You are viewing an old version of this page. View the current version.

Compare with Current View Page History

Version 1 Current »

Hi all

I've been running OpenIFS with nice, even numbers of NPROC, e.g. NPROC=200, 400, so far, but now tried to run with NPROC=199 and got this error

522 ABOR1 CALLED
522 sumpini: nprtrw (approx square value) > nspecresmin

Digging into the source code, this seems to be OpenIFS trying to distribute zonal waves and vertical levels across MPI tasks and failing to reach a good number. But I'm at a loss as to why NPROC=199 is a bad number and NPROC=200 is good. 

Does anyone have a good "rule of thumb" on how NPROC should be set for a given configuration, e.g. T159L91 or T639L137? 
Must NPROC always be an even number? Or a product of any two numbers? Or a multiple of the T or L numbers? 

Best regards
Joakim


  • No labels