Re: [AMBER] about error with "Must have more residues than processors!"

From: Biao Ma <jackyma1981.gmail.com>
Date: Thu, 3 Oct 2013 12:47:49 +0900

Thank you Jason.

I will note the number of processor to use.

Jacky


On Wed, Oct 2, 2013 at 1:26 AM, Jason Swails <jason.swails.gmail.com> wrote:

> On Tue, Oct 1, 2013 at 12:00 PM, Biao Ma <jackyma1981.gmail.com> wrote:
>
> > Hi, all.
> >
> > I plan to run my CpHMD job on 208 processors at 26 nodes, but it's
> stopped
> > with Error "Must have more residues than processors!"
> > Now I redo the job on 80 processors and no problem.
> >
> > Could anyone explain the reason of the error?
> > Weather the number of processors cannot more than the numbers of system's
> > residues?
> >
>
> This is only sander -- pmemd does not have this limitation. [1] Sander
> splits up the workload for different processors on a per-residue basis. If
> you have more processors than residues, some processors would be assigned 0
> residues. It is an implementation detail, but I would not expect sander to
> scale to that many processors even with GB.
>
> Also, I suggest that you run some short benchmarking calculations with
> Amber before deciding on the number of processors that you are going to
> use. More is not always better, and these short simulations can help you
> decide how to make the best use of your resources.
>
> HTH,
> Jason
>
> [1] pmemd requires that the number of atoms be at least 10x the number of
> processors (so each processor is responsible for at least 10 atoms).
>
> --
> Jason M. Swails
> BioMaPS,
> Rutgers University
> Postdoctoral Researcher
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Oct 02 2013 - 21:00:03 PDT
Custom Search