AMBER Archive (2005)

Subject: Re: AMBER: Failed sander runs for large systems

From: Guanglei Cui (cuigl_at_csb.sunysb.edu)
Date: Fri Jul 29 2005 - 23:33:27 CDT

Hi, Pratul

I think I know the reason, but correct me if I'm wrong. I have run big
systems (the biggest is about 1.9 million, coarse grain model) with
modified sander8, and come across the same problem (slightly different
flags reported). The problem is in amber8/src/lib/nxtsec.f, which limits
the max number of lines (I think it's 9999999) that your topology file
can have. There are a couple of other places I needed to change to get
the systems going, but in general, they're very slow. Reading/writing
prmtop and inpcrd take forever (binary is better), and the parallel
performance is poor. But this is only my limited experience. I hope this
helps. Good luck!

Regards, 

--
Guanglei

Pratul K. Agarwal wrote:
> Problem: Failed sander runs with following error messages. Please note, 
> these messages are from 2 different runs:
> ERROR: Flag "IROTAT" not found in PARM file
> ERROR: Flag "ANGLES_WITHOUT_HYDROGEN" not found in PARM file
> 
> Details:
> Large system (520,000 and 1,000,000 atoms). Patches by Vlad applied for 
> system beyond 1 million atoms 
> (http://structbio.vanderbilt.edu/archives/amber-archive/2003/0469.php). 
> Runs on 16 processor machine (Cray XT3). Compiler: PGI v6.0 with MPICH.
> 
> Any help with be highly appreciated.
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu