> > Are you saying the np 2 is on 2 different machines - and same with > np4,8?
It really depends on how I run the parallel tasks. If I run local to a machine, rather, outside of the SGE, then "-np 2" is two processes on a single machine. If I run by submitting into our SGE grid, rather, the part of it currently set up with a parallel environment, then it is two processes on two different machines, neither of which will be the "local machine" from the above case. You'll recall I ventured the information that OpenMPI was compiled so as to be aware of SGE. OpenMPI "knows" it is running within SGE. But anyway, don't try and debug OpenMPI for me. I'll carry on with that myself and get back the PETSc list if I can get PETSc to fail in a different way as a result of any OpenMPI changes. Hopefuly, of course, it won't fail because I'll unearth problems with OpenMPI! There are a couple of things I want to try with regard to the NetBSD port of OpenMPI, actually things that I had meant to tidy up in the wake of the developments on that NetBSD port over the last few weeks. I do not actually think they affect this issue but you never know and if I remove them from the equation, then they certainly can't. Thanks for the PETSc insight, Kevin -- Kevin M. Buckley Room: CO327 School of Engineering and Phone: +64 4 463 5971 Computer Science Victoria University of Wellington New Zealand
