[gmx-users] Re: request for help.
feenstra at few.vu.nl
Mon May 30 12:19:50 CEST 2005
aldo jongejan wrote:
> Hi Diego,
> I am happy the SGI MPI mdrun executable works!
> I have no idea why there is no improvement when
> running with only 2 processors..I was just happy
> I got it up and running and I assumed everything
> worken and would give me some improvement.
> The size of my system also pushed me to running on
> a Linux cluster (CPU > 8) and I never pursued this
> optimizing this executable. I have also send this
> mail to a collegue of mine, how is much more experienced
> in these matters. I hope, he is able to help you..
I'll try ;-)
Depending on what system you are simulating, scaling (i.e. performance
increase with increasing numbers of processors used) can be very
different. In general, Gromacs does not scale very well, and it can be
bad for (very) small systems. But the behaviour you report, with no
improvement from 1 to 2 CPU's, but good scaling from 2 to 4, I have not
If you can tell me what type of system you run, e.g. # atoms, size, type
of molecules, and simulation parameters like cutoff or ewald, I may be
able to tell you more.
I've cc'd this mail (yes - now really!) to the Gromacs mailing list, but
you need to become a member before you can send mails to or receive them
from the list.
(I'll monitor replies to this mail and forward them to you...)
* NOTE: New Phone & Fax numbers (below) *
| | |
| _ _ ___,| K. Anton Feenstra |
| / \ / \'| | | Dept. of Pharmacochem. - Vrije Universiteit Amsterdam |
|( | )| | | De Boelelaan 1083 - 1081 HV Amsterdam - Netherlands |
| \_/ \_/ | | | Tel: +31 20 59 87608 - Fax: +31 20 59 87610 |
| | Feenstra at few.vu.nl - www.few.vu.nl/~feenstra/ |
| | "If You See Me Getting High, Knock Me Down" (RHCP) |
More information about the gromacs.org_gmx-users