Skip to content

GitLab

"vscode:/vscode.git/clone" did not exist on "0d26477a864fcaf6f1dfe709c9f8421e5305b933"
Commit 63b152d6 authored by danyao12's avatar danyao12
Browse files

Merge branch 'develop' into ck_tile/fa_bwd_v3

parents ae2d7d2b 14c3cfb1
Hide whitespace changes
Inline Side-by-side
Showing with 379 additions and 192 deletions
CMakeLists.txt
View file @ 63b152d6
......@@ -97,10 +97,9 @@ if(DL_KERNELS)
add_definitions(-DDL_KERNELS)
set(CK_ENABLE_DL_KERNELS "ON")
endif()
if(INSTANCES_ONLY)
add_definitions(-DINSTANCES_ONLY)
set(CK_ENABLE_INSTANCES_ONLY "ON")
option(CK_USE_CODEGEN "Enable codegen library" OFF)
if(CK_USE_CODEGEN)
add_definitions(-DCK_USE_CODEGEN)
endif()
include(getopt)
......@@ -127,7 +126,17 @@ rocm_setup_version(VERSION ${version})
list(APPEND CMAKE_PREFIX_PATH ${CMAKE_INSTALL_PREFIX} ${CMAKE_INSTALL_PREFIX}/llvm ${CMAKE_INSTALL_PREFIX}/hip /opt/rocm /opt/rocm/llvm /opt/rocm/hip "$ENV{ROCM_PATH}" "$ENV{HIP_PATH}")
message("GPU_TARGETS= ${GPU_TARGETS}")
message("GPU_ARCHS= ${GPU_ARCHS}")
if(GPU_ARCHS)
#disable GPU_TARGETS to avoid conflicts, this needs to happen before we call hip package
unset(GPU_TARGETS CACHE)
unset(AMDGPU_TARGETS CACHE)
endif()
if(GPU_TARGETS)
set(USER_GPU_TARGETS 1)
else()
set(USER_GPU_TARGETS 0)
endif()
find_package(hip)
# No assumption that HIP kernels are launched with uniform block size for backward compatibility
# SWDEV-413293 and https://reviews.llvm.org/D155213
......@@ -135,56 +144,39 @@ math(EXPR hip_VERSION_FLAT "(${hip_VERSION_MAJOR} * 1000 + ${hip_VERSION_MINOR})
message("hip_version_flat=${hip_VERSION_FLAT}")
message("checking which targets are supported")
#This is the list of targets to be used in case GPU_TARGETS is not set on command line
#These targets will be filtered and only supported ones will be used
#Setting GPU_TARGETS on command line will override this list
if(NOT PROFILER_ONLY)
if(NOT ENABLE_ASAN_PACKAGING)
#build CK for all supported targets
if(NOT WIN32 AND ${hip_VERSION_FLAT} LESS 600300000)
# WORKAROUND: compiler does not yet fully support gfx12 targets, need to fix version above
rocm_check_target_ids(DEFAULT_GPU_TARGETS
TARGETS "gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102")
else()
rocm_check_target_ids(DEFAULT_GPU_TARGETS
TARGETS "gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201")
endif()
#In order to build just the CK library (without tests and examples) for all supported GPU targets
#use -D GPU_ARCHS="gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201"
#the GPU_TARGETS flag will be reset in this case in order to avoid conflicts.
#
#In order to build CK along with all tests and examples it should be OK to set GPU_TARGETS to just 1 or 2 similar architectures.
if(NOT ENABLE_ASAN_PACKAGING)
if(NOT WIN32 AND ${hip_VERSION_FLAT} LESS 600300000)
# WORKAROUND: compiler does not yet fully support gfx12 targets, need to fix version above
set(CK_GPU_TARGETS "gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102")
else()
#build CK only for xnack-supported targets
rocm_check_target_ids(DEFAULT_GPU_TARGETS
TARGETS "gfx908:xnack+;gfx90a:xnack+;gfx940:xnack+;gfx941:xnack+;gfx942:xnack+")
set(GPU_TARGETS "${DEFAULT_GPU_TARGETS}" CACHE STRING " " FORCE)
set(CK_GPU_TARGETS "gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201")
endif()
else()
add_definitions(-DPROFILER_ONLY)
set(GPU_TARGETS "" CACHE STRING "" FORCE)
if(GPU_TARGETS)
message(FATAL_ERROR "For PROFILE_ONLY build, please do not set GPU_TARGETS, use GPU_ARCH = gfx90, gfx94, gfx10, gfx11 or gfx12")
endif()
if(GPU_ARCH MATCHES "gfx90")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx908;gfx90a")
elseif(GPU_ARCH MATCHES "gfx94")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx940;gfx941;gfx942")
elseif(GPU_ARCH MATCHES "gfx10")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx1030")
elseif(GPU_ARCH MATCHES "gfx11")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx1100;gfx1101;gfx1102")
elseif(GPU_ARCH MATCHES "gfx12")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx1200;gfx1201")
else()
message(FATAL_ERROR "For PROFILE_ONLY build, please specify GPU_ARCH as gfx90, gfx94, gfx10, gfx11 or gfx12")
endif()
set(GPU_TARGETS "${DEFAULT_GPU_TARGETS}" CACHE STRING " " FORCE)
#build CK only for xnack-supported targets when using ASAN
set(CK_GPU_TARGETS "gfx908:xnack+;gfx90a:xnack+;gfx940:xnack+;gfx941:xnack+;gfx942:xnack+")
endif()
message("Supported GPU_TARGETS= ${DEFAULT_GPU_TARGETS}")
if(GPU_TARGETS)
message("Building CK for the following targets: ${GPU_TARGETS}")
#if user set GPU_ARCHS on the cmake command line, overwrite default target list with user's list
#otherwise, if user set GPU_TARGETS, use that set of targets
if(GPU_ARCHS)
set(CK_GPU_TARGETS ${GPU_ARCHS})
else()
message("Building CK for the default targets: ${DEFAULT_GPU_TARGETS}")
if(USER_GPU_TARGETS)
set(CK_GPU_TARGETS ${GPU_TARGETS})
endif()
endif()
#make sure all the targets on the list are actually supported by the current compiler
rocm_check_target_ids(SUPPORTED_GPU_TARGETS
TARGETS ${CK_GPU_TARGETS})
message("Building CK for the following targets: ${SUPPORTED_GPU_TARGETS}")
if (GPU_TARGETS)
if (GPU_TARGETS MATCHES "gfx9")
add_definitions(-DCK_USE_XDL)
......@@ -557,8 +549,7 @@ ENDFOREACH()
add_custom_target(instances DEPENDS utility;${CK_DEVICE_INSTANCES} SOURCES ${INSTANCE_FILES})
add_subdirectory(library)
if(NOT DEFINED INSTANCES_ONLY)
if(NOT DEFINED PROFILER_ONLY)
if(NOT GPU_ARCHS AND USER_GPU_TARGETS)
rocm_package_setup_component(tests
LIBRARY_NAME composablekernel
PACKAGE_NAME tests # Prevent -static suffix on package name
......@@ -569,24 +560,18 @@ if(NOT DEFINED INSTANCES_ONLY)
PACKAGE_NAME examples
)
add_subdirectory(example)
add_subdirectory(test)
rocm_package_setup_component(profiler
LIBRARY_NAME composablekernel
PACKAGE_NAME ckprofiler
)
add_subdirectory(profiler)
else()
#When building PROFILER_ONLY, label the package with GPU_ARCH
rocm_package_setup_component(profiler
LIBRARY_NAME composablekernel
PACKAGE_NAME ckprofiler_${GPU_ARCH}
)
add_subdirectory(profiler)
endif()
if(BUILD_TESTING)
add_subdirectory(test)
endif()
endif()
if(NOT DEFINED PROFILER_ONLY AND (GPU_TARGETS MATCHES "gfx9" OR DEFINED INSTANCES_ONLY))
rocm_package_setup_component(profiler
LIBRARY_NAME composablekernel
PACKAGE_NAME ckprofiler
)
add_subdirectory(profiler)
if(CK_USE_CODEGEN AND (GPU_TARGETS MATCHES "gfx9" OR GPU_ARCHS))
add_subdirectory(codegen)
endif()
......
Jenkinsfile
View file @ 63b152d6
......@@ -100,7 +100,15 @@ def getDockerImage(Map conf=[:]){
dockerArgs = dockerArgs + " --no-cache "
}
echo "Docker Args: ${dockerArgs}"
def image = getDockerImageName()
def image
if ( params.BUILD_LEGACY_OS && conf.get("docker_name", "") != "" ){
image = conf.get("docker_name", "")
echo "Using legacy docker: ${image}"
}
else{
image = getDockerImageName()
echo "Using default docker: ${image}"
}
//Check if image exists
def retimage
try
......@@ -125,7 +133,9 @@ def buildDocker(install_prefix){
def image_name = getDockerImageName()
echo "Building Docker for ${image_name}"
def dockerArgs = "--build-arg BUILDKIT_INLINE_CACHE=1 --build-arg PREFIX=${install_prefix} --build-arg CK_SCCACHE='${env.CK_SCCACHE}' --build-arg compiler_version='${params.COMPILER_VERSION}' --build-arg compiler_commit='${params.COMPILER_COMMIT}' --build-arg ROCMVERSION='${params.ROCMVERSION}' --build-arg DISABLE_CACHE='git rev-parse ${params.COMPILER_VERSION}' "
if(params.COMPILER_VERSION == "amd-staging" || params.COMPILER_VERSION == "amd-mainline-open" || params.COMPILER_COMMIT != ""){
dockerArgs = dockerArgs + " --no-cache "
}
echo "Build Args: ${dockerArgs}"
try{
if(params.BUILD_DOCKER){
......@@ -259,6 +269,7 @@ def cmake_build(Map conf=[:]){
""")
sh cmd3
}
// reduce parallelism when compiling, clang uses too much memory
def nt = nthreads()
def cmd
......@@ -273,7 +284,7 @@ def cmake_build(Map conf=[:]){
}
else{
setup_cmd = conf.get("setup_cmd", "${cmake_envs} cmake ${setup_args} .. ")
build_cmd = conf.get("build_cmd", "${build_envs} dumb-init make -j${nt} ${config_targets}")
build_cmd = conf.get("build_cmd", "${build_envs} make -j${nt} ${config_targets}")
}
cmd = conf.get("cmd", """
${setup_cmd}
......@@ -292,8 +303,8 @@ def cmake_build(Map conf=[:]){
dir("build"){
//build CK
sh cmd
//run tests
if(!setup_args.contains("NO_CK_BUILD")){
//run tests except when NO_CK_BUILD or BUILD_LEGACY_OS are set
if(!setup_args.contains("NO_CK_BUILD") && !params.BUILD_LEGACY_OS){
if (setup_args.contains("gfx90a") && params.NINJA_BUILD_TRACE){
sh "/ninjatracing/ninjatracing .ninja_log > ck_build_trace.json"
archiveArtifacts "ck_build_trace.json"
......@@ -309,7 +320,7 @@ def cmake_build(Map conf=[:]){
if (package_build == true && (env.BRANCH_NAME == "develop" || env.BRANCH_NAME == "amd-master")) {
archiveArtifacts artifacts: "build/*.deb", allowEmptyArchive: true, fingerprint: true
}
if (params.RUN_CK_TILE_TESTS){
if (params.RUN_CK_TILE_FMHA_TESTS){
try{
archiveArtifacts "perf_fmha_fwd_*.log"
archiveArtifacts "perf_fmha_bwd_*.log"
......@@ -330,11 +341,19 @@ def buildHipClangJob(Map conf=[:]){
env.HSA_ENABLE_SDMA=0
checkout scm
def image = getDockerImageName()
def image
if ( params.BUILD_LEGACY_OS && conf.get("docker_name", "") != "" ){
image = conf.get("docker_name", "")
echo "Using legacy docker: ${image}"
}
else{
image = getDockerImageName()
echo "Using default docker: ${image}"
}
def prefixpath = conf.get("prefixpath", "/opt/rocm")
// Jenkins is complaining about the render group
def dockerOpts="--rm --device=/dev/kfd --device=/dev/dri --group-add video --group-add render --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
def dockerOpts="--device=/dev/kfd --device=/dev/dri --group-add video --group-add render --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
if (conf.get("enforce_xnack_on", false)) {
dockerOpts = dockerOpts + " --env HSA_XNACK=1 "
}
......@@ -352,7 +371,7 @@ def buildHipClangJob(Map conf=[:]){
def retimage
(retimage, image) = getDockerImage(conf)
gitStatusWrapper(credentialsId: "${status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
gitStatusWrapper(credentialsId: "${env.ck_git_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
withDockerContainer(image: image, args: dockerOpts + ' -v=/var/jenkins/:/var/jenkins') {
timeout(time: 48, unit: 'HOURS')
{
......@@ -393,7 +412,7 @@ def runCKProfiler(Map conf=[:]){
def prefixpath = conf.get("prefixpath", "/opt/rocm")
// Jenkins is complaining about the render group
def dockerOpts="--rm --device=/dev/kfd --device=/dev/dri --group-add video --group-add render --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
def dockerOpts="--device=/dev/kfd --device=/dev/dri --group-add video --group-add render --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
if (conf.get("enforce_xnack_on", false)) {
dockerOpts = dockerOpts + " --env HSA_XNACK=1 "
}
......@@ -407,7 +426,7 @@ def runCKProfiler(Map conf=[:]){
def variant = env.STAGE_NAME
def retimage
gitStatusWrapper(credentialsId: "${status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
gitStatusWrapper(credentialsId: "${env.ck_git_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
try {
(retimage, image) = getDockerImage(conf)
withDockerContainer(image: image, args: dockerOpts) {
......@@ -512,11 +531,20 @@ def Build_CK(Map conf=[:]){
env.DOCKER_BUILDKIT=1
checkout scm
def image = getDockerImageName()
def image
if ( params.BUILD_LEGACY_OS && conf.get("docker_name", "") != "" ){
image = conf.get("docker_name", "")
echo "Using legacy docker: ${image}"
}
else{
image = getDockerImageName()
echo "Using default docker: ${image}"
}
def prefixpath = conf.get("prefixpath", "/opt/rocm")
// Jenkins is complaining about the render group
def dockerOpts="--rm --device=/dev/kfd --device=/dev/dri --group-add video --group-add render --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
def dockerOpts="--device=/dev/kfd --device=/dev/dri --group-add video --group-add render --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
if (conf.get("enforce_xnack_on", false)) {
dockerOpts = dockerOpts + " --env HSA_XNACK=1 "
}
......@@ -524,6 +552,9 @@ def Build_CK(Map conf=[:]){
if (params.COMPILER_VERSION == "amd-staging" || params.COMPILER_VERSION == "amd-mainline-open" || params.COMPILER_COMMIT != ""){
dockerOpts = dockerOpts + " --env HIP_CLANG_PATH='/llvm-project/build/bin' "
}
if(params.BUILD_LEGACY_OS){
dockerOpts = dockerOpts + " --env LD_LIBRARY_PATH='/opt/Python-3.8.13/lib' "
}
def video_id = sh(returnStdout: true, script: 'getent group video | cut -d: -f3')
def render_id = sh(returnStdout: true, script: 'getent group render | cut -d: -f3')
dockerOpts = dockerOpts + " --group-add=${video_id} --group-add=${render_id} "
......@@ -532,7 +563,7 @@ def Build_CK(Map conf=[:]){
def variant = env.STAGE_NAME
def retimage
gitStatusWrapper(credentialsId: "${env.status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
gitStatusWrapper(credentialsId: "${env.ck_git_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
try {
(retimage, image) = getDockerImage(conf)
withDockerContainer(image: image, args: dockerOpts) {
......@@ -629,7 +660,7 @@ def process_results(Map conf=[:]){
def prefixpath = "/opt/rocm"
// Jenkins is complaining about the render group
def dockerOpts="--rm --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
def dockerOpts="--cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
if (conf.get("enforce_xnack_on", false)) {
dockerOpts = dockerOpts + " --env HSA_XNACK=1 "
}
......@@ -637,7 +668,7 @@ def process_results(Map conf=[:]){
def variant = env.STAGE_NAME
def retimage
gitStatusWrapper(credentialsId: "${env.status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
gitStatusWrapper(credentialsId: "${env.ck_git_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
try {
(retimage, image) = getDockerImage(conf)
}
......@@ -651,7 +682,7 @@ def process_results(Map conf=[:]){
timeout(time: 1, unit: 'HOURS'){
try{
dir("script"){
if (params.RUN_CK_TILE_TESTS){
if (params.RUN_CK_TILE_FMHA_TESTS){
try{
unstash "perf_fmha_fwd_gfx942.log"
unstash "perf_fmha_bwd_gfx942.log"
......@@ -704,10 +735,11 @@ def process_results(Map conf=[:]){
//launch develop branch daily at 23:00 UT in FULL_QA mode and at 19:00 UT with latest staging compiler version
CRON_SETTINGS = BRANCH_NAME == "develop" ? '''0 23 * * * % RUN_FULL_QA=true;ROCMVERSION=6.2;RUN_CK_TILE_FMHA_TESTS=true;RUN_CK_TILE_GEMM_TESTS=true
0 21 * * * % ROCMVERSION=6.2;hipTensor_test=true
0 21 * * * % ROCMVERSION=6.2;hipTensor_test=true;RUN_CODEGEN_TESTS=true
0 19 * * * % BUILD_DOCKER=true;DL_KERNELS=true;COMPILER_VERSION=amd-staging;BUILD_COMPILER=/llvm-project/build/bin/clang++;BUILD_GFX12=true;USE_SCCACHE=false;NINJA_BUILD_TRACE=true
0 17 * * * % BUILD_DOCKER=true;DL_KERNELS=true;COMPILER_VERSION=amd-mainline-open;BUILD_COMPILER=/llvm-project/build/bin/clang++;BUILD_GFX12=true;USE_SCCACHE=false;NINJA_BUILD_TRACE=true
0 15 * * * % BUILD_INSTANCES_ONLY=true;RUN_CODEGEN_TESTS=false;RUN_PERFORMANCE_TESTS=false;USE_SCCACHE=false''' : ""
0 15 * * * % BUILD_INSTANCES_ONLY=true;RUN_PERFORMANCE_TESTS=false;USE_SCCACHE=false
0 13 * * * % BUILD_LEGACY_OS=true''' : ""
pipeline {
agent none
......@@ -774,6 +806,10 @@ pipeline {
name: "RUN_GROUPED_CONV_LARGE_CASES_TESTS",
defaultValue: false,
description: "Run the grouped conv large cases tests (default: OFF)")
booleanParam(
name: "RUN_CODEGEN_TESTS",
defaultValue: false,
description: "Run codegen tests (default: OFF)")
booleanParam(
name: "RUN_CK_TILE_FMHA_TESTS",
defaultValue: false,
......@@ -794,6 +830,10 @@ pipeline {
name: "NINJA_BUILD_TRACE",
defaultValue: false,
description: "Generate a ninja build trace (default: OFF)")
booleanParam(
name: "BUILD_LEGACY_OS",
defaultValue: false,
description: "Try building CK with legacy OS dockers: RHEL8 and SLES15 (default: OFF)")
}
environment{
dbuser = "${dbuser}"
......@@ -802,7 +842,7 @@ pipeline {
dbsshport = "${dbsshport}"
dbsshuser = "${dbsshuser}"
dbsshpassword = "${dbsshpassword}"
status_wrapper_creds = "${status_wrapper_creds}"
ck_git_creds = "${ck_git_creds}"
gerrit_cred="${gerrit_cred}"
DOCKER_BUILDKIT = "1"
}
......@@ -890,7 +930,30 @@ pipeline {
execute_args = """ ../script/cmake-ck-dev.sh ../ gfx90a && \
make -j64 test_grouped_convnd_fwd_large_cases_xdl && \
./bin/test_grouped_convnd_fwd_large_cases_xdl"""
}
}
steps{
buildHipClangJobAndReboot(setup_args:setup_args, no_reboot:true, build_type: 'Release', execute_cmd: execute_args)
cleanWs()
}
}
}
}
stage("Run Codegen Tests")
{
parallel
{
stage("Run Codegen Tests on gfx90a")
{
when {
beforeAgent true
expression { params.RUN_CODEGEN_TESTS.toBoolean() }
}
agent{ label rocmnode("gfx90a")}
environment{
setup_args = "NO_CK_BUILD"
execute_args = """ CXX=/opt/rocm/llvm/bin/clang++ cmake ../codegen && \
make -j64 check"""
}
steps{
buildHipClangJobAndReboot(setup_args:setup_args, no_reboot:true, build_type: 'Release', execute_cmd: execute_args)
cleanWs()
......@@ -915,7 +978,7 @@ pipeline {
make -j64 tile_example_fmha_fwd tile_example_fmha_bwd && \
cd ../ &&
example/ck_tile/01_fmha/script/run_full_test.sh "CI_${params.COMPILER_VERSION}" "${env.BRANCH_NAME}" "${NODE_NAME}" gfx90a """
}
}
steps{
buildHipClangJobAndReboot(setup_args:setup_args, no_reboot:true, build_type: 'Release', execute_cmd: execute_args)
cleanWs()
......@@ -934,7 +997,7 @@ pipeline {
make -j64 tile_example_fmha_fwd tile_example_fmha_bwd && \
cd ../ &&
example/ck_tile/01_fmha/script/run_full_test.sh "CI_${params.COMPILER_VERSION}" "${env.BRANCH_NAME}" "${NODE_NAME}" gfx942 """
}
}
steps{
buildHipClangJobAndReboot(setup_args:setup_args, no_reboot:true, build_type: 'Release', execute_cmd: execute_args)
cleanWs()
......@@ -946,7 +1009,6 @@ pipeline {
{
parallel
{
stage("Run CK_TILE_GEMM Tests on gfx90a")
{
when {
......@@ -960,12 +1022,11 @@ pipeline {
make -j64 tile_example_gemm_basic && \
cd ../ &&
example/ck_tile/03_gemm/script/run_full_test.sh "CI_${params.COMPILER_VERSION}" "${env.BRANCH_NAME}" "${NODE_NAME}" gfx90a """
}
}
steps{
buildHipClangJobAndReboot(setup_args:setup_args, no_reboot:true, build_type: 'Release', execute_cmd: execute_args)
cleanWs()
}
}
stage("Run CK_TILE_GEMM Tests on gfx942")
{
......@@ -980,7 +1041,7 @@ pipeline {
make -j64 tile_example_gemm_basic && \
cd ../ &&
example/ck_tile/03_gemm/script/run_full_test.sh "CI_${params.COMPILER_VERSION}" "${env.BRANCH_NAME}" "${NODE_NAME}" gfx942 """
}
}
steps{
buildHipClangJobAndReboot(setup_args:setup_args, no_reboot:true, build_type: 'Release', execute_cmd: execute_args)
cleanWs()
......@@ -988,15 +1049,54 @@ pipeline {
}
}
}
stage("Build CK and run Tests")
{
parallel
{
stage("Build CK with RHEL8")
{
when {
beforeAgent true
expression { params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx90a") }
environment{
def docker_name = "${env.CK_DOCKERHUB_PRIVATE}:ck_rhel8_rocm6.3"
setup_args = """ -DGPU_TARGETS="gfx942" \
-DCMAKE_CXX_FLAGS=" -O3 " \
-DCK_USE_ALTERNATIVE_PYTHON=/opt/Python-3.8.13/bin/python3.8 """
execute_args = " "
}
steps{
Build_CK_and_Reboot(setup_args: setup_args, config_targets: " ", no_reboot:true, build_type: 'Release', docker_name: docker_name)
cleanWs()
}
}
stage("Build CK with SLES15")
{
when {
beforeAgent true
expression { params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx90a") }
environment{
def docker_name = "${env.CK_DOCKERHUB_PRIVATE}:ck_sles15_rocm6.3"
setup_args = """ -DGPU_TARGETS="gfx942" \
-DCMAKE_CXX_FLAGS=" -O3 " \
-DCK_USE_ALTERNATIVE_PYTHON=/opt/Python-3.8.13/bin/python3.8 """
execute_args = " "
}
steps{
Build_CK_and_Reboot(setup_args: setup_args, config_targets: " ", no_reboot:true, build_type: 'Release', docker_name: docker_name)
cleanWs()
}
}
stage("Build CK for all gfx9 targets")
{
when {
beforeAgent true
expression { params.RUN_FULL_QA.toBoolean() }
expression { params.RUN_FULL_QA.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx90a") }
environment{
......@@ -1018,7 +1118,7 @@ pipeline {
{
when {
beforeAgent true
expression { params.RUN_FULL_QA.toBoolean() }
expression { params.RUN_FULL_QA.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx942") }
environment{
......@@ -1038,7 +1138,7 @@ pipeline {
{
when {
beforeAgent true
expression { !params.RUN_FULL_QA.toBoolean() && !params.BUILD_INSTANCES_ONLY.toBoolean() }
expression { !params.RUN_FULL_QA.toBoolean() && !params.BUILD_INSTANCES_ONLY.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx90a") }
environment{
......@@ -1058,16 +1158,16 @@ pipeline {
{
when {
beforeAgent true
expression { params.BUILD_INSTANCES_ONLY.toBoolean() && !params.RUN_FULL_QA.toBoolean() }
expression { params.BUILD_INSTANCES_ONLY.toBoolean() && !params.RUN_FULL_QA.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx90a") }
environment{
execute_args = """ cmake -D CMAKE_PREFIX_PATH=/opt/rocm \
-D CMAKE_CXX_COMPILER="${build_compiler()}" \
-D CMAKE_BUILD_TYPE=Release \
-D INSTANCES_ONLY=ON \
-DCMAKE_CXX_FLAGS=" -O3 " .. && make -j64 """
}
-D GPU_ARCHS="gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102" \
-D CMAKE_CXX_FLAGS=" -O3 " .. && make -j64 """
}
steps{
buildHipClangJobAndReboot(setup_cmd: "", build_cmd: "", no_reboot:true, build_type: 'Release', execute_cmd: execute_args)
cleanWs()
......@@ -1077,7 +1177,7 @@ pipeline {
{
when {
beforeAgent true
expression { !params.RUN_FULL_QA.toBoolean() && !params.BUILD_INSTANCES_ONLY.toBoolean() }
expression { !params.RUN_FULL_QA.toBoolean() && !params.BUILD_INSTANCES_ONLY.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx1030") }
environment{
......@@ -1097,7 +1197,7 @@ pipeline {
{
when {
beforeAgent true
expression { !params.RUN_FULL_QA.toBoolean() && !params.BUILD_INSTANCES_ONLY.toBoolean() }
expression { !params.RUN_FULL_QA.toBoolean() && !params.BUILD_INSTANCES_ONLY.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx1101") }
environment{
......@@ -1117,7 +1217,7 @@ pipeline {
{
when {
beforeAgent true
expression { params.BUILD_GFX12.toBoolean() && !params.RUN_FULL_QA.toBoolean() && !params.BUILD_INSTANCES_ONLY.toBoolean() }
expression { params.BUILD_GFX12.toBoolean() && !params.RUN_FULL_QA.toBoolean() && !params.BUILD_INSTANCES_ONLY.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
agent{ label rocmnode("gfx1201") }
environment{
......@@ -1144,7 +1244,7 @@ pipeline {
{
when {
beforeAgent true
expression { params.RUN_PERFORMANCE_TESTS.toBoolean() }
expression { params.RUN_PERFORMANCE_TESTS.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
options { retry(1) }
agent{ label rocmnode("gfx90a")}
......@@ -1165,7 +1265,7 @@ pipeline {
stage("Process results"){
when {
beforeAgent true
expression { params.RUN_PERFORMANCE_TESTS.toBoolean() }
expression { params.RUN_PERFORMANCE_TESTS.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
}
agent { label 'mici' }
steps{
......
README.md
View file @ 63b152d6
......@@ -90,7 +90,13 @@ Docker images are available on [DockerHub](https://hub.docker.com/r/rocm/composa
```
If you don't set `GPU_TARGETS` on the cmake command line, CK is built for all GPU targets
supported by the current compiler (this may take a long time).
supported by the current compiler (this may take a long time).
Tests and examples will only get built if the GPU_TARGETS is set by the user on the cmake command line.
NOTE: If you try setting `GPU_TARGETS` to a list of architectures, the build will only work if the
architectures are similar, e.g., `gfx908;gfx90a`, or `gfx1100;gfx1101;gfx11012`. Otherwise, if you
want to build the library for a list of different architectures,
you should use the `GPU_ARCHS` build argument, for example `GPU_ARCHS=gfx908;gfx1030;gfx1100;gfx942`.
4. Build the entire CK library:
......@@ -137,10 +143,6 @@ crash. In such cases, you can reduce the number of threads to 32 by using `-j32`
Additional cmake flags can be used to significantly speed-up the build:
* `INSTANCES_ONLY` (default is OFF) must be set to ON in order to build only the instances and library
while skipping all tests, examples, and profiler. This is useful in cases when you plan to use CK as a
dependency and don't plan to run any examples or tests.
* `DTYPES` (default is not set) can be set to any subset of "fp64;fp32;fp16;fp8;bf16;int8" to build
instances of select data types only. The main default data types are fp32 and fp16; you can safely skip
other data types.
......
cmake/Embed.cmake
View file @ 63b152d6
......@@ -233,6 +233,8 @@ function(add_embed_library EMBED_NAME)
else()
target_sources(${EMBED_NAME} INTERFACE $<TARGET_OBJECTS:${INTERNAL_EMBED_LIB}>)
endif()
target_include_directories(${EMBED_NAME} INTERFACE "${EMBED_DIR}/include")
target_include_directories(${EMBED_NAME} INTERFACE
$<BUILD_INTERFACE:${EMBED_DIR}/include>
$<INSTALL_INTERFACE:include/ck>)
endfunction()
codegen/CMakeLists.txt
View file @ 63b152d6
cmake_minimum_required(VERSION 3.16)
project(composable_kernel_host)
set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
......@@ -5,49 +8,51 @@ set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
set(CK_ROOT ${CMAKE_CURRENT_SOURCE_DIR}/..)
add_compile_options(-std=c++17)
find_package(hip)
add_custom_target(codegen)
find_package(ROCM)
include(ROCMInstallTargets)
include(ROCMTest)
# add include directories
include_directories(BEFORE
${PROJECT_BINARY_DIR}/include
${PROJECT_SOURCE_DIR}/include
${PROJECT_SOURCE_DIR}/library/include
${HIP_INCLUDE_DIRS}
)
rocm_setup_version(VERSION 1.0)
list(APPEND CMAKE_MODULE_PATH ${CK_ROOT}/cmake)
include(Embed)
file(GLOB_RECURSE KERNEL_FILES CONFIGURE_DEPENDS
${CK_ROOT}/include/ck/*.hpp)
#printouts fot debug purposes
#message(STATUS "KERNEL_FILES: ${KERNEL_FILES}")
#message(STATUS "RELATIVE: ${CK_ROOT}/include")
${CK_ROOT}/include/ck/*.hpp)
# printouts fot debug purposes
# message(STATUS "KERNEL_FILES: ${KERNEL_FILES}")
# message(STATUS "RELATIVE: ${CK_ROOT}/include")
add_embed_library(ck_headers ${KERNEL_FILES} RELATIVE ${CK_ROOT}/include)
file(GLOB SOURCES CONFIGURE_DEPENDS src/*.cpp)
add_compile_options(-std=c++17)
##message(STATUS "SOURCE_FILES: ${SOURCES}")
file(GLOB SOURCES CONFIGURE_DEPENDS src/*.cpp)
# TODO: Use object library
add_library(ck_host STATIC ${SOURCES})
target_link_libraries(ck_host PRIVATE ck_headers)
set_target_properties(ck_host PROPERTIES
LINKER_LANGUAGE CXX
POSITION_INDEPENDENT_CODE ON)
set_target_properties(ck_host PROPERTIES
LINKER_LANGUAGE CXX
POSITION_INDEPENDENT_CODE ON)
target_include_directories(ck_host PUBLIC
$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>
)
# target_include_directories(ck_host PUBLIC
# $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>
# )
add_executable(ck-template-driver driver/main.cpp)
target_link_libraries(ck-template-driver ck_host)
rocm_install(
rocm_install_targets(
TARGETS ck_host ck_headers
EXPORT ck_hostTargets
EXPORT ck_host_targets
INCLUDE include
PRIVATE
)
rocm_install(DIRECTORY include/ck DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
rocm_export_targets(
EXPORT ck_host_targets
NAMESPACE composable_kernel::
)
if(BUILD_TESTING)
add_subdirectory(test)
endif()
add_subdirectory(test)
codegen/test/CMakeLists.txt
View file @ 63b152d6
list(APPEND CMAKE_PREFIX_PATH /opt/rocm)
add_subdirectory(rtc)
file(GLOB TEST_SRCS CONFIGURE_DEPENDS *.cpp)
if(NOT INSTANCES_ONLY)
foreach(TEST_SRC ${TEST_SRCS})
set_source_files_properties(${TEST_SRC} PROPERTIES LANGUAGE HIP)
get_filename_component(BASE_NAME ${TEST_SRC} NAME_WE)
add_executable(codegen_test_${BASE_NAME} ${TEST_SRC})
if(CK_USE_ALTERNATIVE_PYTHON)
target_link_options(codegen_test_${BASE_NAME} PRIVATE -lstdc++fs)
endif()
add_dependencies(codegen codegen_test_${BASE_NAME})
add_dependencies(tests codegen_test_${BASE_NAME})
add_dependencies(check codegen_test_${BASE_NAME})
add_test(NAME codegen_test_${BASE_NAME} COMMAND codegen_test_${BASE_NAME})
message("adding test codegen_test_${BASE_NAME}")
target_link_libraries(codegen_test_${BASE_NAME} ck_rtc ck_host)
target_include_directories(codegen_test_${BASE_NAME} PUBLIC ${CK_ROOT}/codegen/test/include)
# TODO: These tests need to be refactored to remove dependency on main ck
# headers and device compilation.
set(TESTS_REQUIRE_DEVICE_COMPILE
grouped_conv_fwd_multiple_d_v1
grouped_conv_fwd_multiple_d_v2
grouped_conv_fwd_multiple_d_v3
grouped_conv_fwd_multiple_d_v4
)
find_package(hip)
foreach(TEST_SRC ${TEST_SRCS})
get_filename_component(BASE_NAME ${TEST_SRC} NAME_WE)
rocm_add_test_executable(codegen_test_${BASE_NAME} ${TEST_SRC})
target_link_libraries(codegen_test_${BASE_NAME} ck_rtc ck_host)
target_include_directories(codegen_test_${BASE_NAME} PUBLIC include)
if(BASE_NAME IN_LIST TESTS_REQUIRE_DEVICE_COMPILE)
target_link_libraries(codegen_test_${BASE_NAME} hip::device)
target_include_directories(codegen_test_${BASE_NAME} PUBLIC ${CK_ROOT}/include)
target_include_directories(codegen_test_${BASE_NAME} PUBLIC ${CK_ROOT}/library/include)
endforeach()
endif()
endif()
endforeach()
codegen/test/rtc/CMakeLists.txt
View file @ 63b152d6
find_package(hip)
file(GLOB RTC_SOURCES CONFIGURE_DEPENDS src/*.cpp)
add_library(ck_rtc ${RTC_SOURCES})
target_include_directories(ck_rtc PUBLIC include)
target_link_libraries(ck_rtc PUBLIC hip::host)
target_link_libraries(ck_rtc PUBLIC -lstdc++fs)
codegen/test/rtc/include/rtc/compile_kernel.hpp
View file @ 63b152d6
......@@ -2,14 +2,14 @@
#define GUARD_HOST_TEST_RTC_INCLUDE_RTC_COMPILE_KERNEL
#include <rtc/kernel.hpp>
#include <ck/filesystem.hpp>
#include <rtc/filesystem.hpp>
#include <string>
namespace rtc {
struct src_file
{
CK::fs::path path;
fs::path path;
std::string_view content;
};
......
codegen/test/rtc/include/rtc/filesystem.hpp 0 → 100644
View file @ 63b152d6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#ifndef GUARD_TEST_HOST_RTC_FILESYSTEM_HPP
#define GUARD_TEST_HOST_RTC_FILESYSTEM_HPP
#include <string>
#include <string_view>
// clang-format off
#if defined(CPPCHECK)
#define RTC_HAS_FILESYSTEM 1
#define RTC_HAS_FILESYSTEM_TS 1
#elif defined(_WIN32)
#if _MSC_VER >= 1920
#define RTC_HAS_FILESYSTEM 1
#define RTC_HAS_FILESYSTEM_TS 0
#elif _MSC_VER >= 1900
#define RTC_HAS_FILESYSTEM 0
#define RTC_HAS_FILESYSTEM_TS 1
#else
#define RTC_HAS_FILESYSTEM 0
#define RTC_HAS_FILESYSTEM_TS 0
#endif
#elif defined(__has_include)
#if __has_include(<filesystem>) && __cplusplus >= 201703L
#define RTC_HAS_FILESYSTEM 1
#else
#define RTC_HAS_FILESYSTEM 0
#endif
#if __has_include(<experimental/filesystem>) && __cplusplus >= 201103L
#define RTC_HAS_FILESYSTEM_TS 1
#else
#define RTC_HAS_FILESYSTEM_TS 0
#endif
#else
#define RTC_HAS_FILESYSTEM 0
#define RTC_HAS_FILESYSTEM_TS 0
#endif
// clang-format on
#if RTC_HAS_FILESYSTEM
#include <filesystem>
#elif RTC_HAS_FILESYSTEM_TS
#include <experimental/filesystem>
#else
#error "No filesystem include available"
#endif
namespace rtc {
#if RTC_HAS_FILESYSTEM
namespace fs = ::std::filesystem;
#elif RTC_HAS_FILESYSTEM_TS
namespace fs = ::std::experimental::filesystem;
#endif
} // namespace rtc
#endif // GUARD_RTC_FILESYSTEM_HPP_
codegen/test/rtc/include/rtc/tmp_dir.hpp
View file @ 63b152d6
......@@ -2,13 +2,13 @@
#define GUARD_HOST_TEST_RTC_INCLUDE_RTC_TMP_DIR
#include <string>
#include <ck/filesystem.hpp>
#include <rtc/filesystem.hpp>
namespace rtc {
struct tmp_dir
{
CK::fs::path path;
fs::path path;
tmp_dir(const std::string& prefix = "");
void execute(const std::string& cmd) const;
......
codegen/test/rtc/src/compile_kernel.cpp
View file @ 63b152d6
#include "rtc/hip.hpp"
#include <rtc/hip.hpp>
#include <rtc/compile_kernel.hpp>
#include <rtc/tmp_dir.hpp>
#include <stdexcept>
......@@ -70,9 +70,9 @@ kernel compile_kernel(const std::vector<src_file>& srcs, compile_options options
for(const auto& src : srcs)
{
CK::fs::path full_path = td.path / src.path;
CK::fs::path parent_path = full_path.parent_path();
CK::fs::create_directories(parent_path);
fs::path full_path = td.path / src.path;
fs::path parent_path = full_path.parent_path();
fs::create_directories(parent_path);
write_string(full_path.string(), src.content);
if(src.path.extension().string() == ".cpp")
{
......@@ -86,7 +86,7 @@ kernel compile_kernel(const std::vector<src_file>& srcs, compile_options options
td.execute(compiler() + options.flags);
auto out_path = td.path / out;
if(not CK::fs::exists(out_path))
if(not fs::exists(out_path))
throw std::runtime_error("Output file missing: " + out);
auto obj = read_buffer(out_path.string());
......
codegen/test/rtc/src/tmp_dir.cpp
View file @ 63b152d6
......@@ -31,10 +31,10 @@ std::string unique_string(const std::string& prefix)
}
tmp_dir::tmp_dir(const std::string& prefix)
: path(CK::fs::temp_directory_path() /
: path(fs::temp_directory_path() /
unique_string(prefix.empty() ? "ck-rtc" : "ck-rtc-" + prefix))
{
CK::fs::create_directories(this->path);
fs::create_directories(this->path);
}
void tmp_dir::execute(const std::string& cmd) const
......@@ -43,6 +43,6 @@ void tmp_dir::execute(const std::string& cmd) const
std::system(s.c_str());
}
tmp_dir::~tmp_dir() { CK::fs::remove_all(this->path); }
tmp_dir::~tmp_dir() { fs::remove_all(this->path); }
} // namespace rtc
docs/reference/API_Reference_Guide.rst
View file @ 63b152d6
......@@ -12,12 +12,6 @@ API reference guide
This document contains details of the APIs for the Composable Kernel (CK) library and introduces
some of the key design principles that are used to write new classes that extend CK functionality.
=================
Using CK API
=================
This section describes how to use the CK library API.
=================
CK Datatypes
=================
......
docs/sphinx/requirements.in
View file @ 63b152d6
rocm-docs-core==1.8.1
rocm-docs-core==1.8.2
sphinxcontrib-bibtex==2.6.3
docs/sphinx/requirements.txt
View file @ 63b152d6
......@@ -103,7 +103,7 @@ requests==2.32.3
# via
# pygithub
# sphinx
rocm-docs-core==1.8.1
rocm-docs-core==1.8.2
# via -r requirements.in
six==1.16.0
# via pybtex
......
example/01_gemm/common.hpp
View file @ 63b152d6
......@@ -21,6 +21,7 @@
#include "ck/library/utility/host_tensor_generator.hpp"
#include "ck/library/utility/literals.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
#include "ck/library/reference_tensor_operation/gpu/reference_gemm.hpp"
struct ProblemSize final
{
......@@ -28,9 +29,9 @@ struct ProblemSize final
ck::index_t N = 4096;
ck::index_t K = 4096;
ck::index_t StrideA = 4096;
ck::index_t StrideB = 4096;
ck::index_t StrideC = 4096;
ck::index_t StrideA = 0;
ck::index_t StrideB = 0;
ck::index_t StrideC = 0;
};
struct ProblemSizeStreamK final
......@@ -39,9 +40,9 @@ struct ProblemSizeStreamK final
ck::index_t N = 4096;
ck::index_t K = 4096;
ck::index_t StrideA = 4096;
ck::index_t StrideB = 4096;
ck::index_t StrideC = 4096;
ck::index_t StrideA = 0;
ck::index_t StrideB = 0;
ck::index_t StrideC = 0;
ck::index_t NumSKBlocks = -1;
};
......@@ -51,9 +52,9 @@ struct ProblemSizeStreamK_universal final
ck::index_t N = 4096;
ck::index_t K = 4096;
ck::index_t StrideA = 4096;
ck::index_t StrideB = 4096;
ck::index_t StrideC = 4096;
ck::index_t StrideA = 0;
ck::index_t StrideB = 0;
ck::index_t StrideC = 0;
ck::index_t Grid_size = -1; // defaults to max occupancy
ck::index_t Streamk_sel = 1; // defaults to 1-tile SK
......@@ -65,9 +66,9 @@ struct ProblemSizeSplitK final
ck::index_t N = 4096;
ck::index_t K = 4096;
ck::index_t StrideA = 4096;
ck::index_t StrideB = 4096;
ck::index_t StrideC = 4096;
ck::index_t StrideA = 0;
ck::index_t StrideB = 0;
ck::index_t StrideC = 0;
ck::index_t KBatch = 1;
};
......@@ -125,7 +126,7 @@ bool parse_cmd_args<ProblemSize>(int argc,
}
else
{
std::cerr << "arg1: verification (0=no, 1=yes)" << std::endl
std::cerr << "arg1: verification (0=no, 1=CPU and GPU)" << std::endl
<< "arg2: initialization (0=no init, 1=integer value, 2=decimal value)"
<< std::endl
<< "arg3: time kernel (0=no, 1=yes)" << std::endl
......@@ -175,7 +176,7 @@ bool parse_cmd_args<ProblemSizeStreamK_universal>(int argc,
else
{
std::cerr
<< "arg1: verification (0=no, 1=yes)" << std::endl
<< "arg1: verification (0=no, 1=CPU and GPU)" << std::endl
<< "arg2: initialization (0=no init, 1=integer value, 2=decimal value)" << std::endl
<< "arg3: time kernel (0=no, 1=yes)" << std::endl
<< "arg4 to 9: M (256x), N(128x), K(32x), StrideA, StrideB, StrideC" << std::endl
......@@ -224,7 +225,7 @@ bool parse_cmd_args<ProblemSizeStreamK>(int argc,
}
else
{
std::cerr << "arg1: verification (0=no, 1=yes)" << std::endl
std::cerr << "arg1: verification (0=no, 1=CPU and GPU)" << std::endl
<< "arg2: initialization (0=no init, 1=integer value, 2=decimal value)"
<< std::endl
<< "arg3: time kernel (0=no, 1=yes)" << std::endl
......@@ -274,7 +275,7 @@ bool parse_cmd_args<ProblemSizeSplitK>(int argc,
}
else
{
std::cerr << "arg1: verification (0=no, 1=yes)" << std::endl
std::cerr << "arg1: verification (0=no, 1=CPU and GPU)" << std::endl
<< "arg2: initialization (0=no init, 1=integer value, 2=decimal value)"
<< std::endl
<< "arg3: time kernel (0=no, 1=yes)" << std::endl
......
example/01_gemm/gemm_dl_fp16.cpp
View file @ 63b152d6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"
......@@ -32,6 +32,17 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmDl
using ReferenceGemmInstance = ck::tensor_operation::host::
ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;
using ReferenceGemmInstanceGPU = ck::tensor_operation::device::ReferenceGemm<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AccDataType,
AElementOp,
BElementOp,
CElementOp>;
#include "run_gemm_example.inc"
int main(int argc, char* argv[]) { return !run_gemm_example(argc, argv); }
example/01_gemm/gemm_dl_fp32.cpp
View file @ 63b152d6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"
......@@ -32,6 +32,17 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmDl
using ReferenceGemmInstance = ck::tensor_operation::host::
ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;
using ReferenceGemmInstanceGPU = ck::tensor_operation::device::ReferenceGemm<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AccDataType,
AElementOp,
BElementOp,
CElementOp>;
#include "run_gemm_example.inc"
int main(int argc, char* argv[]) { return !run_gemm_example(argc, argv); }
example/01_gemm/gemm_dl_int8.cpp
View file @ 63b152d6
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"
......@@ -32,6 +32,17 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmDl
using ReferenceGemmInstance = ck::tensor_operation::host::
ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;
using ReferenceGemmInstanceGPU = ck::tensor_operation::device::ReferenceGemm<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AccDataType,
AElementOp,
BElementOp,
CElementOp>;
#include "run_gemm_example.inc"
int main(int argc, char* argv[]) { return !run_gemm_example(argc, argv); }
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment