GSI Forum
GSI Helmholtzzentrum für Schwerionenforschung

Home » PANDA » PandaRoot » Bugs, Fixes, Releases » First results of TPC code profiling
Re: First results of TPC code profiling [message #11421 is a reply to message #11419] Sat, 08 January 2011 11:32 Go to previous messageGo to previous message
Felix Boehmer is currently offline  Felix Boehmer
Messages: 149
Registered: May 2007
Location: Munich
first-grade participant

From: *adsl.alicedsl.de
Hi Stefano,

I first tried to use TMemStat, but for some reason I could not even get it to work. I hoped that in this way I would be able to directly associate events and memory consumption.

The plot you see was created with the google performance tools and pprof. It works by replacing the normal libcmalloc for the program you want to examine, thus enabling heap analysis. This can be done by a LD_PRELOAD environment variable, so no re-compiling or re-linking is necessary.

However, it takes a while...

I have not tried DPM yet simply for time reasons. The generator used will of course not change the task behavior. But I do believe that with DPM you will get a lot more TPC objects (all the slow protons!) and will probably run into these kind of memory problems much quicker / more often.

I hope I could help to at least make things a little clearer.


Cheers,

Felix
 
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic: Pandaroot crash due to TParticlePDG.h
Next Topic: error while making pandaroot
Goto Forum:
  


Current Time: Sat Jun 15 22:29:36 CEST 2024

Total time taken to generate the page: 0.01120 seconds