AMBER Archive (2000)

Subject: Re: Is PME MPI-parallelized in v.6.0?

From: Marios Philippopoulos (mphilip_at_sickkids.on.ca)
Date: Tue Feb 15 2000 - 09:58:31 CST


Many thanks to Michael Crowley and Dave Case for their replies to my
enquiry.

I have been able to run most test routines successfully using mpi
parallelization. In BASH the command to use for running mpi-parallel
jobs (say on 4 processors) is:

DO_PARALLEL='mpirun -np 4' <executable>

The unsuccessful tests with MPI were:

LES_CUT
test.gibbs (I got the message "MPI version of gibbs is currently
disabled")
test.sander_classic

test.leap was partly successful (some routines indicated "possible
FAILURE").

It may be that some of the unsuccessful tests are simply not
(MPI-)parallelized (?). Since I'm new to AMBER, I don't know the answer
to that at this point.

Cheers,

Marios Philippopoulos

-------------------------

Michael Crowley wrote:

> Dear Narios,
> Although you will get other answers I send this in case you do not.
> Sander is only mpi pararallelized in amber6.
> On the origin, you need to compile with -DMPI in the
> MACHINEFLAGS (use the Machine.sgi_mpi).
>
> Then you must use mpirun to run on the origin.
> To run the tests in parallel,
>
> setenv DO_PARALLEL "mpirun -np 4 "
>
> before trying to run the tests.
> See if that helps.
> Mike

--
Marios Philippopoulos, PhD
Structural Biology and Biochemistry Programme
Hospital for Sick Children
555 University Avenue
Toronto Ontario Canada M5G 1X8

mphilip_at_sickkids.on.ca