diff --git a/dev-util/Tensile/Manifest b/dev-util/Tensile/Manifest index 857f3e7743ff1..8f731fb9e93c9 100644 --- a/dev-util/Tensile/Manifest +++ b/dev-util/Tensile/Manifest @@ -1,3 +1,4 @@ DIST rocm-Tensile-6.3.3.tar.gz 12723968 BLAKE2B e8c14f85dd44767b685595180974d683d759f84fc044e98974beae107f78ec97d43de30232ea0bd3578aa6cd40e17309d603ce583925ed4dd65cc146241e5b7d SHA512 836e17aab29ace976595c5b3fbc6b1742c774eb0420488a1449e04649c424e5a877168cf71de96fddcfbad337e295206ec456aea7649fe9e67a8802739ddbabe DIST rocm-Tensile-6.4.3.tar.gz 12731564 BLAKE2B b679ecf84cd023eecec8cef58729709f60b77615a76bc3215607c4e605b6057fa1c00e230d8c0ec9b24ca7f31c8e2e99dbffdfc732c3268958ea33d60962e256 SHA512 3d358d212ad52216520d391968578e3d6d7040f23c469e42b9fe6eea0a33034074df45d4a39a8978cc13f822f75a4cdfa850aa85c772ff4443276f9c4b9397a8 DIST rocm-Tensile-7.0.2.tar.gz 12056926 BLAKE2B 4e0af4789505b9fd2c813e7250eb7c1d877dbcce58dcf92828b1d5fc55fe9162aca65c3b8adfca541904fdaa660a620f579991868e3b3379428bf5e9f1f443de SHA512 1ad54d984d9334a4a2da0f10c268d72508e9e471a3bbc252bda7568eaadff83356a7c03726ca4908963ebd8e4bfebd227469766821ffc8418e3e68792f8e9fa1 +DIST rocm-Tensile-7.1.0.tar.gz 12402002 BLAKE2B d6cbb5d18fa5770fe2d5f10a0268bdfe173a1721b6f94db98d39dbaa84d89612e60a832c84f8071794da1766176b6dd2bc6dac7a4404b4711f1546ab34e7d6af SHA512 9e2c011819622f0b292cf450edb65a32e661d44d37bb7b0478ac4e1cdeb1c9cceee4a594fb8466b8e8e94086f1b9eaaf8f746db3ffb78f9ef3f047146e7825ee diff --git a/dev-util/Tensile/Tensile-7.1.0.ebuild b/dev-util/Tensile/Tensile-7.1.0.ebuild new file mode 100644 index 0000000000000..5ddfc216d6b19 --- /dev/null +++ b/dev-util/Tensile/Tensile-7.1.0.ebuild @@ -0,0 +1,157 @@ +# Copyright 1999-2025 Gentoo Authors +# Distributed under the terms of the GNU General Public License v2 + +EAPI=8 + +PYTHON_COMPAT=( python3_{10..13} ) +DISTUTILS_USE_PEP517=setuptools +ROCM_VERSION=${PV} +LLVM_COMPAT=( 20 ) + +inherit cmake distutils-r1 llvm-r1 prefix rocm + +DESCRIPTION="A tool for creating a benchmark-driven GEMMs and tensor contractions code" +HOMEPAGE="https://rocm.docs.amd.com/projects/Tensile/en/latest/src/index.html" + +if [[ "${PV}" == 9999 ]] ; then + inherit git-r3 + EGIT_REPO_URI="https://github.com/ROCm/rocm-libraries.git" + EGIT_BRANCH="develop" + S="${WORKDIR}/${P}/shared/tensile" + SLOT="0/7.0" +else + SRC_URI="https://github.com/ROCm/Tensile/archive/rocm-${PV}.tar.gz -> rocm-Tensile-${PV}.tar.gz" + S="${WORKDIR}/${PN}-rocm-${PV}" + SLOT="0/$(ver_cut 1-2)" + KEYWORDS="~amd64" +fi + +LICENSE="MIT" +IUSE="client test" +REQUIRED_USE="client? ( ${ROCM_REQUIRED_USE} )" + +# tests can freeze machine depending on gpu/kernel +RESTRICT="test" + +RDEPEND="${PYTHON_DEPS} + client? ( dev-libs/boost ) + >=dev-cpp/msgpack-cxx-6.0.0 + dev-python/pyyaml[${PYTHON_USEDEP}] + dev-python/msgpack[${PYTHON_USEDEP}] + dev-python/joblib[${PYTHON_USEDEP}] + dev-util/hip:${SLOT} + dev-util/rocm-smi:${SLOT} + $(llvm_gen_dep " + llvm-core/clang:\${LLVM_SLOT} + ") +" +DEPEND="${RDEPEND}" +BDEPEND=" + test? ( + dev-python/pytest-forked[${PYTHON_USEDEP}] + dev-python/pytest-xdist[${PYTHON_USEDEP}] + dev-python/filelock[${PYTHON_USEDEP}] + dev-python/joblib[${PYTHON_USEDEP}] + ) +" + +distutils_enable_tests pytest + +PATCHES=( + "${FILESDIR}"/${PN}-5.4.2-fix-arch-parse.patch + "${FILESDIR}"/${PN}-6.0.2-expand-isa-compatibility.patch + "${FILESDIR}"/${PN}-7.0.1-fix-install.patch +) + +CMAKE_USE_DIR="${S}/${PN}/Source" + +src_prepare() { + distutils-r1_src_prepare + sed -e "s,\@LLVM_PATH\@,$(get_llvm_prefix),g" \ + "${FILESDIR}/${PN}-5.7.1-gentoopath.patch" > "${S}"/gentoopath.patch || die + eapply $(prefixify_ro "${S}"/gentoopath.patch) + + pushd "${PN}" || die + + sed -e "/ROCM_SMI_ROOT/s,lib,$(get_libdir)," \ + -i Source/cmake/FindROCmSMI.cmake || die + sed -r -e "/TENSILE_USE_LLVM/s/ON/OFF/" \ + -i Source/CMakeLists.txt || die + + # ${Tensile_ROOT}/bin does not exists; call command directly + sed -e "s,\${Tensile_ROOT}/bin/,,g" -i cmake/TensileConfig.cmake || die + + local Tensile_share_dir="\"${EPREFIX}/usr/share/${PN}\"" + sed -e "/HipClangVersion/s/0.0.0/$(hipconfig -v)/" -i Common.py || die + + sed -e "s,os.path.dirname(os.path.realpath(__file__)),${Tensile_share_dir},g" \ + -i ReplacementKernels.py Common.py "${PN}.py" || die + + sed -e "s|os\.path\.dirname.*$|\"${EPREFIX}/usr/share/Tensile/Source\", end='')|" -i __init__.py || die + + # bug 949817: fix v_dot4_i32_i8 syntax for clang-20 + sed 's/ op_sel:\[0,0\] op_sel_hi:\[1,1\]//' -i Components/MAC_I8X4.py || die + + # Fix compiler "validation" + rocm_use_clang + sed "s/amdclang/$(basename "$CC")/g" -i Utilities/Toolchain.py || die + + popd || die + + use client && PATCHES='' cmake_src_prepare # do not apply patches again in cmake_src_prepare +} + +src_configure() { + rocm_use_clang + + distutils-r1_src_configure + if use client; then + local mycmakeargs=( + -DCMAKE_SKIP_RPATH=ON + -DTENSILE_USE_MSGPACK=ON + -DTENSILE_USE_LLVM=ON + -DTensile_LIBRARY_FORMAT=msgpack + -DAMDGPU_TARGETS="$(get_amdgpu_flags)" + ) + cmake_src_configure + fi +} + +src_compile() { + distutils-r1_src_compile + use client && cmake_src_compile +} + +python_install() { + distutils-r1_python_install + + python_moduleinto Tensile + pushd Tensile || die + python_domodule Components + python_domodule Utilities + python_domodule TensileCreateLib +} + +src_install() { + distutils-r1_src_install + + pushd "${PN}" || die + insinto "/usr/share/${PN}" + doins -r Configs Perf Source CustomKernels + insinto "/usr/$(get_libdir)/cmake/${PN}" + doins cmake/*.cmake + + if use client; then + pushd "${BUILD_DIR}" || die + dobin client/tensile_client + fi + + # Remove extra copy + rm -rf "${ED}"/usr/cmake || die +} + +# Test suite fails to start without this +python_test() { + export ROCM_PATH="${EPREFIX}/usr" + epytest +}