[AMBER-Developers] RE: candidate amber 11 from 3/31: problem with pmemd.cuda?

From: Ross Walker <ross.rosswalker.co.uk>
Date: Mon, 5 Apr 2010 13:03:08 -0700

Hi Jodi,

Can you use the configure in AmberTools instead. Do not use the ones in the
PMEMD directory anymore. They are there for historical / debugging reasons.
So do, from a clean untar:

cd amber11/AmberTools/src/
./configure -cuda gnu
cd ../../src
make cuda
cd ../test

And you should be good with pmemd.cuda in $AMBERHOME/bin/

Instructions / manual entry coming soon.

All the best

> -----Original Message-----
> From: hadden.jodi.gmail.com [mailto:hadden.jodi.gmail.com] On Behalf Of
> Jodi Hadden
> Sent: Monday, April 05, 2010 12:58 PM
> To: Ross Walker
> Subject: candidate amber 11 from 3/31: problem with pmemd.cuda?
> Hi Ross,
> I was compiling the candidate Amber11 build from 3/31
> (Amber11.31mar10.tar.bz2) for testing, and noticed a problem with
> pmemd.cuda.
> After I configure (./configure linux_em64t_cuda gfortran nopar) and
> run make, everything goes fine, except for the fact that the
> executable that is created and moved to $AMBERHOME/bin is called pmemd
> instead of pmemd.cuda.
> I tried recompling a previous amber11 version with support for pmemd
> on gpus, and the pmemd.cuda executable was created and moved to
> $AMBERHOME/bin just as I would have expected. Going back to the
> candidate build from 3/31, I also tried compiling regular pmemd for
> cpu first, then compiling the gpu version, and it seems to just
> overwrite the cpu version since the executables have the same name.
> I could not figure out any mistake I could be making with my commands
> to cause the executable to be named incorrectly, so I thought I would
> let you know in case there really was some change that needed to be
> made to the code.
> Thanks,
> Jodi
> Jodi Hadden
> Chemistry Graduate Student
> Woods Group
> University of Georgia

AMBER-Developers mailing list
Received on Mon Apr 05 2010 - 13:30:03 PDT
Custom Search