[gate-users] Re: How to make uncorrelated random number seeds?

Ananth Ravi ananth.ravi at sw.ca
Mon Dec 20 11:36:23 CET 2004

Hi Theo,

Thanks for you assistance, your method seems to be working like a 
charm.  I am now running my four simulations in tandem, and the first 
thousand events are completely unique.


>I have read all of your previous discussions on controlling the random
>number seed to ensure that
>each run is independant and have tried to follow these instructions to only
>have my system perform
>four identical runs.
>I am running 4 parallel simulations each at different time intervals of GATE
>on my dual processor machine.  I have dedicated a separate folder to each
>simulation within which i have placed beginOfRun.rndm,  endOFRun.rndm, the
>macro file, and the GATE executable.
>In each copy of my macro file I have the following code:
>/gate/output/rootsetSaveRndmFlag 1
>/random/setSavingFlag 1
>In my output, I find the exact same event numbers in each of the four hits
>files, the only difference is the time of each event, which are just offset
>by the starting point of the interval for that run.
>This is disconcerting, since I have run several simulations only to recently
>discover that none of them are really admissable.
>You help is always appreciated.
>Ananth Ravi
>Message: 2
>Date: Fri, 17 Dec 2004 18:08:11 +0100 (MET)
>From: "Mutanga Theodore" <muathe02 at student.umu.se>
>Subject: Re: [gate-users] How to make uncorrelated random number
>	seeds?
>To: "GATE feedback and helpline for Users"
>	<gate-users at lphe1pet1.epfl.ch>
>Message-ID: <40016. at>
>Content-Type: text/plain;charset=iso-8859-1
>Hello There,
>Please refer to some of the posts I made earlier. I solved this by making
>a mock simulation with a very weak source like 10Bq or even less. Then I
>made a script that runs this simulation say 100 times (four times in your
>case), in each instance, I get the file endOfrun.rndm , rename it to
>status00i.rndm where i goes from 1 - 100 or whatever number of randoms you
>want to generate. ( in your case four maybe)
>Then use this status00i.rndm files and force your simulations to use them
>by using the comand /random/resetEngineFrom .
>Atleast with this method you have different seeds to begin your
>simulations with, however if they are uncorrleted or not is another issue,
>I made a post about using this method but no one commented about it so Im
>using it for now and Im getting some results
>hope this helps

More information about the Gate-users mailing list