Message: Re: Unique Events Not Logged In (login)
 Next-in-Thread Next-in-Thread
 Next-in-Forum Next-in-Forum

Note Re: Unique Events 

Forum: Multithreading
Re: None Unique Events
Date: 06 Aug, 2016
From: Andrea Dotti <Andrea Dotti>

Hello Anders, your description of the logic is correct. A clarification though: the "global" seed is used to generate (randomly) the list of seeds for each event. Thus you will have repetition only after you reach the end of the periodicity of the random-number-generator engine (that in general is not the size of the seed). I believe that all of the engines we provide have a very large period. Clearly (random) repetition in this list of seeds may happen. However since random number engines usually require 2 seeds, the probability that two events have exactly the same couple of seeds is small (never observed myself). Finally we keep the seeds as "long" (64 bits on Linux and Mac 64bits OS). It is possible to overwrite all this by your own user classes, but it is not trivial. If you want to change this behavior I'll provide more information.

For reproducibility: As you correctly stated only if an event does not depend on the history you get strong reproducibility. Unfortunately in current G4 releases there is still an issue with Radioactive Decay Module that breaks strong reproducibility (we are working on a solution for next release). Note that we believe physics is correct. Finally note that we do not guarantee reproducibility of merged results (scorers, histograms and so on) since floating point operations are not in general associative.

Some more info is here: https://twiki.cern.ch/twiki/bin/view/Geant4/MultiThreadingTaskForce

Andrea

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

1 None: Re: Unique Events   (herr_apa - 08 Aug, 2016)
 Add Message Add Message
to: "Re: Unique Events"

 Subscribe Subscribe

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