AMBER Archive (2009)

Subject: Re: Re: Re: [AMBER] Not able to compile pmemd with openmpi in Amber9

From: case (case_at_biomaps.rutgers.edu)
Date: Wed Nov 25 2009 - 15:06:53 CST


On Wed, Nov 25, 2009, Ross Walker wrote:
>
> MKL Helps with performance somewhat, although more with implicit solvent
> than explicit solvent.

Let me just underscore what Ross is saying, and augment it a bit. In your
situation, I would personally just not try to link to MKL at all. It will
save lots of grief, and won't reduce performance very much (especially for
explicit solvent).

If you *are* really in need to the best performance, you should work on
upgrading to Amber10, which will both make MKL integration easier, and boost
performance way more than spending time trying to get the very latest library
from Intel to work with a version of Amber that is 3-1/2 years old. As Ross
has pointed out, there is no way that we could have anticipated back in 2006
all the things that Intel would decide to do.

....dac

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber