Skip to content

Omp num threads force 1

John Douglas Veitch requested to merge john-veitch/lalsuite:omp_num_threads_force_1 into master

Description

When running on a condor pool with dynamic slots (such as my test box at Glasgow) condor sets OMP_NUM_THREADS to whatever the value of RequestCPU is in the .sub file. This is too clever, since for the MPI MCMC jobs we only want 1 OpenMP thread, but multiple MPI processes. Annoyingly condor over-rides OMP_NUM_THREADS even when it's specified in the .sub file with environment = OMP_NUM_THREADS=1 A workaround is to set this in the wrapper script. This shouldn't affect the CIT cluster but may be responsible for problems on other clusters. There seems to be a race condition when using margdist with >1 OpenMP thread in MCMC.

API Changes and Justification

Backwards Compatible Changes

  • This change introduces no API changes
  • This change adds new API calls

Backwards Incompatible Changes

  • This change modifies an existing API
  • This change removes an existing API

If any of the Backwards Incompatible check boxes are ticked please provide a justification why this change is necessary and why it needs to be done in a backwards incompatible way.

Review Status

Please provide details on any reviews related to this change and and the associated reviewers.

Merge request reports