RE: [fluka-discuss]: How to increase NMAX value in source_TwoSteps.f. Error linking.

From: Joachim Vollaire <joachim.vollaire_at_cern.ch>
Date: Wed, 18 Oct 2017 14:38:41 +0000

Dear Adam

As you notice it seems that you are reaching a memory limit imposed by the compiler and on your hardware.

Looking at your source routine I have the impression that you want to load first a large number of particles in an array and use that you then select randomly particles in that array as primary. What you could do is to pre-process the data file and split it in 10 (as you need to gain one order of magnitude) sub-files which you can then use to do 10 independent runs (you anyway need several runs to derive the statistical error).
 
Another approach if your sample size is ten time larger than the dimension you can store would be to introduce a random rejection when loading the particle information. In that case you should reject the particles in 90 % of the cases. With this approach I would still suggest running several independent calculations rather than having a number of primary (START card) exceeding many times the maximum number of particles you load. That would ensure that you use your sample from the 1st step in a more homogeneous way....

Finally, I have notice that you reject some particles based on the position. Not sure if you first step calculation use biasing or not, but you may introduce a bias in the normalization of the second step by rejecting some particles ?

Hoping this help
Joachim
   

-----Original Message-----
From: owner-fluka-discuss_at_mi.infn.it [mailto:owner-fluka-discuss_at_mi.infn.it] On Behalf Of adwasil
Sent: 18 October 2017 10:17
To: fluka-discuss_at_fluka.org
Subject: [fluka-discuss]: How to increase NMAX value in source_TwoSteps.f. Error linking.

Dear FLUKA experts.

I've got trouble with linking FLUKA while crossing NMAX above a certain limit of 3E+7, which value probably depends on the version of the gfortran compiler used. I would need NMAX larger at least by the order of magnitude. Do you have any idea how to solve this problem?

I've got access to Ubuntu 17.04, 4.10.0-37-generic, x86_64 GNU/Linux with gfortran compiler from gcc package version 6.3.0 20170406. Error linking appears while NMAX=3.1E+7, all works fine with NMAX=3.0E+7.
I've got access to Scientific Linux CERN SLC release 6.9 (Carbon), 2.6.32-642.6.2.el6.x86_64 GNU/Linux with gfortran compiler from gcc package version 4.9.4 (GCC). Error linking appears while NMAX=2.26E+7, all works fine with NMAX=2.25E+7.

The compilation in Ubuntu gives a list of messages attached in the file Ubuntu.Err.txt.
The compilation in SLC gives a list of messages attached in the file SLC.Err.txt.

Best regards
Adam Wasilewski
                                                       Ў(╣╞зЩй┌Быз▓цьr╕й╢*'~║&╢мЩиеКx%К╦Z╢mж╧   0¤∙nСпшо▀Цщ■Шi■'i╟ Кwш
Received on Wed Oct 18 2017 - 18:16:00 CEST

This archive was generated by hypermail 2.3.0 : Wed Oct 18 2017 - 18:16:04 CEST