AMBER Archive (2006)

Subject: Re: AMBER: Amber9 test error on SGI Altix

From: David A. Case (
Date: Mon May 08 2006 - 13:07:17 CDT

On Mon, May 08, 2006, wrote:

> Compilation went peacefully(both serial and parallel 64 bit
> compilation) on SGI Altix using intel fortran compiler
> 9.0.031. I am not able to figure out the error while
> running test.serial.
> %%%%%%%%%%%%%% test.serial error %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> cd dmp; ./Run.dmp
> diffing with mdout.dmp
> possible FAILURE: check mdout.dmp.dif
> ==============================================================
> cd adenine; ./Run.adenine
> diffing with mdout.adenine
> possible FAILURE: check mdout.adenine.dif
> ==============================================================
> cd cytosine; ./Run.cytosine
> ./Run.cytosine: Program error
> make: *** [test.sander.BASIC] Error 1
> On openning cytosine.out I got the following error
> vlimit exceeded for step 3; vmax = 217.6443

Your build of sander has serious problems, but it is not clear exactly how to
find them.

We have used ifort version 9.0.021 and 9.0.026 on sgi_altix. It is possible
but unlikely that 9.0.031 would introduce such serious problems.

1. What arguments did you use for "configure"? (It should be just "sgi_altix").
2. Be sure that your AMBERHOME variable points to the correct (amber9) location.
3. Try "make test.nmode" or "make test.anal". If these fail, it would point
   to some generic compiler/linker problem, and not to anything particular
   to the sander code.
3. Do you have amber8? does it work in this environment?
4. (Long shot): try compiling and testing pmemd; it will handle sgi_altix
   in a little different way. But what you are seeing is a failure in the
   very basic sander functionality, which is pretty unusual.

Beyond this, I don't have much to offer. This may require a a guru like
Roberto to figure out.


The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to