Re: [AMBER] Parallel compilation of AmberTools1.5 failed

From: David A Case <case.biomaps.rutgers.edu>
Date: Wed, 15 Jun 2011 08:07:43 -0400

On Wed, Jun 15, 2011, Xioling Chuang wrote:
>
> However, for AmberTools, I can't compile as parallel.

In addition to the comments Jason made, be sure that you have a need for
a parallel build of AmberTools. This is a fairly specialized option:

1. the parallel version of ptraj requires a high-end computer with a parallel
file system to give much benefit.

2. the parallel version of NAB can be quite useful, but be sure that you are
planning to do large-scale NAB simulations. Even then, the openmp version of
NAB might be better than the mpi version -- depends on your hardware
configuration, but openmp is what I always use.

> ================mpif90 -c -ip -O3 -axSTPW -o dsarpack.o dsarpack.f

For Jason and other developers: are we trying to be too clever by using the
SSE_TYPES environment variable in the configure script? Is there not a danger
that this will be set incorrectly? Do we know that it really provides a
significant performance gain?

Also, do we know that -axSTPW is correct for *all* intel compiler versions
prior to version 11? (This is what the configure script seems to imply).
Again, are we getting any real performance boost here?

....dac


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jun 15 2011 - 05:30:03 PDT
Custom Search