Benchmark WG Meeting

Wednesday 16 Apr 2008, 17:00 → 18:00 Europe/Zurich

1/1-025 (CERN)

Helge Meinhard (CERN-IT)

To join both the web and audio conference (recommended), click here: https://audioconf.cern.ch/call/0160359 If you only want to participate in the audio conference, you may dial in with the following numbers: Call: +41227676000 (Main) Enter access code: 0160359

CPU Benchmark WG Meeting Notes, April 16, 2008


Present:
Alex Iribarren, Helge Meinhard, CERN/IT
Franco Brasolin, CERN/ATLAS
Manfred Alef, FZK Karlsruhe
Peter Hristov CERN/ALICE,
Peter Wegner, DESY Zeuthen
Michele Michelotto, INFN Padova
Ian Gable, U Victoria
Sandy Philpot, JLab

1. Comments on minutes from April 8

• no comments

2. Opening discussions

• Helge's top priority is a more rigorous statistical analysis of the results. We need a fit over all and one for each experiment.
  
• Helge wants to determine how well the results fit a straight line and what the scaling factor is.
• Peter Hristov volunteers to perform the analysis and will distribute his ROOT macros.
• Helge clarified that Jeff Templon's comments were that he wanted to see a better analysis and that normalizing to a machine with insufficient memory is a poor choice.
• Helge reiterated that we are looking for a way to measure machine performance that is "Precise enough, but still practical."

3. Review of experiment progress

Peter Wegner commented that the he has heard that ATLAS release 14 will be made package such that it's easily portable to other machines. He hopes the same for CMS.

Franco Brasolin (ATLAS)

• waiting on release 14, which he expects the end of this week. He noted that there was a problem with the event generators that has caused the delay.

Peter Hristov

• Peter also ran on one of Michele's machine (need clarification of which one).

4. AOB

• We have agreed to a fate-to-face meeting during the Friday afternoon of  of the HEPiX Meeting(May 9).
• There was interest in seeing a presentation of profiling SPEC code during the Benchmarking track of HEPiX.
• Helge brought up how we might best present our results at HEPiX:  One short talk from each experiment and and summary or just a summary. The issue remains open

5. Action items

• Franco to rerun benchmarks on ATLAS release 14 when available
• Michele to rerun benchmarks on his machines using the gcc 3.4 compiler.
• Peter Hristov will run the statistical analysis of the results and distribute his ROOT macro.

6.Next Meeting

• Next meeting will be held April 23, 15:00 UTC.