* [gentoo-dev] [PATCH 1/2] rocm.eclass: new eclass
2022-08-08 14:15 [gentoo-dev] [PATCH 0/2] rocm.eclass for ROCm packages written in HIP Yiyang Wu
@ 2022-08-08 14:15 ` Yiyang Wu
2022-08-08 15:40 ` Ulrich Mueller
2022-08-08 14:15 ` [gentoo-dev] [PATCH 2/2] profiles/desc: add amdgpu_targets.desc for USE_EXPAND Yiyang Wu
2022-08-08 15:04 ` [gentoo-dev] Re: [PATCH 0/2] rocm.eclass for ROCm packages written in HIP wuyy
2 siblings, 1 reply; 6+ messages in thread
From: Yiyang Wu @ 2022-08-08 14:15 UTC (permalink / raw
To: gentoo-dev; +Cc: Benda Xu, YiyangWu
Closes: https://bugs.gentoo.org/810619
Signed-off-by: YiyangWu <xgreenlandforwyy@gmail.com>
---
eclass/rocm.eclass | 278 ++++++++++++++++++++++++++++++++++++
profiles/base/make.defaults | 2 +-
2 files changed, 279 insertions(+), 1 deletion(-)
create mode 100644 eclass/rocm.eclass
diff --git a/eclass/rocm.eclass b/eclass/rocm.eclass
new file mode 100644
index 000000000000..e4b448f7b894
--- /dev/null
+++ b/eclass/rocm.eclass
@@ -0,0 +1,278 @@
+# Copyright 2022 Gentoo Authors
+# Distributed under the terms of the GNU General Public License v2
+
+# @ECLASS: rocm.eclass
+# @MAINTAINER:
+# Gentoo Science Project <sci@gentoo.org>
+# @AUTHOR:
+# Yiyang Wu <xgreenlandforwyy@gmail.com>
+# @SUPPORTED_EAPIS: 7 8
+# @BLURB: Common functions and variables for ROCm packages written in HIP
+# @DESCRIPTION:
+# ROCm packages such as sci-libs/<roc|hip>* can utilize functions in this eclass.
+# Currently, it handles the AMDGPU_TARGETS variable via USE_EXPAND, so user can
+# use USE flag to control which GPU architecture to compile, and ensure coherence
+# among dependencies. It also specify CXX=hipcc, to let hipcc compile. Another
+# important function is src_test, which checks whether a valid KFD device exists
+# for testing, and then execute the test program.
+#
+# Most ROCm packages use cmake as build system, so this eclass does not export
+# phase functions which overwrites the phase functions in cmake.eclass. Ebuild
+# should explicitly call rocm_src_* in src_configure and src_test.
+#
+# @EXAMPLE:
+# # Example for ROCm packages in https://github.com/ROCmSoftwarePlatform
+# inherit cmake rocm
+# SRC_URI="https://github.com/ROCmSoftwarePlatform/${PN}/archive/rocm-${PV}.tar.gz -> ${P}.tar.gz"
+# SLOT="0/$(ver_cut 1-2)"
+# IUSE="test"
+# REQUIRED_USE="${ROCM_REQUIRED_USE}"
+# RESTRICT="!test? ( test )"
+#
+# RDEPEND="
+# dev-util/hip
+# sci-libs/rocBLAS:${SLOT}[${ROCM_USEDEP}]
+# "
+#
+# S=${WORKDIR}/${PN}-rocm-${PV}
+#
+# src_configure() {
+# local mycmakeargs=(
+# -DBUILD_CLIENTS_TESTS=$(usex test ON OFF)
+# )
+# rocm_src_configure
+# }
+#
+# src_test() {
+# rocm_src_test
+# }
+#
+# # Example for packages depend on ROCm libraries -- a package depend on
+# # rocBLAS, and use comma seperated ${HCC_AMDGPU_TARGET} to determine GPU
+# # architecture to compile. Requires ROCm version >5.
+# ROCM_VERSION=5
+# inherit rocm
+# IUSE="rocm"
+# REQUIRED_USE="rocm? ( ${ROCM_REQUIRED_USE} )"
+# DEPEND="rocm? ( >=dev-util/hip-${ROCM_VERSION}
+# >=sci-libs/rocBLAS-${ROCM_VERSION}[${ROCM_USEDEP}] )"
+# ....
+# src_configure() {
+# if use rocm; then
+# local AMDGPU_FLAGS=$(get_amdgpu_flags)
+# export HCC_AMDGPU_TARGET=${AMDGPU_FLAGS//;/,}
+# fi
+# default
+# }
+
+if [[ ! ${_ROCM_ECLASS} ]]; then
+
+case ${EAPI} in
+ 0|1|2|3|4|5|6)
+ die "${ECLASS}: unsupported EAPI=${EAPI:-0} (too old)"
+ ;;
+ 7|8)
+ ;;
+ *)
+ die "${ECLASS}: unsupported EAPI=${EAPI} (unknown)"
+ ;;
+esac
+
+inherit edo
+
+
+# @ECLASS_VARIABLE: ROCM_VERSION
+# @DEFAULT_UNSET
+# @PRE_INHERIT
+# @DESCRIPTION:
+# The ROCm version of current package. Default is ${PV}, but for other packages
+# that depend on ROCm libraries, this can be set to match the version of
+# required ROCm libraries.
+
+# @ECLASS_VARIABLE: ALL_AMDGPU_TARGETS
+# @INTERNAL
+# @DESCRIPTION:
+# The list of USE flags corresponding to all AMDGPU targets in this ROCm
+# version. The value depends on ${PV}. Architectures and devices map:
+# https://www.coelacanth-dream.com/posts/2019/12/30/did-rid-product-matome-p2
+
+# @ECLASS_VARIABLE: OFFICIAL_AMDGPU_TARGETS
+# @INTERNAL
+# @DESCRIPTION:
+# The list of USE flags corresponding to all officlially supported AMDGPU
+# targets in this ROCm version, documented at
+# https://docs.amd.com/bundle/ROCm-Installation-Guide-v${PV}/page/Prerequisite_Actions.html.
+# USE flag of these architectures will be default on. Depends on ${PV}.
+
+# @ECLASS_VARIABLE: ROCM_REQUIRED_USE
+# @OUTPUT_VARIABLE
+# @DESCRIPTION:
+# Requires at least one AMDGPU target to be compiled.
+# Example use for ROCm libraries:
+# REQUIRED_USE="${ROCM_REQUIRED_USE}"
+# Example use for packages that depend on ROCm libraries
+# IUSE="rocm"
+# REQUIRED_USE="rocm? ( ${ROCM_REQUIRED_USE} )"
+
+# @ECLASS_VARIABLE: ROCM_USEDEP
+# @OUTPUT_VARIABLE
+# @DESCRIPTION:
+# This is an eclass-generated USE-dependency string which can be used to
+# depend on another ROCm package being built for the same AMDGPU architecture.
+#
+# The generated USE-flag list is compatible with packages using rocm.eclass.
+#
+# Example use:
+# @CODE
+# DEPEND="sci-libs/rocBLAS[${ROCM_USEDEP}]"
+# @CODE
+
+# @FUNCTION: _rocm_set_globals
+# @DESCRIPTION:
+# Set global variables used by the eclass: ALL_AMDGPU_TARGETS,
+# OFFICIAL_AMDGPU_TARGETS, ROCM_REQUIRED_USE, and ROCM_USEDEP
+_rocm_set_globals() {
+ case ${ROCM_VERSION:-${PV}} in
+ 4*)
+ ALL_AMDGPU_TARGETS=(
+ gfx803 gfx900 gfx906 gfx908
+ gfx1010 gfx1011 gfx1012 gfx1030
+ )
+ OFFICIAL_AMDGPU_TARGETS=(
+ gfx906 gfx908
+ )
+ ;;
+ 5*)
+ ALL_AMDGPU_TARGETS=(
+ gfx803 gfx900 gfx906 gfx908 gfx90a
+ gfx1010 gfx1011 gfx1012 gfx1030 gfx1031
+ )
+ OFFICIAL_AMDGPU_TARGETS=(
+ gfx906 gfx908 gfx90a gfx1030
+ )
+ ;;
+ *)
+ die "Unknown ROCm major version! Please update rocm.eclass before bumping to new ebuilds"
+ ;;
+ esac
+
+ ROCM_REQUIRED_USE+=" || ("
+ for gpu_target in ${ALL_AMDGPU_TARGETS[@]}; do
+ if [[ " ${OFFICIAL_AMDGPU_TARGETS[*]} " =~ " ${gpu_target} " ]]; then
+ IUSE+=" ${gpu_target/#/+amdgpu_targets_}"
+ else
+ IUSE+=" ${gpu_target/#/amdgpu_targets_}"
+ fi
+ ROCM_REQUIRED_USE+=" ${gpu_target/#/amdgpu_targets_}"
+ done
+ ROCM_REQUIRED_USE+=" ) "
+
+ local flags=( "${ALL_AMDGPU_TARGETS[@]/#/amdgpu_targets_}" )
+ local optflags=${flags[@]/%/(-)?}
+ ROCM_USEDEP=${optflags// /,}
+}
+_rocm_set_globals
+unset -f _rocm_set_globals
+
+
+# @FUNCTION: get_amdgpu_flags
+# @USAGE: get_amdgpu_flags
+# @DESCRIPTION:
+# Convert specified use flag of amdgpu_targets to compilation flags.
+# Append default target feature to GPU arch. See
+# https://llvm.org/docs/AMDGPUUsage.html#target-features
+get_amdgpu_flags() {
+ local AMDGPU_TARGET_FLAGS
+ for gpu_target in ${ALL_AMDGPU_TARGETS[@]}; do
+ local target_feature=
+ if use amdgpu_targets_${gpu_target}; then
+ case ${gpu_target} in
+ gfx906|gfx908)
+ target_feature=:xnack-
+ ;;
+ gfx90a)
+ target_feature=:xnack+
+ ;;
+ *)
+ ;;
+ esac
+ AMDGPU_TARGET_FLAGS+="${gpu_target}${target_feature};"
+ fi
+ done
+ echo ${AMDGPU_TARGET_FLAGS}
+}
+
+# @FUNCTION: check_rw_permission
+# @USAGE: check_rw_permission <file>
+# @DESCRIPTION:
+# check read and write permissions on specific files.
+# allow using wildcard, for example check_rw_permission /dev/dri/render*
+check_rw_permission() {
+ [[ -r "$1" ]] && [[ -w "$1" ]] || die \
+ "${PORTAGE_USERNAME} do not have read or write permissions on $1! \n Make sure ${PORTAGE_USERNAME} is in render group and check the permissions."
+}
+
+# == phase functions ==
+
+# @FUNCTION: rocm_src_configure
+# @DESCRIPTION:
+# configure rocm packages, and setting common cmake arguments
+rocm_src_configure() {
+ # allow acces to hardware
+ addpredict /dev/kfd
+ addpredict /dev/dri/
+
+ mycmakeargs+=(
+ -DCMAKE_INSTALL_PREFIX="${EPREFIX}/usr"
+ -DAMDGPU_TARGETS="$(get_amdgpu_flags)"
+ -DCMAKE_SKIP_RPATH=TRUE
+ )
+
+ CXX="hipcc" cmake_src_configure
+}
+
+# @FUNCTION: rocm_src_test
+# @DESCRIPTION:
+# Test whether valid GPU device is present. If so, find how to, and execute test.
+# ROCm packages can have to test mechanism:
+# 1. cmake_src_test. Set MAKEOPTS="-j1" to make sure only one test on GPU at a time;
+# 2. one single gtest binary called "${PN,,}"-test;
+# 3. Some package like rocFFT have alternative test like rocfft-selftest;
+# 4. Custome testing binaries like dev-libs/rccl. Use ${ROCM_TESTS} to specify.
+rocm_src_test() {
+ addwrite /dev/kfd
+ addwrite /dev/dri/
+
+ # check permissions on /dev/kfd and /dev/dri/render*
+ check_rw_permission /dev/kfd
+ check_rw_permission /dev/dri/render*
+
+ : ${LD_LIBRARY_PATH:="${BUILD_DIR}/clients:${BUILD_DIR}/src:${BUILD_DIR}/library:${BUILD_DIR}/library/src:${BUILD_DIR}/library/src/device"}
+ export LD_LIBRARY_PATH
+ if grep -q 'build test:' "${BUILD_DIR}"/build.ninja; then
+ MAKEOPTS="-j1" cmake_src_test
+ elif [[ -d "${BUILD_DIR}"/clients/staging ]]; then
+ cd "${BUILD_DIR}/clients/staging" || die "Test directory not found!"
+ for test_program in "${PN,,}-"*test; do
+ if [[ -x ${test_program} ]]; then
+ edob ./${test_program}
+ else
+ die "The test program ${test_program} does not exist or cannot be excuted!"
+ fi
+ done
+ elif [[ ! -z "${ROCM_TESTS}" ]]; then
+ for test_program in "${ROCM_TESTS}"; do
+ cd "${BUILD_DIR}" || die
+ if [[ -x ${test_program} ]]; then
+ edob ./${test_program}
+ else
+ die "The test program ${test_program} does not exist or cannot be excuted!"
+ fi
+ done
+ else
+ die "There is no cmake tests, no \${ROCM_TESTS} executable provided, nor ${BUILD_DIR}/clients/staging where test program might be located."
+ fi
+}
+
+_ROCM_ECLASS=1
+fi
diff --git a/profiles/base/make.defaults b/profiles/base/make.defaults
index 326cb28de537..2c288d12d103 100644
--- a/profiles/base/make.defaults
+++ b/profiles/base/make.defaults
@@ -13,7 +13,7 @@ USE_EXPAND_VALUES_USERLAND="BSD GNU"
# Env vars to expand into USE vars. Modifying this requires prior
# discussion on gentoo-dev@lists.gentoo.org.
-USE_EXPAND="ABI_MIPS ABI_S390 ABI_X86 ADA_TARGET ALSA_CARDS APACHE2_MODULES APACHE2_MPMS CALLIGRA_FEATURES CAMERAS COLLECTD_PLUGINS CPU_FLAGS_ARM CPU_FLAGS_PPC CPU_FLAGS_X86 CURL_SSL ELIBC FFTOOLS GPSD_PROTOCOLS GRUB_PLATFORMS INPUT_DEVICES KERNEL L10N LCD_DEVICES LIBREOFFICE_EXTENSIONS LLVM_TARGETS LUA_SINGLE_TARGET LUA_TARGETS MONKEYD_PLUGINS NGINX_MODULES_HTTP NGINX_MODULES_MAIL NGINX_MODULES_STREAM OFFICE_IMPLEMENTATION OPENMPI_FABRICS OPENMPI_OFED_FEATURES OPENMPI_RM PHP_TARGETS POSTGRES_TARGETS PYTHON_SINGLE_TARGET PYTHON_TARGETS QEMU_SOFTMMU_TARGETS QEMU_USER_TARGETS ROS_MESSAGES RUBY_TARGETS SANE_BACKENDS USERLAND UWSGI_PLUGINS VIDEO_CARDS VOICEMAIL_STORAGE XTABLES_ADDONS"
+USE_EXPAND="ABI_MIPS ABI_S390 ABI_X86 ADA_TARGET ALSA_CARDS AMDGPU_TARGETS APACHE2_MODULES APACHE2_MPMS CALLIGRA_FEATURES CAMERAS COLLECTD_PLUGINS CPU_FLAGS_ARM CPU_FLAGS_PPC CPU_FLAGS_X86 CURL_SSL ELIBC FFTOOLS GPSD_PROTOCOLS GRUB_PLATFORMS INPUT_DEVICES KERNEL L10N LCD_DEVICES LIBREOFFICE_EXTENSIONS LLVM_TARGETS LUA_SINGLE_TARGET LUA_TARGETS MONKEYD_PLUGINS NGINX_MODULES_HTTP NGINX_MODULES_MAIL NGINX_MODULES_STREAM OFFICE_IMPLEMENTATION OPENMPI_FABRICS OPENMPI_OFED_FEATURES OPENMPI_RM PHP_TARGETS POSTGRES_TARGETS PYTHON_SINGLE_TARGET PYTHON_TARGETS QEMU_SOFTMMU_TARGETS QEMU_USER_TARGETS ROS_MESSAGES RUBY_TARGETS SANE_BACKENDS USERLAND UWSGI_PLUGINS VIDEO_CARDS VOICEMAIL_STORAGE XTABLES_ADDONS"
# USE_EXPAND variables whose contents are not shown in package manager
# output. Changes need discussion on gentoo-dev.
--
2.34.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [gentoo-dev] [PATCH 2/2] profiles/desc: add amdgpu_targets.desc for USE_EXPAND
2022-08-08 14:15 [gentoo-dev] [PATCH 0/2] rocm.eclass for ROCm packages written in HIP Yiyang Wu
2022-08-08 14:15 ` [gentoo-dev] [PATCH 1/2] rocm.eclass: new eclass Yiyang Wu
@ 2022-08-08 14:15 ` Yiyang Wu
2022-08-08 16:37 ` [gentoo-dev] " Benda Xu
2022-08-08 15:04 ` [gentoo-dev] Re: [PATCH 0/2] rocm.eclass for ROCm packages written in HIP wuyy
2 siblings, 1 reply; 6+ messages in thread
From: Yiyang Wu @ 2022-08-08 14:15 UTC (permalink / raw
To: gentoo-dev; +Cc: Benda Xu
Signed-off-by: Yiyang Wu <xgreenlandforwyy@gmail.com>
---
profiles/desc/amdgpu_targets.desc | 15 +++++++++++++++
1 file changed, 15 insertions(+)
create mode 100644 profiles/desc/amdgpu_targets.desc
diff --git a/profiles/desc/amdgpu_targets.desc b/profiles/desc/amdgpu_targets.desc
new file mode 100644
index 000000000000..942ab51356fb
--- /dev/null
+++ b/profiles/desc/amdgpu_targets.desc
@@ -0,0 +1,15 @@
+# Copyright 1999-2022 Gentoo Authors.
+# Distributed under the terms of the GNU General Public License v2
+
+# Copied from https://www.coelacanth-dream.com/posts/2019/12/30/did-rid-product-matome-p2/#fn:67
+
+gfx803 - Fiji GPU, codename fiji, including Radeon R9 Nano/Fury/FuryX, Radeon Pro Duo, FirePro S9300x2, Radeon Instinct MI8
+gfx900 - Vega GPU, codename vega10, including Radeon Vega Frontier Edition, Radeon RX Vega 56/64, Radeon RX Vega 64 Liquid, Radeon Pro Vega 48/56/64/64X, Radeon Pro WX 8200/9100, Radeon Pro V320/V340/SSG, Radeon Instinct MI25
+gfx906 - Vega GPU, codename vega20, including Radeon (Pro) VII, Radeon Instinct MI50/MI60
+gfx908 - CDNA Accelerator, codename arcturus, including AMD Instinct MI100 Accelerator
+gfx90a - CDNA2 Accelerator, codename aldebaran, including AMD Instinct MI200 series Accelerators
+gfx1010 - RDNA GPU, codename navi10, including Radeon RX 5700XT/5700/5700M/5700B/5700XTB/5600XT/5600/5600M, Radeon Pro 5700Xt/5700, Radeon Pro W5700X/W5700
+gfx1011 - RDNA GPU, codename navi12, including Radeon Pro 5600M/V520
+gfx1012 - RDNA GPU, codename navi14, including Radeon RX 5500XT/5500/5500M/5500XTB/5300/5300M, Radeon Pro 5500XT/5500M/5300/5300M, Radeon Pro W5500X/W5500/W5500M/W5300M
+gfx1030 - RDNA2 GPU, codename navi21/sienna cichlid, including Radeon RX 6950XT/6900XT/6800XT/6800, Radeon Pro W6800
+gfx1031 - RDNA2 GPU, codename navi22/navy flounder, including Radeon RX 6750XT/6700XT/6800M/6700M
--
2.34.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [gentoo-dev] Re: [PATCH 0/2] rocm.eclass for ROCm packages written in HIP
2022-08-08 14:15 [gentoo-dev] [PATCH 0/2] rocm.eclass for ROCm packages written in HIP Yiyang Wu
2022-08-08 14:15 ` [gentoo-dev] [PATCH 1/2] rocm.eclass: new eclass Yiyang Wu
2022-08-08 14:15 ` [gentoo-dev] [PATCH 2/2] profiles/desc: add amdgpu_targets.desc for USE_EXPAND Yiyang Wu
@ 2022-08-08 15:04 ` wuyy
2 siblings, 0 replies; 6+ messages in thread
From: wuyy @ 2022-08-08 15:04 UTC (permalink / raw
To: gentoo-dev
On Mon, Aug 08, 2022 at 10:15:25PM +0800, Yiyang Wu wrote:
> This eclass provides utilities for ROCm libraries in
> https://github.com/ROCmSoftwarePlatform, e.g. rocBLAS, rocFFT.
> It contains a USE_EXPAND, amdgpu_targets_*, which handles the GPU
> architecture to compile, and keep targets coherent among dependencies.
> Packages that depend on ROCm libraries, like cupy, can also make use of
> this eclass, mainly specify GPU architecture and it's corresponding
> dependencies via USE_EXPAND.
>
> This is the core delivery in GSoC 2022 project "Refining ROCm Packages
> in Gentoo"
>
> Yiyang Wu (2):
> rocm.eclass: new eclass
> profiles/desc: add amdgpu_targets.desc for USE_EXPAND
>
> eclass/rocm.eclass | 278 ++++++++++++++++++++++++++++++
> profiles/base/make.defaults | 2 +-
> profiles/desc/amdgpu_targets.desc | 15 ++
> 3 files changed, 294 insertions(+), 1 deletion(-)
> create mode 100644 eclass/rocm.eclass
> create mode 100644 profiles/desc/amdgpu_targets.desc
>
> --
> 2.34.1
>
I have created a PR of this patch at https://github.com/gentoo/gentoo/pull/26784.
--
Yiyang Wu
^ permalink raw reply [flat|nested] 6+ messages in thread