%CERTIFY%
GuilhermeLimaSandbox
Understanding memory usage reports
I am trying to understand a little better the memory usage numbers reported by Athena, as part of the inert material validation. My idea was to switch some systems on/off, and look at the effects in terms of memory usage.
I start with Muon system only (MSOnly), but having a full detector as a reference (-full-). MSfull includes both chambers and inert material. old- means that I am using the old description for inert material (
GeoModel-based) --- I will add AGDD-based volumes soon, but I am trying to understand a few puzzles first.
So here is a first batch of numbers:
Configuration |
Muon |
Field |
Physics |
G4init |
Geant4 |
GeoModel |
Total |
|
( c0) empty |
0.000 |
0.032 |
0.320 |
0.000 |
0.352 |
21.848 |
364.200 |
( c1) old-MSonly-inertOnly |
0.000 |
1.464 |
0.320 |
5.884 |
7.668 |
21.944 |
371.612 |
( c2) old-MSonly-chambersOnly |
20.488 |
1.464 |
0.452 |
38.456 |
60.860 |
37.148 |
440.008 |
( c3) old-MSonly-MSfull |
21.176 |
1.456 |
0.320 |
74.532 |
97.484 |
37.264 |
476.748 |
|
( cf) old-full-MSfull |
75.160 |
1.468 |
45.564 |
163.492 |
285.684 |
13.772 |
641.456 |
Open Questions
- (c1) shows that inert volumes do not contribute to Muon memory usage. There is some increase to G4init and Geant4 memory only.
- (c3 vs. c1+c2) suggests that for Geant4, MSfull is much more than chambers + inert Material. Does anyone understands why?
- (cf-c3) GeoModel requires less memory for full Atlas than for MSonly?!?
At this point it may well be that there are problems with my analysis scripts...
--
GuilhermeLima - 02 Mar 2009