AMBER Archive (2004)

Subject: Re: AMBER: Problems in Amber MPI version (64 bits machine)

From: David A. Case (case_at_scripps.edu)
Date: Wed Jun 23 2004 - 12:58:07 CDT

On Wed, Jun 23, 2004, Haijun Yang wrote:
>
> We successfully compiled the AMBER7 using MPICH_GM MPI and G77 compiler on
> AMD 64 OPTERON (Suse OS) system. And it can run some benchmark jobs well.
> At the same time, we can run another GB job smoothly . We didn't use
> configure (this is for AMBER8?).
>
> The weird thing is that: when we recompiled AMBER7 using MPICH and G77 on
> this X86-64 system, and ran benchmark jobs, for all benchmark jobs, it is
> okay if we used 1 cpu ( "mpirun -np 1 ../../sander ....."). But if we used
> 2 cpus, gb_mb could go through, and other benchmark jobs failed due to the
> exceeding of vlimit (like my original post).

You don't say whether or not the test cases worked or not. You should run
all of the sander test cases (both single-cpu and with MPI) and see what works
and what doesn't.

Note that the OS/machine combination (Suse + X86-64) didn't really exist
when Amber 7 was written, so you may have to do some experimentation. Be sure
that you can run correctly simple mpi jobs using your configuration as well.

...dac

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu