Closed cookpa closed 1 year ago
Hi,
Thanks for making this nice software! This PR addresses a performance issue mostly affecting HPC users. Thanks @valeriejill and @mattcieslak for finding the problem
I think it is better to be consistent with the use of CPU cores in the whole application. Thank you @cookpa, @valeriejill and @mattcieslak for the fix! I will update the binaries later too. This will go to v0.9 release.
The tracking uses the number of threads set on the command line (
GENERAL::numberOfThreads
), but the FOD discretization usesMT::maxNumberOfThreads
, which doesn't get changed from its default of all CPUs. On an HPC cluster, this causes a spike in resource uses, leading to jobs being terminated.