RE: [fluka-discuss]: running time

From: Vasilis Vlachoudis <Vasilis.Vlachoudis_at_cern.ch>
Date: Tue, 10 Jun 2014 07:40:53 +0000

Dear Yanyuanlin,

this is a VERY complicated topic and there is no straight forward answer.

What is invariant is the number of calculation your CPU does per second. Therefore you have to "guide" your
simulation towards the region (in the whole phase-space) of interest and avoid unnecessary calculations elsewhere
without compromising the results.

The best way to learn that is to follow our advanced course :) in the mean time please give a look on the

- Biasing: https://indico.cern.ch/event/296741/contribution/20/material/slides/0.pdf
   Increase the importance towards the region of interest and decrease everywhere else (BIASING). How?
   Bias interactions with LAM-BIAS if needed and use the EMF-BIAS

- Transport: https://indico.cern.ch/event/296741/contribution/24/material/slides/0.pdf
  Set appropriate thresholds depending your problem. Remember that everything ends up in electromagnetic
  showers, which they will eat most of your CPU time. So set the appropriate threshold there (EMF-CUT, and don't forget the EMF-BIAS)

- Geometry: optimize the geometry for faster tracking: https://indico.cern.ch/event/296741/contribution/7/material/slides/0.pdf
   Check the concept of zones, use the simplest possible definition of zones, larger in size and possibly overlapping WITHIN the same region only.

- Statistics and Sampling: https://indico.cern.ch/event/296741/contribution/15/material/slides/0.pdf
  variance * time = constant

- Brute force: Use more in number and faster CPU's. (Check the spawn feature in flair)

All the above (apart the last one) are not physics but an art and cannot be explained in an email.

Vasilis.


________________________________
From: owner-fluka-discuss_at_mi.infn.it [owner-fluka-discuss_at_mi.infn.it] on behalf of ãÆÔ¨ÁÖ [yanyuanlin_at_impcas.ac.cn]
Sent: 10 June 2014 03:41
To: fluka-discuss_at_fluka.org
Subject: [fluka-discuss]: running time


Dear FLUKA experts,

How can reduce the running time? When the erergy is high and the USRBIN scoring step is small, my simulation is as long as about 20 days.The card
START' what(3) can't be set up. If the what(6) is reduced, the result seems wrong.Is there any settings can reduce the running time?

Best Wish!
Yanyuanlin



Received on Tue Jun 10 2014 - 12:10:31 CEST

This archive was generated by hypermail 2.3.0 : Tue Jun 10 2014 - 12:10:32 CEST