Skip to content

GitLab

Commit 7420d70b authored by illsilin's avatar illsilin
Browse files

turn on full qa only on gfx90a, use int initialization

parent 85978e02
Hide whitespace changes
Inline Side-by-side
Showing with 82 additions and 76 deletions
Jenkinsfile
View file @ 7420d70b
......@@ -268,7 +268,7 @@ def runCKProfiler(Map conf=[:]){
archiveArtifacts "perf_gemm_${gpu_arch}.log"
archiveArtifacts "perf_resnet50_N256_${gpu_arch}.log"
archiveArtifacts "perf_resnet50_N4_${gpu_arch}.log"
archiveArtifacts "perf_bathced_gemm_${gpu_arch}.log"
archiveArtifacts "perf_batched_gemm_${gpu_arch}.log"
archiveArtifacts "perf_grouped_gemm_${gpu_arch}.log"
archiveArtifacts "perf_fwd_conv_${gpu_arch}.log"
archiveArtifacts "perf_bwd_conv_${gpu_arch}.log"
......@@ -278,7 +278,7 @@ def runCKProfiler(Map conf=[:]){
stash name: "perf_gemm_${gpu_arch}.log"
stash name: "perf_resnet50_N256_${gpu_arch}.log"
stash name: "perf_resnet50_N4_${gpu_arch}.log"
stash name: "perf_bathced_gemm_${gpu_arch}.log"
stash name: "perf_batched_gemm_${gpu_arch}.log"
stash name: "perf_grouped_gemm_${gpu_arch}.log"
stash name: "perf_fwd_conv_${gpu_arch}.log"
stash name: "perf_bwd_conv_${gpu_arch}.log"
......@@ -362,7 +362,7 @@ def process_results(Map conf=[:]){
unstash "perf_gemm_${gpu_arch}.log"
unstash "perf_resnet50_N256_${gpu_arch}.log"
unstash "perf_resnet50_N4_${gpu_arch}.log"
unstash "perf_bathced_gemm_${gpu_arch}.log"
unstash "perf_batched_gemm_${gpu_arch}.log"
unstash "perf_grouped_gemm_${gpu_arch}.log"
unstash "perf_fwd_conv_${gpu_arch}.log"
unstash "perf_bwd_conv_${gpu_arch}.log"
......@@ -389,13 +389,13 @@ def process_results(Map conf=[:]){
}
//launch develop branch daily at 23:00 in FULL_QA mode
//CRON_SETTINGS = BRANCH_NAME == "develop" ? '''0 23 * * * % RUN_FULL_QA=true;USE_9110=true''' : ""
CRON_SETTINGS = BRANCH_NAME == "lwpck-367" ? '''0 23 * * * % RUN_FULL_QA=true;USE_9110=true''' : ""
pipeline {
agent none
//triggers {
// cron(CRON_SETTINGS)
//}
triggers {
parameterizedCron(CRON_SETTINGS)
}
options {
parallelsAlwaysFailFast()
}
......@@ -451,89 +451,95 @@ pipeline {
}
}
}
stage("Tests")
stage("Performance Tests")
{
parallel
{
stage("Run Tests: gfx908")
stage("Run ckProfiler: gfx908")
{
when {
expression { return params.RUN_FULL_QA != 'true'; }
}
agent{ label rocmnode("gfx908")}
environment{
setup_args = """ -D CMAKE_CXX_FLAGS=" --offload-arch=gfx908 -O3 " -DBUILD_DEV=On """
}
setup_args = """ -D CMAKE_CXX_FLAGS="--offload-arch=gfx908 -O3 " -DBUILD_DEV=On """
}
steps{
buildHipClangJobAndReboot(setup_args:setup_args, config_targets: "check", no_reboot:true, build_type: 'Release', gpu_arch: "gfx908")
runPerfTest(setup_args:setup_args, config_targets: "ckProfiler", no_reboot:true, build_type: 'Release', gpu_arch: "gfx908")
}
}
stage("Run Tests: gfx90a")
stage("Run ckProfiler: gfx90a")
{
agent{ label rocmnode("gfx90a")}
environment{
setup_args = """ -D CMAKE_CXX_FLAGS="--offload-arch=gfx90a -O3 " -DBUILD_DEV=On """
}
}
steps{
buildHipClangJobAndReboot(setup_args:setup_args, config_targets: "check", no_reboot:true, build_type: 'Release', gpu_arch: "gfx90a")
runPerfTest(setup_args:setup_args, config_targets: "ckProfiler", no_reboot:true, build_type: 'Release', gpu_arch: "gfx90a")
}
}
}
}
stage("Client App")
stage("Process Performance Test Results")
{
parallel
{
stage("Run Client App")
{
agent{ label rocmnode("gfx908")}
environment{
setup_args = """ -D -DBUILD_DEV=Off -DCMAKE_INSTALL_PREFIX=../install CMAKE_CXX_FLAGS="--offload-arch=gfx908 -O3 " """
execute_args = """ cd ../client_example && rm -rf build && mkdir build && cd build && cmake -DCMAKE_PREFIX_PATH="${env.WORKSPACE}/install;/opt/rocm" -DCMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc .. && make -j """
stage("Process results for gfx908"){
when {
expression { return params.RUN_FULL_QA != 'true'; }
}
agent { label 'mici' }
steps{
buildHipClangJobAndReboot(setup_args: setup_args, config_targets: "install", no_reboot:true, build_type: 'Release', execute_cmd: execute_args, prefixpath: '/usr/local')
process_results(gpu_arch: "gfx908")
}
}
stage("Process results for gfx90a"){
agent { label 'mici' }
steps{
process_results(gpu_arch: "gfx90a")
}
}
}
}
stage("Performance Tests")
stage("Tests")
{
parallel
{
stage("Run ckProfiler: gfx908")
stage("Run Tests: gfx908")
{
agent{ label rocmnode("gfx908")}
environment{
setup_args = """ -D CMAKE_CXX_FLAGS="--offload-arch=gfx908 -O3 " -DBUILD_DEV=On """
}
setup_args = """ -D CMAKE_CXX_FLAGS=" --offload-arch=gfx908 -O3 " -DBUILD_DEV=On """
}
steps{
runPerfTest(setup_args:setup_args, config_targets: "ckProfiler", no_reboot:true, build_type: 'Release', gpu_arch: "gfx908")
buildHipClangJobAndReboot(setup_args:setup_args, config_targets: "check", no_reboot:true, build_type: 'Release', gpu_arch: "gfx908")
}
}
stage("Run ckProfiler: gfx90a")
stage("Run Tests: gfx90a")
{
agent{ label rocmnode("gfx90a")}
environment{
setup_args = """ -D CMAKE_CXX_FLAGS="--offload-arch=gfx90a -O3 " -DBUILD_DEV=On """
}
}
steps{
runPerfTest(setup_args:setup_args, config_targets: "ckProfiler", no_reboot:true, build_type: 'Release', gpu_arch: "gfx90a")
buildHipClangJobAndReboot(setup_args:setup_args, config_targets: "check", no_reboot:true, build_type: 'Release', gpu_arch: "gfx90a")
}
}
}
}
stage("Process Performance Test Results")
stage("Client App")
{
parallel
{
stage("Process results for gfx908"){
agent { label 'mici' }
steps{
process_results(gpu_arch: "gfx908")
stage("Run Client App")
{
agent{ label rocmnode("gfx908")}
environment{
setup_args = """ -D -DBUILD_DEV=Off -DCMAKE_INSTALL_PREFIX=../install CMAKE_CXX_FLAGS="--offload-arch=gfx908 -O3 " """
execute_args = """ cd ../client_example && rm -rf build && mkdir build && cd build && cmake -DCMAKE_PREFIX_PATH="${env.WORKSPACE}/install;/opt/rocm" -DCMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc .. && make -j """
}
}
stage("Process results for gfx90a"){
agent { label 'mici' }
steps{
process_results(gpu_arch: "gfx90a")
buildHipClangJobAndReboot(setup_args: setup_args, config_targets: "install", no_reboot:true, build_type: 'Release', execute_cmd: execute_args, prefixpath: '/usr/local')
}
}
}
......
script/run_full_performance_tests.sh
View file @ 7420d70b
......@@ -60,62 +60,62 @@ print_log_header $gemm_log $env_type $branch $host_name
#run resnet50 tests
export resnet256_log="perf_resnet50_N256_${gpu_arch}.log"
print_log_header $resnet256_log $env_type $branch $host_name
./profile_resnet50.sh conv_fwd_bias_relu 1 1 1 1 $verify 2 0 1 256 | tee -a $resnet256_log
./profile_resnet50.sh conv_fwd_bias_relu 1 1 1 1 $verify 1 0 1 256 | tee -a $resnet256_log
export resnet4_log="perf_resnet50_N4_${gpu_arch}.log"
print_log_header $resnet4_log $env_type $branch $host_name
./profile_resnet50.sh conv_fwd_bias_relu 1 1 1 1 $verify 2 0 1 4 | tee -a $resnet4_log
./profile_resnet50.sh conv_fwd_bias_relu 1 1 1 1 $verify 1 0 1 4 | tee -a $resnet4_log
#run batched_gemm tests
export batched_gemm_log="perf_batched_gemm_${gpu_arch}.log"
print_log_header $batched_gemm_log $env_type $branch $host_name
./profile_batched_gemm.sh batched_gemm 0 0 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 0 1 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 0 2 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 0 3 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 1 0 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 1 1 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 1 2 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 1 3 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 2 0 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 2 1 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 2 2 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 2 3 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 3 0 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 3 1 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 3 2 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 3 3 $verify 2 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 0 0 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 0 1 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 0 2 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 0 3 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 1 0 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 1 1 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 1 2 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 1 3 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 2 0 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 2 1 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 2 2 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 2 3 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 3 0 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 3 1 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 3 2 $verify 1 0 5 | tee -a $batched_gemm_log
./profile_batched_gemm.sh batched_gemm 3 3 $verify 1 0 5 | tee -a $batched_gemm_log
#run grouped_gemm tests
export grouped_gemm_log="perf_grouped_gemm_${gpu_arch}.log"
print_log_header $grouped_gemm_log $env_type $branch $host_name
./profile_grouped_gemm.sh grouped_gemm 1 0 $verify 2 0 5 | tee -a $grouped_gemm_log
./profile_grouped_gemm.sh grouped_gemm 1 1 $verify 2 0 5 | tee -a $grouped_gemm_log
./profile_grouped_gemm.sh grouped_gemm 1 2 $verify 2 0 5 | tee -a $grouped_gemm_log
./profile_grouped_gemm.sh grouped_gemm 1 3 $verify 2 0 5 | tee -a $grouped_gemm_log
./profile_grouped_gemm.sh grouped_gemm 1 0 $verify 1 0 5 | tee -a $grouped_gemm_log
./profile_grouped_gemm.sh grouped_gemm 1 1 $verify 1 0 5 | tee -a $grouped_gemm_log
./profile_grouped_gemm.sh grouped_gemm 1 2 $verify 1 0 5 | tee -a $grouped_gemm_log
./profile_grouped_gemm.sh grouped_gemm 1 3 $verify 1 0 5 | tee -a $grouped_gemm_log
#run fwd_conv tests
export fwd_conv_log="perf_fwd_conv_${gpu_arch}.log"
print_log_header $fwd_conv_log $env_type $branch $host_name
./profile_conv.sh conv_fwd 0 1 $verify 2 0 5 2 256 | tee -a $fwd_conv_log
./profile_conv.sh conv_fwd 1 1 $verify 2 0 5 2 256 | tee -a $fwd_conv_log
./profile_conv.sh conv_fwd 2 1 $verify 2 0 5 2 256 | tee -a $fwd_conv_log
./profile_conv.sh conv_fwd 3 1 $verify 2 0 5 2 256 | tee -a $fwd_conv_log
./profile_conv.sh conv_fwd 0 1 $verify 1 0 5 2 256 | tee -a $fwd_conv_log
./profile_conv.sh conv_fwd 1 1 $verify 1 0 5 2 256 | tee -a $fwd_conv_log
./profile_conv.sh conv_fwd 2 1 $verify 1 0 5 2 256 | tee -a $fwd_conv_log
./profile_conv.sh conv_fwd 3 1 $verify 1 0 5 2 256 | tee -a $fwd_conv_log
#run bwd_conv tests
export bwd_conv_log="perf_bwd_conv_${gpu_arch}.log"
print_log_header $bwd_conv_log $env_type $branch $host_name
./profile_conv.sh conv2d_bwd_data 0 1 1 1 $verify 2 0 5 128 | tee -a $bwd_conv_log
./profile_conv.sh conv2d_bwd_data 1 1 1 1 $verify 2 0 5 128 | tee -a $bwd_conv_log
./profile_conv.sh conv2d_bwd_data 2 1 1 1 $verify 2 0 5 128 | tee -a $bwd_conv_log
./profile_conv.sh conv2d_bwd_data 3 1 1 1 $verify 2 0 5 128 | tee -a $bwd_conv_log
./profile_conv.sh conv2d_bwd_data 0 1 1 1 $verify 1 0 5 128 | tee -a $bwd_conv_log
./profile_conv.sh conv2d_bwd_data 1 1 1 1 $verify 1 0 5 128 | tee -a $bwd_conv_log
./profile_conv.sh conv2d_bwd_data 2 1 1 1 $verify 1 0 5 128 | tee -a $bwd_conv_log
./profile_conv.sh conv2d_bwd_data 3 1 1 1 $verify 1 0 5 128 | tee -a $bwd_conv_log
#run fusion tests
export fusion_log="perf_fusion_${gpu_arch}.log"
print_log_header $fusion_log $env_type $branch $host_name
./profile_gemm_bilinear.sh gemm_bilinear 1 0 $verify 2 0 1 | tee -a $fusion_log
./profile_gemm_bilinear.sh gemm_bilinear 1 1 $verify 2 0 1 | tee -a $fusion_log
./profile_gemm_bilinear.sh gemm_bilinear 1 2 $verify 2 0 1 | tee -a $fusion_log
./profile_gemm_bilinear.sh gemm_bilinear 1 3 $verify 2 0 1 | tee -a $fusion_log
./profile_gemm_bilinear.sh gemm_bilinear 1 0 $verify 1 0 1 | tee -a $fusion_log
./profile_gemm_bilinear.sh gemm_bilinear 1 1 $verify 1 0 1 | tee -a $fusion_log
./profile_gemm_bilinear.sh gemm_bilinear 1 2 $verify 1 0 1 | tee -a $fusion_log
./profile_gemm_bilinear.sh gemm_bilinear 1 3 $verify 1 0 1 | tee -a $fusion_log
#run reduction tests
export reduction_log="perf_reduction_${gpu_arch}.log"
......
script/run_performance_tests.sh
View file @ 7420d70b
......@@ -56,7 +56,7 @@ print_log_header $gemm_log $env_type $branch $host_name
#run resnet50 test
export resnet256_log="perf_resnet50_N256_${gpu_arch}.log"
print_log_header $resnet256_log $env_type $branch $host_name
./profile_resnet50.sh conv_fwd_bias_relu 1 1 1 1 $verify 2 0 1 256 | tee -a $resnet256_log
./profile_resnet50.sh conv_fwd_bias_relu 1 1 1 1 $verify 1 0 1 256 | tee -a $resnet256_log
export resnet4_log="perf_resnet50_N4_${gpu_arch}.log"
print_log_header $resnet4_log $env_type $branch $host_name
./profile_resnet50.sh conv_fwd_bias_relu 1 1 1 1 $verify 2 0 1 4 | tee -a $resnet4_log
./profile_resnet50.sh conv_fwd_bias_relu 1 1 1 1 $verify 1 0 1 4 | tee -a $resnet4_log
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