From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Mon Mar 31 2014 - 04:13:21 CDT
Hi Francesco,
> -----Ursprüngliche Nachricht-----
> Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im
> Auftrag von Francesco Pietra
> Gesendet: Montag, 31. März 2014 09:41
> An: Norman Geist
> Cc: Namd Mailing List
> Betreff: Re: namd-l: remd across gpus
>
> From what you say, I can only conclude that we failed to set the
> correct
> flags on that IBM cluster (which is largely unbalances as to the
> GPU/CPU
> ratio).
Well I guess you need the following conditions to be fulfilled:
1. You need at least num_replicas*2 cores. The 1st two processes of each replica must be on the same node. Therefore
Your machine file may not have unique fashion, but one entry for every slot.
So NOT:
C01
C02
C03
BUT:
C01
C01
C01
C01
C02
etc.
2. Job scheduler should only give you "full" unshared nodes. Not your chosen amount of cores distributed all over the machines. See 1.
3. num_cores_per_node / (num_cores/num_replicas) should be integer.
4. Regarding the performance, you need to simply see, that each replica is a independently started simulation with own resources. So it will run like every canonical md using the same resources.
Good luck.
>
> Could you please reveal details on the hardware / flags / size of the
> system that might help me to find the correct conditions on the IBM
> iDataPlex DX360M3? In particular, for a T-REMD on a peptide of ca 500
> atoms under GB conditions, do GPUs really help? Or are you carrying out
> T-REMD in a water peridic box, where, because of the much larger size
> of
> the system, certainly GPUs could help?
>
> thanks
> francesco pietra
>
>
> On Mon, Mar 31, 2014 at 7:39 AM, Norman Geist <
> norman.geist_at_uni-greifswald.de> wrote:
>
> > Hmm O_o ??
> >
> >
> >
> > I’m running REMD without problems over our four Dual-GPU-Nodes with
> NAMD
> > 2.10 NB from many months ago. Why should this be a problem?
> >
> >
> >
> > Norman Geist.
> >
> >
> >
> > *Von:* owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] *Im
> > Auftrag von *Francesco Pietra
> > *Gesendet:* Samstag, 29. März 2014 08:29
> > *An:* NAMD
> > *Betreff:* namd-l: remd across gpus
> >
> >
> >
> > Hello:
> >
> >
> >
> > On Tue, May 10, 2011 at 4:29 PM, Massimiliano Porrini
> >
> <M.Porrini_at_ed.ac.uk<M.Porrini_at_ed.ac.uk?Subject=Re:%20%20REMD%20ac
> ross%20gpus>>
> > wrote:
> > *> Dear Axel, *
> > *> *
> > *> Thanks for the useful reply. *
> > *> *
> > *> Only one clarification: *
> > *> *
> > *> Your word "mess" was referred only to the hybrid (gpu/cpu)
> simulation *
> > *> or also to the REMD across multiple GPUs ? *
> >
> > mostly.
> >
> > *> I mean, is REMD across GPUs easily doable? *
> >
> > i don't know. as far as i know, the NAMD implementation
> > is entirely based on Tcl script code using a socket interface.
> > that is an elegant way of implementing REMD without having
> > the rewrite the executable. i did something similar a while ago
> > in python using pypar and HOOMD and it was fun to see it
> > work, but somehow a bit unsatisfactory.
> >
> > i know of other codes that have such features and other
> > multi-replica method implemented on a lower level and
> > they appear to me more elegant to use. that may be
> > the price to pay.
> >
> > cheers,
> > axel.
> >
> >
> >
> > translated into the present, we found it difficult to run t-remd
> across
> > cpu/gpus with either namd2.9 or namd2.10 (remd hard cored into the
> latter).
> > For example, on IBM DX360M3. Difficult, means that it failed in our
> hands=
--- Diese E-Mail ist frei von Viren und Malware, denn der avast! Antivirus Schutz ist aktiv. http://www.avast.com
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:16 CST