From: Martin, Erik W (Erik.Martin_at_stjude.org)
Date: Fri Mar 08 2013 - 15:44:50 CST
Info: Benchmark time: 1 CPUs 0.934806 s/step 10.8195 days/ns 3.23184e+06 MB memory
Info: Benchmark time: 1 CPUs 0.944691 s/step 10.9339 days/ns 3.23184e+06 MB memory
Here is an example of what I'm concerned about. These benchmarks are from the log file of one of my replicas. This is after something like 100 exchange attempts. According to this output I am using 3.2 TB of memory. This can't possibly be true. The resources I'm using only has 8GB per core. Even if this benchmark was somehow reporting on all the replicas, I'm only using 100 processors. Can anyone explain this to me? Its really confusing at present. It also is a bit of a problem since I'm using these calculations to essentially measure what sort of resources I'll need for an actual production run of this calculation.
Thanks a lot,
PS. I'm sorry if I sent a number of blank emails to the list - it seems something was odd with my webmail
From: owner-namd-l_at_ks.uiuc.edu [owner-namd-l_at_ks.uiuc.edu] On Behalf Of Martin, Erik W [Erik.Martin_at_stjude.org]
Sent: Thursday, March 07, 2013 6:27 PM
To: Chris Harrison
Cc: Namd Mailing List
Subject: RE: namd-l: memory demands of REMD. .
I'm using NAMD 2.9 on a very large SGI Altix shared memory machine.
From: Chris Harrison [charris5_at_gmail.com]
Sent: Thursday, March 07, 2013 6:19 PM
To: Martin, Erik W
Cc: Namd Mailing List
Subject: Re: namd-l: memory demands of REMD
You need to clarify which version of namd you're using and whether
you're using a shared memory machine or distributed cluster.
On Thu, Mar 7, 2013 at 4:16 PM, Martin, Erik W <Erik.Martin_at_stjude.org> wrote:
> Could someone who has experience running replica exchange simulations comment on the memory requirements of these type of simulations? Does it purely scale linearly with the number of replicas? For example, if it requires 450MB to run a traditional MD simulation on the structure, can I assume that it would take 4.5e4MB to run 100 replicas of this system? I can't seem to find any information searching on how namd handles this, but since it seems to just drive x number of different simultaneous simulations, this seems logical.
> Email Disclaimer: www.stjude.org/emaildisclaimer
> Consultation Disclaimer: www.stjude.org/consultationdisclaimer
-- Chris Harrison, Ph.D. NIH Center for Macromolecular Modeling and Bioinformatics Theoretical and Computational Biophysics Group Beckman Institute for Advanced Science and Technology University of Illinois, 405 N. Mathews Ave., Urbana, IL 61801 http://www.ks.uiuc.edu/Research/namd Voice: 773-570-6078 http://www.ks.uiuc.edu/~char Fax: 217-244-6078
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:21:01 CST