Unverified Commit bdcd0374 authored by Illia Silin's avatar Illia Silin Committed by GitHub
Browse files

Re-enable the performance tracking in CI. (#1203)

* test CK with rocm6.1 RC2

* add docker credentials for pull

* update the performance db name

* use environment variable for db name

* add rocm-llvm-dev package to ck docker

* turn off verification for daily performance runs

* do not stash ckProfiler on MI300 node

* add processing of mixed gemms to qa, fix parsing of splitk gemm logs

* fix the splitk gemm log file name

* turn the timing on for splitk gemm performance
parent e626d520
......@@ -16,17 +16,17 @@ RUN apt-get install -y --allow-unauthenticated apt-utils wget gnupg2 curl
ENV APT_KEY_DONT_WARN_ON_DANGEROUS_USAGE=DontWarn
RUN curl -fsSL https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor -o /etc/apt/trusted.gpg.d/rocm-keyring.gpg
RUN if [ "$ROCMVERSION" != "6.0.1" ]; then \
RUN if [ "$ROCMVERSION" != "6.1" ]; then \
sh -c "wget https://repo.radeon.com/amdgpu-install/6.0/ubuntu/focal/amdgpu-install_6.0.60000-1_all.deb --no-check-certificate" && \
apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated ./amdgpu-install_6.0.60000-1_all.deb && \
wget -qO - http://repo.radeon.com/rocm/rocm.gpg.key | apt-key add - && \
sh -c "echo deb [arch=amd64 signed-by=/etc/apt/trusted.gpg.d/rocm-keyring.gpg] $DEB_ROCM_REPO focal main > /etc/apt/sources.list.d/rocm.list" && \
sh -c 'echo deb [arch=amd64 signed-by=/etc/apt/trusted.gpg.d/rocm-keyring.gpg] https://repo.radeon.com/amdgpu/$ROCMVERSION/ubuntu focal main > /etc/apt/sources.list.d/amdgpu.list'; \
elif [ "$ROCMVERSION" = "6.0.1" ] && [ "$compiler_version" = "rc1" ]; then \
sh -c "wget http://artifactory-cdn.amd.com/artifactory/list/amdgpu-deb/amdgpu-install-internal_6.0-20.04-1_all.deb --no-check-certificate" && \
apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install dialog && DEBIAN_FRONTEND=noninteractive apt-get install ./amdgpu-install-internal_6.0-20.04-1_all.deb && \
sh -c 'echo deb [arch=amd64 trusted=yes] http://compute-artifactory.amd.com/artifactory/list/rocm-release-archive-20.04-deb/ 6.0.1 rel-95 > /etc/apt/sources.list.d/rocm-build.list' && \
amdgpu-repo --amdgpu-build=1704947; \
elif [ "$ROCMVERSION" = "6.1" ] && [ "$compiler_version" = "rc2" ]; then \
sh -c "wget http://artifactory-cdn.amd.com/artifactory/list/amdgpu-deb/amdgpu-install-internal_6.1-20.04-1_all.deb --no-check-certificate" && \
apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install dialog && DEBIAN_FRONTEND=noninteractive apt-get install ./amdgpu-install-internal_6.1-20.04-1_all.deb && \
sh -c 'echo deb [arch=amd64 trusted=yes] http://compute-artifactory.amd.com/artifactory/list/rocm-release-archive-20.04-deb/ 6.1 rel-48 > /etc/apt/sources.list.d/rocm-build.list' && \
amdgpu-repo --amdgpu-build=1736298; \
fi
RUN sh -c "echo deb http://mirrors.kernel.org/ubuntu focal main universe | tee -a /etc/apt/sources.list"
......@@ -41,6 +41,7 @@ chmod +x ${SCCACHE_INSTALL_LOCATION}/sccache
ENV PATH=$PATH:${SCCACHE_INSTALL_LOCATION}
# Install dependencies
# hipTensor requires rocm-llvm-dev for rocm versions > 6.0.1
RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated \
build-essential \
cmake \
......@@ -60,6 +61,7 @@ RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-
python3-dev \
python3-pip \
redis \
rocm-llvm-dev \
sshpass \
stunnel \
software-properties-common \
......@@ -73,6 +75,9 @@ RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-
apt-get clean && \
rm -rf /var/lib/apt/lists/*
# Update the cmake to version 3.27.5
RUN pip install --upgrade cmake==3.27.5
#Install latest ccache
RUN git clone https://github.com/ccache/ccache.git && \
cd ccache && mkdir build && cd build && cmake .. && make install
......@@ -82,8 +87,6 @@ RUN wget -qO /usr/local/bin/ninja.gz https://github.com/ninja-build/ninja/releas
RUN gunzip /usr/local/bin/ninja.gz
RUN chmod a+x /usr/local/bin/ninja
RUN git clone https://github.com/nico/ninjatracing.git
# Update the cmake to the latest version
RUN pip install --upgrade cmake==3.27.5
#Install latest cppcheck
RUN git clone https://github.com/danmar/cppcheck.git && \
......
......@@ -38,7 +38,7 @@ def getDockerImageName(){
img = "${params.USE_CUSTOM_DOCKER}"
}
else{
if (params.ROCMVERSION != "6.0.1"){
if (params.ROCMVERSION != "6.1"){
if (params.COMPILER_VERSION == "") {
img = "${env.CK_DOCKERHUB}:ck_ub20.04_rocm${params.ROCMVERSION}"
}
......@@ -117,8 +117,10 @@ def getDockerImage(Map conf=[:]){
{
echo "Pulling down image: ${image}"
retimage = docker.image("${image}")
withDockerRegistry([ credentialsId: "docker_test_cred", url: "" ]) {
retimage.pull()
}
}
catch(Exception ex)
{
error "Unable to locate image: ${image}"
......@@ -406,7 +408,7 @@ def runCKProfiler(Map conf=[:]){
dir("script"){
if (params.RUN_FULL_QA){
sh "./run_full_performance_tests.sh 1 QA_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME}"
sh "./run_full_performance_tests.sh 0 QA_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME}"
archiveArtifacts "perf_gemm.log"
archiveArtifacts "perf_resnet50_N256.log"
archiveArtifacts "perf_resnet50_N4.log"
......@@ -416,9 +418,9 @@ def runCKProfiler(Map conf=[:]){
archiveArtifacts "perf_conv_bwd_data.log"
archiveArtifacts "perf_gemm_bilinear.log"
archiveArtifacts "perf_reduction.log"
archiveArtifacts "perf_splitK_gemm_verify.log"
archiveArtifacts "perf_splitK_gemm.log"
archiveArtifacts "perf_onnx_gemm.log"
archiveArtifacts "perf_mixed_gemm.log"
// stash perf files to master
stash name: "perf_gemm.log"
stash name: "perf_resnet50_N256.log"
......@@ -431,6 +433,7 @@ def runCKProfiler(Map conf=[:]){
stash name: "perf_reduction.log"
stash name: "perf_splitK_gemm.log"
stash name: "perf_onnx_gemm.log"
stash name: "perf_mixed_gemm.log"
//we will process results on the master node
}
else{
......@@ -493,9 +496,6 @@ def Build_CK(Map conf=[:]){
def variant = env.STAGE_NAME
def retimage
def navi_node = 0
def mi300_node = 0
gitStatusWrapper(credentialsId: "${env.status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
try {
(retimage, image) = getDockerImage(conf)
......@@ -508,14 +508,6 @@ def Build_CK(Map conf=[:]){
else{
echo "GPU is OK"
}
if ( runShell('grep -n "gfx1030" rocminfo.log') || runShell('grep -n "gfx1101" rocminfo.log') ){
navi_node = 1
echo "This is a Navi node"
}
if ( runShell('grep -n "gfx942" rocminfo.log') ){
mi300_node = 1
echo "This is MI300 node"
}
}
}
}
......@@ -526,15 +518,27 @@ def Build_CK(Map conf=[:]){
withDockerContainer(image: image, args: dockerOpts + ' -v=/var/jenkins/:/var/jenkins') {
timeout(time: 24, unit: 'HOURS')
{
//check whether running on Navi or MI300 node
def navi_node = 0
def mi300_node = 0
sh 'rocminfo | tee rocminfo.log'
if ( runShell('grep -n "gfx1030" rocminfo.log') || runShell('grep -n "gfx1101" rocminfo.log') ){
navi_node = 1
echo "This is a Navi node"
}
if ( runShell('grep -n "gfx942" rocminfo.log') ){
mi300_node = 1
echo "This is MI300 node"
}
cmake_build(conf)
dir("build"){
//run tests and examples
sh 'make -j check'
if (navi_node == 0 ){
if (params.RUN_PERFORMANCE_TESTS && navi_node == 0 && mi300_node == 0 ){
//we only need the ckProfiler to run the performance tests, so we pack and stash it
//do not stash profiler on Navi nodes
//do not stash profiler on Navi or MI300 nodes
sh 'tar -zcvf ckProfiler.tar.gz bin/ckProfiler'
stash "ckProfiler.tar.gz"
stash name: "ckProfiler.tar.gz"
}
if (params.RUN_FULL_QA && mi300_node == 0 ){
// build deb packages for all MI100/200/300 targets and prepare to export
......@@ -542,7 +546,7 @@ def Build_CK(Map conf=[:]){
archiveArtifacts artifacts: 'composablekernel-ckprofiler_*.deb'
archiveArtifacts artifacts: 'composablekernel-tests_*.deb'
sh 'mv composablekernel-ckprofiler_*.deb ckprofiler_0.2.0_amd64.deb'
stash "ckprofiler_0.2.0_amd64.deb"
stash name: "ckprofiler_0.2.0_amd64.deb"
}
}
if (params.hipTensor_test && navi_node == 0 ){
......@@ -629,6 +633,7 @@ def process_results(Map conf=[:]){
unstash "perf_reduction.log"
unstash "perf_splitK_gemm.log"
unstash "perf_onnx_gemm.log"
unstash "perf_mixed_gemm.log"
sh "./process_qa_data.sh"
unstash "ckprofiler_0.2.0_amd64.deb"
sh "sshpass -p ${env.ck_deb_pw} scp -o StrictHostKeyChecking=no ckprofiler_0.2.0_amd64.deb ${env.ck_deb_user}@${env.ck_deb_ip}:/var/www/html/composable_kernel/"
......@@ -716,8 +721,8 @@ pipeline {
description: "Run the cppcheck static analysis (default: OFF)")
booleanParam(
name: "RUN_PERFORMANCE_TESTS",
defaultValue: false,
description: "Run the performance tests (default: OFF)")
defaultValue: true,
description: "Run the performance tests (default: ON)")
booleanParam(
name: "RUN_CODEGEN_TESTS",
defaultValue: true,
......
......@@ -133,11 +133,16 @@ def parse_logfile(logfile):
if 'Best Perf' in line:
lst=line.split()
res.append(lst[4])
elif 'onnx_gemm' in logfile or 'splitK_gemm' in logfile or 'mixed_gemm' in logfile:
elif 'onnx_gemm' in logfile or 'mixed_gemm' in logfile:
for line in open(logfile):
if 'Best Perf' in line:
lst=line.split()
res.append(lst[33])
elif 'splitK_gemm' in logfile:
for line in open(logfile):
if 'Best Perf' in line:
lst=line.split()
res.append(lst[36])
return res
......@@ -231,7 +236,7 @@ def main():
sql_hostname = '127.0.0.1'
sql_username = os.environ["dbuser"]
sql_password = os.environ["dbpassword"]
sql_main_database = 'miopen_perf'
sql_main_database = os.environ["ck_perf_db"]
sql_port = 3306
ssh_host = os.environ["dbsship"]
ssh_user = os.environ["dbsshuser"]
......
......@@ -121,26 +121,16 @@ print_log_header $reduction_log $env_type $branch $host_name
./profile_reduce_no_index.sh $verify 2 10 --half 2>&1 | tee -a $reduction_log
#run splitK_gemm tests, first correctness verification, then performance
export splitK_gemm_ver_log="perf_splitK_gemm_verify.log"
print_log_header $splitK_gemm_ver_log $env_type $branch $host_name
./profile_splitK_gemm.sh gemm_splitk 0 0 $verify 1 0 0 4 2>&1 | tee -a $splitK_gemm_ver_log
./profile_splitK_gemm.sh gemm_splitk 0 1 $verify 1 0 0 4 2>&1 | tee -a $splitK_gemm_ver_log
./profile_splitK_gemm.sh gemm_splitk 0 2 $verify 1 0 0 4 2>&1 | tee -a $splitK_gemm_ver_log
./profile_splitK_gemm.sh gemm_splitk 0 3 $verify 1 0 0 4 2>&1 | tee -a $splitK_gemm_ver_log
./profile_splitK_gemm.sh gemm_splitk 1 0 $verify 1 0 0 4 2>&1 | tee -a $splitK_gemm_ver_log
./profile_splitK_gemm.sh gemm_splitk 1 1 $verify 1 0 0 4 2>&1 | tee -a $splitK_gemm_ver_log
./profile_splitK_gemm.sh gemm_splitk 1 2 $verify 1 0 0 4 2>&1 | tee -a $splitK_gemm_ver_log
./profile_splitK_gemm.sh gemm_splitk 1 3 $verify 1 0 0 4 2>&1 | tee -a $splitK_gemm_ver_log
export splitK_gemm_log="perf_splitK_gemm.log"
print_log_header $splitK_gemm_log $env_type $branch $host_name
./profile_splitK_gemm.sh gemm_splitk 0 0 0 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 0 1 0 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 0 2 0 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 0 3 0 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 1 0 0 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 1 1 0 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 1 2 0 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 1 3 0 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 0 0 $verify 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 0 1 $verify 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 0 2 $verify 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 0 3 $verify 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 1 0 $verify 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 1 1 $verify 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 1 2 $verify 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
./profile_splitK_gemm.sh gemm_splitk 1 3 $verify 1 0 1 4 2>&1 | tee -a $splitK_gemm_log
#run ONNX gemm tests
export onnx_log="perf_onnx_gemm.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