AMBER Archive (2006)

Subject: Re: AMBER: pmemd9 issue

From: Robert Duke (rduke_at_email.unc.edu)
Date: Sun Jul 02 2006 - 06:46:21 CDT


Brent -
Okay, I refreshed my memory. This is likely a "framing error" problem in the prmtop, meaning that the molecule and residue lists do not agree; typically the last residue in one molecule extends into the next molecule. Go ahead and send me your prmtop/inpcrd/mdin, I will enable some debug code and look at the input prmtop data to confirm. Assuming I am correct, you will then need to talk to the [xt]leap experts to figure out how your system is incorrectly specified (might be handy to have your leap log files). At a minimum, I know that the virial won't be right for constant pressure runs with systems like this. I don't know if this is the extent of damage possible or not. The reason that parallel pmemd freaks out over this condition has to do with use of more efficient workload distribution schemes (and reduced data interchange during the run); nonetheless the system is probably not correctly defined and should be fixed.
Regards - Bob Duke
  ----- Original Message -----
  From: Brent Krueger
  To: amber_at_scripps.edu
  Sent: Saturday, July 01, 2006 11:45 PM
  Subject: AMBER: pmemd9 issue

  Hey there AMBER gurus!

  I'm having a strange problem running an MPI simulation with pmemd. When I try to start the calculation it fails with:

  | ERROR: Bad residue/molecule data in prmtop!
  | Residue 131(atoms 1990- 2026) is in multiple molecules.

  But I can successfully run the calculation using a serial version of pmemd, with a serial version of sander, and with a parallel version of sander, so I think the prmtop and rst files must be fine. (I can also read the prmtop file with ptraj or rdparm or VMD without an issue.)

  This is all amber9 built with gcc version 4.0.2 20051125 and ifort version 9.0 and lam 7.1.2. The hardware is opterons with gigE.

  I have run the test suite and everything passes nicely. Any ideas? The prmtop and rst files are a bit large so I hesitate to email them over the whole list, but if someone is interested I'd be happy to send them on.

  Cheers,
  Brent

  __________________________________________________________________

  Brent P. Krueger phone: 616 395 7629

  Assistant Professor fax: 616 395 7118

  Hope College SC 2120

  Department of Chemistry

  Holland, MI 49423

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