vtk-m/.gitlab/ci/crusher.yml

##=============================================================================
##
##  Copyright (c) Kitware, Inc.
##  All rights reserved.
##  See LICENSE.txt for details.
##
##  This software is distributed WITHOUT ANY WARRANTY; without even
##  the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
##  PURPOSE.  See the above copyright notice for more information.
##
##=============================================================================

# Ad-hoc build that runs in the ECP Hardware, concretely in OLCF Spock.
.crusher_gcc_hip:
  variables:
    CCACHE_BASEDIR:          "/lustre/orion/csc331/scratch/"
    CCACHE_DIR:              "/lustre/orion/csc331/scratch/vbolea/ci/vtk-m/ccache"
    CUSTOM_CI_BUILDS_DIR:    "/lustre/orion/csc331/scratch/vbolea/ci/vtk-m/runtime"

    # -isystem= is not affected by CCACHE_BASEDIR, thus we must ignore it
    CCACHE_IGNOREOPTIONS:    "-isystem=*"
    CCACHE_NOHASHDIR:        "true"
    CCACHE_INSTALL_DIR:      "$CI_PROJECT_DIR/build"

    CMAKE_BUILD_TYPE:        "RelWithDebInfo"
    CMAKE_GENERATOR:         "Ninja"
    CMAKE_PREFIX_PATH:       "$CI_BUILDS_DIR/kokkos_install"

    # We do not want to use the user's ~/.gitconfig
    GIT_CONFIG_GLOBAL:       "true"

    KOKKOS_OPTS: >-
      -DCMAKE_INSTALL_PREFIX:PATH=$CI_BUILDS_DIR/kokkos_install
      -DCMAKE_CXX_COMPILER:FILEPATH=/opt/rocm-5.4.3/hip/bin/hipcc
      -DKokkos_ARCH_VEGA90A:BOOL=ON

    # The user default module list should not exist
    # craype;rocm;gcc should be loaded first
    JOB_MODULES: >-
      craype-accel-amd-gfx90a
      gcc/12
      cmake
      rocm/5.4.3
      git
      git-lfs
      ninja
      zstd
    VTKM_SETTINGS: kokkos+hip+gfx90a+crusher+ccache+no_rendering
  interruptible: true

.setup_env_ecpci: &setup_env_ecpci |
  module purge
  module load ${JOB_MODULES}
  module list
  export PATH="${CCACHE_INSTALL_DIR}/ccache:$PATH"

build:crusher_gcc_hip:
  stage: build
  tags: [crusher, shell]
  extends:
    - .crusher_gcc_hip
    - .run_crusher_ci
  before_script:
    - *setup_env_ecpci
    - mkdir -p "$CCACHE_INSTALL_DIR"
    - cmake --version
    - cmake -VV -P .gitlab/ci/config/ccache.cmake
    - ccache -z
    - ccache -s
    - .gitlab/ci/config/kokkos.sh "$CI_BUILDS_DIR" "3.7.01" $KOKKOS_OPTS

    - git remote add lfs https://gitlab.kitware.com/vtk/vtk-m.git
    - git fetch lfs
    - git-lfs install
    - git-lfs pull lfs
  script:
    - cmake -V -P .gitlab/ci/config/gitlab_ci_setup.cmake
    - ctest -VV -S .gitlab/ci/ctest_configure.cmake
  artifacts:
    expire_in: 24 hours
    when: always
    paths:
      - build/
      - .gitlab/ccache/ccache

test:crusher_gcc_hip:
  stage: test
  tags: [crusher, slurm]
  extends:
    - .crusher_gcc_hip
    - .run_crusher_ci
  needs:
    - build:crusher_gcc_hip
  dependencies:
    - build:crusher_gcc_hip
  variables:
    SCHEDULER_PARAMETERS: "-ACSC331_crusher -pbatch -t120 --nice=0 -c32 --gpus=4 -N 1"
    # We need this to skip ctest_submit from being run inside a jsrun job
    GITLAB_CI_EMULATION: 1
    # Tests errors to address due to different env in Spock
    # Refer to issue: https://gitlab.kitware.com/vtk/vtk-m/-/issues/743
    CTEST_EXCLUSIONS: >-
      UnitTestMIRFilter
      UnitTestDistributedBranchDecompositionFilter
      UnitTestWorkletParticleAdvection
      UnitTestLagrangianFilter
  before_script:
    - *setup_env_ecpci
  script:
    - CTEST_MAX_PARALLELISM=32 ctest -VV -S .gitlab/ci/ctest_build.cmake
    - CTEST_MAX_PARALLELISM=4  ctest -VV -S .gitlab/ci/ctest_test.cmake
  after_script:
    - *setup_env_ecpci
    - ccache -s
    - ctest -VV -S .gitlab/ci/ctest_submit_build.cmake
    - ctest -VV -S .gitlab/ci/ctest_submit_test.cmake
  timeout: 150 minutes