Skip to content

Sync meeting on EESSI test suite (2023 05 31)

Jump to bottom
Kenneth Hoste edited this page May 31, 2023 · 1 revision

EESSI test suite sync meetings

Previous meetings

Notes for 2023-05-31 meeting

  • merged PRs
    • PR #28 by Sam: extend scales + add constants
      • list of supported scales should be documented in README?
      • extra 1_gpu, 2_gpu scales?
        • core count can be picked based on available cores/gpus per node
      • 1_4_node didn't work for Kenneth on Hortense to get 1 GPU runs
        • was using hortense branch not properly updated with latest main :man-facepalming:
      • hardcoded limit of 30min in GROMACS tests (see open issue)
    • PR #33 by Xin: update README
    • PR #42 by Sam: add more comments, docs to the gromacs test
  • open PRs
    • PR #24 by Sam+Kenneth add configuration file for VSC Tier-1 Hortense
      • was updated to have all partitions on Hortense (GPU + CPU)
      • on GPU nodes, both CPU-only and GPU test variants are submitted
        • Do we want this?
        • We can not include cpu in features for the GPU partitions
      • we should also specify num_sockets along with num_cpus for all partitions
    • PR #36 by Caspar: Expand developer instructions
  • WIP PRs
    • PR #38 by Caspar: WIP TensorFlow
      • how to specify binding behaviour in a portable way?
      • options depend on launcher being used (mpirun, srun, etc.)
      • something to ask ReFrame developers?
      • can we set environment variables that control bindings for all known launchers?
    • PR #44 by Kenneth: add GitHub Actions workflow to run EESSI test suite
    • PR #45 by Kenneth: refactor namespace to eessi.testsuite.* (WIP)
      • remove __init__.py files? (maybe be needed for 'pip install' => add CI for that)
      • eessi/testsuite/utils/constants.py -> eessi/testsuite/constants.py
      • example configuration files in config dir directly in test-suite repo
      • config/vsc_hortense.py, config/surf_snellius.py, config/aws_eessi_citc.py, config/github_actions.py, config/local_cpu.py, config/local_cpu_gpu.py, config/local_gpu.py, config/eessi.py
      • need to also update README!
  • Satish is stuck with problem with configuration file (issue #43)
    • maybe due to hostnames?
  • consider using remote detection support of ReFrame in example configuration file
  • Caspar: running tests with EESSI
    • requires "source /cvmfs/pilot.eessi-hpc.org/latest/init/bash" prepare_cmds in partition
    • but also still requires to run the "source" command in the shell session where 'reframe' is run
  • [Kenneth] from README to docs (https://eessi.github.io/docs/software_testing/ or https://eessi.github.io/docs/test-suite)
  • should we be relying on the hpctestlib provided by ReFrame
  • next steps
Clone this wiki locally