From: Giuseppe Battistoni (Giuseppe.Battistoni@mi.infn.it)
Date: Thu May 05 2005 - 09:13:28 CEST
Yes! Now it works and this mechanism is actually already in use
by many of us.
There is an alternative, that is to prepare a set of different random
seeds (analogous to the random.dat file that you have in the $FLUPRO
directory) to start jobs on different machines.
This would avoid to loose some time in job initialization,
but sometimes people get confused in doing this, therefore
I would say that the use of different WHAT(2) values
in RANDOMIZE command is a less error-prone procedure.
On Wed, 4 May 2005, LEE, KERRY T. (JSC-SF) (UHCL) wrote:
> Date: Wed, 4 May 2005 15:23:52 -0500
> From: "LEE, KERRY T. (JSC-SF) (UHCL)" <firstname.lastname@example.org>
> To: "'email@example.com'" <firstname.lastname@example.org>
> Subject: Parallel Monte Carlo with FLUKA
> I would like to run different FLUKA runs on multiple processors/machines,
> but want to avoid the overlapping of random seed sequences. I saw the
> following in the fluka-discuss archive.
> Alfredo responded to this message saying that this mechanism was not in
> place yet but would be there in the next release. Since this was almost 3
> years ago I assume this mechanism exists now in the April 2004 release, is
> that correct?
> If I understand what Alberto said in his mail, I should be able to put in
> any number less than 9.E8 in what(2) of RANDOMIZE and not have to worry
> about repeating random sequences for different runs as long as I use
> different values in what(2) of RANDOMIZE. Is that correct?
This archive was generated by hypermail 2.1.6 : Thu May 05 2005 - 09:18:49 CEST