I recently tried to launch a new processing with commit: https://github.com/sct-pipeline/csa-atrophy/commit/64d99c1f3cddec9f1a75da2df2ece6209bf2b09e. After a few minutes, i received a mail from Compute Canada entitled "Too Many Threads for Graham Jobs". It explained that the "submitted jobs asked for 32 CPU cores but it was starting many more threads than that".
The proposed solution is to add in job_template.sh after any #SBATCH commands and module loads the following lines:
export OMP_NUM_THREADS=1
export MKL_NUM_THREADS=1
For now, i have cancelled the submitted jobs.
Because 1 subject = 1 CPU (no multiprocessing on each subject), i think we should implement this improvement.
I recently tried to launch a new processing with commit: https://github.com/sct-pipeline/csa-atrophy/commit/64d99c1f3cddec9f1a75da2df2ece6209bf2b09e. After a few minutes, i received a mail from Compute Canada entitled "Too Many Threads for Graham Jobs". It explained that the "submitted jobs asked for 32 CPU cores but it was starting many more threads than that". The proposed solution is to add in job_template.sh after any #SBATCH commands and module loads the following lines: export OMP_NUM_THREADS=1 export MKL_NUM_THREADS=1
For now, i have cancelled the submitted jobs. Because 1 subject = 1 CPU (no multiprocessing on each subject), i think we should implement this improvement.