AMBER Archive (2008)

Subject: Re: AMBER: nmode

From: David A. Case (case_at_scripps.edu)
Date: Mon Apr 21 2008 - 16:05:20 CDT

On Mon, Apr 21, 2008, jacopo.sgrignani_at_unifi.it wrote:

> i would like to perform an entropy extimation on a system aroud 7500
> atom, but I usually get same different errors.
> My ws has 8GB of RAM, the first time I tought that i Had to better
> minimize my protein but also with a more expensive minimization (10000
> cycles and 0.00001 RMS) I can only get segmentation fault error in the
> ws with 8GB of RAM, and ASSERTion ' ier == 0 ' failed in nmode.f at
> line 106, in another ws with 4GB of ram.
> Are the other softwares to perfrom this type of calculations ? Or same
> news in AMBER10?

nmode is being retired in Amber10, in favor of doing normal mode calculations
with NAB. But these are big systems. A second derivative matrix for
7500 atoms requires 8*(7500*3)^2 = 4 Gbytes just to store the matrix, and
you certainly need 64-bit OS and pointers to hold a single array that is that
large. The assertion failure at line 106 is just telling you that your system
can't allocate enough memory. I don't know where the segfault is coming from,
but your workspace must have a 64-bit OS.

You might want to print the value of memusd_x that you are trying to allocate,
and write a simple fortran program to allocate that many double-precision
words, then try that out on your system. That might help pinpoint whether the
problem is in the nmode code, or in your environment.

For these sorts of systems (about 450 amino acids, I would guess) you may well
need to investigate other ways of estimating entropies.

...regards...dac

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