Closed alchem0x2A closed 1 year ago
Many of the failed tests should be related to the bug that _find_mpi_process
cannot locate the actual mpi process, if srun
is used.
srun
controlled jobs may need other mechanism to control pause / resume by scancel
. But wee need to find which jobid.stepid
is used
Currenly unit tests work on cori haswell & knl vasp builds, while perlmutter doesn't work (yet). Will keep track on this issue and create a table of compatible vasp builds section in README
Some unit tests fail on NERSC with their native builds (shifter image works). Need to investigate the reason and apply patches if needed
srun
on slurm systems (may need another PR)