[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-11476":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":9,"totalLinesOfCode":9,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":9,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":18,"compositeScore":20,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":42,"readmeContent":43,"aiSummary":44,"trendingCount":16,"starSnapshotCount":16,"syncStatus":45,"lastSyncTime":46,"discoverSource":47},11476,"oneDNN","uxlfoundation\u002FoneDNN","uxlfoundation","oneAPI Deep Neural Network Library (oneDNN)",null,"https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002FoneDNN","C++",4005,1146,164,27,0,4,12,15,31.18,false,"main",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41],"onednn","oneapi","deep-learning","deep-neural-networks","performance","cpp","openmp","tbb","x86-64","x64","aarch64","avx512","amx","xe-architecture","library","bfloat16","sycl","vnni","2026-06-12 02:02:32","[![UXL Foundation Logo](https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002Fartwork\u002Fblob\u002Fmain\u002Ffoundation\u002Fuxl-foundation-logo-horizontal-color.png)][UXL Foundation]\n\n# oneAPI Deep Neural Network Library (oneDNN)\n\n[![OpenSSF Best Practices](https:\u002F\u002Fwww.bestpractices.dev\u002Fprojects\u002F8762\u002Fbadge)](https:\u002F\u002Fwww.bestpractices.dev\u002Fprojects\u002F8762)\n[![OpenSSF Scorecard](https:\u002F\u002Fapi.securityscorecards.dev\u002Fprojects\u002Fgithub.com\u002Fuxlfoundation\u002FoneDNN\u002Fbadge)](https:\u002F\u002Fsecurityscorecards.dev\u002Fviewer\u002F?uri=github.com\u002Fuxlfoundation\u002FoneDNN)\n\noneAPI Deep Neural Network Library (oneDNN) is an open-source cross-platform\nperformance library of basic building blocks for deep learning applications.\noneDNN project is part of the [UXL Foundation] and is an implementation\nof the [oneAPI specification] for oneDNN component.\n\nThe library is optimized for Intel 64\u002FAMD64 architecture based processors,\nArm(R) 64-bit Architecture (AArch64)-based processors, and Intel Graphics.\noneDNN has experimental support for the following architectures: NVIDIA\\* GPU,\nAMD\\* GPU, OpenPOWER\\* Power ISA (PPC64), IBMz\\* (s390x), and RISC-V.\n\noneDNN is intended for deep learning applications and framework\ndevelopers interested in improving application performance on CPUs and GPUs.\n\nDeep learning practitioners should use one of the applications enabled with oneDNN:\n\n* [Apache SINGA](https:\u002F\u002Fsinga.apache.org)\n* [DeepLearning4J\\*](https:\u002F\u002Fdeeplearning4j.konduit.ai)\n* [Flashlight\\*](https:\u002F\u002Fgithub.com\u002Fflashlight\u002Fflashlight)\n* [llama.cpp](https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp)\n* [ONNX Runtime](https:\u002F\u002Fonnxruntime.ai)\n* [OpenNMT CTranslate2](https:\u002F\u002Fgithub.com\u002FOpenNMT\u002FCTranslate2)\n* [OpenVINO(TM) toolkit](https:\u002F\u002Fgithub.com\u002Fopenvinotoolkit\u002Fopenvino)\n* [PaddlePaddle\\*](http:\u002F\u002Fwww.paddlepaddle.org)\n* [PyTorch\\*](https:\u002F\u002Fpytorch.org)\n* [Tensorflow\\*](https:\u002F\u002Fwww.tensorflow.org)\n\n[UXL Foundation]: http:\u002F\u002Fwww.uxlfoundation.org\n[oneAPI specification]: https:\u002F\u002Foneapi-spec.uxlfoundation.org\u002Fspecifications\u002Foneapi\u002Flatest\u002Felements\u002Fonednn\u002Fsource\u002F\n\n## Table of Contents\n\n- [Documentation](#documentation)\n- [System Requirements](#system-requirements)\n- [Installation](#installation)\n- [Validated Configurations](#validated-configurations)\n- [Governance](#governance)\n- [Support](#support)\n- [Contributing](#contributing)\n- [License](#license)\n- [Security](#security)\n- [Trademark Information](#trademark-information)\n\n## Documentation\n\n* [oneDNN Developer Guide and Reference] explains the programming\n  model, supported functionality, implementation details, and includes\n  annotated examples.\n* [API Reference] provides a comprehensive reference of the library\n  API.\n* [Release Notes] explain the new features, performance\n  optimizations, and improvements implemented in each version of\n  oneDNN.\n\n[oneDNN Developer Guide and Reference]: https:\u002F\u002Fuxlfoundation.github.io\u002FoneDNN\n[API Reference]: https:\u002F\u002Fuxlfoundation.github.io\u002FoneDNN\u002Fgroup_dnnl_api.html\n[Release Notes]: https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002FoneDNN\u002Freleases\n\n## System Requirements\n\noneDNN supports platforms based on the following architectures:\n- [Intel 64 or AMD64](https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FX86-64),\n- [Arm 64-bit Architecture (AArch64)](https:\u002F\u002Fdeveloper.arm.com\u002Farchitectures\u002Fcpu-architecture\u002Fa-profile).\n- [OpenPOWER](https:\u002F\u002Fopenpowerfoundation.org\u002F) \u002F [IBM Power ISA](https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FPower_ISA).\n- [IBMz z\u002FArchitecture (s390x)](https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FZ\u002FArchitecture).\n- [RISC-V 64-bit (RV64)](https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FRISC-V).\n\n> **WARNING**\n>\n> Power ISA (PPC64), IBMz (s390x), and RISC-V (RV64) support is\n> **experimental** with limited testing validation.\n\nThe library is optimized for the following CPUs:\n* Intel 64\u002FAMD64 architecture\n  * Intel Atom(R) processor (at least Intel SSE4.1 support is required)\n  * Intel Core(TM) processor (at least Intel SSE4.1 support is required)\n  * Intel Xeon(R) processor E3, E5, and E7 family (formerly Sandy Bridge,\n    Ivy Bridge, Haswell, and Broadwell)\n  * Intel Xeon Scalable processor (formerly Skylake, Cascade Lake, Cooper\n    Lake, Ice Lake, Sapphire Rapids, and Emerald Rapids)\n  * Intel Xeon CPU Max Series (formerly Sapphire Rapids HBM)\n  * Intel Core Ultra processors (formerly Meteor Lake, Arrow Lake,\n    Lunar Lake, and Panther Lake)\n  * Intel Xeon 6 processors (formerly Sierra Forest and Granite Rapids)\n  * future Intel Core processor with Intel AVX10.2 instruction set support\n    (code name Nova Lake)\n  * future Intel Xeon processor with Intel AVX10.2 instruction set support\n    (code name Diamond Rapids)\n* AArch64 architecture\n  * Arm Neoverse(TM) N1 and V1 processors\n\nOn a CPU based on Intel 64 or on AMD64 architecture, oneDNN detects\nthe instruction set architecture (ISA) at runtime and uses just-in-time (JIT)\ncode generation to deploy the code optimized for the latest supported ISA.\nFuture ISAs may have initial support in the library disabled by default and\nrequire the use of run-time controls to enable them. See\n[CPU dispatcher control] for more details.\n\n\n> **WARNING**\n>\n> On macOS, applications that use oneDNN may need to request special\n> entitlements if they use the hardened runtime. See the\n> [Linking Guide] for more details.\n\nThe library is optimized for the following GPUs:\n* Intel discrete GPUs:\n  * Intel Iris Xe MAX Graphics (formerly DG1)\n  * Intel Arc(TM) A-Series Graphics (formerly Alchemist)\n  * Intel Data Center GPU Flex Series (formerly Arctic Sound)\n  * Intel Data Center GPU Max Series (formerly Ponte Vecchio)\n  * Intel Arc B-Series Graphics and Intel Arc Pro B-Series Graphics\n   (formerly Battlemage)\n  * future discrete GPUs based on Xe3p-XPC architecture (code name Crescent Island)\n* Intel Graphics integrated with:\n  * 11th-14th Generation Intel Core Processors\n  * Intel Graphics for Intel Core Ultra Series 1 processors (formerly Meteor Lake)\n  * Intel Graphics for Intel Core Ultra Series 2 processors (formerly Arrow Lake and Lunar Lake)\n  * Intel Graphics for Intel Core Ultra Series 3 processors (formerly Panther Lake)\n  * Intel Graphics for Intel Core Series 3 processors (formerly Wildcat Lake)\n  * Intel Graphics for future Intel Core Ultra processors (code name Nova Lake)\n\n[CPU dispatcher control]: https:\u002F\u002Fuxlfoundation.github.io\u002FoneDNN\u002Fdev_guide_cpu_dispatcher_control.html\n[Linking Guide]: https:\u002F\u002Fuxlfoundation.github.io\u002FoneDNN\u002Fdev_guide_link.html\n\n### Requirements for Building from Source\n\noneDNN supports systems meeting the following requirements:\n* Operating system with Intel 64\u002FAMD64, AArch 64, PPC64, or s390x architecture support\n* C++ compiler with C++11 standard support\n* [CMake] 3.13 or later\n\nThe following tools are required to build oneDNN documentation:\n* [Doxygen] 1.8.5 or later\n* [Doxyrest] 2.1.2 or later\n* [Sphinx] 6.2.1 or later\n* [sphinx-book-theme] 1.1.4 or later\n* [sphinx-copybutton] 0.5.2 or later\n* [graphviz] 2.40.1\n\nConfigurations of CPU and GPU engines may introduce additional build time\ndependencies.\n\n[CMake]: https:\u002F\u002Fcmake.org\u002Fdownload\u002F\n[Doxygen]: http:\u002F\u002Fwww.doxygen.nl\u002Fdownload.html#srcbin\n[Doxyrest]: https:\u002F\u002Fgithub.com\u002Fvovkos\u002Fdoxyrest\n[Sphinx]: https:\u002F\u002Fwww.sphinx-doc.org\u002Fen\u002Fmaster\u002Fusage\u002Finstallation.html\n[sphinx-book-theme]: https:\u002F\u002Fsphinx-book-theme.readthedocs.io\u002Fen\u002Flatest\n[sphinx-copybutton]:https:\u002F\u002Fsphinx-copybutton.readthedocs.io\u002Fen\u002Flatest\u002F\n[graphviz]:https:\u002F\u002Fwww.linuxfromscratch.org\u002Fblfs\u002Fview\u002F8.2\u002Fgeneral\u002Fgraphviz.html\n\n#### CPU Engine\n\noneDNN CPU engine is used to execute primitives on Intel 64\u002FAMD64 based processors,\n64-bit Arm Architecture (AArch64) processors, 64-bit Power ISA (PPC64) processors,\nIBMz (s390x), and compatible devices.\n\nThe CPU engine is built by default but can be disabled at build time by setting\n`ONEDNN_CPU_RUNTIME` to `NONE`. In this case, GPU engine must be enabled.\nThe CPU engine can be configured to use the OpenMP, TBB or SYCL runtime.\nThe following additional requirements apply:\n* OpenMP runtime requires C++ compiler with OpenMP 2.0 or later\n  standard support\n* TBB runtime requires [Threading Building Blocks (TBB)] 2017 or later.\n* SYCL runtime requires\n  * [Intel oneAPI DPC++\u002FC++ Compiler]\n  * [Threading Building Blocks (TBB)]\n\nSome implementations rely on OpenMP 4.0 SIMD extensions. For the best\nperformance results on Intel Architecture Processors we recommend using the\nIntel C++ Compiler.\n\n[Threading Building Blocks (TBB)]: https:\u002F\u002Fwww.threadingbuildingblocks.org\u002F\n[Intel oneAPI DPC++\u002FC++ Compiler]: https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdeveloper\u002Ftools\u002Foneapi\u002Fdpc-compiler.html\n\nOn a CPU based on Arm AArch64 architecture, oneDNN CPU engine can be built with\n[Arm Compute Library (ACL)] integration. ACL is an open-source library for\nmachine learning applications and provides AArch64 optimized implementations\nof core functions. This functionality currently requires that ACL is downloaded\nand built separately. See [Build from Source] section of the Developer Guide for\ndetails. The minimum supported version of ACL is 52.7.0.\n\n[Arm Compute Library (ACL)]: https:\u002F\u002Fgithub.com\u002Farm-software\u002FComputeLibrary\n\n#### GPU Engine\n\noneDNN GPU engine is used to execute primitives on various accelerators\nincluding Intel integrated and discrete GPUs, NVIDIA GPUs, AMD GPUs, and\nother devices supporting SYCL programming language. The GPU engine is disabled\nin the default build configuration and can be enabled by setting\n`ONEDNN_GPU_RUNTIME` build option to value other than `NONE`. Target accelerator\nvendor must be selected at build time using `ONEDNN_GPU_VENDOR` build option.\n\n> **WARNING**\n>\n> Linux will reset GPU when kernel runtime exceeds several seconds. The user\n> can prevent this behavior by [disabling hangcheck] for Intel GPU driver.\n> Windows has built-in [timeout detection and recovery] mechanism that results\n> in similar behavior. The user can prevent this behavior by increasing the\n> [TdrDelay] value.\n\nThe following additional requirements apply for Intel integrated and discrete\nGPUs:\n* With OpenCL(TM) runtime:\n  * OpenCL SDK (with OpenCL 1.2 support)\n  * Intel Graphics Driver with support for OpenCL C 2.0, Intel subgroups\n  support, and USM extensions support\n* With SYCL runtime:\n  * [Intel oneAPI DPC++\u002FC++ Compiler]\n  * OpenCL SDK (with OpenCL 3.0 support)\n  * [oneAPI Level Zero]\n  * Intel Graphics Driver with support for OpenCL C 2.0, Intel subgroups\n    support, and USM extensions support\n\nThe following additional requirements apply for NVIDIA GPUs:\n* [oneAPI DPC++ Compiler with support for CUDA] or [oneAPI for NVIDIA GPUs]\n* NVIDIA CUDA\\* driver\n* cuBLAS 10.1 or later\n* cuDNN 7.6 or later\n\n> **WARNING**\n>\n> NVIDIA GPU support is experimental. General information, build instructions,\n> and implementation limitations are available in the\n> [NVIDIA backend readme](src\u002Fgpu\u002Fnvidia\u002FREADME.md).\n\nThe following additional requirements apply for AMD GPUs:\n* [oneAPI DPC++ Compiler with support for HIP AMD] or [oneAPI for AMD GPUs]\n* [AMD ROCm] version 5.3 or later\n* [MIOpen] version 2.18 or later (optional if AMD ROCm includes\n  the required version of MIOpen)\n* [rocBLAS] version 2.45.0 or later (optional if AMD ROCm includes\n  the required version of rocBLAS)\n\n> **WARNING**\n>\n> AMD GPU support is experimental. General information, build instructions,\n> and implementation limitations are available in the\n> [AMD backend readme](src\u002Fgpu\u002Famd\u002FREADME.md).\n\nOther devices supporting SYCL programming model require\noneAPI DPC++\u002FC++ Compiler that supports the target GPU. Refer to\n[generic GPU vendor] documentation for additional details.\n\n[oneAPI Level Zero]: https:\u002F\u002Fgithub.com\u002Foneapi-src\u002Flevel-zero\n[oneAPI DPC++ Compiler with support for CUDA]: https:\u002F\u002Fgithub.com\u002Fintel\u002Fllvm\u002Fblob\u002Fsycl\u002Fsycl\u002Fdoc\u002FGetStartedGuide.md#build-dpc-toolchain-with-support-for-nvidia-cuda\n[oneAPI for NVIDIA GPUs]: https:\u002F\u002Fdeveloper.codeplay.com\u002Fproducts\u002Foneapi\u002Fnvidia\u002Fhome\n[oneAPI DPC++ Compiler with support for HIP AMD]: https:\u002F\u002Fgithub.com\u002Fintel\u002Fllvm\u002Fblob\u002Fsycl\u002Fsycl\u002Fdoc\u002FGetStartedGuide.md#build-dpc-toolchain-with-support-for-hip-amd\n[oneAPI for AMD GPUs]: https:\u002F\u002Fdeveloper.codeplay.com\u002Fproducts\u002Foneapi\u002Famd\u002Fhome\u002F\n[AMD ROCm]: https:\u002F\u002Fgithub.com\u002FRadeonOpenCompute\u002FROCm\n[MIOpen]: https:\u002F\u002Fgithub.com\u002FROCmSoftwarePlatform\u002FMIOpen\n[rocBLAS]: https:\u002F\u002Fgithub.com\u002FROCmSoftwarePlatform\u002FrocBLAS\n[disabling hangcheck]: https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdocs\u002Foneapi\u002Finstallation-guide-linux\u002F2023-0\u002Fgpu-disable-hangcheck.html\n[timeout detection and recovery]: https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fwindows-hardware\u002Fdrivers\u002Fdisplay\u002Ftimeout-detection-and-recovery\n[TdrDelay]: https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fwindows-hardware\u002Fdrivers\u002Fdisplay\u002Ftdr-registry-keys#tdrdelay\n[generic GPU vendor](src\u002Fgpu\u002Fgeneric\u002Fsycl\u002FREADME.md)\n\n### Runtime Dependencies\n\nWhen oneDNN is built from source, the library runtime dependencies and specific\nversions are defined by the build environment.\n\n#### Linux\n\nCommon dependencies:\n* GNU C Library (`libc.so`)\n* GNU Standard C++ Library v3 (`libstdc++.so`)\n* Dynamic Linking Library (`libdl.so`)\n* C Math Library (`libm.so`)\n* POSIX Threads Library (`libpthread.so`)\n\nRuntime-specific dependencies:\n\n| Runtime configuration      | Compiler                      | Dependency\n| :------------------------- | :---------------------------- | :---------\n| `ONEDNN_CPU_RUNTIME=OMP`   | GCC                           | GNU OpenMP runtime (`libgomp.so`)\n| `ONEDNN_CPU_RUNTIME=OMP`   | Intel C\u002FC++ Compiler          | Intel OpenMP runtime (`libiomp5.so`)\n| `ONEDNN_CPU_RUNTIME=OMP`   | Clang                         | Intel OpenMP runtime (`libiomp5.so`)\n| `ONEDNN_CPU_RUNTIME=TBB`   | any                           | TBB (`libtbb.so`)\n| `ONEDNN_CPU_RUNTIME=SYCL`  | Intel oneAPI DPC++ Compiler   | Intel oneAPI DPC++ Compiler runtime (`libsycl.so`), TBB (`libtbb.so`), OpenCL loader (`libOpenCL.so`)\n| `ONEDNN_GPU_RUNTIME=OCL`   | any                           | OpenCL loader (`libOpenCL.so`)\n| `ONEDNN_GPU_RUNTIME=SYCL`  | Intel oneAPI DPC++ Compiler   | Intel oneAPI DPC++ Compiler runtime (`libsycl.so`), OpenCL loader (`libOpenCL.so`), oneAPI Level Zero loader (`libze_loader.so`)\n| `ONEDNN_GPU_RUNTIME=ZE`    | any                           | oneAPI Level Zero loader (`libze_loader.so`)\n\n#### Windows\n\nCommon dependencies:\n* Microsoft Visual C++ Redistributable (`msvcrt.dll`)\n\nRuntime-specific dependencies:\n\n| Runtime configuration      | Compiler                      | Dependency\n| :------------------------- | :---------------------------- | :---------\n| `ONEDNN_CPU_RUNTIME=OMP`   | Microsoft Visual C++ Compiler | No additional requirements\n| `ONEDNN_CPU_RUNTIME=OMP`   | Intel C\u002FC++ Compiler          | Intel OpenMP runtime (`iomp5.dll`)\n| `ONEDNN_CPU_RUNTIME=TBB`   | any                           | TBB (`tbb.dll`)\n| `ONEDNN_CPU_RUNTIME=SYCL`  | Intel oneAPI DPC++ Compiler   | Intel oneAPI DPC++ Compiler runtime (`sycl.dll`), TBB (`tbb.dll`), OpenCL loader (`OpenCL.dll`)\n| `ONEDNN_GPU_RUNTIME=OCL`   | any                           | OpenCL loader (`OpenCL.dll`)\n| `ONEDNN_GPU_RUNTIME=SYCL`  | Intel oneAPI DPC++ Compiler   | Intel oneAPI DPC++ Compiler runtime (`sycl.dll`), OpenCL loader (`OpenCL.dll`), oneAPI Level Zero loader (`ze_loader.dll`)\n| `ONEDNN_GPU_RUNTIME=ZE`    | any                           | oneAPI Level Zero loader (`ze_loader.dll`)\n\n#### macOS\n\nCommon dependencies:\n* System C\u002FC++ runtime (`libc++.dylib`, `libSystem.dylib`)\n\nRuntime-specific dependencies:\n\n| Runtime configuration    | Compiler                      | Dependency\n| :----------------------- | :---------------------------- | :---------\n| `ONEDNN_CPU_RUNTIME=OMP` | Intel C\u002FC++ Compiler          | Intel OpenMP runtime (`libiomp5.dylib`)\n| `ONEDNN_CPU_RUNTIME=TBB` | any                           | TBB (`libtbb.dylib`)\n\n## Installation\n\nYou can download and install the oneDNN library using one of the following options:\n\n- Binary Distribution: You can download pre-built binary packages from\n  the following sources:\n    - [conda-forge]: If the configuration you need is not available on\n      the conda-forge channel, you can build the library using the\n      Source Distribution.\n    - Intel oneAPI:\n       - [Intel® oneAPI Base Toolkit](https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdeveloper\u002Ftools\u002Foneapi\u002Fbase-toolkit-download.htm)\n       - [Intel® oneDNN standalone package](https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdeveloper\u002Ftools\u002Foneapi\u002Fonednn-download.html)\n\n- Source Distribution: You can build the library from source by\n  following the instructions on the [Build from Source] page.\n\n[conda-forge]: https:\u002F\u002Fanaconda.org\u002Fconda-forge\u002Fonednn\n[System Requirements]: #system-requirements\n[Build Options]: https:\u002F\u002Fuxlfoundation.github.io\u002FoneDNN\u002Fdev_guide_build_options.html\n[Build from Source]: https:\u002F\u002Fuxlfoundation.github.io\u002FoneDNN\u002Fdev_guide_build.html\n\n## Validated Configurations\n\nx86-64 CPU engine was validated on RedHat\\* Enterprise Linux 8 with\n* GNU Compiler Collection 8.5, 9.5, 11.1, 11.3\n* Clang\\* 11.0, 14.0.6\n* [Intel oneAPI DPC++\u002FC++ Compiler] 2025.1\n\non Windows Server\\* 2019 with\n* Microsoft Visual Studio 2022 with MSVC 19.43\n* [Intel oneAPI DPC++\u002FC++ Compiler] 2025.1\n\non macOS 14 (Sonoma) with\n* Apple LLVM version 15.0\n\nAArch64 CPU engine was validated on Ubuntu 22.04 with\n* GNU Compiler Collection 10.0, 13.0\n* Clang\\* 17.0\n* [Arm Compiler for Linux] 24.04\n* [Arm Compute Library (ACL)] built for armv8-a arch, latest stable version\navailable at the time of release\n\non macOS 14 (Sonoma) with\n* Apple LLVM version 15.0\n\nGPU engine was validated on Ubuntu\\* 22.04 with\n* GNU Compiler Collection 8.5, and 9.5\n* Clang\\* 11.0\n* [Intel oneAPI DPC++\u002FC++ Compiler] 2025.1\n* [Intel Software for General Purpose GPU capabilities] latest stable version\navailable at the time of release\n\non Windows Server\\* 2019 with\n* Microsoft Visual Studio 2022 with MSVC 19.43\n* [Intel oneAPI DPC++\u002FC++ Compiler] 2025.1\n* [Intel Arc & Iris Xe Graphics Driver] latest stable version available at the\ntime of release\n\n[Intel Software for General Purpose GPU capabilities]: https:\u002F\u002Fdgpu-docs.intel.com\u002Findex.html\n[Intel Arc & Iris Xe Graphics Driver]: https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdownload\u002F785597\u002Fintel-arc-iris-xe-graphics-windows.html\n[Arm Compiler for Linux]: https:\u002F\u002Fdeveloper.arm.com\u002FTools%20and%20Software\u002FArm%20Compiler%20for%20Linux\n\n## Support\n\nSubmit questions, feature requests, and bug reports on the\n[GitHub issues] page.\n\nYou can also contact oneDNN developers via [UXL Foundation Slack] using\n[#onednn] channel.\n\n[Github issues]: https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002FoneDNN\u002Fissues\n[UXL Foundation Slack]: https:\u002F\u002Fslack-invite.uxlfoundation.org\u002F\n[#onednn]: https:\u002F\u002Fuxlfoundation.slack.com\u002Fchannels\u002Fonednn\n\n## Governance\n\noneDNN project is governed by the [UXL Foundation] and you can get involved in\nthis project in multiple ways. It is possible to join the [AI Special Interest\nGroup (SIG)] meetings where the groups discuss and demonstrate work using this\nproject. Members can also join the Open Source and Specification Working Group\nmeetings.\n\nYou can also join the [mailing lists for the UXL Foundation] to be informed\nof when meetings are happening and receive the latest information and\ndiscussions.\n\n[AI Special Interest Group (SIG)]: https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002Ffoundation\n[mailing lists for the UXL Foundation]: https:\u002F\u002Flists.uxlfoundation.org\u002Fg\u002Fmain\u002Fsubgroups\n\n## Contributing\n\nWe welcome community contributions to oneDNN. You can find the oneDNN release\nschedule and work already in progress towards future milestones in Github's\n[Milestones] section. If you are looking for a specific task to start,\nconsider selecting from issues that are marked with the [help wanted] label.\n\n\nSee [contribution guidelines](CONTRIBUTING.md) to start contributing\nto oneDNN. You can also contact oneDNN developers and maintainers via\n[UXL Foundation Slack] using [#onednn] channel.\n\nThis project is intended to be a safe, welcoming space for\ncollaboration, and contributors are expected to adhere to the\n[Contributor Covenant](CODE_OF_CONDUCT.md) code of conduct.\n\n[RFC pull request]: https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002FoneDNN\u002Ftree\u002Frfcs\n[code contribution guidelines]: CONTRIBUTING.md#code-contribution-guidelines\n[coding standards]: CONTRIBUTING.md#coding-standards\n[pull request]: https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002FoneDNN\u002Fpulls\n[Milestones]: https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002FoneDNN\u002Fmilestones\n[help wanted]: https:\u002F\u002Fgithub.com\u002Fuxlfoundation\u002FoneDNN\u002Fissues?q=is%3Aissue+is%3Aopen+label%3A%22help+wanted%22\n\n## License\n\noneDNN is licensed under [Apache License Version 2.0](LICENSE). Refer\nto the \"[LICENSE](LICENSE)\" file for the full license text and\ncopyright notice.\n\nThis distribution includes third party software governed by separate\nlicense terms.\n\n3-clause BSD license:\n* [Xbyak](https:\u002F\u002Fgithub.com\u002Fherumi\u002Fxbyak)\n* [gtest](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fgoogletest)\n* [Instrumentation and Tracing Technology API\n(ITT API)](https:\u002F\u002Fgithub.com\u002Fintel\u002Fittapi)\n* [CMake](https:\u002F\u002Fgithub.com\u002FKitware\u002FCMake)\n\n2-clause BSD license:\n* [Sphinx](https:\u002F\u002Fwww.sphinx-doc.org\u002F)\n\nApache License Version 2.0:\n* [Xbyak_aarch64](https:\u002F\u002Fgithub.com\u002Ffujitsu\u002Fxbyak_aarch64)\n* [OpenCL\u003Csup>TM\u003C\u002Fsup> API Headers](https:\u002F\u002Fgithub.com\u002FKhronosGroup\u002FOpenCL-Headers)\n\nBoost Software License, Version 1.0:\n* [Boost C++ Libraries](https:\u002F\u002Fwww.boost.org\u002F)\n\nMIT License:\n* [Intel Graphics Compute Runtime for oneAPI Level Zero\nand OpenCL Driver](https:\u002F\u002Fgithub.com\u002Fintel\u002Fcompute-runtime)\n* [Intel Graphics Compiler](https:\u002F\u002Fgithub.com\u002Fintel\u002Fintel-graphics-compiler)\n* [oneAPI Level Zero](https:\u002F\u002Fgithub.com\u002Foneapi-src\u002Flevel-zero)\n* [Doxyrest](https:\u002F\u002Fgithub.com\u002Fvovkos\u002Fdoxyrest)\n* [Intel Metrics Discovery Application Programming\nInterface](https:\u002F\u002Fgithub.com\u002Fintel\u002Fmetrics-discovery)\n* [spdlog](https:\u002F\u002Fgithub.com\u002Fgabime\u002Fspdlog)\n* [sphinx-copybutton](https:\u002F\u002Fgithub.com\u002Fexecutablebooks\u002Fsphinx-copybutton)\n\nThis third-party software, even if included with the distribution of\nother software, may be governed by separate license terms,\nincluding without limitation, third party license terms, other\nsoftware license terms, and open source software license terms. These\nseparate license terms govern your use of the third party programs as\nset forth in the \"[THIRD-PARTY-PROGRAMS](THIRD-PARTY-PROGRAMS)\" file.\n\n## Security\n\n[Security Policy](SECURITY.md) outlines our guidelines and procedures\nfor ensuring the highest level of security and trust for our users\nwho consume oneDNN.\n\n-------------------------------------------------------------------------------\n\n[Legal Information](doc\u002Flegal_information.md)\n","oneAPI Deep Neural Network Library (oneDNN) 是一个开源的跨平台性能库，为深度学习应用提供基础构建块。该项目使用C++开发，针对Intel 64\u002FAMD64、Arm 64位架构处理器以及Intel图形处理器进行了优化，并且实验性地支持NVIDIA GPU、AMD GPU等其他架构。它通过利用OpenMP、TBB和SYCL等技术来提高计算效率。oneDNN适合于希望在CPU和GPU上提升深度学习应用性能的开发者，特别是那些正在使用如PyTorch、TensorFlow或ONNX Runtime等流行框架的人士。",2,"2026-06-11 03:31:57","trending"]