From 1331637681273dc8391d7a7caa6175405dceb735 Mon Sep 17 00:00:00 2001 From: V3n3RiX Date: Fri, 1 Mar 2024 23:41:30 +0000 Subject: gentoo auto-resync : 01:03:2024 - 23:41:29 --- sci-libs/Manifest.gz | Bin 44612 -> 44773 bytes sci-libs/caffe2/Manifest | 3 +- sci-libs/caffe2/caffe2-2.1.2-r5.ebuild | 243 ------------------------ sci-libs/caffe2/caffe2-2.1.2-r6.ebuild | 242 +++++++++++++++++++++++ sci-libs/ideep/Manifest | 6 + sci-libs/ideep/files/ideep-3.1.1-20230825.patch | 36 ++++ sci-libs/ideep/ideep-3.1.1.ebuild | 22 +++ sci-libs/ideep/ideep-3.3.2.ebuild | 20 ++ sci-libs/ideep/metadata.xml | 11 ++ 9 files changed, 338 insertions(+), 245 deletions(-) delete mode 100644 sci-libs/caffe2/caffe2-2.1.2-r5.ebuild create mode 100644 sci-libs/caffe2/caffe2-2.1.2-r6.ebuild create mode 100644 sci-libs/ideep/Manifest create mode 100644 sci-libs/ideep/files/ideep-3.1.1-20230825.patch create mode 100644 sci-libs/ideep/ideep-3.1.1.ebuild create mode 100644 sci-libs/ideep/ideep-3.3.2.ebuild create mode 100644 sci-libs/ideep/metadata.xml (limited to 'sci-libs') diff --git a/sci-libs/Manifest.gz b/sci-libs/Manifest.gz index 273eb7cfbaa1..0eb679798d6c 100644 Binary files a/sci-libs/Manifest.gz and b/sci-libs/Manifest.gz differ diff --git a/sci-libs/caffe2/Manifest b/sci-libs/caffe2/Manifest index 6c372a2d88be..72c2b7f09a0b 100644 --- a/sci-libs/caffe2/Manifest +++ b/sci-libs/caffe2/Manifest @@ -11,11 +11,10 @@ AUX caffe2-2.1.1-cudaExtra.patch 1336 BLAKE2B f12d29a624f9345452227076108b9ded4c AUX caffe2-2.1.1-gentoo.patch 6329 BLAKE2B c002f1605888ac79bed292cf1967eedd4d177193947dbafea8443958e498b371865a8f7548cc7e2c572d7fff69418344286071d64f4742691e943e81b73d811c SHA512 8eb2b2f89de809ff93e0deaa4bd6116f73bed55642904e689ba97e5bf21ba5754935a8d0c1f632b1f262400f0c809398524056f7691f7787c154a5b7f6f1b7b2 AUX caffe2-2.1.2-fix-openmp-link.patch 483 BLAKE2B dfe9ae0978b71449486d12f769c3c962001c138bc53b0605ed6f2215fe732356da90c5084488bc4d9f54c88aab2a426b8c935543f7ba0890783bb58da30a5cb4 SHA512 a9cee8ae2a8477a21e92bd4ef286af68efd90283593db7ed4641a826b19b4266bcab9c131c93be2b6c2f13d5cd545c766612791cf2e23cff9a731a03f3ad7648 AUX caffe2-2.1.2-fix-rpath.patch 510 BLAKE2B 6585e8089acc6e7fe69713ad04a64263ab2e9d2d5cf6d3a844dbbc917383e4b7a0688ebdfb3142e3c42108c8c14a6435f48ae4ce3adee8db338286d60ed7b503 SHA512 ede770559b487fff736aded0cf4b077d8308c2b85c5cd139150b04bcb8b72a0d78af6a2b74fbb153b75acff71df0832d8f139231d1c18558f5b5479af348e6a0 -DIST ideep-6f4d653802bd43bc4eda515460df9f90353dbebe.tar.gz 57109 BLAKE2B 698ccc197e27a4812fc9df8d6821f8544e0528d9f925af97b7acf558cd91a5fa6ecb66012dda07e58d18d42c0aebb3f00baf5f947f01c851d6ecb2d9823eca0c SHA512 5310f15d7dfd45fc166d6b14a95ec35aa9b8e5fc3f5a5c5d2ab11666bce45ca5f23087780dccdcc60459d55e3b58fa13e3ea45bc0a51f890c0603660213e97f1 DIST pytorch-1.13.1.tar.gz 108279745 BLAKE2B 75de03b74dfdaf8d8fb5ea743fcc0c1b0e408a714ad4160c487921220a7b1755e5fa6e587e6bbc8c9f34dd75e096d2e6dd69c80d24821835fff6c833314434d3 SHA512 f16f89d027efade11d057245cad5b69a390e88b458398310ae30de2dbff7c8fd7f1165be7b8da7ea989c81ac3f5a66c5cb9050610e441a97c83fb8aa28c0bd62 DIST pytorch-2.0.1.tar.gz 111335778 BLAKE2B 7a10cc2b2d5e2422aef7e060a0c3a62ca5c7460c6e0b9becade9b98939501975c74ed5a175a653731f43ca824d2c9bd31f41d1f633c2b139779ab23d5331e9ce SHA512 2309a22b3be3ccdb36d8d9781a59a7bdcc2fdb8d95ada205702ec77862480f0cbb12cd5d6b8cd3114d01a6e33b7743d0fe9de93debf37138ca5c14403cdb0c43 DIST pytorch-2.1.2.tar.gz 116316469 BLAKE2B c5a55ee264bc3477d3556ba6376b5591117e992e56e0dd0c9ba93d12526e2727f7840f6f1e0730a38223b6492c9556840c4ebf22ffd220e97225c2abff303747 SHA512 a8961d78ad785b13c959a0612563a60e0de17a7c8bb9822ddea9a24072796354d07e81c47b6cc8761b21a6448845b088cf80e1661d9e889b0ed5474d3dc76756 EBUILD caffe2-1.13.1-r6.ebuild 5244 BLAKE2B d7f5f16e1f1122604a6df64f16c62552fb8b4b0de67bd231036b4835a5a71c58da02f0e1df64f3bd22c2a282a150d7f5a803c87cdf792e3e97ec8f518e055191 SHA512 58b1a09e5e3814d9475d4fe0e46aa837477843e09ed1b0c803c2ac3190e5c819d4216e33e3003652194e90a9f3f35146657eb25f30a4419bf16b2067e5a4b027 EBUILD caffe2-2.0.1-r5.ebuild 5868 BLAKE2B c17450d01ff68d42188c9da9e9c7a6d6469fe5c8b72c91e4ea4456eb4fe9d08cf30619fbb06f3dae21add641329f3ce8ced24bca93eb05900340639c042a1cdb SHA512 d32fa4c194c51efb76b143f22f783efa7a1415902b03ddea1d3f145d632455ebfe672a1b05b1203f2ba752f3331831455ae631a66bb409f45d6c695c68f39c3b -EBUILD caffe2-2.1.2-r5.ebuild 6851 BLAKE2B e45978a9e2ae0bf1f5278d3084cfb5a7ba720c9ae268f1950d8383bc72d8c11d0d8e660ce729d994142db681ff86bae51062613d6ddea5f7e2b4a0ef8f93f6c3 SHA512 c733da7efe19ab707e1ca38408feaa1c16929ec8441d14b3e902c869f558e2b2a607238edd60bc68d50a9f429ee41a1a34521de4f531f8eb7e24efabcdc86eb0 +EBUILD caffe2-2.1.2-r6.ebuild 6672 BLAKE2B dd4e772f6a1b0e049c25550ddbe8270d2f56390cdac5ccd6fe23bbcb5824a64085d4d8c0fa076779681cbc055f6f00191d8dcb485b91b9706bde45747a6223c1 SHA512 c71321f4e76fbb8ab3fb2af46560b0d689b2fdfb951ed1ead83d6ca8b30042a4080ec98695f43112b97a07aa748ec62c08e40a95e3b37440fb2306839f0b9847 MISC metadata.xml 1161 BLAKE2B 77145d6b17a38da3fc791b85ec6d1d8a4faa5f08485f7b8d7918f301342c9d95b9b9db9147334788ffa5137526365d0161a5e1420eabafb2058e1d85a5fa52bb SHA512 44fa18ac5e1abcfb021e8fc48db1bb9c0f61bc115484ae6f293f38c48d1f42704524490c8e3977eec8dccb728837e1fcb3ce2e892986e55044af3a15e82a61e9 diff --git a/sci-libs/caffe2/caffe2-2.1.2-r5.ebuild b/sci-libs/caffe2/caffe2-2.1.2-r5.ebuild deleted file mode 100644 index 91444530fe84..000000000000 --- a/sci-libs/caffe2/caffe2-2.1.2-r5.ebuild +++ /dev/null @@ -1,243 +0,0 @@ -# Copyright 2022-2024 Gentoo Authors -# Distributed under the terms of the GNU General Public License v2 - -EAPI=8 - -PYTHON_COMPAT=( python3_{9..12} ) -inherit python-single-r1 cmake cuda flag-o-matic prefix - -MYPN=pytorch -MYP=${MYPN}-${PV} -IDEEP_VERSION="6f4d653802bd43bc4eda515460df9f90353dbebe" - -DESCRIPTION="A deep learning framework" -HOMEPAGE="https://pytorch.org/" -SRC_URI="https://github.com/pytorch/${MYPN}/archive/refs/tags/v${PV}.tar.gz -> ${MYP}.tar.gz -onednn? ( https://github.com/intel/ideep/archive/${IDEEP_VERSION}.tar.gz -> ideep-${IDEEP_VERSION}.tar.gz ) -" - -LICENSE="BSD" -SLOT="0" -KEYWORDS="~amd64" -IUSE="cuda distributed fbgemm ffmpeg gloo mkl mpi nnpack +numpy onednn openblas opencl opencv openmp qnnpack tensorpipe xnnpack" -RESTRICT="test" -REQUIRED_USE=" - ${PYTHON_REQUIRED_USE} - ffmpeg? ( opencv ) - mpi? ( distributed ) - tensorpipe? ( distributed ) - distributed? ( tensorpipe ) - gloo? ( distributed ) -" # ?? ( cuda rocm ) - -# CUDA 12 not supported yet: https://github.com/pytorch/pytorch/issues/91122 -RDEPEND=" - ${PYTHON_DEPS} - dev-cpp/gflags:= - >=dev-cpp/glog-0.5.0 - dev-libs/cpuinfo - dev-libs/libfmt - dev-libs/protobuf:= - dev-libs/pthreadpool - dev-libs/sleef - virtual/lapack - >=sci-libs/onnx-1.12.0 - =dev-libs/cudnn-frontend-0.9.2:0/8 - dev-util/nvidia-cuda-toolkit:=[profiler] - ) - fbgemm? ( >=dev-libs/FBGEMM-2023.11.02 ) - ffmpeg? ( media-video/ffmpeg:= ) - gloo? ( sci-libs/gloo[cuda?] ) - mpi? ( virtual/mpi ) - nnpack? ( sci-libs/NNPACK ) - numpy? ( $(python_gen_cond_dep ' - dev-python/numpy[${PYTHON_USEDEP}] - ') ) - onednn? ( dev-libs/oneDNN ) - opencl? ( virtual/opencl ) - opencv? ( media-libs/opencv:= ) - qnnpack? ( sci-libs/QNNPACK ) - tensorpipe? ( sci-libs/tensorpipe[cuda?] ) - xnnpack? ( >=sci-libs/XNNPACK-2022.12.22 ) - mkl? ( sci-libs/mkl ) - openblas? ( sci-libs/openblas ) -" -DEPEND=" - ${RDEPEND} - cuda? ( >=dev-libs/cutlass-3.1.0 ) - dev-libs/psimd - dev-libs/FP16 - dev-libs/FXdiv - dev-libs/pocketfft - dev-libs/flatbuffers - >=sci-libs/kineto-0.4.0_p20231031 - $(python_gen_cond_dep ' - dev-python/pyyaml[${PYTHON_USEDEP}] - dev-python/pybind11[${PYTHON_USEDEP}] - ') -" - -S="${WORKDIR}"/${MYP} - -PATCHES=( - "${FILESDIR}"/${PN}-2.1.1-gentoo.patch - "${FILESDIR}"/${PN}-1.13.0-install-dirs.patch - "${FILESDIR}"/${PN}-1.12.0-glog-0.6.0.patch - "${FILESDIR}"/${PN}-1.13.1-tensorpipe.patch - "${FILESDIR}"/${PN}-2.0.0-gcc13.patch - "${FILESDIR}"/${PN}-2.0.0-cudnn_include_fix.patch - "${FILESDIR}"/${PN}-2.1.1-cudaExtra.patch - "${FILESDIR}"/${PN}-2.1.2-fix-rpath.patch - "${FILESDIR}"/${PN}-2.1.2-fix-openmp-link.patch -) - -src_prepare() { - filter-lto #bug 862672 - sed -i \ - -e "/third_party\/gloo/d" \ - cmake/Dependencies.cmake \ - || die - cmake_src_prepare - pushd torch/csrc/jit/serialization || die - flatc --cpp --gen-mutable --scoped-enums mobile_bytecode.fbs || die - popd - # prefixify the hardcoded paths, after all patches are applied - hprefixify \ - aten/CMakeLists.txt \ - caffe2/CMakeLists.txt \ - cmake/Metal.cmake \ - cmake/Modules/*.cmake \ - cmake/Modules_CUDA_fix/FindCUDNN.cmake \ - cmake/Modules_CUDA_fix/upstream/FindCUDA/make2cmake.cmake \ - cmake/Modules_CUDA_fix/upstream/FindPackageHandleStandardArgs.cmake \ - cmake/public/LoadHIP.cmake \ - cmake/public/cuda.cmake \ - cmake/Dependencies.cmake \ - torch/CMakeLists.txt \ - CMakeLists.txt -} - -src_configure() { - if use cuda && [[ -z ${TORCH_CUDA_ARCH_LIST} ]]; then - ewarn "WARNING: caffe2 is being built with its default CUDA compute capabilities: 3.5 and 7.0." - ewarn "These may not be optimal for your GPU." - ewarn "" - ewarn "To configure caffe2 with the CUDA compute capability that is optimal for your GPU," - ewarn "set TORCH_CUDA_ARCH_LIST in your make.conf, and re-emerge caffe2." - ewarn "For example, to use CUDA capability 7.5 & 3.5, add: TORCH_CUDA_ARCH_LIST=7.5 3.5" - ewarn "For a Maxwell model GPU, an example value would be: TORCH_CUDA_ARCH_LIST=Maxwell" - ewarn "" - ewarn "You can look up your GPU's CUDA compute capability at https://developer.nvidia.com/cuda-gpus" - ewarn "or by running /opt/cuda/extras/demo_suite/deviceQuery | grep 'CUDA Capability'" - fi - - local mycmakeargs=( - -DBUILD_CUSTOM_PROTOBUF=OFF - -DBUILD_SHARED_LIBS=ON - - -DUSE_CCACHE=OFF - -DUSE_CUDA=$(usex cuda) - -DUSE_CUDNN=$(usex cuda) - -DTORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST:-3.5 7.0}" - -DBUILD_NVFUSER=$(usex cuda) - -DUSE_DISTRIBUTED=$(usex distributed) - -DUSE_MPI=$(usex mpi) - -DUSE_FAKELOWP=OFF - -DUSE_FBGEMM=$(usex fbgemm) - -DUSE_FFMPEG=$(usex ffmpeg) - -DUSE_GFLAGS=ON - -DUSE_GLOG=ON - -DUSE_GLOO=$(usex gloo) - -DUSE_KINETO=OFF # TODO - -DUSE_LEVELDB=OFF - -DUSE_MAGMA=OFF # TODO: In GURU as sci-libs/magma - -DUSE_MKLDNN=$(usex onednn) - -DUSE_NCCL=OFF # TODO: NVIDIA Collective Communication Library - -DUSE_NNPACK=$(usex nnpack) - -DUSE_QNNPACK=$(usex qnnpack) - -DUSE_XNNPACK=$(usex xnnpack) - -DUSE_SYSTEM_XNNPACK=$(usex xnnpack) - -DUSE_TENSORPIPE=$(usex tensorpipe) - -DUSE_PYTORCH_QNNPACK=OFF - -DUSE_NUMPY=$(usex numpy) - -DUSE_OPENCL=$(usex opencl) - -DUSE_OPENCV=$(usex opencv) - -DUSE_OPENMP=$(usex openmp) - -DUSE_ROCM=OFF # TODO - -DUSE_SYSTEM_CPUINFO=ON - -DUSE_SYSTEM_PYBIND11=ON - -DUSE_UCC=OFF - -DUSE_VALGRIND=OFF - -DPYBIND11_PYTHON_VERSION="${EPYTHON#python}" - -DPYTHON_EXECUTABLE="${PYTHON}" - -DUSE_ITT=OFF - -DUSE_SYSTEM_PTHREADPOOL=ON - -DUSE_SYSTEM_FXDIV=ON - -DUSE_SYSTEM_FP16=ON - -DUSE_SYSTEM_GLOO=ON - -DUSE_SYSTEM_ONNX=ON - -DUSE_SYSTEM_SLEEF=ON - -DUSE_METAL=OFF - - -Wno-dev - -DTORCH_INSTALL_LIB_DIR="${EPREFIX}"/usr/$(get_libdir) - -DLIBSHM_INSTALL_LIB_SUBDIR="${EPREFIX}"/usr/$(get_libdir) - ) - - if use mkl; then - mycmakeargs+=(-DBLAS=MKL) - elif use openblas; then - mycmakeargs+=(-DBLAS=OpenBLAS) - else - mycmakeargs+=(-DBLAS=Generic -DBLAS_LIBRARIES=) - fi - - if use cuda; then - addpredict "/dev/nvidiactl" # bug 867706 - addpredict "/dev/char" - - mycmakeargs+=( - -DCMAKE_CUDA_FLAGS="$(cuda_gccdir -f | tr -d \")" - ) - fi - - if use onednn; then - mycmakeargs+=( - -DUSE_MKLDNN=ON - -DMKLDNN_FOUND=ON - -DMKLDNN_LIBRARIES=dnnl - -DMKLDNN_INCLUDE_DIR="${ESYSROOT}/usr/include/oneapi/dnnl;${WORKDIR}/ideep-${IDEEP_VERSION}/include" - ) - fi - - cmake_src_configure -} - -src_install() { - cmake_src_install - - insinto "/var/lib/${PN}" - doins "${BUILD_DIR}"/CMakeCache.txt - - rm -rf python - mkdir -p python/torch/include || die - mv "${ED}"/usr/lib/python*/site-packages/caffe2 python/ || die - if use cuda; then - mv "${ED}${S}"/nvfuser python/nvfuser || die - mv "${ED}"/usr/$(get_libdir)/nvfuser.so python/nvfuser/_C.so || die - fi - cp torch/version.py python/torch/ || die - python_domodule python/caffe2 - python_domodule python/torch - ln -s ../../../../../include/torch \ - "${D}$(python_get_sitedir)"/torch/include/torch || die # bug 923269 - if use cuda; then - python_domodule python/nvfuser - fi - rm -rf "${ED}"/var/tmp - find "${ED}" -empty -delete -} diff --git a/sci-libs/caffe2/caffe2-2.1.2-r6.ebuild b/sci-libs/caffe2/caffe2-2.1.2-r6.ebuild new file mode 100644 index 000000000000..a8d3fb624cf2 --- /dev/null +++ b/sci-libs/caffe2/caffe2-2.1.2-r6.ebuild @@ -0,0 +1,242 @@ +# Copyright 2022-2024 Gentoo Authors +# Distributed under the terms of the GNU General Public License v2 + +EAPI=8 + +PYTHON_COMPAT=( python3_{9..12} ) +inherit python-single-r1 cmake cuda flag-o-matic prefix + +MYPN=pytorch +MYP=${MYPN}-${PV} + +DESCRIPTION="A deep learning framework" +HOMEPAGE="https://pytorch.org/" +SRC_URI="https://github.com/pytorch/${MYPN}/archive/refs/tags/v${PV}.tar.gz + -> ${MYP}.tar.gz" + +LICENSE="BSD" +SLOT="0" +KEYWORDS="~amd64" +IUSE="cuda distributed fbgemm ffmpeg gloo mkl mpi nnpack +numpy onednn openblas opencl opencv openmp qnnpack tensorpipe xnnpack" +RESTRICT="test" +REQUIRED_USE=" + ${PYTHON_REQUIRED_USE} + ffmpeg? ( opencv ) + mpi? ( distributed ) + tensorpipe? ( distributed ) + distributed? ( tensorpipe ) + gloo? ( distributed ) +" # ?? ( cuda rocm ) + +# CUDA 12 not supported yet: https://github.com/pytorch/pytorch/issues/91122 +RDEPEND=" + ${PYTHON_DEPS} + dev-cpp/gflags:= + >=dev-cpp/glog-0.5.0 + dev-libs/cpuinfo + dev-libs/libfmt + dev-libs/protobuf:= + dev-libs/pthreadpool + dev-libs/sleef + virtual/lapack + >=sci-libs/onnx-1.12.0 + =dev-libs/cudnn-frontend-0.9.2:0/8 + dev-util/nvidia-cuda-toolkit:=[profiler] + ) + fbgemm? ( >=dev-libs/FBGEMM-2023.11.02 ) + ffmpeg? ( media-video/ffmpeg:= ) + gloo? ( sci-libs/gloo[cuda?] ) + mpi? ( virtual/mpi ) + nnpack? ( sci-libs/NNPACK ) + numpy? ( $(python_gen_cond_dep ' + dev-python/numpy[${PYTHON_USEDEP}] + ') ) + onednn? ( dev-libs/oneDNN ) + opencl? ( virtual/opencl ) + opencv? ( media-libs/opencv:= ) + qnnpack? ( sci-libs/QNNPACK ) + tensorpipe? ( sci-libs/tensorpipe[cuda?] ) + xnnpack? ( >=sci-libs/XNNPACK-2022.12.22 ) + mkl? ( sci-libs/mkl ) + openblas? ( sci-libs/openblas ) +" +DEPEND=" + ${RDEPEND} + cuda? ( >=dev-libs/cutlass-3.1.0 ) + onednn? ( sci-libs/ideep ) + dev-libs/psimd + dev-libs/FP16 + dev-libs/FXdiv + dev-libs/pocketfft + dev-libs/flatbuffers + >=sci-libs/kineto-0.4.0_p20231031 + $(python_gen_cond_dep ' + dev-python/pyyaml[${PYTHON_USEDEP}] + dev-python/pybind11[${PYTHON_USEDEP}] + ') +" + +S="${WORKDIR}"/${MYP} + +PATCHES=( + "${FILESDIR}"/${PN}-2.1.1-gentoo.patch + "${FILESDIR}"/${PN}-1.13.0-install-dirs.patch + "${FILESDIR}"/${PN}-1.12.0-glog-0.6.0.patch + "${FILESDIR}"/${PN}-1.13.1-tensorpipe.patch + "${FILESDIR}"/${PN}-2.0.0-gcc13.patch + "${FILESDIR}"/${PN}-2.0.0-cudnn_include_fix.patch + "${FILESDIR}"/${PN}-2.1.1-cudaExtra.patch + "${FILESDIR}"/${PN}-2.1.2-fix-rpath.patch + "${FILESDIR}"/${PN}-2.1.2-fix-openmp-link.patch +) + +src_prepare() { + filter-lto #bug 862672 + sed -i \ + -e "/third_party\/gloo/d" \ + cmake/Dependencies.cmake \ + || die + cmake_src_prepare + pushd torch/csrc/jit/serialization || die + flatc --cpp --gen-mutable --scoped-enums mobile_bytecode.fbs || die + popd + # prefixify the hardcoded paths, after all patches are applied + hprefixify \ + aten/CMakeLists.txt \ + caffe2/CMakeLists.txt \ + cmake/Metal.cmake \ + cmake/Modules/*.cmake \ + cmake/Modules_CUDA_fix/FindCUDNN.cmake \ + cmake/Modules_CUDA_fix/upstream/FindCUDA/make2cmake.cmake \ + cmake/Modules_CUDA_fix/upstream/FindPackageHandleStandardArgs.cmake \ + cmake/public/LoadHIP.cmake \ + cmake/public/cuda.cmake \ + cmake/Dependencies.cmake \ + torch/CMakeLists.txt \ + CMakeLists.txt +} + +src_configure() { + if use cuda && [[ -z ${TORCH_CUDA_ARCH_LIST} ]]; then + ewarn "WARNING: caffe2 is being built with its default CUDA compute capabilities: 3.5 and 7.0." + ewarn "These may not be optimal for your GPU." + ewarn "" + ewarn "To configure caffe2 with the CUDA compute capability that is optimal for your GPU," + ewarn "set TORCH_CUDA_ARCH_LIST in your make.conf, and re-emerge caffe2." + ewarn "For example, to use CUDA capability 7.5 & 3.5, add: TORCH_CUDA_ARCH_LIST=7.5 3.5" + ewarn "For a Maxwell model GPU, an example value would be: TORCH_CUDA_ARCH_LIST=Maxwell" + ewarn "" + ewarn "You can look up your GPU's CUDA compute capability at https://developer.nvidia.com/cuda-gpus" + ewarn "or by running /opt/cuda/extras/demo_suite/deviceQuery | grep 'CUDA Capability'" + fi + + local mycmakeargs=( + -DBUILD_CUSTOM_PROTOBUF=OFF + -DBUILD_SHARED_LIBS=ON + + -DUSE_CCACHE=OFF + -DUSE_CUDA=$(usex cuda) + -DUSE_CUDNN=$(usex cuda) + -DTORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST:-3.5 7.0}" + -DBUILD_NVFUSER=$(usex cuda) + -DUSE_DISTRIBUTED=$(usex distributed) + -DUSE_MPI=$(usex mpi) + -DUSE_FAKELOWP=OFF + -DUSE_FBGEMM=$(usex fbgemm) + -DUSE_FFMPEG=$(usex ffmpeg) + -DUSE_GFLAGS=ON + -DUSE_GLOG=ON + -DUSE_GLOO=$(usex gloo) + -DUSE_KINETO=OFF # TODO + -DUSE_LEVELDB=OFF + -DUSE_MAGMA=OFF # TODO: In GURU as sci-libs/magma + -DUSE_MKLDNN=$(usex onednn) + -DUSE_NCCL=OFF # TODO: NVIDIA Collective Communication Library + -DUSE_NNPACK=$(usex nnpack) + -DUSE_QNNPACK=$(usex qnnpack) + -DUSE_XNNPACK=$(usex xnnpack) + -DUSE_SYSTEM_XNNPACK=$(usex xnnpack) + -DUSE_TENSORPIPE=$(usex tensorpipe) + -DUSE_PYTORCH_QNNPACK=OFF + -DUSE_NUMPY=$(usex numpy) + -DUSE_OPENCL=$(usex opencl) + -DUSE_OPENCV=$(usex opencv) + -DUSE_OPENMP=$(usex openmp) + -DUSE_ROCM=OFF # TODO + -DUSE_SYSTEM_CPUINFO=ON + -DUSE_SYSTEM_PYBIND11=ON + -DUSE_UCC=OFF + -DUSE_VALGRIND=OFF + -DPYBIND11_PYTHON_VERSION="${EPYTHON#python}" + -DPYTHON_EXECUTABLE="${PYTHON}" + -DUSE_ITT=OFF + -DUSE_SYSTEM_PTHREADPOOL=ON + -DUSE_SYSTEM_FXDIV=ON + -DUSE_SYSTEM_FP16=ON + -DUSE_SYSTEM_GLOO=ON + -DUSE_SYSTEM_ONNX=ON + -DUSE_SYSTEM_SLEEF=ON + -DUSE_METAL=OFF + + -Wno-dev + -DTORCH_INSTALL_LIB_DIR="${EPREFIX}"/usr/$(get_libdir) + -DLIBSHM_INSTALL_LIB_SUBDIR="${EPREFIX}"/usr/$(get_libdir) + ) + + if use mkl; then + mycmakeargs+=(-DBLAS=MKL) + elif use openblas; then + mycmakeargs+=(-DBLAS=OpenBLAS) + else + mycmakeargs+=(-DBLAS=Generic -DBLAS_LIBRARIES=) + fi + + if use cuda; then + addpredict "/dev/nvidiactl" # bug 867706 + addpredict "/dev/char" + + mycmakeargs+=( + -DCMAKE_CUDA_FLAGS="$(cuda_gccdir -f | tr -d \")" + ) + fi + + if use onednn; then + mycmakeargs+=( + -DUSE_MKLDNN=ON + -DMKLDNN_FOUND=ON + -DMKLDNN_LIBRARIES=dnnl + -DMKLDNN_INCLUDE_DIR="${ESYSROOT}/usr/include/oneapi/dnnl" + ) + fi + + cmake_src_configure +} + +src_install() { + cmake_src_install + + insinto "/var/lib/${PN}" + doins "${BUILD_DIR}"/CMakeCache.txt + + rm -rf python + mkdir -p python/torch/include || die + mv "${ED}"/usr/lib/python*/site-packages/caffe2 python/ || die + if use cuda; then + mv "${ED}${S}"/nvfuser python/nvfuser || die + mv "${ED}"/usr/$(get_libdir)/nvfuser.so python/nvfuser/_C.so || die + fi + cp torch/version.py python/torch/ || die + python_domodule python/caffe2 + python_domodule python/torch + ln -s ../../../../../include/torch \ + "${D}$(python_get_sitedir)"/torch/include/torch || die # bug 923269 + if use cuda; then + python_domodule python/nvfuser + fi + rm -rf "${ED}"/var/tmp + find "${ED}" -empty -delete +} diff --git a/sci-libs/ideep/Manifest b/sci-libs/ideep/Manifest new file mode 100644 index 000000000000..2055840a1493 --- /dev/null +++ b/sci-libs/ideep/Manifest @@ -0,0 +1,6 @@ +AUX ideep-3.1.1-20230825.patch 1390 BLAKE2B 2787650021a21073aff0e200d8cdd6e47768b863d2b32275b5108f41d1d97f6caf40b40c4b2ebc18d92dbe2136684891f88e19365984aa1a91f5074633ee4279 SHA512 f2608fc21d64b6aaf59daf3af8003fad529decf5011601190bead62e0fe8e19d353b7e2522b9b7dba0ac3ce8ca2322b9bea1f2735f0144a747296b84566ac6b1 +DIST ideep-3.1.1.tar.gz 57015 BLAKE2B 054bb0363b5a6e8c40bf1192d557163f2e90195cbf3387562d5bf05a18eb0f3f2a45e5398ed0895ecc07807e753ab4f1a6375b12ede06aea33cc22e1e1d9cdb4 SHA512 1d636072512d90c1c47c0a976787a1fdaecdbd993ee319c40a2fbbc702143a0a752128e76e99f483d5628f64814684aa39b7355b2df675a4247037ddeeff03e5 +DIST ideep-3.3.2.tar.gz 57631 BLAKE2B 57ec733a3bf9e6e8ebdebdeb63000adcef8e9ecda00118c88bca24a80fd71602b5dc3d46892410d0231d2ecf2111d5a08c3a4e95bb785964cfc4c5d6c81537f0 SHA512 e0a9ca99f25b488e38b8887099c015ed0e67cc5c350c922b85f746e25c88e07a3291f8f5d71aef5f37c72814f32e00d81ec16853a425e0fe735c71c5fccc6a41 +EBUILD ideep-3.1.1.ebuild 470 BLAKE2B 947996aafac554f72bdd01451ec3dd29519eab8ff2906a3da7ed0750643e586f78ea4722a7094a8568cf9d529e3abcb0e11a2dd26bd6174043ce708fcf1485a8 SHA512 b09b6d6cb7447d0475f372dcba0cdaf62e146a55019a6268af237e5cf38e03fc4d6bda831097e0c4f9fae426067728a5a3cceae286466a56c7eef7328d175581 +EBUILD ideep-3.3.2.ebuild 423 BLAKE2B c100045c7f17e2cdb2f8e8e01ced4ec8513cd1c02425488ae94020ae9d3d0d71eb8d6b762f868e8f171aaf5c5511cc98c4f65c0375ce8f103d87af6063afaea8 SHA512 e064358c5264aa2eeaaf13b5904a20c126b317cc13351569d099ead1ff544af249647850b0ab61f0a6a543aa854d12f93926f73bcb3a060ecf998b3c297e0fcd +MISC metadata.xml 324 BLAKE2B 64ab7e6f3c76c6627e6d86be6650cd725a10b4bb8c886aeda603c94817c949a55579a4bf630dfaf82757f1245c9362a5378685378ee2f725e591e3f28b6703f2 SHA512 ff287f7f251a1c27d35876a89f2adf622c43a2bf02569d5035a28ba44a5381afc1ca9e801c628a07cfec776bdd53810948116eaacdf1dcc3e7dc1b073e29eb6e diff --git a/sci-libs/ideep/files/ideep-3.1.1-20230825.patch b/sci-libs/ideep/files/ideep-3.1.1-20230825.patch new file mode 100644 index 000000000000..a8625e758b00 --- /dev/null +++ b/sci-libs/ideep/files/ideep-3.1.1-20230825.patch @@ -0,0 +1,36 @@ +From 6f4d653802bd43bc4eda515460df9f90353dbebe Mon Sep 17 00:00:00 2001 +From: leslie-fang-intel +Date: Fri, 25 Aug 2023 13:37:20 +0800 +Subject: [PATCH] reland remove conv weight scale reciprocal (#226) + +--- + include/ideep.hpp | 2 +- + include/ideep/operators/conv.hpp | 3 --- + 2 files changed, 1 insertion(+), 4 deletions(-) + +diff --git a/include/ideep.hpp b/include/ideep.hpp +index 7e02c366..a11c3553 100644 +--- a/include/ideep.hpp ++++ b/include/ideep.hpp +@@ -48,6 +48,6 @@ + #define IDEEP_VERSION_MAJOR DNNL_VERSION_MAJOR + #define IDEEP_VERSION_MINOR DNNL_VERSION_MINOR + #define IDEEP_VERSION_PATCH DNNL_VERSION_PATCH +-#define IDEEP_VERSION_REVISION 0 ++#define IDEEP_VERSION_REVISION 1 + + #endif +diff --git a/include/ideep/operators/conv.hpp b/include/ideep/operators/conv.hpp +index 49677ff6..27ee84b6 100644 +--- a/include/ideep/operators/conv.hpp ++++ b/include/ideep/operators/conv.hpp +@@ -161,9 +161,6 @@ struct conv_deconv_utils { + weights_scales_in.size(), oc_per_group, groups, is_deconv); + auto wei_scales = weights_scales_in; + if (!std::all_of(wei_scales.begin(), wei_scales.end(), [](float i){ return i == 1.0f; })) { +- for (auto& s : wei_scales) { +- s = 1.0 / s; +- } + op_attr.set_scales(DNNL_ARG_WEIGHTS, wei_scale_mask, wei_scales); + } + if (dst_scales_in[0] != 1.0f) { diff --git a/sci-libs/ideep/ideep-3.1.1.ebuild b/sci-libs/ideep/ideep-3.1.1.ebuild new file mode 100644 index 000000000000..8f892eea0ed5 --- /dev/null +++ b/sci-libs/ideep/ideep-3.1.1.ebuild @@ -0,0 +1,22 @@ +# Copyright 2024 Gentoo Authors +# Distributed under the terms of the GNU General Public License v2 + +EAPI=8 + +MYP=pytorch-rls-v${PV} +DESCRIPTION="Intel® Optimization for Chainer" +HOMEPAGE="https://github.com/intel/ideep" +SRC_URI="https://github.com/intel/${PN}/archive/refs/tags/${MYP}.tar.gz + -> ${PF}.tar.gz" + +LICENSE="MIT" +SLOT="0" +KEYWORDS="~amd64" + +PATCHES=( "${FILESDIR}"/${P}-20230825.patch ) + +S="${WORKDIR}"/ideep-${MYP} + +src_install() { + doheader -r include/* +} diff --git a/sci-libs/ideep/ideep-3.3.2.ebuild b/sci-libs/ideep/ideep-3.3.2.ebuild new file mode 100644 index 000000000000..58a69716df8e --- /dev/null +++ b/sci-libs/ideep/ideep-3.3.2.ebuild @@ -0,0 +1,20 @@ +# Copyright 2024 Gentoo Authors +# Distributed under the terms of the GNU General Public License v2 + +EAPI=8 + +MYP=pytorch-rls-v${PV} +DESCRIPTION="Intel® Optimization for Chainer" +HOMEPAGE="https://github.com/intel/ideep" +SRC_URI="https://github.com/intel/${PN}/archive/refs/tags/${MYP}.tar.gz + -> ${PF}.tar.gz" + +LICENSE="MIT" +SLOT="0" +KEYWORDS="~amd64" + +S="${WORKDIR}"/ideep-${MYP} + +src_install() { + doheader -r include/* +} diff --git a/sci-libs/ideep/metadata.xml b/sci-libs/ideep/metadata.xml new file mode 100644 index 000000000000..953b04a60d15 --- /dev/null +++ b/sci-libs/ideep/metadata.xml @@ -0,0 +1,11 @@ + + + + + tupone@gentoo.org + Tupone Alfredo + + + intel/ideep + + -- cgit v1.2.3