[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: PetSc

Adam C Powell IV wrote:
> > For performance of mpich and lam, I cannot comment on that.
> I've heard rumors of lam's advantages, but nothing convincing.

The only way to verify such a thing is to run standard benchmarks
on a true cluster and check the wall clock. LAM has some nice optimizations
which I don't think mpich implements on clusters. Though, mpich has some
legacy code for optimized stuff on legacy supercomputers... It's quite
easy for us to run such a benchmark against these libs; we have both
libs installed at the moment. We leave it to the programmers to use one
of them, but I actively test only LAM.

Such a study will also be valuable for recommending the software to
use to parallel code programmers here, and on other similar Beowulf
class systems.


Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: erayo@cs.bilkent.edu.tr
www: http://www.cs.bilkent.edu.tr/~erayo

Reply to: