From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Fri Aug 16 2013 - 00:32:09 CDT
Hi,
Measure the r/w speeds for the volumes that hold the simulation inputs,
that's what came to my mind at 1st. You can use dd for that. Example:
dd if=/my/volume/apoa1/whatever.psf of=/dev/null bs=1M <-- measure read
performance
dd if=/dev/zero of=/my/volume/mytestfile.img bs=1M count=100 <-- write
performance
Give us the outputs of that, afterwards we can dig deeper.
Regards
Norman Geist.
> -----Ursprüngliche Nachricht-----
> Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im
> Auftrag von Jan Fredin
> Gesendet: Donnerstag, 15. August 2013 23:25
> An: namd-l_at_ks.uiuc.edu
> Cc: Jan Fredin
> Betreff: namd-l: apoa1 slow Phase 1
>
> Hi,
>
> I am working with a customer using an SGI shared memory UV machine and
> an SGI MPT build of NAMD following the MPI build method. They are
> seeing the standard 500 step apoa1 test run with slow WallClock time
> but the seconds/step is basically correct. Can you tell me what
> contributes to the Startup phase 1 part of the computation and what
> might be causing the excess time usage there?
>
> Customer site: Info: Finished startup at 194.537 s, 82.7031 MB of
> memory in use
> Standard SGI: Info: Finished startup at 1.56274 s, 82.7031 MB of
> memory in use
>
> Customer site: Info: Startup phase 1 took 186.643 s, 64.5078 MB of
> memory in use
> Standard SGI: Info: Startup phase 1 took 0.175269 s, 64.5078 MB of
> memory in use
>
> Customer site: TIMING: 500 CPU: 25.7678, 0.0468977/step Wall:
> 25.7678, 0.0468977/step, 0 hours remaining, 82.703125 MB of memory in
> use.
> Standard SGI: TIMING: 500 CPU: 25.8591, 0.0466062/step Wall:
> 25.8591, 0.0466062/step, 0 hours remaining, 82.703125 MB of memory in
> use.
>
> Thanks,
> Jan
>
> --
> Dr. Jan Fredin
> SGI, Member Technical Staff - Technical Lead
> Austin, TX
> 512-331-2860
> jfredin_at_sgi.com
This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:23:35 CST