Hi Kevin,
Were the runs single processor or multiprocessor? Were the executables used
identical? Built with the same compilers, compilation options etc? Also are
you certain the system was dedicated for the run? If you are sharing a
system with other users (i.e. not using a properly configured queuing
system) then you will see massive variation in the wallclock time.
Also, was the test case large enough such that the benchmarks are
statistically relevant? Typically you want a total runtime of at least 10
seconds or so before you can get reproducible numbers. E.g. comparing runs
that complete in milliseconds will give you huge error bars on the wallclock
time.
All the best
Ross
/\
\/
|\oss Walker
| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
|
http://www.rosswalker.co.uk <
http://www.rosswalker.co.uk/> | PGP Key
available on request |
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.
_____
From: owner-amber.scripps.edu [mailto:owner-amber.scripps.edu] On Behalf Of
Glass, Kevin A
Sent: Thursday, August 17, 2006 17:18
To: amber.scripps.edu
Subject: AMBER: Execution time discrepancies
A user on our system made some test runs with Amber 8 and found significant
(50%) discrepancies in the execution times of her code. The two runs used
identical decks. Is there a programmatic reason why this happened?
Thanks,
Kevin
Kevin A. Glass, Ph.D.
MSCF - Visualization and User Services
Environmental Molecular Sciences Laboratory
Pacific Northwest National Laboratory
P.O. Box 999 MS:K8-91
3335 Q Ave. Phone: 1-509-376-1543
Richland, WA 99352 Fax: 1-509-376-0420
Email: kevin.glass.pnl.gov
Web: <http://mscf.emsl.pnl.gov> http://mscf.emsl.pnl.gov
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Aug 20 2006 - 06:07:16 PDT
