Andreas Svrcek-Seiler wrote:
>> Thanks for the info. Just to make sure I have this right, this is
>> for NAB molecular dynamics, not AMBER, correct?
> ...yes, but the most important features of AMBER/sander with respect to
> implicit solvent are available in NAB, too. So, given the same options,
> both should give equally "valid" trajectories.
Oh yes, I agree. I just wanted to clarify for the list, mostly.
Some subscribers (most?) may not be aware of the existence of NAB.
>
>> Is this the regular 5.1.2 version of NAB? I had forgotten that
>> NAB MD was parallelized, I guess.
> ...I tested the cvs version from July 16 (and right now ver. 5.1.2,
> since you asked -> same scaling behavior).
>
That is nice to know.
>> Wonder why the simulation engine in NAB seemingly parallelizes a
>> little better than AMBER?
> ...NAB (as I tested it) uses openmp parallelization, which is intended
> for SMP machines. So one might expect this scales better than MPI on
> such a machine (though I know hardly anything about the latter, and did
> my first 'mpirun' only today). Besides, I used the newest intel compiler
> (10.0.025), which might also have helped(?).
>
I was interested due to a series of posts over the last several
months (of which I am sure your are aware) on how well various MD
calculations scale on small scale clusters of contemporary
commercial computers, the kinds of machines that are likely to be
found in the labs of individual PIs, not in supercomputer centers.
Your posting is a valuable data point for people who are primarily
using implicit solvent methods, it seems to me.
> good luck,
> Andreas
--
M. L. Dodson
Email: mldodson-at-houston-dot-rr-dot-com
Phone: eight_three_two-56_three-386_one
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Aug 15 2007 - 06:07:29 PDT
