From: "V. Daniel Elvira" 
Date: May 12, 2011 9:14:45 PM GMT+02:00
Cc: "geant4-sb (Geant4 Steering Board)" 
Subject: Info for the SB meeting

Dear all,

Since I will not be able to attend the SB meeting tomorrow Friday, I am sending some information related to a couple of action items on my plate:

1- Validation Framework
Andreas Schaelicke is moving from DESY to Edinburgh and there would be nobody to develop and maintain the validation framework database there. We requested and successfully obtained resources here at FNAL to host it. Hans has already created the database and is currently testing the infrastructure. Now we need a web server for the web interface. Incidentally, Hans is at DESY Zeuthen at the moment and he intends to copy the DB contents and export to FNAL

2- G4 memory profiling
Please see below my report on the special meeting on this topic we had after last SB meeting. Since then, Philippe, Krzysztof, and Julia met and decided to select the novice ExampleN04 as the simple application to profile for time and memory. Remember that it is not clear we can always profile CMS because of potential inconsistencies between the experiment software and the G4 reference releases. The idea is to modify ExampleN04 and make it more suitable for useful time and memory studies. This is in progress.

Cheers, Daniel
A. Dotti, G. Folger, G. Cosmo, and Ph. Canal met on Tues 19th of April to discuss G4 memory profiling.
• Goal: perform memory profiling of relevant G4 simulation applications with some frequency. For example, (almost) every time there is a reference release.
• Applications: the best candidate is the CMS simulation application. The problem is that it is not guaranteed that CMSSW will work with each new reference release. Candidate samples to profile are ttbar (semi-leptonic), high pT jets, Zprime. A. Dotti’s simple calorimeter with no magnetic field is not very useful to profile memory but he will use it to profile time (and memory anyway) and submit this info to the FNAL team. His findings may trigger or not the need of profiling CMS for cpu time.
• Memory tools: Ph. Canal suggests the use igprof for CMS to take memory snapshots . CMSSW can be configured to take these snapshots somewhere at the end of each event. We should look at the total number of memory alloc and dealloc and the memory size for a process. We should also look at how much memory is used at one moment in time.

Action items:
- The FNAL team will discuss how to integrate the time and memory profiling exercises for the CMS simulation application. K. Genser, Ph. Canal, J. Yarba will meet to discuss details in the week of April 25th.
- A. Dotti will profile for CPU time (and memory) the simple calorimeter example without magnetic field for every G4 reference release.