\";\n // \n // label2\n // \n this.label2.AutoSize = true;\n this.label2.Location = new System.Drawing.Point(410, 0);\n this.label2.Name = \"label2\";\n this.label2.Size = new System.Drawing.Size(107, 13);\n this.label2.TabIndex = 3;\n this.label2.Text = \"XNA Compiler Output\";\n // \n // label3\n // \n this.label3.AutoSize = true;\n this.label3.Location = new System.Drawing.Point(817, 0);\n this.label3.Name = \"label3\";\n this.label3.Size = new System.Drawing.Size(191, 13);\n this.label3.TabIndex = 4;\n this.label3.Text = \"xenia-gpu-shader-compiler Disassembly\";\n // \n // translationComboBox\n // \n this.translationComboBox.Anchor = ((System.Windows.Forms.AnchorStyles)(((System.Windows.Forms.AnchorStyles.Top | System.Windows.Forms.AnchorStyles.Left) \n | System.Windows.Forms.AnchorStyles.Right)));\n this.translationComboBox.DropDownStyle = System.Windows.Forms.ComboBoxStyle.DropDownList;\n this.translationComboBox.FormattingEnabled = true;\n this.translationComboBox.Items.AddRange(new object[] {\n \"DXBC (render target RB)\",\n \"DXBC (rasterizer-ordered view RB)\",\n \"SPIR-V (framebuffer RB)\",\n \"SPIR-V (fragment shader interlock RB)\"});\n this.translationComboBox.Location = new System.Drawing.Point(1224, 0);\n this.translationComboBox.Margin = new System.Windows.Forms.Padding(3, 0, 3, 0);\n this.translationComboBox.Name = \"translationComboBox\";\n this.translationComboBox.Size = new System.Drawing.Size(404, 21);\n this.translationComboBox.TabIndex = 6;\n // \n // Editor\n // \n this.AutoScaleDimensions = new System.Drawing.SizeF(6F, 13F);\n this.AutoScaleMode = System.Windows.Forms.AutoScaleMode.Font;\n this.ClientSize = new System.Drawing.Size(1655, 806);\n this.Controls.Add(this.tableLayoutPanel1);\n this.Controls.Add(this.wordsTextBox);\n this.Name = \"Editor\";\n this.Text = \"Shader Playground\";\n this.tableLayoutPanel1.ResumeLayout(false);\n this.tableLayoutPanel1.PerformLayout();\n this.ResumeLayout(false);\n this.PerformLayout();\n\n }\n\n #endregion\n private System.Windows.Forms.TextBox wordsTextBox;\n private System.Windows.Forms.TableLayoutPanel tableLayoutPanel1;\n private System.Windows.Forms.TextBox sourceCodeTextBox;\n private System.Windows.Forms.TextBox outputTextBox;\n private System.Windows.Forms.TextBox compilerUcodeTextBox;\n private System.Windows.Forms.Label label1;\n private System.Windows.Forms.Label label2;\n private System.Windows.Forms.Label label3;\n private System.Windows.Forms.TextBox compilerTranslatedTextBox;\n private System.Windows.Forms.ComboBox translationComboBox;\n private System.Windows.Forms.ComboBox vertexShaderComboBox;\n }\n}\n\n" }, { "path": "tools/shader-playground/Editor.cs", "content": "using Microsoft.Xna.Framework.Graphics;\nusing System;\nusing System.Collections.Generic;\nusing System.ComponentModel;\nusing System.Diagnostics;\nusing System.Drawing;\nusing System.IO;\nusing System.Runtime.InteropServices;\nusing System.Text;\nusing System.Windows.Forms;\n\nnamespace shader_playground {\n public partial class Editor : Form {\n string compilerPath_ = @\"..\\..\\..\\..\\..\\build\\bin\\Windows\\Debug\\xenia-gpu-shader-compiler.exe\";\n\n FileSystemWatcher compilerWatcher_;\n bool pendingTimer_ = false;\n\n public Editor() {\n InitializeComponent();\n\n var scrollUpdateTimer = new Timer();\n scrollUpdateTimer.Interval = 200;\n scrollUpdateTimer.Tick += (object sender, EventArgs e) => {\n UpdateScrollStates();\n };\n scrollUpdateTimer.Start();\n\n var compilerBinPath = Path.Combine(Directory.GetCurrentDirectory(),\n Path.GetDirectoryName(compilerPath_));\n compilerWatcher_ = new FileSystemWatcher(compilerBinPath, \"*.exe\");\n compilerWatcher_.NotifyFilter = NotifyFilters.LastWrite | NotifyFilters.Size;\n compilerWatcher_.Changed += (object sender, FileSystemEventArgs e) => {\n if (e.Name == Path.GetFileName(compilerPath_)) {\n Invoke((MethodInvoker)delegate {\n if (pendingTimer_) {\n return;\n }\n pendingTimer_ = true;\n var timer = new Timer();\n timer.Interval = 1000;\n timer.Tick += (object sender1, EventArgs e1) => {\n pendingTimer_ = false;\n timer.Dispose();\n Process(sourceCodeTextBox.Text);\n };\n timer.Start();\n });\n }\n };\n compilerWatcher_.EnableRaisingEvents = true;\n\n wordsTextBox.Click += (object sender, EventArgs e) => {\n wordsTextBox.SelectAll();\n wordsTextBox.Copy();\n };\n\n sourceCodeTextBox.Click += (object sender, EventArgs e) => {\n Process(sourceCodeTextBox.Text);\n };\n sourceCodeTextBox.TextChanged += (object sender, EventArgs e) => {\n Process(sourceCodeTextBox.Text);\n };\n\n translationComboBox.SelectedIndex = 0;\n translationComboBox.SelectedIndexChanged += (object sender, EventArgs e) => {\n Process(sourceCodeTextBox.Text);\n };\n vertexShaderComboBox.SelectedIndex = 0;\n vertexShaderComboBox.SelectedIndexChanged += (object sender, EventArgs e) => {\n Process(sourceCodeTextBox.Text);\n };\n\n sourceCodeTextBox.Text = string.Join(\n Environment.NewLine, new string[] {\n\"xps_3_0\",\n\"dcl_texcoord1 r0\",\n\"dcl_color r1.xy\",\n\"exec\",\n\"alloc colors\",\n\"exec\",\n\"tfetch1D r2, r0.y, tf0, FetchValidOnly=false\",\n\"tfetch1D r2, r0.x, tf2\",\n\"tfetch2D r3, r3.wx, tf13\",\n\"tfetch2D r[aL+3], r[aL+5].wx, tf13, FetchValidOnly=false, UnnormalizedTextureCoords=true, MagFilter=linear, MinFilter=linear, MipFilter=point, AnisoFilter=max1to1, UseRegisterGradients=true, UseComputedLOD=false, UseRegisterLOD=true, OffsetX=-1.5, OffsetY=1.0\",\n\"tfetch3D r31.w_01, r0.xyw, tf15\",\n\"tfetchCube r5, r1.xyw, tf31\",\n\" setTexLOD r1.z\",\n\" setGradientH r1.zyx\",\n\"(!p0) setGradientV r1.zyx\",\n\" getGradients r5, r1.xy, tf3\",\n\" mad oC0, r0, r1.yyyy, c0\",\n\" mad oC0._, r0, r1.yyyy, c0\",\n\" mad oC0.x1_, r0, r1.yyyy, c0\",\n\" mad oC0.x10w, r0, r1.yyyy, c0\",\n\" mul r4.xyz, r1.xyzz, c5.xyzz\",\n\" mul r4.xyz, r1.xyzz, c[0 + aL].xyzz\",\n\" mul r4.xyz, r1.xyzz, c[6 + aL].xyzz\",\n\" mul r4.xyz, r1.xyzz, c[0 + a0].xyzz\",\n\" mul r4.xyz, r1.xyzz, c[8 + a0].xyzz\",\n\" + adds r5.w, r0.xz\",\n\" cos r6.w, r0.x\",\n\" adds r5.w, r0.zx\",\n\" mul r4.xyz, r[aL+1].xyzz, c[8 + a0].xyzz\",\n\" adds r5.w, r[aL+0].zx\",\n\" jmp l5\",\n\"ccall b1, l5\",\n\"nop\",\n\" label l5\",\n\"(!p0) exec\",\n\"cexec b5, Yield=true\",\n\"cexec !b6\",\n\" mulsc r3.w, c1.z, r1.w\",\n\"loop i7, L4\",\n\" label L3\",\n\" exec\",\n\" setp_eq r15, c[aL].w\",\n\" (!p0) add r0, r0, c[aL]\",\n\"(p0) endloop i7, L3\",\n\"label L4\",\n\"exece\",\n\" mulsc r3.w, c3.z, r6.x\",\n\" mulsc r3.w, c200.z, r31.x\",\n\" mov oDepth.x, c3.w\",\n\" cnop\",\n });\n }\n\n class NopIncludeHandler : CompilerIncludeHandler {\n public override Stream Open(CompilerIncludeHandlerType includeType,\n string filename) {\n throw new NotImplementedException();\n }\n }\n\n void Process(string shaderSourceCode) {\n if (shaderSourceCode.IndexOf(\"xvs_3_0\") != -1 || shaderSourceCode.IndexOf(\"xps_3_0\") != -1) {\n shaderSourceCode += \"\\ncnop\";\n shaderSourceCode += \"\\ncnop\";\n }\n var preprocessorDefines = new CompilerMacro[2];\n preprocessorDefines[0].Name = \"XBOX\";\n preprocessorDefines[1].Name = \"XBOX360\";\n var includeHandler = new NopIncludeHandler();\n var options = CompilerOptions.None;\n var compiledShader = ShaderCompiler.AssembleFromSource(\n shaderSourceCode, preprocessorDefines, includeHandler, options,\n Microsoft.Xna.Framework.TargetPlatform.Xbox360);\n\n var disassembledSourceCode = compiledShader.ErrorsAndWarnings;\n disassembledSourceCode = disassembledSourceCode.Replace(\"\\n\", Environment.NewLine);\n if (disassembledSourceCode.IndexOf(\"// PDB hint 00000000-00000000-00000000\") == -1) {\n UpdateTextBox(outputTextBox, disassembledSourceCode, false);\n UpdateTextBox(compilerUcodeTextBox, \"\", false);\n UpdateTextBox(wordsTextBox, \"\", false);\n return;\n }\n var prefix = disassembledSourceCode.Substring(\n 0, disassembledSourceCode.IndexOf(\n ':', disassembledSourceCode.IndexOf(':') + 1));\n disassembledSourceCode =\n disassembledSourceCode.Replace(prefix + \": \", \"\");\n disassembledSourceCode = disassembledSourceCode.Replace(\n \"// PDB hint 00000000-00000000-00000000\" + Environment.NewLine, \"\");\n var firstLine = disassembledSourceCode.IndexOf(\"//\");\n var warnings = \"// \" +\n disassembledSourceCode.Substring(0, firstLine)\n .Replace(Environment.NewLine, Environment.NewLine + \"// \");\n disassembledSourceCode =\n warnings + disassembledSourceCode.Substring(firstLine + 3);\n disassembledSourceCode = disassembledSourceCode.Trim();\n UpdateTextBox(outputTextBox, disassembledSourceCode, true);\n\n string shaderType =\n shaderSourceCode.IndexOf(\"vs_\") == -1 ? \"ps\" : \"vs\";\n var ucodeWords = ExtractAndDumpWords(shaderType, compiledShader.GetShaderCode());\n if (ucodeWords != null) {\n TryCompiler(shaderType, ucodeWords);\n } else {\n UpdateTextBox(compilerUcodeTextBox, \"\", false);\n }\n\n if (compilerUcodeTextBox.Text.Length > 0) {\n var sourcePrefix = disassembledSourceCode.Substring(0, disassembledSourceCode.IndexOf(\"/*\"));\n TryRoundTrip(sourcePrefix, compilerUcodeTextBox.Text, compiledShader.GetShaderCode());\n }\n }\n\n void TryCompiler(string shaderType, uint[] ucodeWords) {\n string ucodePath = Path.Combine(Path.GetTempPath(), \"shader_playground_ucode.bin.\" + shaderType);\n string ucodeDisasmPath = Path.Combine(Path.GetTempPath(), \"shader_playground_disasm.ucode.txt\");\n string translatedDisasmPath = Path.Combine(Path.GetTempPath(), \"shader_playground_disasm.translated.txt\");\n if (File.Exists(ucodePath)) {\n File.Delete(ucodePath);\n }\n if (File.Exists(ucodeDisasmPath)) {\n File.Delete(ucodeDisasmPath);\n }\n if (File.Exists(translatedDisasmPath)) {\n File.Delete(translatedDisasmPath);\n }\n\n byte[] ucodeBytes = new byte[ucodeWords.Length * 4];\n Buffer.BlockCopy(ucodeWords, 0, ucodeBytes, 0, ucodeWords.Length * 4);\n File.WriteAllBytes(ucodePath, ucodeBytes);\n\n if (!File.Exists(compilerPath_)) {\n UpdateTextBox(compilerUcodeTextBox, \"Compiler not found: \" + compilerPath_, false);\n return;\n }\n\n var startInfo = new ProcessStartInfo(compilerPath_);\n startInfo.Arguments = string.Join(\" \", new string[]{\n \"--shader_input=\" + ucodePath,\n \"--shader_input_type=\" + shaderType,\n \"--shader_output=\" + ucodeDisasmPath,\n \"--shader_output_type=ucode\",\n });\n startInfo.WindowStyle = ProcessWindowStyle.Hidden;\n startInfo.CreateNoWindow = true;\n try {\n using (var process = System.Diagnostics.Process.Start(startInfo)) {\n process.WaitForExit();\n }\n string disasmText = File.ReadAllText(ucodeDisasmPath);\n UpdateTextBox(compilerUcodeTextBox, disasmText.Replace(\"\\n\", Environment.NewLine), true);\n } catch {\n UpdateTextBox(compilerUcodeTextBox, \"COMPILER FAILURE\", false);\n }\n\n string outputType = \"ucode\";\n switch (translationComboBox.SelectedIndex) {\n case 0:\n case 1:\n outputType = \"dxbctext\";\n break;\n case 2:\n case 3:\n outputType = \"spirvtext\";\n break;\n }\n\n string vertexShaderType = \"vertex\";\n switch (vertexShaderComboBox.SelectedIndex) {\n case 1:\n vertexShaderType = \"linedomaincp\";\n break;\n case 2:\n vertexShaderType = \"linedomainpatch\";\n break;\n case 3:\n vertexShaderType = \"triangledomaincp\";\n break;\n case 4:\n vertexShaderType = \"triangledomainpatch\";\n break;\n case 5:\n vertexShaderType = \"quaddomaincp\";\n break;\n case 6:\n vertexShaderType = \"quaddomainpatch\";\n break;\n }\n\n List startArguments = new List{\n \"--shader_input=\" + ucodePath,\n \"--shader_input_type=\" + shaderType,\n \"--shader_output=\" + translatedDisasmPath,\n \"--shader_output_type=\" + outputType,\n \"--vertex_shader_output_type=\" + vertexShaderType,\n \"--dxbc_source_map=true\",\n };\n if (translationComboBox.SelectedIndex == 1 ||\n translationComboBox.SelectedIndex == 3) {\n startArguments.Add(\"--shader_output_pixel_shader_interlock=true\");\n }\n\n startInfo = new ProcessStartInfo(compilerPath_);\n startInfo.Arguments = string.Join(\" \", startArguments.ToArray());\n startInfo.WindowStyle = ProcessWindowStyle.Hidden;\n startInfo.CreateNoWindow = true;\n try {\n using (var process = System.Diagnostics.Process.Start(startInfo)) {\n process.WaitForExit();\n }\n string disasmText = File.ReadAllText(translatedDisasmPath);\n UpdateTextBox(compilerTranslatedTextBox, disasmText.Replace(\"\\n\", Environment.NewLine), true);\n } catch {\n UpdateTextBox(compilerTranslatedTextBox, \"COMPILER FAILURE\", false);\n }\n }\n\n void TryRoundTrip(string sourcePrefix, string compilerSource, byte[] expectedBytes) {\n var shaderSourceCode = sourcePrefix + compilerSource;\n var preprocessorDefines = new CompilerMacro[2];\n preprocessorDefines[0].Name = \"XBOX\";\n preprocessorDefines[1].Name = \"XBOX360\";\n var includeHandler = new NopIncludeHandler();\n var options = CompilerOptions.None;\n var compiledShader = ShaderCompiler.AssembleFromSource(\n shaderSourceCode, preprocessorDefines, includeHandler, options,\n Microsoft.Xna.Framework.TargetPlatform.Xbox360);\n var compiledBytes = compiledShader.GetShaderCode();\n if (compiledBytes == null ||\n compiledBytes.Length != expectedBytes.Length ||\n !MemCmp(compiledBytes, expectedBytes)) {\n compilerUcodeTextBox.BackColor = System.Drawing.Color.Red;\n } else {\n compilerUcodeTextBox.BackColor = System.Drawing.SystemColors.Control;\n }\n }\n\n void UpdateScrollStates() {\n foreach (var handle in scrollPreserve_.Keys) {\n if (scrollPreserve_[handle]) {\n var scrollInfo = new ScrollInfo();\n scrollInfo.cbSize = Marshal.SizeOf(scrollInfo);\n scrollInfo.fMask = (uint)ScrollInfoMask.SIF_TRACKPOS;\n bool hasScrollInfo = GetScrollInfo(handle, SB_VERT, ref scrollInfo);\n scrollPositions_[handle] = scrollInfo.nTrackPos;\n }\n }\n }\n\n Dictionary scrollPreserve_ = new Dictionary();\n Dictionary scrollPositions_ = new Dictionary();\n void UpdateTextBox(TextBox textBox, string value, bool preserveScroll) {\n scrollPreserve_[textBox.Handle] = preserveScroll;\n\n textBox.Text = value;\n\n int previousScroll;\n if (!scrollPositions_.TryGetValue(textBox.Handle, out previousScroll)) {\n previousScroll = 0;\n }\n var scrollInfo = new ScrollInfo();\n scrollInfo.cbSize = Marshal.SizeOf(scrollInfo);\n scrollInfo.fMask = (uint)ScrollInfoMask.SIF_TRACKPOS;\n scrollInfo.nTrackPos = previousScroll;\n SetScrollInfo(textBox.Handle, SB_VERT, ref scrollInfo, 1);\n\n var ptrWparam = new IntPtr(SB_THUMBPOSITION | previousScroll << 16);\n SendMessage(textBox.Handle, WM_VSCROLL, ptrWparam, IntPtr.Zero);\n }\n\n private const int SB_VERT = 1;\n private const uint SB_THUMBPOSITION = 4;\n private const uint WM_VSCROLL = 0x115;\n public enum ScrollInfoMask : uint {\n SIF_RANGE = 0x1,\n SIF_PAGE = 0x2,\n SIF_POS = 0x4,\n SIF_DISABLENOSCROLL = 0x8,\n SIF_TRACKPOS = 0x10,\n SIF_ALL = (SIF_RANGE | SIF_PAGE | SIF_POS | SIF_TRACKPOS),\n }\n private struct ScrollInfo {\n public int cbSize;\n public uint fMask;\n public int nMin;\n public int nMax;\n public int nPage;\n public int nPos;\n public int nTrackPos;\n }\n [DllImport(\"user32.dll\", SetLastError = true)]\n private static extern bool GetScrollInfo(IntPtr hwnd, int bar, ref ScrollInfo scrollInfo);\n [DllImport(\"user32.dll\", SetLastError = true)]\n private static extern bool SetScrollInfo(IntPtr hwnd, int bar, ref ScrollInfo scrollInfo, int redraw);\n [DllImport(\"user32.dll\", CharSet = CharSet.Auto)]\n static extern IntPtr SendMessage(IntPtr hWnd, UInt32 Msg, IntPtr wParam, IntPtr lParam);\n\n bool MemCmp(byte[] a1, byte[] b1) {\n if (a1 == null || b1 == null) {\n return false;\n }\n int length = a1.Length;\n if (b1.Length != length) {\n return false;\n }\n while (length > 0) {\n length--;\n if (a1[length] != b1[length]) {\n return false;\n }\n }\n return true;\n }\n\n uint[] ExtractAndDumpWords(string shaderType, byte[] shaderCode) {\n if (shaderCode == null || shaderCode.Length == 0) {\n UpdateTextBox(wordsTextBox, \"\", false);\n return null;\n }\n\n // Find shader code.\n int byteOffset = (shaderCode[4] << 24) | (shaderCode[5] << 16) |\n (shaderCode[6] << 8) | (shaderCode[7] << 0);\n int wordOffset = byteOffset / 4;\n\n uint[] shaderDwords = new uint[(shaderCode.Length - wordOffset) / sizeof(uint)];\n Buffer.BlockCopy(shaderCode, wordOffset * 4, shaderDwords, 0, shaderCode.Length - wordOffset * 4);\n\n var sb = new StringBuilder();\n sb.Append(\"const uint32_t shader_dwords[] = {\");\n for (int i = 0; i < shaderDwords.Length; ++i) {\n sb.AppendFormat(\"0x{0:X8}, \", SwapByte(shaderDwords[i]));\n }\n sb.Append(\"};\" + Environment.NewLine);\n sb.Append(\"shader_type = ShaderType::\" + (shaderType == \"vs\" ? \"kVertex\" : \"kPixel\") + \";\" + Environment.NewLine);\n UpdateTextBox(wordsTextBox, sb.ToString(), true);\n wordsTextBox.SelectAll();\n\n return shaderDwords;\n }\n\n uint SwapByte(uint x) {\n return ((x & 0x000000ff) << 24) +\n ((x & 0x0000ff00) << 8) +\n ((x & 0x00ff0000) >> 8) +\n ((x & 0xff000000) >> 24);\n }\n }\n}\n" }, { "path": "tools/shader-playground/Editor.resx", "content": "\n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n text/microsoft-resx\n \n \n 2.0\n \n \n System.Resources.ResXResourceReader, System.Windows.Forms, Version=2.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089\n \n \n System.Resources.ResXResourceWriter, System.Windows.Forms, Version=2.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089\n \n" }, { "path": "tools/shader-playground/Program.cs", "content": "using System;\nusing System.Collections.Generic;\nusing System.Reflection;\nusing System.Runtime.InteropServices;\nusing System.Windows.Forms;\n\n// General Information about an assembly is controlled through the following\n// set of attributes. Change these attribute values to modify the information\n// associated with an assembly.\n[assembly: AssemblyTitle(\"shader-playground\")]\n[assembly: AssemblyDescription(\"\")]\n[assembly: AssemblyConfiguration(\"\")]\n[assembly: AssemblyCompany(\"\")]\n[assembly: AssemblyProduct(\"shader-playground\")]\n[assembly: AssemblyCopyright(\"Copyright © 2015\")]\n[assembly: AssemblyTrademark(\"\")]\n[assembly: AssemblyCulture(\"\")]\n\n// Setting ComVisible to false makes the types in this assembly not visible\n// to COM components. If you need to access a type in this assembly from\n// COM, set the ComVisible attribute to true on that type.\n[assembly: ComVisible(false)]\n\n// The following GUID is for the ID of the typelib if this project is exposed to COM\n[assembly: Guid(\"05d27ad4-ef23-4fa1-ad5a-7cfd587fd093\")]\n\n// Version information for an assembly consists of the following four values:\n//\n// Major Version\n// Minor Version\n// Build Number\n// Revision\n//\n// You can specify all the values or you can default the Build and Revision Numbers\n// by using the '*' as shown below:\n// [assembly: AssemblyVersion(\"1.0.*\")]\n[assembly: AssemblyVersion(\"1.0.0.0\")]\n[assembly: AssemblyFileVersion(\"1.0.0.0\")]\n\nnamespace shader_playground {\n static class Program {\n [STAThread]\n static void Main() {\n Application.EnableVisualStyles();\n Application.SetCompatibleTextRenderingDefault(false);\n Application.Run(new Editor());\n }\n }\n}\n" }, { "path": "tools/shader-playground/README.md", "content": "This requires XNA Game Studio 3.1 to be installed (not just the redist):\nhttps://www.microsoft.com/en-us/download/details.aspx?id=39\n\nIt's not really compatible with modern VS', but you can open the downloaded\n`XNAGS31_setup.exe` with 7zip and run the included `redists.msi` directly.\n\nIf installed correctly you should have this file:\n`C:\\Program Files (x86)\\Microsoft XNA\\XNA Game Studio\\v3.1\\References\\Windows\\x86\\Microsoft.Xna.Framework.dll`\n\nXNA is only compatible with 32-bit x86 .NET, so ensure Visual Studio is set to\ntarget that (not `Any CPU`).\n" }, { "path": "tools/shader-playground/shader-playground.csproj", "content": "\r\n\r\n \r\n \r\n Debug\r\n AnyCPU\r\n {05D27AD4-EF23-4FA1-AD5A-7CFD587FD093}\r\n WinExe\r\n Properties\r\n shader_playground\r\n shader-playground\r\n v2.0\r\n 512\r\n \r\n \r\n \r\n true\r\n bin\\x86\\Debug\\\r\n DEBUG;TRACE\r\n full\r\n x86\r\n prompt\r\n MinimumRecommendedRules.ruleset\r\n true\r\n \r\n \r\n bin\\x86\\Release\\\r\n TRACE\r\n true\r\n pdbonly\r\n x86\r\n prompt\r\n MinimumRecommendedRules.ruleset\r\n true\r\n \r\n \r\n \r\n \r\n \r\n \r\n \r\n \r\n \r\n Form\r\n \r\n \r\n Editor.cs\r\n \r\n \r\n \r\n Editor.cs\r\n \r\n \r\n \r\n \r\n \r\n \r\n \r\n \r\n \r\n \r\n" }, { "path": "tools/shader-playground/shader-playground.sln", "content": "\r\nMicrosoft Visual Studio Solution File, Format Version 12.00\r\n# Visual Studio 14\r\nVisualStudioVersion = 14.0.23107.0\r\nMinimumVisualStudioVersion = 10.0.40219.1\r\nProject(\"{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}\") = \"shader-playground\", \"shader-playground.csproj\", \"{05D27AD4-EF23-4FA1-AD5A-7CFD587FD093}\"\r\nEndProject\r\nGlobal\r\n\tGlobalSection(SolutionConfigurationPlatforms) = preSolution\r\n\t\tDebug|x86 = Debug|x86\r\n\t\tRelease|x86 = Release|x86\r\n\tEndGlobalSection\r\n\tGlobalSection(ProjectConfigurationPlatforms) = postSolution\r\n\t\t{05D27AD4-EF23-4FA1-AD5A-7CFD587FD093}.Debug|x86.ActiveCfg = Debug|x86\r\n\t\t{05D27AD4-EF23-4FA1-AD5A-7CFD587FD093}.Debug|x86.Build.0 = Debug|x86\r\n\t\t{05D27AD4-EF23-4FA1-AD5A-7CFD587FD093}.Release|x86.ActiveCfg = Release|x86\r\n\t\t{05D27AD4-EF23-4FA1-AD5A-7CFD587FD093}.Release|x86.Build.0 = Release|x86\r\n\tEndGlobalSection\r\n\tGlobalSection(SolutionProperties) = preSolution\r\n\t\tHideSolutionNode = FALSE\r\n\tEndGlobalSection\r\nEndGlobal\r\n" }, { "path": "tools/vswhere/LICENSE.txt", "content": "The MIT License (MIT) \nCopyright (C) Microsoft Corporation. All rights reserved.\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE." }, { "path": "tools/xenosci/local_runner.py", "content": "#!/usr/bin/env python3\n# Copyright 2017 Justin Moore. All Rights Reserved.\nimport argparse\nimport glob\nimport itertools\nfrom multiprocessing.dummy import Pool\nimport os\nimport time\nimport subprocess\nimport sys\n\n\ndef parallel_arg_type(value):\n ivalue = int(value)\n if ivalue < 1:\n raise argparse.ArgumentError(\n \"%s is an invalid number of processes (>= 1)\" % ivalue)\n\n return ivalue\n\n\ndef file_type(value):\n if not os.path.isfile(value):\n raise argparse.ArgumentError(\"%s is an invalid filename\" % value)\n\n return value\n\n\ndef run_dumper(exe_path, input_name):\n start_time = time.perf_counter()\n\n p = subprocess.Popen([\n exe_path,\n input_name,\n '--log_file=stdout',\n ], stdout=subprocess.PIPE, stderr=subprocess.PIPE)\n p.wait()\n elapsed_time = time.perf_counter() - start_time\n print(\"%.3f seconds, code %2d (%s)\" %\n (elapsed_time, p.returncode, input_name))\n\n\ndef main(argv):\n parser = argparse.ArgumentParser()\n parser.add_argument(\n 'exepath', help='path to the trace dump executable', type=file_type)\n parser.add_argument('files', nargs='*')\n parser.add_argument('-j', type=parallel_arg_type,\n help='number of parallel processes', default=1)\n args = parser.parse_args(argv[1:])\n exepath = args.exepath\n\n files = []\n for wildcard in args.files:\n files.extend(glob.glob(wildcard))\n\n if len(files) == 0:\n print(\"No input files found!\")\n return 1\n\n # Test the executable first.\n valid = False\n try:\n p = subprocess.Popen([\n exepath,\n '--log_file=stdout',\n ], stdout=subprocess.PIPE, stderr=subprocess.PIPE)\n p.wait()\n if p.returncode == 5:\n # code 5 = invalid trace file / trace file unspecified\n valid = True\n except OSError:\n pass\n\n if not valid:\n print(\"Trace executable invalid!\")\n return 1\n\n print(\"Processing...\")\n pool = Pool(args.j)\n\n start_time = time.perf_counter()\n results = pool.starmap(run_dumper, zip(itertools.repeat(exepath), files))\n pool.close()\n pool.join()\n\n elapsed_time = time.perf_counter() - start_time\n print(\"entire runtime took %.3f seconds\" % elapsed_time)\n return 0\n\n\nif __name__ == '__main__':\n sys.exit(main(sys.argv))\n" }, { "path": "xb.bat", "content": "@ECHO OFF\r\nREM Copyright 2022 Ben Vanik. All Rights Reserved.\r\n\r\nSET \"DIR=%~dp0\"\r\n\r\nREM ============================================================================\r\nREM Environment Validation\r\nREM ============================================================================\r\n\r\nSET \"PYTHON_MINIMUM_VERSION[0]=3\"\r\nSET \"PYTHON_MINIMUM_VERSION[1]=6\"\r\nCALL :check_python\r\nIF %_RESULT% NEQ 0 (\r\n ECHO.\r\n ECHO Python %PYTHON_MINIMUM_VERSION[0]%.%PYTHON_MINIMUM_VERSION[1]%+ must be installed and on PATH:\r\n ECHO https://www.python.org/\r\n GOTO :eof\r\n)\r\n\r\n\r\nREM ============================================================================\r\nREM Trampoline into xenia-build\r\nREM ============================================================================\r\n\r\n\"%PYTHON_EXE%\" \"%DIR%\\xenia-build\" %*\r\nEXIT /b %ERRORLEVEL%\r\n\r\n\r\nREM ============================================================================\r\nREM Utilities\r\nREM ============================================================================\r\n\r\n:check_python\r\nSETLOCAL ENABLEDELAYEDEXPANSION\r\n\r\nSET FOUND_PATH=\"\"\r\n\r\nSET \"CANDIDATE_PATHS[0]=C:\\python310\\python.exe\"\r\nSET \"CANDIDATE_PATHS[1]=C:\\python39\\python.exe\"\r\nSET \"CANDIDATE_PATHS[2]=C:\\python38\\python.exe\"\r\nSET \"CANDIDATE_PATHS[3]=C:\\python37\\python.exe\"\r\nSET \"CANDIDATE_PATHS[4]=C:\\python%PYTHON_MINIMUM_VERSION[0]%%PYTHON_MINIMUM_VERSION[1]%\\python.exe\"\r\nSET OUTPUT_INDEX=5\r\n\r\nFOR /F \"usebackq delims=\" %%L IN (`2^>NUL where python3`) DO (\r\n IF %%~zL NEQ 0 (\r\n SET \"CANDIDATE_PATHS[!OUTPUT_INDEX!]=%%L\"\r\n SET /A OUTPUT_INDEX+=1\r\n )\r\n)\r\nFOR /F \"usebackq delims=\" %%L IN (`2^>NUL where python`) DO (\r\n IF %%~zL NEQ 0 (\r\n SET \"CANDIDATE_PATHS[!OUTPUT_INDEX!]=%%L\"\r\n SET /A OUTPUT_INDEX+=1\r\n )\r\n)\r\n\r\nSET CANDIDATE_INDEX=0\r\n:check_candidate_loop\r\nIF NOT DEFINED CANDIDATE_PATHS[%CANDIDATE_INDEX%] (\r\n GOTO :found_python\r\n)\r\nCALL SET CANDIDATE_PATH=%%CANDIDATE_PATHS[%CANDIDATE_INDEX%]%%\r\nIF NOT EXIST \"%CANDIDATE_PATH%\" (\r\n SET /A CANDIDATE_INDEX+=1\r\n GOTO :check_candidate_loop\r\n)\r\nSET \"FOUND_PATH=%CANDIDATE_PATH%\"\r\n\r\n:found_python\r\nIF \"%FOUND_PATH%\"==\"\" (\r\n ECHO ERROR: no Python executable found on PATH.\r\n ECHO Make sure you can run 'python' or 'python3' in a Command Prompt.\r\n ENDLOCAL & SET _RESULT=1\r\n GOTO :eof\r\n)\r\n\r\nCMD /C \"\"%FOUND_PATH%\" -c \"import sys; sys.exit(1 if not sys.version_info[:2] ^>= (%PYTHON_MINIMUM_VERSION[0]%, %PYTHON_MINIMUM_VERSION[1]%) else 0)\"\r\nIF %ERRORLEVEL% NEQ 0 (\r\n ECHO ERROR: Python version mismatch, not at least %PYTHON_MINIMUM_VERSION[0]%.%PYTHON_MINIMUM_VERSION[1]%.\r\n ECHO Found Python executable was \"%FOUND_PATH%\".\r\n ENDLOCAL & SET _RESULT=1\r\n GOTO :eof\r\n)\r\n\r\nENDLOCAL & (\r\n SET _RESULT=0\r\n SET \"PYTHON_EXE=%FOUND_PATH%\"\r\n)\r\nGOTO :eof\r\n" }, { "path": "xb.ps1", "content": "function Write-FatalError($message) {\n [Console]::ForegroundColor = 'red'\n [Console]::Error.WriteLine($message)\n [Console]::ResetColor()\n Exit 1\n}\n\n$pythonExecutables = 'python', 'python3'\nforeach ($pythonExecutable in $pythonExecutables) {\n if (!$pythonPath) {\n $pythonPath = powershell -NoLogo -NonInteractive \"(Get-Command -ErrorAction SilentlyContinue $pythonexecutable).Definition\" # Hack to not give command suggestion\n } else {\n break\n }\n}\n# Neither found, error and exit\n$pythonMinimumVer = 3,6\nif (!$pythonPath) {\n Write-FatalError \"ERROR: Python $($pythonMinimumVer[0]).$($pythonMinimumVer[1])+ must be installed and on PATH:`nhttps://www.python.org/\"\n}\n\n& $pythonPath -c \"import sys; sys.exit(1 if not sys.version_info[:2] >= ($($pythonMinimumVer[0]), $($pythonMinimumVer[1])) else 0)\"\nif ($LASTEXITCODE -gt 0) {\n Write-FatalError \"ERROR: Python version mismatch, not at least $($pythonMinimumVer[0]).$($pythonMinimumVer[1]).`nFound Python executable was `\"$($pythonPath)`\".\"\n}\n\n& $pythonPath \"$($PSScriptRoot)/xenia-build\" $args\n" }, { "path": "xenia-build", "content": "#!/usr/bin/env python3\n\n# Copyright 2022 Ben Vanik. All Rights Reserved.\n\n\"\"\"Main build script and tooling for xenia.\n\nRun with --help or no arguments for possible commands.\n\"\"\"\nfrom __future__ import print_function\nfrom datetime import datetime\nfrom multiprocessing import Pool\nfrom functools import partial\nimport argparse\nimport glob\nimport json\nimport os\nimport re\nimport shutil\nimport subprocess\nimport sys\n\n__author__ = 'ben.vanik@gmail.com (Ben Vanik)'\n\n\nself_path = os.path.dirname(os.path.abspath(__file__))\n\n\n# Detect if building on Android via Termux.\nhost_linux_platform_is_android = False\nif sys.platform == 'linux':\n try:\n host_linux_platform_is_android = subprocess.Popen(\n ['uname', '-o'], stdout=subprocess.PIPE, stderr=subprocess.DEVNULL,\n universal_newlines=True).communicate()[0] == 'Android\\n'\n except Exception:\n pass\n\n\ndef import_subprocess_environment(args):\n popen = subprocess.Popen(\n args, shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, universal_newlines=True)\n variables, _ = popen.communicate()\n envvars_to_save = (\n 'devenvdir',\n 'include',\n 'lib',\n 'libpath',\n 'path',\n 'pathext',\n 'systemroot',\n 'temp',\n 'tmp',\n 'windowssdkdir',\n )\n for line in variables.splitlines():\n for envvar in envvars_to_save:\n if re.match(envvar + '=', line.lower()):\n var, setting = line.split('=', 1)\n if envvar == 'path':\n setting = os.path.dirname(sys.executable) + os.pathsep + setting\n os.environ[var.upper()] = setting\n break\n\n\ndef import_vs_environment():\n \"\"\"Finds the installed Visual Studio version and imports\n interesting environment variables into os.environ.\n\n Returns:\n A version such as 2015 or None if no installation is found.\n \"\"\"\n\n if sys.platform != 'win32':\n return None\n\n version = 0\n install_path = None\n env_tool_args = None\n\n vswhere = subprocess.check_output(\n 'tools/vswhere/vswhere.exe -version \"[15,)\" -latest -prerelease -format json -utf8 -products '\n \"Microsoft.VisualStudio.Product.Enterprise \"\n \"Microsoft.VisualStudio.Product.Professional \"\n \"Microsoft.VisualStudio.Product.Community \"\n \"Microsoft.VisualStudio.Product.BuildTools\",\n shell=False,\n universal_newlines=True,\n encoding=\"utf-8\",\n )\n if vswhere:\n vswhere = json.loads(vswhere)\n if vswhere and len(vswhere) > 0:\n version = int(vswhere[0].get(\"catalog\", {}).get(\"productLineVersion\", 2017))\n install_path = vswhere[0].get(\"installationPath\", None)\n\n if version < 2017:\n if 'VS140COMNTOOLS' in os.environ:\n version = 2015\n vcvars_path = os.environ['VS140COMNTOOLS']\n vcvars_path = os.path.join(vcvars_path, '..\\\\..\\\\vc\\\\vcvarsall.bat')\n env_tool_args = [vcvars_path, 'x64', '&&', 'set']\n else:\n vsdevcmd_path = os.path.join(install_path, 'Common7\\\\Tools\\\\VsDevCmd.bat')\n if os.path.isfile(vsdevcmd_path) and os.access(vsdevcmd_path, os.X_OK):\n env_tool_args = [vsdevcmd_path, '-arch=amd64', '-host_arch=amd64', '&&', 'set']\n else:\n vcvars_path = os.path.join(install_path, 'VC\\\\Auxiliary\\\\Build\\\\vcvarsall.bat')\n env_tool_args = [vcvars_path, 'x64', '&&', 'set']\n\n if version == 0:\n return None\n\n import_subprocess_environment(env_tool_args)\n os.environ['VSVERSION'] = str(version)\n return version\n\n\nvs_version = import_vs_environment()\n\n\ndef main():\n # Add self to the root search path.\n sys.path.insert(0, self_path)\n\n # Augment path to include our fancy things.\n os.environ['PATH'] += os.pathsep + os.pathsep.join([\n self_path,\n os.path.abspath(os.path.join('tools', 'build')),\n ])\n\n # Check git exists.\n if not has_bin('git'):\n print('WARNING: Git should be installed and on PATH. Version info will be omitted from all binaries!')\n print('')\n elif not git_is_repository():\n print('WARNING: The source tree is unversioned. Version info will be omitted from all binaries!')\n print('')\n\n # Check python version.\n python_minimum_ver=3,6\n if not sys.version_info[:2] >= (python_minimum_ver[0], python_minimum_ver[1]):\n print('ERROR: Python ', python_minimum_ver[0], '.', python_minimum_ver[1], '+ must be installed and on PATH', sep='')\n sys.exit(1)\n\n # Grab Visual Studio version and execute shell to set up environment.\n if sys.platform == 'win32' and vs_version is None:\n print('WARNING: Visual Studio not found!')\n print('Building for Windows will not be supported.')\n print('Please refer to the building guide:')\n print('https://github.com/xenia-project/xenia/blob/master/docs/building.md')\n\n # Setup main argument parser and common arguments.\n parser = argparse.ArgumentParser(prog='xenia-build')\n\n # Grab all commands and populate the argument parser for each.\n subparsers = parser.add_subparsers(title='subcommands',\n dest='subcommand')\n commands = discover_commands(subparsers)\n\n # If the user passed no args, die nicely.\n if len(sys.argv) == 1:\n parser.print_help()\n sys.exit(1)\n\n # Gather any arguments that we want to pass to child processes.\n command_args = sys.argv[1:]\n pass_args = []\n try:\n pass_index = command_args.index('--')\n pass_args = command_args[pass_index + 1:]\n command_args = command_args[:pass_index]\n except Exception:\n pass\n\n # Parse command name and dispatch.\n args = vars(parser.parse_args(command_args))\n command_name = args['subcommand']\n try:\n command = commands[command_name]\n return_code = command.execute(args, pass_args, os.getcwd())\n except Exception:\n raise\n sys.exit(return_code)\n\n\ndef print_box(msg):\n \"\"\"Prints an important message inside a box\n \"\"\"\n print(\n '┌{0:─^{2}}╖\\n'\n '│{1: ^{2}}║\\n'\n '╘{0:═^{2}}╝\\n'\n .format('', msg, len(msg) + 2))\n\n\ndef has_bin(binary):\n \"\"\"Checks whether the given binary is present.\n\n Args:\n binary: binary name (without .exe, etc).\n\n Returns:\n True if the binary exists.\n \"\"\"\n bin_path = get_bin(binary)\n if not bin_path:\n return False\n return True\n\n\ndef get_bin(binary):\n \"\"\"Checks whether the given binary is present and returns the path.\n\n Args:\n binary: binary name (without .exe, etc).\n\n Returns:\n Full path to the binary or None if not found.\n \"\"\"\n for path in os.environ['PATH'].split(os.pathsep):\n path = path.strip('\"')\n exe_file = os.path.join(path, binary)\n if os.path.isfile(exe_file) and os.access(exe_file, os.X_OK):\n return exe_file\n exe_file += '.exe'\n if os.path.isfile(exe_file) and os.access(exe_file, os.X_OK):\n return exe_file\n return None\n\n\ndef shell_call(command, throw_on_error=True, stdout_path=None, stderr_path=None, shell=False):\n \"\"\"Executes a shell command.\n\n Args:\n command: Command to execute, as a list of parameters.\n throw_on_error: Whether to throw an error or return the status code.\n stdout_path: File path to write stdout output to.\n stderr_path: File path to write stderr output to.\n\n Returns:\n If throw_on_error is False the status code of the call will be returned.\n \"\"\"\n stdout_file = None\n if stdout_path:\n stdout_file = open(stdout_path, 'w')\n stderr_file = None\n if stderr_path:\n stderr_file = open(stderr_path, 'w')\n result = 0\n try:\n if throw_on_error:\n result = 1\n subprocess.check_call(command, shell=shell, stdout=stdout_file, stderr=stderr_file)\n result = 0\n else:\n result = subprocess.call(command, shell=shell, stdout=stdout_file, stderr=stderr_file)\n finally:\n if stdout_file:\n stdout_file.close()\n if stderr_file:\n stderr_file.close()\n return result\n\n\ndef generate_version_h():\n \"\"\"Generates a build/version.h file that contains current git info.\n \"\"\"\n header_file = 'build/version.h'\n pr_number = 0\n pr_repo_name = \"\"\n pr_branch_name = \"\"\n pr_commit = \"\"\n pr_commit_short = \"\"\n if os.getenv('APPVEYOR') == 'True':\n branch_name = os.getenv('APPVEYOR_REPO_BRANCH')\n commit = os.getenv('APPVEYOR_REPO_COMMIT')\n commit_short = commit[:9]\n pr_number = os.getenv('APPVEYOR_PULL_REQUEST_NUMBER')\n if not pr_number:\n pr_number = 0\n else:\n pr_repo_name = os.getenv('APPVEYOR_PULL_REQUEST_HEAD_REPO_NAME')\n pr_branch_name = os.getenv('APPVEYOR_PULL_REQUEST_HEAD_REPO_BRANCH')\n pr_commit = os.getenv('APPVEYOR_PULL_REQUEST_HEAD_COMMIT')\n pr_commit_short = pr_commit[:9]\n elif git_is_repository():\n (branch_name, commit, commit_short) = git_get_head_info()\n else:\n branch_name = 'tarball'\n commit = ':(-dont-do-this'\n commit_short = ':('\n\n # header\n contents_new = '''// Autogenerated by `xb premake`.\n#ifndef GENERATED_VERSION_H_\n#define GENERATED_VERSION_H_\n#define XE_BUILD_BRANCH \"{}\"\n#define XE_BUILD_COMMIT \"{}\"\n#define XE_BUILD_COMMIT_SHORT \"{}\"\n#define XE_BUILD_DATE __DATE__\n'''.format(branch_name, commit, commit_short)\n\n # PR info (if available)\n if pr_number != 0:\n contents_new += '''#define XE_BUILD_IS_PR\n#define XE_BUILD_PR_NUMBER \"{}\"\n#define XE_BUILD_PR_REPO \"{}\"\n#define XE_BUILD_PR_BRANCH \"{}\"\n#define XE_BUILD_PR_COMMIT \"{}\"\n#define XE_BUILD_PR_COMMIT_SHORT \"{}\"\n'''.format(pr_number, pr_repo_name, pr_branch_name, pr_commit, pr_commit_short)\n\n # footer\n contents_new += '''#endif // GENERATED_VERSION_H_\n'''\n\n contents_old = None\n if os.path.exists(header_file) and os.path.getsize(header_file) < 1024:\n with open(header_file, 'r') as f:\n contents_old = f.read()\n\n if contents_old != contents_new:\n with open(header_file, 'w') as f:\n f.write(contents_new)\n\n\ndef generate_source_class(path):\n header_path = '{}.h'.format(path)\n source_path = '{}.cc'.format(path)\n\n if os.path.isfile(header_path) or os.path.isfile(source_path):\n print('ERROR: Target file already exists')\n return 1\n\n if generate_source_file(header_path) > 0:\n return 1\n if generate_source_file(source_path) > 0:\n # remove header if source file generation failed\n os.remove(os.path.join(source_root, header_path))\n return 1\n\n return 0\n\ndef generate_source_file(path):\n \"\"\"Generates a source file at the specified path containing copyright notice\n \"\"\"\n copyright = '''/**\n ******************************************************************************\n * Xenia : Xbox 360 Emulator Research Project *\n ******************************************************************************\n * Copyright {} Ben Vanik. All rights reserved. *\n * Released under the BSD license - see LICENSE in the root for more details. *\n ******************************************************************************\n */'''.format(datetime.now().year)\n\n if os.path.isfile(path):\n print('ERROR: Target file already exists')\n return 1\n try:\n with open(path, 'w') as f:\n f.write(copyright)\n except Exception as e:\n print('ERROR: Could not write to file [path {}]'.format(path))\n return 1\n\n return 0\n\n\n\ndef git_get_head_info():\n \"\"\"Queries the current branch and commit checksum from git.\n\n Returns:\n (branch_name, commit, commit_short)\n If the user is not on any branch the name will be 'detached'.\n \"\"\"\n p = subprocess.Popen([\n 'git',\n 'symbolic-ref',\n '--short',\n '-q',\n 'HEAD',\n ], stdout=subprocess.PIPE, stderr=subprocess.PIPE)\n (stdout, stderr) = p.communicate()\n branch_name = stdout.decode('ascii').strip() or 'detached'\n p = subprocess.Popen([\n 'git',\n 'rev-parse',\n 'HEAD',\n ], stdout=subprocess.PIPE, stderr=subprocess.PIPE)\n (stdout, stderr) = p.communicate()\n commit = stdout.decode('ascii').strip() or 'unknown'\n p = subprocess.Popen([\n 'git',\n 'rev-parse',\n '--short',\n 'HEAD',\n ], stdout=subprocess.PIPE, stderr=subprocess.PIPE)\n (stdout, stderr) = p.communicate()\n commit_short = stdout.decode('ascii').strip() or 'unknown'\n return branch_name, commit, commit_short\n\n\ndef git_is_repository():\n \"\"\"Checks if git is available and this source tree is versioned.\n \"\"\"\n if not has_bin('git'):\n return False\n return shell_call([\n 'git',\n 'rev-parse',\n '--is-inside-work-tree',\n ], throw_on_error=False, stdout_path=os.devnull, stderr_path=os.devnull) == 0\n\n\ndef git_submodule_update():\n \"\"\"Runs a git submodule init and update.\n \"\"\"\n shell_call([\n 'git',\n '-c',\n 'fetch.recurseSubmodules=on-demand',\n 'submodule',\n 'update',\n '--init',\n ])\n\n\ndef get_clang_format_binary():\n \"\"\"Finds a clang-format binary. Aborts if none is found.\n\n Returns:\n A path to the clang-format executable.\n \"\"\"\n attempts = [\n 'C:\\\\Program Files\\\\LLVM\\\\bin\\\\clang-format.exe',\n 'C:\\\\Program Files (x86)\\\\LLVM\\\\bin\\\\clang-format.exe',\n 'clang-format-14',\n 'clang-format-13',\n 'clang-format',\n ]\n for binary in attempts:\n if has_bin(binary):\n return binary\n print('ERROR: clang-format is not on PATH')\n print('LLVM is available from https://llvm.org/releases/download.html')\n print('At least version 13 is required.')\n print('See docs/style_guide.md for instructions on how to get it.')\n sys.exit(1)\n\n\ndef get_premake_target_os(target_os_override=None):\n \"\"\"Gets the target --os to pass to premake, either for the current platform\n or for the user-specified cross-compilation target.\n\n Args:\n target_os_override: override specified by the user for cross-compilation,\n or None to target the host platform.\n\n Returns:\n Target --os to pass to premake. If a return value of this function valid\n for the current configuration is passed to it again, the same value will\n be returned.\n \"\"\"\n if sys.platform == 'darwin':\n target_os = 'macosx'\n elif sys.platform == 'win32':\n target_os = 'windows'\n elif host_linux_platform_is_android:\n target_os = 'android'\n else:\n target_os = 'linux'\n if target_os_override is not None and target_os_override != target_os:\n if target_os_override == 'android':\n target_os = target_os_override\n else:\n print(\n 'ERROR: cross-compilation is only supported for Android target')\n sys.exit(0)\n return target_os\n\n\ndef run_premake(target_os, action, cc=None):\n \"\"\"Runs premake on the main project with the given format.\n\n Args:\n target_os: target --os to pass to premake.\n action: action to preform.\n \"\"\"\n args = [\n sys.executable,\n os.path.join('tools', 'build', 'premake'),\n '--file=premake5.lua',\n '--os=%s' % target_os,\n '--test-suite-mode=combined',\n '--verbose',\n action,\n ]\n if cc:\n args.insert(4, '--cc=%s' % cc)\n\n ret = subprocess.call(args, shell=False)\n\n if ret == 0:\n generate_version_h()\n\n return ret\n\n\ndef run_platform_premake(target_os_override=None, cc='clang', devenv=None):\n \"\"\"Runs all gyp configurations.\n \"\"\"\n target_os = get_premake_target_os(target_os_override)\n if devenv is None:\n if target_os == 'macosx':\n devenv = 'xcode4'\n elif target_os == 'windows':\n vs_version = '2015'\n if 'VSVERSION' in os.environ:\n vs_version = os.environ['VSVERSION']\n devenv = 'vs' + vs_version\n elif target_os == 'android':\n devenv = 'androidndk'\n else:\n devenv = 'gmake2'\n if target_os != 'linux':\n cc = None\n return run_premake(target_os=target_os, action=devenv, cc=cc)\n\n\ndef get_build_bin_path(args):\n \"\"\"Returns the path of the bin/ path with build results based on the\n configuration specified in the parsed arguments.\n\n Args:\n args: Parsed arguments.\n\n Returns:\n A full path for the bin folder.\n \"\"\"\n if sys.platform == 'darwin':\n platform = 'macosx'\n elif sys.platform == 'win32':\n platform = 'windows'\n else:\n platform = 'linux'\n return os.path.join(self_path, 'build', 'bin', platform.capitalize(), args['config'].capitalize())\n\n\ndef create_clion_workspace():\n \"\"\"Creates some basic workspace information inside the .idea directory for first start.\n \"\"\"\n if os.path.exists('.idea'):\n # No first start\n return False\n print('Generating CLion workspace files...')\n # Might become easier in the future: https://youtrack.jetbrains.com/issue/CPP-7911\n\n # Set the location of the CMakeLists.txt\n os.mkdir('.idea')\n with open(os.path.join('.idea', 'misc.xml'), 'w') as f:\n f.write(\"\"\"\n\n \n \n \n\n\"\"\")\n\n # Set available configurations\n # TODO Find a way to trigger a cmake reload\n with open(os.path.join('.idea', 'workspace.xml'), 'w') as f:\n f.write(\"\"\"\n\n \n \n \n \n \n \n \n\"\"\")\n\n return True\n\n\ndef discover_commands(subparsers):\n \"\"\"Looks for all commands and returns a dictionary of them.\n In the future commands could be discovered on disk.\n\n Args:\n subparsers: Argument subparsers parent used to add command parsers.\n\n Returns:\n A dictionary containing name-to-Command mappings.\n \"\"\"\n commands = {\n 'setup': SetupCommand(subparsers),\n 'pull': PullCommand(subparsers),\n 'premake': PremakeCommand(subparsers),\n 'build': BuildCommand(subparsers),\n 'buildshaders': BuildShadersCommand(subparsers),\n 'devenv': DevenvCommand(subparsers),\n 'gentests': GenTestsCommand(subparsers),\n 'test': TestCommand(subparsers),\n 'gputest': GpuTestCommand(subparsers),\n 'clean': CleanCommand(subparsers),\n 'nuke': NukeCommand(subparsers),\n 'lint': LintCommand(subparsers),\n 'format': FormatCommand(subparsers),\n 'style': StyleCommand(subparsers),\n 'tidy': TidyCommand(subparsers),\n 'stub': StubCommand(subparsers),\n }\n return commands\n\n\nclass Command(object):\n \"\"\"Base type for commands.\n \"\"\"\n\n def __init__(self, subparsers, name, help_short=None, help_long=None,\n *args, **kwargs):\n \"\"\"Initializes a command.\n\n Args:\n subparsers: Argument subparsers parent used to add command parsers.\n name: The name of the command exposed to the management script.\n help_short: Help text printed alongside the command when queried.\n help_long: Extended help text when viewing command help.\n \"\"\"\n self.name = name\n self.help_short = help_short\n self.help_long = help_long\n\n self.parser = subparsers.add_parser(name,\n help=help_short,\n description=help_long)\n self.parser.set_defaults(command_handler=self)\n\n def execute(self, args, pass_args, cwd):\n \"\"\"Executes the command.\n\n Args:\n args: Arguments hash for the command.\n pass_args: Arguments list to pass to child commands.\n cwd: Current working directory.\n\n Returns:\n Return code of the command.\n \"\"\"\n return 1\n\n\nclass SetupCommand(Command):\n \"\"\"'setup' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(SetupCommand, self).__init__(\n subparsers,\n name='setup',\n help_short='Setup the build environment.',\n *args, **kwargs)\n self.parser.add_argument(\n '--target_os', default=None,\n help='Target OS passed to premake, for cross-compilation')\n\n def execute(self, args, pass_args, cwd):\n print('Setting up the build environment...')\n print('')\n\n # Setup submodules.\n print('- git submodule init / update...')\n if git_is_repository():\n git_submodule_update()\n else:\n print('WARNING: Git not available or not a repository. Dependencies may be missing.')\n print('')\n\n print('- running premake...')\n ret = run_platform_premake(target_os_override=args['target_os'])\n print('')\n print('Success!' if ret == 0 else 'Error!')\n\n return ret\n\n\nclass PullCommand(Command):\n \"\"\"'pull' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(PullCommand, self).__init__(\n subparsers,\n name='pull',\n help_short='Pulls the repo and all dependencies and rebases changes.',\n *args, **kwargs)\n self.parser.add_argument(\n '--merge', action='store_true',\n help='Merges on master instead of rebasing.')\n self.parser.add_argument(\n '--target_os', default=None,\n help='Target OS passed to premake, for cross-compilation')\n\n def execute(self, args, pass_args, cwd):\n print('Pulling...')\n print('')\n\n print('- switching to master...')\n shell_call([\n 'git',\n 'checkout',\n 'master',\n ])\n print('')\n\n print('- pulling self...')\n if args['merge']:\n shell_call([\n 'git',\n 'pull',\n ])\n else:\n shell_call([\n 'git',\n 'pull',\n '--rebase',\n ])\n print('')\n\n print('- pulling dependencies...')\n git_submodule_update()\n print('')\n\n print('- running premake...')\n if run_platform_premake(target_os_override=args['target_os']) == 0:\n print('')\n print('Success!')\n\n return 0\n\n\nclass PremakeCommand(Command):\n \"\"\"'premake' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(PremakeCommand, self).__init__(\n subparsers,\n name='premake',\n help_short='Runs premake to update all projects.',\n *args, **kwargs)\n self.parser.add_argument(\n '--cc', default='clang', help='Compiler toolchain passed to premake')\n self.parser.add_argument(\n '--devenv', default=None, help='Development environment')\n self.parser.add_argument(\n '--target_os', default=None,\n help='Target OS passed to premake, for cross-compilation')\n\n def execute(self, args, pass_args, cwd):\n # Update premake. If no binary found, it will be built from source.\n print('Running premake...')\n print('')\n ret = run_platform_premake(target_os_override=args['target_os'],\n cc=args['cc'], devenv=args['devenv'])\n print('Success!' if ret == 0 else 'Error!')\n\n return ret\n\n\nclass BaseBuildCommand(Command):\n \"\"\"Base command for things that require building.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(BaseBuildCommand, self).__init__(\n subparsers,\n *args, **kwargs)\n self.parser.add_argument(\n '--cc', default='clang', help='Compiler toolchain passed to premake')\n self.parser.add_argument(\n '--config', choices=['checked', 'debug', 'release'], default='debug',\n type=str.lower, help='Chooses the build configuration.')\n self.parser.add_argument(\n '--target', action='append', default=[],\n help='Builds only the given target(s).')\n self.parser.add_argument(\n '--force', action='store_true',\n help='Forces a full rebuild.')\n self.parser.add_argument(\n '--no_premake', action='store_true',\n help='Skips running premake before building.')\n self.parser.add_argument(\n '-j', default=4, type=int, help='Number of parallel threads')\n\n def execute(self, args, pass_args, cwd):\n if not args['no_premake']:\n print('- running premake...')\n run_platform_premake(cc=args['cc'])\n print('')\n\n threads = args['j']\n if threads < 0:\n threads = 0\n\n print('- building (%s):%s...' % (\n 'all' if not len(args['target']) else ', '.join(args['target']),\n args['config']))\n if sys.platform == 'win32':\n if vs_version is None:\n print('ERROR: Visual Studio is not installed.');\n result = 1\n else:\n targets = None\n if len(args['target']):\n targets = '/t:' + ';'.join(\n target + (':Rebuild' if args['force'] else '')\n for target in args['target'])\n else:\n targets = '/t:Rebuild' if args['force'] else None\n\n result = subprocess.call([\n 'msbuild',\n 'build/xenia.sln',\n '/nologo',\n '/m',\n '/v:m',\n '/p:Configuration=' + args['config'],\n ] + ([targets] if targets is not None else []) + pass_args,\n shell=False)\n elif sys.platform == 'darwin':\n schemes = args['target'] if len(args['target']) else ['xenia-app']\n nested_args = [['-scheme', scheme] for scheme in schemes]\n scheme_args = [arg for pair in nested_args for arg in pair]\n result = subprocess.call([\n 'xcodebuild',\n '-workspace',\n 'build/xenia.xcworkspace',\n '-configuration',\n args['config']\n ] + scheme_args + pass_args, shell=False, env=dict(os.environ))\n else:\n result = subprocess.call([\n 'make',\n '-j' if threads == 0 else '-j%d' % threads,\n '-Cbuild/',\n 'config=%s_linux' % (args['config']),\n ] + pass_args + args['target'], shell=False, env=dict(os.environ))\n print('')\n if result != 0:\n print('ERROR: build failed with one or more errors.')\n return result\n return 0\n\n\nclass BuildCommand(BaseBuildCommand):\n \"\"\"'build' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(BuildCommand, self).__init__(\n subparsers,\n name='build',\n help_short='Builds the project with the default toolchain.',\n *args, **kwargs)\n\n def execute(self, args, pass_args, cwd):\n print('Building %s...' % (args['config']))\n print('')\n\n result = super(BuildCommand, self).execute(args, pass_args, cwd)\n if not result:\n print('Success!')\n return result\n\n\nclass BuildShadersCommand(Command):\n \"\"\"'buildshaders' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(BuildShadersCommand, self).__init__(\n subparsers,\n name='buildshaders',\n help_short='Generates shader binaries for inclusion in C++ files.',\n help_long='''\n Generates the shader binaries under src/*/shaders/bytecode/.\n Run after modifying any .hs/vs/ds/gs/ps/cs.glsl/hlsl/xesl files.\n Direct3D shaders can be built only on a Windows host.\n ''',\n *args, **kwargs)\n self.parser.add_argument(\n '--target', action='append', choices=['dxbc', 'spirv'], default=[],\n help='Builds only the given target(s).')\n\n def execute(self, args, pass_args, cwd):\n src_paths = [os.path.join(root, name)\n for root, dirs, files in os.walk('src')\n for name in files\n if (name.endswith('.glsl') or\n name.endswith('.hlsl') or\n name.endswith('.xesl'))]\n targets = args['target']\n all_targets = len(targets) == 0\n\n # XeSL (\"Xenia Shading Language\") means shader files that can be\n # compiled as multiple languages from a single file. Whenever possible,\n # this is achieved without the involvement of the build script, using\n # just conditionals, macros and functions in shaders, however, in some\n # cases, that's necessary (such as to prepend `#version` in GLSL, as\n # well as to enable `#include` in GLSL, to include `xesl.xesli` itself,\n # without writing the same `#if` / `#extension` / `#endif` in every\n # shader). Also, not all shading languages provide a built-in\n # preprocessor definition for identification of them, so\n # `SHADING_LANGUAGE_*_XE` is also defined via the build arguments.\n # `SHADING_LANGUAGE_*_XE` is set regardless of whether the file is XeSL\n # or a raw source file in a specific language, as XeSL headers may be\n # used in language-specific sources.\n\n # Direct3D DXBC.\n if all_targets or 'dxbc' in targets:\n if sys.platform == 'win32':\n print('Building Direct3D 12 Shader Model 5.1 DXBC shaders...')\n\n # Get the path to FXC from the latest Windows SDK version.\n\n # Assume native or emulated 32-bit x86 if not one of the\n # architectures listed here.\n windows_sdk_architecture = 'x86'\n processor_architecture = os.environ.get(\n 'PROCESSOR_ARCHITECTURE')\n if processor_architecture == 'AMD64':\n windows_sdk_architecture = 'x64'\n elif processor_architecture == 'ARM64':\n windows_sdk_architecture = 'arm64'\n program_files_path = os.environ.get('ProgramFiles(x86)')\n if program_files_path is None:\n program_files_path = os.environ.get('ProgramFiles')\n if (program_files_path is None or\n not os.path.exists(program_files_path)):\n print('ERROR: could not find the Program Files folder with '\n 'the Windows SDK')\n return 1\n fxc_paths = glob.glob(\n os.path.join(program_files_path, 'Windows Kits/10/bin/*',\n windows_sdk_architecture, 'fxc.exe'))\n if not fxc_paths:\n print('ERROR: could not find the Windows 10/11 SDK with '\n 'the Effect-Compiler Tool (FXC)')\n return 1\n # TODO(Triang3l): Sort by version number, not alphabetically.\n fxc_paths.sort()\n fxc = fxc_paths[-1]\n\n # Build DXBC.\n\n dxbc_stages = ['vs', 'hs', 'ds', 'gs', 'ps', 'cs']\n for src_path in src_paths:\n src_name = os.path.basename(src_path)\n if ((not src_name.endswith('.hlsl') and\n not src_name.endswith('.xesl')) or\n len(src_name) <= 8 or src_name[-8] != '.'):\n continue\n dxbc_identifier = src_name[:-5].replace('.', '_')\n dxbc_stage = dxbc_identifier[-2:]\n if not dxbc_stage in dxbc_stages:\n continue\n print('- %s > d3d12_5_1' % (src_path))\n dxbc_dir_path = os.path.join(os.path.dirname(src_path),\n 'bytecode/d3d12_5_1')\n os.makedirs(dxbc_dir_path, exist_ok=True)\n dxbc_file_path_base = os.path.join(dxbc_dir_path,\n dxbc_identifier)\n # Not enabling treating warnings as errors (/WX) because it\n # overrides #pragma warning, and the FXAA shader triggers a\n # bug in FXC causing an uninitialized variable warning if\n # early exit from a function is done.\n # FXC writes errors and warnings to stderr, not stdout, but\n # stdout receives generic status messages that only add\n # clutter in this case.\n if subprocess.call([\n fxc,\n '/D', 'SHADING_LANGUAGE_HLSL_XE=1',\n '/Fh', dxbc_file_path_base + '.h',\n '/T', dxbc_stage + '_5_1',\n '/Vn', dxbc_identifier,\n '/nologo',\n src_path,\n ], stdout=subprocess.DEVNULL) != 0:\n print('ERROR: failed to compile a DXBC shader')\n return 1\n else:\n if all_targets:\n print('WARNING: Direct3D DXBC shader building is supported '\n 'only on Windows')\n else:\n print('ERROR: Direct3D DXBC shader building is supported '\n 'only on Windows')\n return 1\n\n # Vulkan SPIR-V.\n if all_targets or 'spirv' in targets:\n print('Building Vulkan SPIR-V shaders...')\n\n # Get the SPIR-V tool paths.\n vulkan_sdk_path = os.environ['VULKAN_SDK']\n if not os.path.exists(vulkan_sdk_path):\n print('ERROR: could not find the Vulkan SDK in $VULKAN_SDK')\n return 1\n # bin is lowercase on Linux (even though it's uppercase on Windows).\n vulkan_bin_path = os.path.join(vulkan_sdk_path, 'bin')\n if not os.path.exists(vulkan_bin_path):\n print('ERROR: could not find the Vulkan SDK binaries')\n return 1\n glslang = os.path.join(vulkan_bin_path, 'glslangValidator')\n if not has_bin(glslang):\n print('ERROR: could not find glslangValidator')\n return 1\n spirv_opt = os.path.join(vulkan_bin_path, 'spirv-opt')\n if not has_bin(spirv_opt):\n print('ERROR: could not find spirv-opt')\n return 1\n spirv_dis = os.path.join(vulkan_bin_path, 'spirv-dis')\n if not has_bin(spirv_dis):\n print('ERROR: could not find spirv-dis')\n return 1\n\n # Build SPIR-V.\n spirv_stages = {\n 'vs': 'vert',\n 'hs': 'tesc',\n 'ds': 'tese',\n 'gs': 'geom',\n 'ps': 'frag',\n 'cs': 'comp',\n }\n # #version and extensions must be before everything else in a GLSL\n # file, can't use a language conditional to add them. Use string\n # interpolation to insert the file name. Using #include also\n # preserves line numbers in error and warning messages.\n spirv_xesl_wrapper = \\\n '#version 460\\n' + \\\n '#extension GL_EXT_control_flow_attributes : require\\n' + \\\n '#extension GL_EXT_samplerless_texture_functions : require\\n' + \\\n '#extension GL_GOOGLE_include_directive : require\\n' + \\\n '#include \"%s\"\\n'\n for src_path in src_paths:\n src_name = os.path.basename(src_path)\n src_is_xesl = src_name.endswith('.xesl')\n if ((not src_is_xesl and not src_name.endswith('.glsl')) or\n len(src_name) <= 8 or src_name[-8] != '.'):\n continue\n spirv_identifier = src_name[:-5].replace('.', '_')\n spirv_stage = spirv_stages.get(spirv_identifier[-2:], None)\n if spirv_stage is None:\n continue\n print('- %s > vulkan_spirv' % (src_path))\n src_dir = os.path.dirname(src_path)\n spirv_dir_path = os.path.join(src_dir, 'bytecode/vulkan_spirv')\n os.makedirs(spirv_dir_path, exist_ok=True)\n spirv_file_path_base = os.path.join(spirv_dir_path,\n spirv_identifier)\n spirv_glslang_file_path = spirv_file_path_base + '.glslang.spv'\n # --stdin must be before -S for some reason.\n glslang_arguments = [glslang,\n '--stdin' if src_is_xesl else src_path,\n '-DSHADING_LANGUAGE_GLSL_XE=1',\n '-S', spirv_stage,\n '-o', spirv_glslang_file_path,\n '-V']\n # When compiling the code from stdin, there's no directory\n # containing the file, add the include directory explicitly.\n if src_is_xesl:\n glslang_arguments.append('-I' + src_dir)\n if subprocess.run(\n glslang_arguments,\n input = (spirv_xesl_wrapper % src_name) if src_is_xesl\n else None,\n universal_newlines = True).returncode != 0:\n print('ERROR: failed to build a SPIR-V shader')\n return 1\n # Optimize the shader, and then improve the compression quality\n # for the application executable via ID canonicalization by\n # making IDs in similar shaders more likely to be the same.\n # Don't strip debug information such as binding and buffer\n # structure member names, however, as it's useful for\n # investigating issues in debuggers such as RenderDoc.\n # spirv-opt input and output files must be different.\n spirv_file_path = spirv_file_path_base + '.spv'\n if subprocess.call([\n spirv_opt,\n '-O',\n '--canonicalize-ids',\n spirv_glslang_file_path,\n '-o', spirv_file_path,\n ]) != 0:\n print('ERROR: failed to optimize a SPIR-V shader')\n return 1\n os.remove(spirv_glslang_file_path)\n spirv_dis_file_path = spirv_file_path_base + '.txt';\n if subprocess.call([\n spirv_dis,\n '-o', spirv_dis_file_path,\n spirv_file_path,\n ]) != 0:\n print('ERROR: failed to disassemble a SPIR-V shader')\n return 1\n # Generate the header from the disassembly and the binary.\n with open(spirv_file_path_base + '.h', 'w') as out_file:\n out_file.write(\n '// Generated with `xb buildshaders`.\\n#if 0\\n')\n with open(spirv_dis_file_path, 'r') as spirv_dis_file:\n spirv_dis_data = spirv_dis_file.read()\n if len(spirv_dis_data) > 0:\n out_file.write(spirv_dis_data)\n if spirv_dis_data[-1] != '\\n':\n out_file.write('\\n')\n out_file.write('#endif\\n\\nconst uint32_t %s[] = {' %\n spirv_identifier)\n with open(spirv_file_path, 'rb') as spirv_file:\n index = 0\n # SPIR-V consists of host-endian 32-bit words.\n c = spirv_file.read(4)\n while len(c) != 0:\n if len(c) != 4:\n print('ERROR: a SPIR-V shader is misaligned')\n return 1\n if index % 6 == 0:\n out_file.write('\\n ')\n else:\n out_file.write(' ')\n index += 1\n out_file.write(\n '0x%08X,' % int.from_bytes(c, sys.byteorder))\n c = spirv_file.read(4)\n out_file.write('\\n};\\n')\n os.remove(spirv_dis_file_path)\n os.remove(spirv_file_path)\n return 0\n\n\nclass TestCommand(BaseBuildCommand):\n \"\"\"'test' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(TestCommand, self).__init__(\n subparsers,\n name='test',\n help_short='Runs automated tests that have been built with `xb build`.',\n help_long='''\n To pass arguments to the test executables separate them with `--`.\n For example, you can run only the instr_foo.s tests with:\n $ xb test -- instr_foo\n ''',\n *args, **kwargs)\n self.parser.add_argument(\n '--no_build', action='store_true',\n help='Don\\'t build before running tests.')\n self.parser.add_argument(\n '--continue', action='store_true',\n help='Don\\'t stop when a test errors, but continue running all.')\n\n def execute(self, args, pass_args, cwd):\n print('Testing...')\n print('')\n\n # The test executables that will be built and run.\n test_targets = args['target'] or [\n 'xenia-base-tests',\n 'xenia-cpu-ppc-tests'\n ]\n args['target'] = test_targets\n\n # Build all targets (if desired).\n if not args['no_build']:\n result = super(TestCommand, self).execute(args, [], cwd)\n if result:\n print('Failed to build, aborting test run.')\n return result\n\n # Ensure all targets exist before we run.\n test_executables = [\n get_bin(os.path.join(get_build_bin_path(args), test_target))\n for test_target in test_targets]\n for i in range(0, len(test_targets)):\n if test_executables[i] is None:\n print('ERROR: Unable to find %s - build it.' % (test_targets[i]))\n return 1\n\n # Run tests.\n any_failed = False\n for test_executable in test_executables:\n print('- %s' % test_executable)\n result = shell_call([test_executable] + pass_args,\n throw_on_error=False)\n if result:\n any_failed = True\n if args['continue']:\n print('ERROR: test failed but continuing due to --continue.')\n else:\n print('ERROR: test failed, aborting, use --continue to keep going.')\n return result\n\n if any_failed:\n print('ERROR: one or more tests failed.')\n result = 1\n return result\n\n\nclass GenTestsCommand(Command):\n \"\"\"'gentests' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(GenTestsCommand, self).__init__(\n subparsers,\n name='gentests',\n help_short='Generates test binaries.',\n help_long='''\n Generates test binaries (under src/xenia/cpu/ppc/testing/bin/).\n Run after modifying test .s files.\n ''',\n *args, **kwargs)\n\n def process_src_file(test_bin, ppc_as, ppc_objdump, ppc_ld, ppc_nm, src_file):\n print('- %s' % src_file)\n\n def make_unix_path(p):\n \"\"\"Forces a unix path separator style, as required by binutils.\n \"\"\"\n return p.replace(os.sep, '/')\n\n src_name = os.path.splitext(os.path.basename(src_file))[0]\n obj_file = os.path.join(test_bin, src_name) + '.o'\n shell_call([\n ppc_as,\n '-a32',\n '-be',\n '-mregnames',\n '-mpower7',\n '-maltivec',\n '-mvsx',\n '-mvmx128',\n '-R',\n '-o%s' % (make_unix_path(obj_file)),\n make_unix_path(src_file),\n ])\n dis_file = os.path.join(test_bin, src_name) + '.dis'\n shell_call([\n ppc_objdump,\n '--adjust-vma=0x100000',\n '-Mpower7',\n '-Mvmx128',\n '-D',\n '-EB',\n make_unix_path(obj_file),\n ], stdout_path=dis_file)\n # Eat the first 4 lines to kill the file path that'll differ across machines.\n with open(dis_file) as f:\n dis_file_lines = f.readlines()\n with open(dis_file, 'w') as f:\n f.writelines(dis_file_lines[4:])\n shell_call([\n ppc_ld,\n '-A powerpc:common32',\n '-melf32ppc',\n '-EB',\n '-nostdlib',\n '--oformat=binary',\n '-Ttext=0x80000000',\n '-e0x80000000',\n '-o%s' % (make_unix_path(os.path.join(test_bin, src_name) + '.bin')),\n make_unix_path(obj_file),\n ])\n shell_call([\n ppc_nm,\n '--numeric-sort',\n make_unix_path(obj_file),\n ], stdout_path=os.path.join(test_bin, src_name) + '.map')\n\n def execute(self, args, pass_args, cwd):\n print('Generating test binaries...')\n print('')\n\n binutils_path = os.path.join('third_party', 'binutils-ppc-cygwin')\n ppc_as = os.path.join(binutils_path, 'powerpc-none-elf-as')\n ppc_ld = os.path.join(binutils_path, 'powerpc-none-elf-ld')\n ppc_objdump = os.path.join(binutils_path, 'powerpc-none-elf-objdump')\n ppc_nm = os.path.join(binutils_path, 'powerpc-none-elf-nm')\n\n test_src = os.path.join('src', 'xenia', 'cpu', 'ppc', 'testing')\n test_bin = os.path.join(test_src, 'bin')\n\n # Ensure the test output path exists.\n if not os.path.exists(test_bin):\n os.mkdir(test_bin)\n\n src_files = [os.path.join(root, name)\n for root, dirs, files in os.walk('src')\n for name in files\n if (name.startswith('instr_') or name.startswith('seq_'))\n and name.endswith(('.s'))]\n\n any_errors = False\n\n pool_func = partial(GenTestsCommand.process_src_file, test_bin, ppc_as, ppc_objdump, ppc_ld, ppc_nm)\n with Pool() as pool:\n pool.map(pool_func, src_files)\n\n\n if any_errors:\n print('ERROR: failed to build one or more tests.')\n return 1\n\n return 0\n\n\nclass GpuTestCommand(BaseBuildCommand):\n \"\"\"'gputest' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(GpuTestCommand, self).__init__(\n subparsers,\n name='gputest',\n help_short='Runs automated GPU diff tests against reference imagery.',\n help_long='''\n To pass arguments to the test executables separate them with `--`.\n ''',\n *args, **kwargs)\n self.parser.add_argument(\n '--no_build', action='store_true',\n help='Don\\'t build before running tests.')\n self.parser.add_argument(\n '--update_reference_files', action='store_true',\n help='Update all reference imagery.')\n self.parser.add_argument(\n '--generate_missing_reference_files', action='store_true',\n help='Create reference files for new traces.')\n\n def execute(self, args, pass_args, cwd):\n print('Testinging...')\n print('')\n\n # The test executables that will be built and run.\n test_targets = args['target'] or [\n 'xenia-gpu-vulkan-trace-dump',\n ]\n args['target'] = test_targets\n\n # Build all targets (if desired).\n if not args['no_build']:\n result = super(GpuTestCommand, self).execute(args, [], cwd)\n if result:\n print('Failed to build, aborting test run.')\n return result\n\n # Ensure all targets exist before we run.\n test_executables = [\n get_bin(os.path.join(get_build_bin_path(args), test_target))\n for test_target in test_targets]\n for i in range(0, len(test_targets)):\n if test_executables[i] is None:\n print('ERROR: Unable to find %s - build it.' % (test_targets[i]))\n return 1\n\n output_path = os.path.join(self_path, 'build', 'gputest')\n if os.path.isdir(output_path):\n shutil.rmtree(output_path)\n os.makedirs(output_path)\n print('Running tests and outputting to %s...' % (output_path))\n\n reference_trace_root = os.path.join(self_path, 'testdata',\n 'reference-gpu-traces')\n\n # Run tests.\n any_failed = False\n result = shell_call([\n sys.executable,\n os.path.join(self_path, 'tools', 'gpu-trace-diff'),\n '--executable=' + test_executables[0],\n '--trace_path=' + os.path.join(reference_trace_root, 'traces'),\n '--output_path=' + output_path,\n '--reference_path=' + os.path.join(reference_trace_root, 'references'),\n ] + (['--generate_missing_reference_files'] if args['generate_missing_reference_files'] else []) +\n (['--update_reference_files'] if args['update_reference_files'] else []) +\n pass_args,\n throw_on_error=False)\n if result:\n any_failed = True\n\n if any_failed:\n print('ERROR: one or more tests failed.')\n result = 1\n print('Check %s/results.html for more details.' % (output_path))\n return result\n\n\nclass CleanCommand(Command):\n \"\"\"'clean' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(CleanCommand, self).__init__(\n subparsers,\n name='clean',\n help_short='Removes intermediate files and build outputs.',\n *args, **kwargs)\n self.parser.add_argument(\n '--target_os', default=None,\n help='Target OS passed to premake, for cross-compilation')\n\n def execute(self, args, pass_args, cwd):\n print('Cleaning build artifacts...')\n print('')\n\n print('- premake clean...')\n run_premake(get_premake_target_os(args['target_os']), 'clean')\n print('')\n\n print('Success!')\n return 0\n\n\nclass NukeCommand(Command):\n \"\"\"'nuke' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(NukeCommand, self).__init__(\n subparsers,\n name='nuke',\n help_short='Removes all build/ output.',\n *args, **kwargs)\n self.parser.add_argument(\n '--target_os', default=None,\n help='Target OS passed to premake, for cross-compilation')\n\n def execute(self, args, pass_args, cwd):\n print('Cleaning build artifacts...')\n print('')\n\n print('- removing build/...')\n if os.path.isdir('build/'):\n shutil.rmtree('build/')\n print('')\n\n print('- git reset to master...')\n shell_call([\n 'git',\n 'reset',\n '--hard',\n 'master',\n ])\n print('')\n\n print('- running premake...')\n run_platform_premake(target_os_override=args['target_os'])\n print('')\n\n print('Success!')\n return 0\n\n\ndef find_xenia_source_files():\n \"\"\"Gets all xenia source files in the project.\n\n Returns:\n A list of file paths.\n \"\"\"\n return [os.path.join(root, name)\n for root, dirs, files in os.walk('src')\n for name in files\n if name.endswith(('.cc', '.c', '.h', '.inl', '.inc'))]\n\n\ndef find_all_source_files():\n \"\"\"Gets all interesting source files in the project.\n\n Returns:\n A list of file paths.\n \"\"\"\n return find_xenia_source_files()\n\n\nclass LintCommand(Command):\n \"\"\"'lint' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(LintCommand, self).__init__(\n subparsers,\n name='lint',\n help_short='Checks for lint errors with clang-format.',\n *args, **kwargs)\n self.parser.add_argument(\n '--all', action='store_true',\n help='Lint all files, not just those changed.')\n self.parser.add_argument(\n '--origin', action='store_true',\n help='Lints all files changed relative to origin/master.')\n\n def execute(self, args, pass_args, cwd):\n clang_format_binary = get_clang_format_binary()\n\n difftemp = '.difftemp.txt'\n\n if args['all']:\n all_files = find_all_source_files()\n all_files.sort()\n print('- linting %d files' % (len(all_files)))\n any_errors = False\n for file_path in all_files:\n if os.path.exists(difftemp): os.remove(difftemp)\n ret = shell_call([\n clang_format_binary,\n '-output-replacements-xml',\n '-style=file',\n file_path,\n ], throw_on_error=False, stdout_path=difftemp)\n with open(difftemp) as f:\n had_errors = ' 0:\n return status\n\n print('Created class \\'{0}\\' at {1}'.format(class_name, target_dir))\n\n elif args['file']:\n path = os.path.normpath(os.path.join(source_root, args['file']))\n target_dir = os.path.dirname(path)\n file_name = os.path.basename(path)\n\n status = generate_source_file(path)\n if status > 0:\n return status\n\n print('Created file \\'{0}\\' at {1}'.format(file_name, target_dir))\n\n else:\n print('ERROR: Please specify a file/class to generate')\n return 1\n\n run_platform_premake(target_os_override=args['target_os'])\n return 0\n\nclass DevenvCommand(Command):\n \"\"\"'devenv' command.\"\"\"\n\n def __init__(self, subparsers, *args, **kwargs):\n super(DevenvCommand, self).__init__(\n subparsers,\n name='devenv',\n help_short='Launches the development environment.',\n *args, **kwargs)\n\n def execute(self, args, pass_args, cwd):\n devenv = None\n show_reload_prompt = False\n if sys.platform == 'win32':\n if vs_version is None:\n print('ERROR: Visual Studio is not installed.');\n return 1\n print('Launching Visual Studio...')\n elif sys.platform == 'darwin':\n print('Launching Xcode...')\n devenv = 'xcode4'\n elif has_bin('clion') or has_bin('clion.sh'):\n print('Launching CLion...')\n show_reload_prompt = create_clion_workspace()\n devenv = 'cmake'\n else:\n print('Launching CodeLite...')\n devenv = 'codelite'\n print('')\n\n print('- running premake...')\n run_platform_premake(devenv=devenv)\n print('')\n\n print('- launching devenv...')\n if show_reload_prompt:\n print_box('Please run \"File ⇒ ↺ Reload CMake Project\" from inside the IDE!')\n if sys.platform == 'win32':\n shell_call([\n 'devenv',\n 'build\\\\xenia.sln',\n ])\n elif sys.platform == 'darwin':\n shell_call([\n 'xed',\n 'build/xenia.xcworkspace',\n ])\n elif has_bin('clion'):\n shell_call([\n 'clion',\n '.',\n ])\n elif has_bin('clion.sh'):\n shell_call([\n 'clion.sh',\n '.',\n ])\n else:\n shell_call([\n 'codelite',\n 'build/xenia.workspace',\n ])\n print('')\n\n return 0\n\n\nif __name__ == '__main__':\n main()\n" }, { "path": "xeniarc", "content": "# Copyright 2015 Ben Vanik. All Rights Reserved.\n#\n# Useful bash aliases and completions.\n\nalias xb='python xenia-build'\n" } ]