Continuous Integration (CI) Testing Maintenance Tasks¶

In Continuous Integration (CI) Testing, we described RAJA CI testing workflows. This section describes how to perform common RAJA CI testing maintenance tasks.

GitLab CI Tasks¶

The tasks in this section apply to GitLab CI testing on Livermore Computing (LC) platforms. LC folks and others maintain Confluence pages with a lot of useful information for setting up and maintaining GitLab CI for a project, mirroring a GitHub to GitLab, etc. Please refer to LC GitLab CI for such information.

Changing build and test configurations¶

The configurations that are tested in RAJA are defined by a Spack spec in one of two places, depending on whether it is shared with other projects or it is specific to RAJA. The details are described in Launching CI pipelines (step 2). Each spec contains information (compiler and version, build variants, etc.) that must be consistent with the build specs defined in the RADIUSS Spack Configs project, which also includes the RAJA Spack package. The RADIUSS Spack Configs project is included as a RAJA submodule in the RAJA/scripts directory.

Removing a configuration¶

To remove a RAJA-specific test configuration, simply delete the entry for it in the RAJA/.gitlab/jobs/<MACHINE>.yml file where it is defined. Here, MACHINE is the name of an LC platform where GitLab CI is run.

To remove a shared configuration, it must be removed from the appropriate gitlab/radiuss-jobs/<MACHINE>.yml file in the RADIUSS Spack Configs project. Create a branch there, remove the job entry, and create a pull request.

Important

The RADIUSS Spack Configs project is used by several other projects. When changing a shared configuration file, please make sure the change is OK with those other projects. Typically, shared configurations are only changed when it makes sense to update compilers for all projects, such as when system default compiler versions change.

Adding a configuration¶

To add a RAJA-specific test configuration, add an entry for it to the RAJA/.gitlab/jobs/<MACHINE>.yml file, where MACHINE is the name of the LC platform where it will be run. When adding a test configuration, it is important to note two items that must be specified properly:

Each jobs must have a unique job label, which identifies it in the machine configuration file and also on a web page for a GitLab CI pipeline

The Spack spec name identifies the compiler and version, compiler flags, build options, etc. must match an existing spec in the RADIUSS Spack Configs project. Also, the build options must be consistent with the variants defined in the RAJA package in that project.

For example, an entry for a build using the clang 12.0.1 compiler with CUDA 11.5.0 on the LC lassen machine would be something like this:

clang_12_0_1_cuda_11_5_0:
  variables:
    SPEC: " ~shared +openmp +tests +cuda cuda_arch=70 %clang@12.0.1 ^cuda@11.5.0"
  extends: .job_on_lassen

Here, we enable OpenMP and CUDA, both of which must be enabled to test those RAJA back-ends, and specify the CUDA target architecture ‘sm_70’.

To add a shared configuration, it must be added to the appropriate gitlab/radiuss-jobs/<MACHINE>.yml file in the RADIUSS Spack Configs project. Create a branch there, add the job entry, and create a pull request.

Important

The RADIUSS Spack Configs project is used by several other projects. When changing a shared configuration file, please make sure the change is OK with those other projects. Typically, shared configurations are only changed when it makes sense to update compilers for all projects, such as when system default compiler versions change.

Modifying a configuration¶

To change an existing configuration, change the relevant information in the configuration in the appropriate RAJA/.gitlab/jobs/<MACHINE>.yml file. Make sure to also modify the job label as needed, so it is descriptive of the configuration is unique with respect to the others that are being run.

To modify a shared configuration, it must be changed in the appropriate gitlab/radiuss-jobs/<MACHINE>.yml file in the RADIUSS Spack Configs project. Create a branch there, modify the job entry, and create a pull request.

Important

Build spec information used in RAJA GitLab CI pipelines must exist in the compilers.yaml file and/or packages.yaml file for the appropriate system type in the RADIUSS Spack Configs repo.

If the desired entry is not there, but exists in a newer version of the RADIUSS Spack Configs project, update the RAJA submodule to use the newer version. If the information does not exist in any version of the RADIUSS Spack Configs project, create a branch there, add the needed spec info, and create a pull request. Then, when that PR is merged, update the RAJA submodule for the RADIUSS Spack Configs project to the new version.

Changing run parameters¶

The parameters for each system/scheduler on which we run GitLab CI for RAJA, such as job time limits, resource allocations, etc. are defined in the RAJA/.gitlab/custom-variables.yml file. Job-specific templates and customizations are defined in RAJA/.gitlab/custom-jobs.yml. This information can remain as is, for the most part, and should not be changed unless absolutely necessary.

For example, sometimes a particular job will take longer to build and run than the default allotted time for jobs on a machine. In this case, the time for the job can be adjusted in the job entry in the associated RAJA/.gitlab/jobs/<MACHINE>.yml file. For example:

gcc_8_1_0:
variables:
  SPEC: " ${PROJECT_DANE_VARIANTS} %gcc@8.1.0 ${PROJECT_DANE_DEPS}"
  DANE_BUILD_AND_TEST_JOB_ALLOC: "--time=60 --nodes=1"
extends: .job_on_dane

This example sets the build and test allocation time to 60 minutes and the the run resource to one node.

Allowing failures¶

Sometimes a shared job configuration is known to fail for RAJA. To allow the job to fail without the CI check associated with it failing, we can annotate the job for this. For example:

ibm_clang_9_0_0:
  variables:
    SPEC: " ${PROJECT_LASSEN_VARIANTS} %clang@ibm.9.0.0 ${PROJECT_LASSEN_DEPS}"
  extends: .job_on_lassen
  allow_failure: true

Important

When a shared job needs to be modified for RAJA specifically, we call that “overriding”. The job label must be kept the same as for the shared job in the gitlab/radiuss-jobs/<MACHINE>.yml file in the RADIUSS Spack Configs, and the RAJA-specific job can be adapted. If you override a shared job, please add a comment to describe the change in the RAJA/.gitlab/jobs/<MACHINE>.yml file where the job is overridden.

Building the Intel clang + SYCL HIP compiler for use in CI¶

To run CI tests for the RAJA SYCL back-end on GitLab, we use the corona system and a custom Intel Clang compiler that we build ourselves to support SYCL for AMD GPUs. This compiler lives in the /usr/workspace/raja-dev/ folder so that it can be accessed by the service user account that we use to run our GitLab CI. Since the Intel compiler does not do releases in the typical sense (they simply update their repo every night), it may become necessary to periodically build a new version of the compiler to ensure that we are using the most up-to-date version available. The steps for building, installing, and running are shown here.

Building the Compiler¶

Important

Because Intel updates their compiler repo daily, it is possible that the head of the SYCL branch will fail to build. If it does not build, try checking out an earlier commit in your local cloned repo. On the Intel/LLVM GitHub Project, you can see which of their commits builds by checking the status badge next to each commit. Look for a recent commit that passes.

On LC machines, please follow the good neighbor policy and do your build on a compute node.

Use an appropriate bank to get an interactive node, e.g on Corona:
```
flux alloc -t 60 -N 1 --bank=wbronze
```
Load the module of the version of GCC headers that you want to use. We typically use the system default, which on corona is currently gcc/10.3.1-magic. Set then environment variable GCC_VERSION to the GCC version, then load the module:
```
module load gcc/${GCC_VERSION}-magic
```
Load the module of the version of ROCm that you want to use. Set the environment variable ROCM_VERSION to the ROCm version, then load the module:
```
module load rocm/${ROCM_VERSION}
```
Load Python module you want to use. The LLVM configure requires at least version 3.7. Set the environment variable PYTHON_VERSION to the Python version, then load the module:
```
module load python/${PYTHON_VERSION}
```

Clone the SYCL branch of Intel’s LLVM compiler:

git clone https://github.com/intel/llvm -b sycl

Go into the LLVM folder and get the Git SHA for the commit hash you are building. The first 12 characters of the hash value are used in the name of the compiler install directory. To get the first 12 characters of the hash value:
```
cd llvm
git rev-parse --short=12 HEAD
```
Then, set the environment variable GIT_SHA to the hash value, and set the environment variable INSTALL_PREFIX to the name of the installation directory, which has the following form: /usr/workspace/raja-dev/clang_sycl_${GIT_SHA}_hip_gcc${GCC_VERSION}_rocm${ROCM_VERSION}
After, the compiler repo code is in place and the build environment is set as described in the previous steps, build and install the compiler.
1. Configure:

  python3 buildbot/configure.py --hip -o buildrocm${ROCM_VERSION} \
  --cmake-gen "Unix Makefiles" \
  --cmake-opt=-DSYCL_BUILD_PI_HIP_ROCM_DIR=/opt/rocm-${ROCM_VERSION} \
  --cmake-opt=-DSYCL_BUILD_PI_HIP_ROCM_INCLUDE_DIR=/opt/rocm-${ROCM_VERSION}/include \
  --cmake-opt=-DSYCL_BUILD_PI_HIP_ROCM_LIB_DIR=/opt/rocm-${ROCM_VERSION}/lib \
  --cmake-opt=-DSYCL_BUILD_PI_HIP_INCLUDE_DIR=/opt/rocm-${ROCM_VERSION}/include \
  --cmake-opt=-DSYCL_BUILD_PI_HIP_HSA_INCLUDE_DIR=/opt/rocm-${ROCM_VERSION}/hsa/include/hsa \
  --cmake-opt=-DSYCL_BUILD_PI_HIP_LIB_DIR=/opt/rocm-${ROCM_VERSION}/lib \
  --cmake-opt=-DUR_HIP_ROCM_DIR=/opt/rocm-${ROCM_VERSION} \
  --cmake-opt=-DUR_HIP_INCLUDE_DIR=/opt/rocm-${ROCM_VERSION}/include \
  --cmake-opt=-DUR_HIP_HSA_INCLUDE_DIR=/opt/rocm-${ROCM_VERSION}/hsa/include/hsa \
  --cmake-opt=-DUR_HIP_LIB_DIR=/opt/rocm-${ROCM_VERSION}/lib

b. Build::

   python buildbot/compile.py -o buildrocm${ROCM_VERSION}

c. Install::

   cp -rp buildrocm${ROCM_VERSION}/install ${INSTALL_PREFIX}
   cd ..

Set the permissions of the installation folder, and everything in it to 750:
```
chmod 750 ${INSTALL_PREFIX} -R
```
Change the group of the folder and everything in it to raja-dev:
```
chgrp raja-dev ${INSTALL_PREFIX} -R
```
Test the compiler

Follow the steps in the Using the compiler section to test the installation

Using the compiler¶

Load the version of ROCm that you used when building the compiler, for example:
```
ROCM_VERSION=6.4.3
module load rocm/${ROCM_VERSION}
```
Navigate to the root of your local RAJA checkout space:
```
cd /path/to/raja
```

Determine where you installed the compiler.

This is the INSTALL_PREFIX used above. For example:

SYCL_INSTALL_PREFIX=/usr/workspace/raja-dev/clang_sycl_16b7bcb09915_hip_gcc10.3.1_rocm6.4.3

Run the test config script:

./scripts/lc-builds/corona_sycl.sh ${SYCL_INSTALL_PREFIX}

As indicated in the output of the corona_sycl.sh script the SYCL compiler libraries need to be on the LD_LIBRARY_PATH:
```
export LD_LIBRARY_PATH=${SYCL_INSTALL_PREFIX}/lib:${SYCL_INSTALL_PREFIX}/lib64:$LD_LIBRARY_PATH
```
cd into the generated build directory:
```
cd build_corona-sycl_${USER}
```
Build the code and run the RAJA tests:
```
make -j
make test
```

GitHub Actions CI Tasks¶

The tasks in this section apply to RAJA GitHub Actions CI testing that was described in GitHub Actions CI

Changing Builds/Container Images¶

The builds we run in GitHub Actions are defined in the RAJA/.github/workflows/build.yml file.

Linux/Docker¶

To update or add a new compiler / job to GitHub Actions CI, we need to edit the RAJA/.github/workflows/build.yml file and the RAJA/Dockerfile, if changes are needed there.

For GitHub Actions, we add the name of the job to the job list in the RAJA/.github/workflows/build.yml file:

jobs:
build_docker:
  strategy:
    matrix:
      target: [..., compilerX]

In the RAJA/Dockerfile file, we add a section that defines the commands for the compilerX job, such as:

FROM ghcr.io/llnl/radiuss:compilerX-ubuntu-22.04 AS compilerX
ENV GTEST_COLOR=1
COPY . /home/raja/workspace
WORKDIR /home/raja/workspace/build
RUN cmake -DCMAKE_CXX_COMPILER=compilerX ... && \
    make -j 6 && \
    ctest -T test --output-on-failure && \
    make clean

Each of our docker builds is built up on a base image maintained in the RADIUSS Docker Project.

The base container images are shared by multiple projects and are rebuilt regularly. If bugs are fixed in the base images, the changes will be automatically propagated to all projects using them in their Docker builds.

Check RADIUSS Docker Project for a list of currently available images.

Windows / MacOS¶

In GitHub Actions, we run our Windows and MacOS builds directly on the provided machine instances. To change the versions, change the appropriate lines in the RAJA/.github/workflows/build.yml file:

build_mac:
  runs-on: macos-latest
...

build_windows:
  runs-on: windows-latest
...

Changing Build/Run Parameters¶

Linux/Docker¶

We can edit the build and run configurations of each Docker build, by editing the appropriate line containing the RUN command in the RAJA/Dockerfile file. For example, we can change CMake options or change the parallel build value of make -j N for adjusting throughput.

Each base image is built using spack. For the most part the container environments are set up to run our CMake and build commands out of the box. However, there are a few exceptions where we may need to load compiler specific environment variables, such as for the Intel LLVM compiler. For example, this may appear as:

RUN /bin/bash -c "source /opt/intel/oneapi/setvars.sh 2>&1 > /dev/null && \
  cmake ..."

In these cases, it is important to include the double quotes in the correct locations.

Windows / MacOS¶

Windows and MacOS build / run parameters can be configured directly in the RAJA/.github/workflows/build.yml file. CMake options can be configured in the workflow file for each job. The parallel build value can also be edited directly in the workflow file for each job.

RAJA Performance Suite CI Tasks¶

The RAJA Performance Suite project CI testing processes, directory/file structure, and dependencies are nearly identical to that for RAJA, which is described in Continuous Integration (CI) Testing. Specifically,

The RAJA Performance Suite GitLab CI process is driven by the RAJAPerf/.gitlab-ci.yml file, which uses GitLab CI Components from radiuss-shared-ci.

The custom-jobs.yml and custom-variables.yml files reside in RAJAPerf/.gitlab.

The build_and_test.sh script resides in the RAJAPerf/scripts/gitlab directory.

The RAJAPerf/Dockerfile drives the GitHub Actions testing pipelines.

The Performance Suite GitLab CI uses the uberenv and radiuss-spack-configs versions located in the RAJA submodule to make the testing consistent across projects and avoid redundancy. This is reflected in the RAJAPerf/.uberenv_config.json file which point at the relevant RAJA submodule locations. That is the paths contain tpl/RAJA/....

Apart from these minor differences, all CI maintenance and development tasks for the RAJA Performance Suite follow the same pattern that is described in Continuous Integration (CI) Testing Maintenance Tasks.