Skip to content

Sync meeting on EESSI test suite (2024 05 03)

Jump to bottom
Kenneth Hoste edited this page May 5, 2024 · 1 revision

EESSI test suite sync meetings

Planning

  • every 2 weeks on Thursday at 14:00 CE(S)T
  • next meetings:
    • Thu 23 May 2024 14:00 CEST (excused: Satish)
    • Thu 13 June 2024 13:00 CEST
    • Thu 27 June 14:00 CEST (TO CONFIRM)

Meeting (2024-05-03)

Attending: Ake, Lara, Satish, Caspar, Kenneth, Sam

  • ESPResSo test
    • Due June'24 (M18) for MultiXscale milestone 3
    • Two test cases (available: )Lennard-Jones + Ionic Liquid) run fine from a manually crafted batch script
    • Not inheriting from any test, so needs to be written from scratch
    • Tolerances for sanity check unclear
    • Test cases has parameters for strong & weak scaling
    • Contacted Jean-Noel for clarification about some warnings & tolerances
    • Some warnings to be looked into, maybe test case is being run with too many MPI ranks?
  • OpenFOAM test
    • Extracted the bike mesh (8M), turns out to be quite large (1GB)
      • Need a big enough mesh to weak-scale to e.g. 16 nodes
    • How to store large files => Create issue for this
      • Look into git-lfs, see if that would be a suitable way to distribute such files
        • E.g. could be nice if we can version those in a synchronous way with the test-suite releases
      • Alternatively, explore if we can host it in cvmfs, e.g. data.eessi.io
        • How should this repo be structured?
        • Data files in there should be "versioned" somehow, maybe based on checksum?
      • or simply in an S3 bucket in AWS?
        • easy to set up through sponsored AWS credits
    • Playing with tolerances to run a more meaningful simulation
    • exaFOAM benchmarks
      • which OpenFOAM version is required for this?
      • is deliverable with details on these benchmarks available already?
      • Lara & Kenneth will probably bump into Martha at ISC'24 in Hamburg
  • problems with TensorFlow test during demo at EUM'24
  • open PRs
    • LAMMPS (https://github.com/EESSI/test-suite/pull/131)
      • Check with Tilen if we can implement a sanity check that checks scientific correctness of the result
    • QuantumESPRESSO
      • Satish will look into scheduling meeting with Davide to discuss his PR (should also involve Caspar & Alan)
      • base test on which this is based in hpctestlib is now included in latest ReFrame release
    • CP2K
      • Sam added README
      • still need to look into hook to set $OMP_NUM_THREADS (common between this PR and LAMMPS PR)
    • PyTorch
      • some trouble getting it to run multi-node
  • issues
    • suggestion to make scale tags more unique (issue #136)

Previous meetings

Clone this wiki locally