AMBER Archive (2007)

Subject: Re: AMBER: SANDER bug

From: David A. Case (case_at_scripps.edu)
Date: Wed Feb 14 2007 - 12:11:12 CST


On Tue, Feb 13, 2007, Ichinkhorloo Erdenebaatar wrote:

> Jagshemash! [means 'Hello Everybody!' in Mongolian]
>
> I hope that everybody in the high-performance computing have
> already heard about the "dynamic memory of AMBER" -- revolutionary
> new concept routinely employed in recent releases of AMBER.

[Note that if your interest is in performance, you should use pmemd, if
possible.]

>
> + ix(1:lasti) = 0 ! hurts performance, @#$!@$#
> +

> Without this
> line 'ix(1:lasti) = 0' the SANDER
> of AMBER does malfunctioning sometimes. In particular, a part of the
> array determines whether apply SHAKE to some bond or not.

Thanks for the report. Do you know what part of the ix() array is the
culprit? I looked through the obvious places where shake() uses the
ix() array, and all those places appeared to be initialized. I ran the
test cases with and without the above patch, and did not see
differences. Do you a particular example that illustrates the problem?
[Also, what compiler/OS are you using?]

...thanks...dac

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