AMBER: amber with MPI

From: <Stefano.Pieraccini.unimi.it>
Date: Mon, 04 Oct 2004 16:40:16 +0200

Hi all,
I use amber7 on a computer cluster for parallel computation. The operative sistem is linux red hat 8.0 and I use MPI for using amber in parallel. It sometimes (rather often actually) happens that, during the run there is a VLIMIT EXCEEDED alert in the outfile, and since then the numbers of energies and so on become NAN. The problem can't be reproduced, because it may happen or not, and at different point of the calculation with the same sistem starting from the same configuration and in the same conditions. It never happens if you work on a single processor. It seems to be a problem connected with parallel calculations only (maybe a LAM problem?). The run gets to normal temination, but all the data in the outfile are NAN and obviously you don't have a final structure.
Has someone had the same problem? How could it be solved?

Thank you in advance.

Stefano Pieraccini

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Mon Oct 04 2004 - 15:53:00 PDT
Custom Search