Re: amber-developers: Extra Points Calculation

From: Lachele Foley <lfoley.ccrc.uga.edu>
Date: Tue, 11 Nov 2008 14:23:35 -0500

Is there any remote chance that having only recompiled the parallel sander using the patch, but not the serial sander, would have made a difference? I think those are independent of each other (hence the requirement for "make clean"), but I've been wrong before.

I'll change min.in and do some serial/parallel comparisons, too. Might be tomorrow before I can get it done, though.


:-) Lachele

--
B. Lachele Foley, PhD '92,'02
Assistant Research Scientist
Complex Carbohydrate Research Center, UGA
706-542-0263
lfoley.ccrc.uga.edu
----- Original Message -----
From: David A. Case
[mailto:case.biomaps.rutgers.edu]
To: amber-developers.scripps.edu
Sent:
Tue, 11 Nov 2008 14:07:28 -0500
Subject: Re: amber-developers: Extra Points
Calculation
> On Tue, Nov 11, 2008, Lachele Foley wrote:
> 
> > > We thought it does, but Lachele discovered more problems yesterday
> > > evening; he was going to take a look today.
> > 
> > I kept having jobs fail...  I could only get them to behave on 2
> processors.
> 
> You should certainly check short runs on a single processor vs parallel.
> Print the results at every step....are you getting (nearly) identical
> results
> independent of the number of cpus?  If not, the problems have to be sorted
> out
> before going on to run longer.
> 
> e.g. with your MIN.in file, in my tests, I change maxcyc from 14000 to 14,
> and
> ntpr from 200 to 1.  I think going slowly here, and checking each step on
> 1,2,4, etc cpus will be essential to establishing correct behavior.
> 
> ...dac
> 
> 

Received on Fri Dec 05 2008 - 14:29:08 PST