Dear Oscar,
There is a fundamental difference between a scoring a la USRxxx and a collision tape:
1) When using a USRxxx scoring you don't store information for each individual event but you populate a spatial mesh or a histogram. In this case, it is essential to merge results from different cycles as this is the only way that FLUKA will provide you an error estimate for the bin entries.
2) In case of a collision tape, you store specific information about each individual event, and it is up to the user to post-process the data according to his/her needs (this requires evidently some customized user code - in Fortran or whatever language/package you want to use). In this sense, simply merging collision tape files from different cycles into one file, i.e. simply appending the output of cycles one after the other, would be rather a cosmetic process (contrary to merging USRxxx files, where the average+error is calculated from the different cycles). Apart from that, FLUKA cannot provide a general script to merge the output from modified mgdraw routines since the output will look different in all cases.
May I ask why you need to have all the collision tapes in one file? I would rather write a post-processing program, which reads in individual collision tapes from different cycles and performs the necessary arithmetic.
Cheers, Anton
________________________________
From: owner-fluka-discuss_at_mi.infn.it [owner-fluka-discuss_at_mi.infn.it] on behalf of Oscar Frasciello [oscar.frasciello_at_lnf.infn.it]
Sent: 12 July 2017 14:39
To: fluka-discuss_at_fluka.org
Cc: Oscar Frasciello
Subject: [fluka-discuss]: merging multiple collision tape files
Dear FLUKA experts,
dealing with "large" runs (billions of primaries) I usually split them in multiple runs with less primaries through the "spawn" option in flair (as many as the available physical cores on the dedicated server), then merging the output of the specific scoring cards (USRXXX). So being able to reach the desired statistics in a more reasonable amount of time. Now I am using a modified mgdraw.f routine, getting the desired collision tape file with requested informations. And the test run (very few primaries, ~1E3) does work the expected way. But it seems to me it is not possible to proceed the same way with the n-collision tape files, i.e., it is not possible to merge them from the n cycles' output.
Do you confirm this? Do I need to run one single core cycle with the necessary number of primaries (~1E9)? Do you have any suggestion?
Thanks in advance for your time and valuable support.
Best regards,
Oscar
--
Oscar Frasciello, PhD
Radiation Protection Group
Istituto Nazionale di Fisica Nucleare
Laboratori Nazionali di Frascati
Via Enrico Fermi 40
00044 Frascati (RM) Italy
Tel. office: +39 06 9403 2104
Tel. Tenocode: +39 06 9403 8114
Fax FISMEL: +39 06 9403 2364
email: oscar.frasciello_at_lnf.infn.it<mailto:oscar.frasciello_at_lnf.infn.it>
__________________________________________________________________________
You can manage unsubscription from this mailing list at https://www.fluka.org/fluka.php?id=acc_info
Received on Fri Jul 14 2017 - 11:10:54 CEST