Message: time is money ;) Not Logged In (login)
 Next-in-Thread Next-in-Thread
 Next-in-Forum Next-in-Forum

Question time is money ;) 

Forum: Run Management
Date: 14 Jan, 2011
From: Mike <Mike>


I am using/learning Geant4 for my graduate project and have run into a time issue. In my simulation each run is 1 event consisting of ~100,000 different particles generated using 'for' loops. Each particle with different initial momentum directions and energies I set in PrimaryGeneratorAction.

The geometry consists of a 10 million voxel logical detector constructed using G4RegularNavigation for optimization per the DICOM example. Most of these voxels are meaningless to me (air), however.

My physics list is also similar to the DICOM example, except for an increase in the defaultCut to 1.0*mm instead of 0.01*micrometer. (These are primarily the Livermore low energy models).

To complete 1 run/event of 100,000 particles it takes approximately 1.5 minutes, and approximately 9 minutes to complete 10 events(beamOn(10)).

Are there some recommendations on further "speed up" techniques I could attempt? Any and all help is greatly appreciated. I am also somewhat of a C++ newbie, so be kind. I am hoping to accomplish 1x10^9 runs total using between 100-200 independent machines running the simulation in "clumps of events", but even with these resources I am looking at months and months of computational time.

Thank you, Mike

Inline Depth:
 1 1
 All All
Outline Depth:
 1 1
 2 2
 All All
Add message: (add)

1 None: Re: time is money ;)   (Mike - 19 Jan, 2011)
 Add Message Add Message
to: "time is money ;)"

 Subscribe Subscribe

This site runs SLAC HyperNews version 1.11-slac-98, derived from the original HyperNews

[ Geant 4 Home | Geant 4 HyperNews | Search | Request New Forum | Feedback ]