From dcb38b66c2d6de1d84675dc1069d990fb05ee9b6 Mon Sep 17 00:00:00 2001
From: Florent Pruvost <florent.pruvost@inria.fr>
Date: Wed, 18 Dec 2024 08:41:11 +0100
Subject: [PATCH] ci: add test_starpu_cuda
---
.gitlab-ci.yml | 1 +
.gitlab/build.sh | 2 +-
.gitlab/build.yml | 26 +++++++++++++------
.gitlab/common.yml | 1 +
.gitlab/sbatch.sh | 29 +++++++++++++++++++++
.gitlab/test.sh | 1 -
.gitlab/test_starpu.yml | 8 +++---
.gitlab/test_starpu_plafrim.yml | 46 +++++++++++++++++++++++++++++++++
8 files changed, 101 insertions(+), 13 deletions(-)
create mode 100755 .gitlab/sbatch.sh
create mode 100644 .gitlab/test_starpu_plafrim.yml
diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
index e01e26e32..270076d74 100644
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -14,6 +14,7 @@ include:
- .gitlab/docker.yml
- .gitlab/build.yml
- .gitlab/test_starpu.yml
+ - .gitlab/test_starpu_plafrim.yml
- .gitlab/test_starpu_simgrid.yml
- .gitlab/test_parsec.yml
- .gitlab/test_quark.yml
diff --git a/.gitlab/build.sh b/.gitlab/build.sh
index 0d17fd144..e631b37f2 100755
--- a/.gitlab/build.sh
+++ b/.gitlab/build.sh
@@ -59,7 +59,7 @@ case $SYSTEM in
esac
# Compile
-eval '${SCAN}cmake --build build-${VERSION} -j 4 > /dev/null'
+eval '${SCAN}cmake --build build-${VERSION} -j ${CMAKE_BUILD_PARALLEL_LEVEL} > /dev/null'
# Install
cmake --install build-${VERSION}
diff --git a/.gitlab/build.yml b/.gitlab/build.yml
index e5d674332..2d449df9d 100644
--- a/.gitlab/build.yml
+++ b/.gitlab/build.yml
@@ -2,6 +2,8 @@
.build_script_common:
stage: build
extends: .only-master-mr
+ variables:
+ CMAKE_BUILD_PARALLEL_LEVEL: 4
artifacts:
name: "$CI_JOB_NAME-$CI_COMMIT_REF_SLUG"
expire_in: 1 week
@@ -10,6 +12,8 @@
.build_script_linux:
extends: .build_script_common
+ variables:
+ SYSTEM: linux
script:
- bash .gitlab/build.sh | tee ${LOGNAME}.log
@@ -56,29 +60,31 @@ build_starpu:
.build_script_guix:
tags: ['plafrim']
extends: .build_script_common
+ variables:
+ SYSTEM: guix
+ CMAKE_BUILD_PARALLEL_LEVEL: 20
script:
- guix describe -f channels
- - guix shell --pure --preserve=SYSTEM --preserve=VERSION --preserve=LOGNAME --preserve=BUILD_OPTIONS
- -D chameleon-${GUIX_CHAMELEON_VARIANT} coreutils bash-minimal --
+ - guix shell --pure --preserve=SYSTEM --preserve=VERSION --preserve=LOGNAME --preserve=BUILD_OPTIONS --preserve=CMAKE_BUILD_PARALLEL_LEVEL
+ -D chameleon-${GPU_BACKEND} ${BLAS} coreutils bash-minimal --
bash .gitlab/build.sh | tee ${LOGNAME}.log
build_starpu_cuda:
extends: .build_script_guix
variables:
- SYSTEM: guix
VERSION: starpu_cuda
- GUIX_CHAMELEON_VARIANT: cuda
+ GPU_BACKEND: cuda
+ BLAS: "--with-input=openblas=intel-oneapi-mkl"
LOGNAME: "chameleon-build-${SYSTEM}-${VERSION}"
- BUILD_OPTIONS: "-DCHAMELEON_USE_CUDA=ON -DCHAMELEON_USE_MPI=ON -DBLA_VENDOR=OpenBLAS"
+ BUILD_OPTIONS: "-DCHAMELEON_USE_CUDA=ON -DCHAMELEON_USE_MPI=OFF -DBLA_VENDOR=Intel10_64lp_seq"
build_starpu_hip:
extends: .build_script_guix
variables:
- SYSTEM: guix
VERSION: starpu_hip
- GUIX_CHAMELEON_VARIANT: hip
+ GPU_BACKEND: hip
LOGNAME: "chameleon-build-${SYSTEM}-${VERSION}"
- BUILD_OPTIONS: "-DCHAMELEON_USE_HIP_ROC=ON -DCHAMELEON_USE_MPI=ON -DBLA_VENDOR=OpenBLAS -DCMAKE_C_COMPILER=gcc -DCMAKE_Fortran_COMPILER=gfortran"
+ BUILD_OPTIONS: "-DCHAMELEON_USE_HIP_ROC=ON -DCHAMELEON_USE_MPI=OFF -DBLA_VENDOR=OpenBLAS -DCMAKE_C_COMPILER=gcc -DCMAKE_Fortran_COMPILER=gfortran"
build_starpu_simgrid:
extends: .build_script_linux
@@ -97,11 +103,13 @@ build_starpu_macosx:
SYSTEM: macosx
VERSION: starpu
LOGNAME: "chameleon-build-${SYSTEM}-${VERSION}"
+ CMAKE_BUILD_PARALLEL_LEVEL: 4
script:
- bash .gitlab/build.sh | tee ${LOGNAME}.log
artifacts:
name: "$CI_JOB_NAME-$CI_COMMIT_REF_SLUG"
expire_in: 1 week
+ when: always
paths:
- ${LOGNAME}.log
cache:
@@ -119,11 +127,13 @@ build_starpu_macosx:
MSYSTEM: UCRT64
VERSION: starpu
LOGNAME: "chameleon-build-${SYSTEM}-${VERSION}"
+ CMAKE_BUILD_PARALLEL_LEVEL: 4
script:
- bash -lc .gitlab/build.sh | tee "$env:LOGNAME.log"
artifacts:
name: "$CI_JOB_NAME-$CI_COMMIT_REF_SLUG"
expire_in: 1 week
+ when: always
paths:
- ${LOGNAME}.log
cache:
diff --git a/.gitlab/common.yml b/.gitlab/common.yml
index 9e679f1c4..f4e30eb4b 100644
--- a/.gitlab/common.yml
+++ b/.gitlab/common.yml
@@ -37,5 +37,6 @@ variables:
name: "$CI_JOB_NAME-$CI_COMMIT_REF_SLUG"
expire_in: 1 week
untracked: true
+ when: always
reports:
junit: ${LOGNAME}-junit.xml
diff --git a/.gitlab/sbatch.sh b/.gitlab/sbatch.sh
new file mode 100755
index 000000000..d202d0686
--- /dev/null
+++ b/.gitlab/sbatch.sh
@@ -0,0 +1,29 @@
+#!/bin/bash
+# sbatch.sh : submit slurm jobs and wait for completion before exiting
+set -x
+
+if [ $# -gt 0 ]
+then
+ JOB_NAME=$1
+fi
+JOB_NAME=${JOB_NAME:-chameleon}
+
+# to get kernels execution on both cpus and gpus
+export STARPU_SCHED=random
+
+# execution commands
+sbatch --wait \
+ --job-name="$JOB_NAME" \
+ --output="$JOB_NAME.out" \
+ --nodes=1 \
+ --exclusive --ntasks-per-node=1 --threads-per-core=1 \
+ --constraint="$SLURM_CONSTRAINTS" \
+ --time=01:00:00 \
+ $(dirname "$0")/test.sh
+# get the error code from the last command: sbatch --wait ...
+err=$?
+
+cat $JOB_NAME.out
+
+# exit with error code from the guix command
+exit $err
diff --git a/.gitlab/test.sh b/.gitlab/test.sh
index 0a4a8639f..27ba5484b 100755
--- a/.gitlab/test.sh
+++ b/.gitlab/test.sh
@@ -1,5 +1,4 @@
#!/usr/bin/env bash
-
set -e
set -x
diff --git a/.gitlab/test_starpu.yml b/.gitlab/test_starpu.yml
index efc3bd783..1adc4997f 100644
--- a/.gitlab/test_starpu.yml
+++ b/.gitlab/test_starpu.yml
@@ -1,5 +1,5 @@
---
-.test_starpu:
+.test_script_starpu:
extends: .test_script_linux
needs: [build_starpu]
variables:
@@ -9,7 +9,7 @@
TESTS_RESTRICTION: "-R _${CATEGORY}_${PRECISION}|example -E sytrf|sysv"
test_starpu_master:
- extends: .test_starpu
+ extends: .test_script_starpu
rules:
- if: ($CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH || $CI_COMMIT_BRANCH =~ /^ci-.*$/) && $CI_PIPELINE_SOURCE != "schedule"
parallel:
@@ -18,7 +18,7 @@ test_starpu_master:
CATEGORY: [shm, mpi]
test_starpu_mr:
- extends: .test_starpu
+ extends: .test_script_starpu
rules:
- if: ($CI_PIPELINE_SOURCE == "merge_request_event" && $CI_MERGE_REQUEST_SOURCE_BRANCH_NAME !~ /^notest-.*$/)
parallel:
@@ -43,6 +43,7 @@ test_starpu_macosx:
artifacts:
name: "$CI_JOB_NAME-$CI_COMMIT_REF_SLUG"
expire_in: 1 week
+ when: always
paths:
- ${LOGNAME}.log
reports:
@@ -69,6 +70,7 @@ test_starpu_macosx:
artifacts:
name: "$CI_JOB_NAME-$CI_COMMIT_REF_SLUG"
expire_in: 1 week
+ when: always
paths:
- ${LOGNAME}.log
reports:
diff --git a/.gitlab/test_starpu_plafrim.yml b/.gitlab/test_starpu_plafrim.yml
new file mode 100644
index 000000000..c5178e6ca
--- /dev/null
+++ b/.gitlab/test_starpu_plafrim.yml
@@ -0,0 +1,46 @@
+---
+.test_script_starpu_plafrim:
+ tags: ['plafrim']
+ stage: test
+ variables:
+ SYSTEM: guix
+ script:
+ - guix describe -f channels | tee guix-channels.scm
+ - guix shell --pure --preserve=SYSTEM --preserve=VERSION --preserve=LOGNAME --preserve=BUILD_OPTIONS --preserve=TESTS_RESTRICTION --preserve=SLURM_CONSTRAINTS --preserve=LD_PRELOAD
+ -D chameleon-${GPU_BACKEND} ${BLAS} slurm coreutils inetutils bash-minimal --
+ bash .gitlab/sbatch.sh ${LOGNAME} | tee ${LOGNAME}.log
+ artifacts:
+ name: "$CI_JOB_NAME-$CI_COMMIT_REF_SLUG"
+ expire_in: 1 week
+ when: always
+ paths:
+ - guix-channels.scm
+ - ${LOGNAME}.log
+ reports:
+ junit: ${LOGNAME}-junit.xml
+
+.test_script_starpu_plafrim_cuda:
+ extends: .test_script_starpu_plafrim
+ needs: [build_starpu_cuda]
+ variables:
+ VERSION: starpu_cuda
+ GPU_BACKEND: cuda
+ BLAS: "--with-input=openblas=intel-oneapi-mkl"
+ LOGNAME: "chameleon-${SYSTEM}-${VERSION}"
+ BUILD_OPTIONS: "-DCHAMELEON_USE_CUDA=ON -DCHAMELEON_USE_MPI=OFF -DBLA_VENDOR=Intel10_64lp_seq"
+ SLURM_CONSTRAINTS: "sirocco"
+ LD_PRELOAD: "/usr/lib64/libcuda.so"
+
+test_starpu_cuda_master:
+ extends: .test_script_starpu_plafrim_cuda
+ rules:
+ - if: ($CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH || $CI_COMMIT_BRANCH =~ /^ci-.*$/) && $CI_PIPELINE_SOURCE != "schedule"
+ variables:
+ TESTS_RESTRICTION: "-R test_shm_gpu"
+
+test_starpu_cuda_mr:
+ extends: .test_script_starpu_plafrim_cuda
+ rules:
+ - if: ($CI_PIPELINE_SOURCE == "merge_request_event" && $CI_MERGE_REQUEST_SOURCE_BRANCH_NAME !~ /^notest-.*$/)
+ variables:
+ TESTS_RESTRICTION: "-R test_shm_gpu_d|test_shm_gpu_c"
--
GitLab