I was referred to this list from the SGE and Rocks cluster lists. If this question is inappropriate for this list I apologize.
-- original post -- I thought someone on this list might have a suggestion for me, if I'm completely off base posting this here please let me know. We are doing some benchmarking of our clusters and in the process of working with our user base and discussing what types of jobs are being run it has become apparent that its difficult to tell what kind of precision many of our apps favor. Essentially we know our hardware capabilities and have decent RPeak values for performance based on single and double precision operations but its hard to tell which way our jobs are leaning. Its either mostly single or double precision operations or some hopefully quantifiable ratio of both. Does any know of any tools available that we might use to help us profile our jobs, their resource usage, single v.s. double precision operation, simple FLOP values etc. I'm not too concerned about the performance weight of the tool as it will only be run in test scenarios to get a feel for what we have and not run with every job. A help would be greatly appreciated. Cheers, Lucas Sheehan _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
