Too long running time

Dear experts,

My input file running time is too long and statistical error was around 9.6%. I tried the run with increased EMFcutoff 10 times for transport and production cards, the number of spawns and primaries but did’t work. It took 3 hours with the number of primaries was 10^6 with high statistical error and 21 hours with 10^7 primaries.
I’ve read someting about long running time in the forum in case of trapping photons in such simulation geometry. Is there a way that I can reduce the running time as well as statistical error?

Best regards,
mg
cyc.inp (3.4 KB)

Dear @mgulmen,

Thanks for your question ! For the transport settings you have chosen and your simulation geometry, the CPU time per primary is reasonable: I have tested on my machine and the order of magnitude is similar. The question is if these settings are really needed and what statistical accuracy you want / need to achieve on the quantity that you have to score. I have some general recommendations that could help you:

  • Some neighbouring regions (for instance around the NaI detector) do not have the same electron/photon energy threshold. You should preserve the electronic equilibrium by setting the same thresholds everywhere (the overall CPU penalty is not excessive)
  • Photons travel much more than electrons. Considering your scoring and geometry it would maybe be better to set a 10 keV e+/e- threshold and a 1 keV photon threshold and not the other way around. With this, I got a reduction in CPU time by a factor ~3.
  • Depending on your scoring needs you may want to increase those as well.

In case of doubts you can have a look at the slides on EM transport of the latest FLUKA Beginner Course.

Best,
Davide