Hello to all,
I have a doubt about the seed selection for the cycles. I was under the impression that the seed for the first cycles was fixed and that the cycle for the next cycles was randomly chosen from the ran file.
But I was runing parallel simulations of the same .inp file on my institutions cluster and the files for cycle #2 all have the same results…
Did the seed selection mechanism change? I have version 4.2.1 installed in the cluster…
Kind Regards,
Duarte
Dear Duarte,
in FLUKA (and in any other computer code) the random numbers are pseudo random. This means, that the random numbers generated by the program are reproducible and will be always the same, if the same starting seed is used.
In case of FLUKA, the “chain” of random numbers is continued though the cycles. This allows to reproduce the whole simulation even if multiple cycles were used.
This mean, if you want to run multiple parallel simulations, you need to specify different random number seed in your input for each separate job on the RANDOMIZ card.
You may want to use Flair’s Spawn option on the Run tab to generate the input files with different seeds, or use the programming language of your choice to write a tool which takes care of it for you.
Cheers,
David
P.S.: I would encourage you to update your FLUKA installation to the latest 4-2.2 version.
Hello Dávid,
Thank you for your answer.
Sorry for the delayed reply.
I tried running the simulations with your advice to create the .inp files with the spwan command in Flair and it worked.
Thank you.
Kind Regards,
Duarte