AMBER Archive (2004)

Subject: Re: AMBER: Problems with amber8 and pgf90

From: Scott Brozell (sbrozell_at_scripps.edu)
Date: Thu Jul 15 2004 - 11:23:44 CDT

On Thu, 15 Jul 2004, Marc Perea wrote:

> I'm having a problem running amber8 compiled with PGI Fortran 90
>
> $ /opt/soft/amber8-pgf90/exe/sander -O -i opd.sander -o opd.sanderout -p
> opd.top -c opd.coor -r opd.rst -inf opd.info -ref opd.refc -x opd.mdcrd
> -v opd.mdvel -e opd.mden
>
> After 1-2 seconds it produces the following output to stderr and exits :
>
> PGFIO-F-231/formatted read/unit=8/error on data
> conversion. File name = opd.top formatted, sequential access record
> = 88216 In source file _rdparm.f, at line number 1286

Did you read the error message ?
If the tests passed then ones confidence that the prmtop was corrupted
would be even greater.

The next obvious step is to isolate the bad section of the prmtop:
try rdparm and see what happens; or look at src/sander/_rdparm.f near 1286
and find the big comment, eg, READ THE H-BOND PARAMETERS;
or use the bisection method on the prmtop (delete the last half of the
file; rerun sander and see if the error is the same; by repeated
application the first exact line of corruption can be found).
If fixing the problem after locating the error is not obvious then
one should goooggle the error to see if help is in the archives.

My point is that users should make an earnest effort to fix their
problems before they send email to more than 400 mailboxes.
Recently, several Amber users have told me that they do not subscribe
to the reflector because of the large bandwidth.
The proper tradeoff between losing potentially valuable scientific
insight and teaching computer 101 is clear to me.

Scott Brozell
Scientific Software Consultant
High Performance Technologies Inc.

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