On Thu, 22 Dec 2011, Torsten Bronger wrote:
Md. Golam Rashed writes:
What is the best platform for High performance Computing?
windows is long gone, i think. So only Unix and Linux stand the chance.
As you people are expert here on Computational Science, Can you tell that expertise on what platform will strongly back my interest on Computational mechanics??
Is http://en.wikipedia.org/wiki/File:Operating_systems_used_on_top_500_supercom... what you are interested in? (The 13th fastest computer of the world resides 600m away from my office, running Linux. ;-)
As far as SfePy is concerned, it doesn't matter really I suppose. I find developing very convenient on Linux, though.
I don't think that it is sensible to try to get supercomputer resources for running an SfePy script. You can get much memory also for rather cheap servers in your basement, and I don't think SfePy can keep 16384 cores busy simultaneously.
Apropos cores, does SfePy use more than one core at all?
If the underlying blas/lapack that is called by numpy or solvers like umfpack supports multiple cores, then yes.
In the git master there is now also a, for the moment, rather naive parallel petsc-based solver. I would like to be able to run sfepy fully in parallel, but will do it as the need for my work arises - any help in this direction that might speed this up would be appreciated.