Re: AMBER: Intel Fortran Compiler 9.1, MPICH, and AMBER 8

From: Andreas Svrcek-Seiler <svrci.tbi.univie.ac.at>
Date: Fri, 23 Feb 2007 12:02:52 +0100 (CET)

Hi,
> It is my impression that it is not
> necessary to use the Intel Cluster OpenMP for parallel AMBER 8 with the
> Intel Fortran compiler 9.1 on Linux systems,
...I'm not aware of any openMP parallelization within amber.
At least a "grep -ilr openmp ." inside my amber8-home
found nothing. Besides, cluster-openMP seems new, intel-proprietary
and quite costly (the cluster-openmp license costs much more than the
rest of the compiler).
> and if I understood a recent
> e-mail on the reflector correctly, the Cluster OpenMP does not provide such
> good performance.
...This would definitely depend on the problem at hand, on the way the
solution is implemented, and of course on your hardware.

In general, my impression is: On average, the intel compilers generate
faster code (even for AMD processors) than their competitors, which claim
AMD-specific optimizations. This is of course only my personal opinion,
based on a (too) limited set of benchmarks to draw general conclusions.

best,
Andreas
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Feb 25 2007 - 06:07:31 PST
Custom Search