I'd decided that Python was to be my next learning project, but it looks like that's going on hold and I'm going to actually learn MPI. My 3D convection code is parallelized. Not massively, 12 processors (6 nodes) is typical, but it does make heavy use of MPI. I should point out that I'm not the one who originally parallelized this code. While I understand and have contributed to the C code, I've never really messed with the MPI routines. I've always been more concerneded in the science than the computing details.
But I'm getting think there's something seriously wrong with the Beowulf cluster (upsand) at work (it's old and nodes are dying and there's no funding to replace them). I don't think the problem lies with my code, but I don't know MPI well enough to really be sure that's not the case.
( Details behind cut, for the morbidly curious and bored )I expect soon to have access to the computing facilities at Berkeley, so if it's a hardware/configuration issue I'm facing, that may be solved soon. But it's clear that I should learn MPI, if I'm to continue in this field. I really do have odd gaps in my education for having the PhD. Any recommendations for resources on learning MPI? Through the UC library system, I can get access to all of the O'Reilly books online, but I welcome other suggestions. I expect soon to have access to the computing facilities at Berkeley, so if it's a hardware/configuration issue I'm facing, that may be solved soon.