AMBER: Division by zero in sander

From: Caleb S Howe (cshowe.Princeton.EDU) <"Caleb>
Date: Thu, 28 Jul 2005 10:57:00 -0400

I'm running amber7 on an amd opteron system running solaris and have run into trouble with sander.
The programs are compiled using the sun studio compilers.

All other programs pass all tests - but sander dumps core on:

cytosine
tip4p
tip5p
ubiquitin

Running it through a debugger reveals that the crash is a division by zero occurring in dynlib.f on the line:
TMN=one/TMA(N)

Suspecting it might be a floating point issue I recompiled without optimizations and using -fprecision=extended. Still no luck.

Examining the contents of the TMA array before any division occurs
reveals every value beyond the first is 0.0. Unfortunately I don't know enough about
FORTRAN to figure out why this happening.

I've looked through the mailing list archives but no one seems to have had a similar problem.

Does anyone have any ideas as to how I might proceed to solve this?

Thanks
Caleb Howe
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Thu Jul 28 2005 - 16:53:00 PDT
Custom Search