RE: "memory" problems in CG?

From: Chris Theis (Christian.Theis@cern.ch)
Date: Mon Jan 22 2007 - 10:33:37 CET

  • Next message: Giuseppe Battistoni: "Re: Issue with USRYIELD results for e+/e-"

    Hello Silvia,

    I quickly checked your geometry and saw that you're using a lot of
    parentheses in the definition of the air region. This could actually be
    the culprit, which is causing FLUKA to crash. I would recommend to check
    the log and err file for messages; don't forget to look at the top (!)
    of the log file. You might find a message there that there is an array
    subscript out of range or something similar. Due to the design of the CG
    kernel in FLUKA the use of parentheses requires normalization and
    expansion of the CSG expression. This could cause an overflow due to
    possible creation of an excessively large number of sub-terms. However,
    without seeing any messages from the log/err file this is only an
    educated guess and I might be wrong! Please check the files & post the
    message that you found to the list, because this information would be
    helpful for the developers to actually track down the problem.

    Just one more general/personal remark which is actually more cosmetics
    than really a requirement. You're writing the union expressions with a
    leading | operator right after the opening parenthesis:

    (|+Xhold1 |+Xhold2 |+Xhold3 |+Xhold4)

    However, in complicated expressions this makes it harder to read and
    check for consistency, at least in my opinion. Thus, I would recommend
    to write simply (+Xhold1 |+Xhold2 |+Xhold3 |+Xhold4). But as I said this
    is not related to your problem as parentheses are given highest operator
    precedence and is purely a cosmetic suggestion.

    Cheers
    Chris

    ------------------------------------------------------------------------

    --
    Chris Theis
    CERN/SC-RP - European Organization for Nuclear Research
    1211 Geneva 23, Switzerland
    Phone: +41 22 767 8069                    Office: 892-2A-015
    e-mail: Christian.Theis@cern.ch           www: http://www.cern.ch/theis
    ------------------------------------------------------------------------
    --  
    > -----Original Message-----
    > From: owner-fluka-discuss@fisica.unimi.it 
    > [mailto:owner-fluka-discuss@fisica.unimi.it] On Behalf Of Silvia Pani
    > Sent: Friday, January 19, 2007 12:12 PM
    > To: fluka-discuss@fluka.org
    > Subject: Re: "memory" problems in CG?
    > 
    > Ciao Alfredo,
    > 
    > sorry, I didn't mean that the package has some problems (I 
    > was suspecting that a couple of slabs of material shouldn't 
    > scare it too much), but that there might be something that I 
    > am missing, such as some option that needed to be enabled.
    > However, I did try the debugger but it gets stuck without any 
    > error messages at the same point as before, i.e., after 
    > reading the body input and before the region data input (and 
    > the debugger output...), with the lines NUMBER OF BODIES 31 
    > LENGTH OF FPD ARRAY 320 With any of the regions from the 
    > subset removed the debugger doesn't report any error (I tried 
    > several grids).
    > Any other suggestion?
    > I went round the problem by modifying the geometry (I 
    > replaced a shielding box with an L, which is probably more 
    > appropriate for describing my system) and now it works, so it 
    > is really just a matter of  curiosity!
    > 
    > Thank you
    > 
    > Silvia
    > 
    > On Jan 18, 2007, at 2:11 PM, Alfredo Ferrari wrote:
    > 
    > > ... about modelling more complex systems: we routinely run systems 
    > > with tens of thousands regions, modelling km's of 
    > accelerators with no 
    > > problems. There is no limitation in the FLUKA geometry 
    > (apart a limit 
    > > of 20000 different "prototype" regions before repetitions, with 
    > > repetitions they can be millions). Again, if you have 
    > problems it is 
    > > almost certain you made a mistake in defining your geometry.
    > > Removing one
    > > or more regions and getting it (apparently) to work, 
    > doesn't prove at 
    > > all it was (and it is after removal) correct. Only running the 
    > > geometry debugger can help, please run it.
    > >
    > >                   Ciao
    > >                  Alfredo
    > >
    > > On Thu, 18 Jan 2007, Silvia Pani wrote:
    > >
    > >> Dear Mattias,
    > >>
    > >> Thank you for your prompt reply. Unfortunately it doesn't work!
    > >>
    > >> Apologies to everybody for not sending an attachment (for 
    > some reason 
    > >> I had thought that e-mails with attachments were 
    > rejected!). Here is 
    > >> the attachment with the change suggested by Mattias in 
    > case someone 
    > >> has some time to look into it.
    > >> I guess I can go round the problem in one way ore another, 
    > but I was 
    > >> a bit curious as to what happens in case I need to model 
    > more complex 
    > >> systems. Has anybody experienced a similar problem?
    > >>
    > >> Regards,
    > >>
    > >> Silvia
    > >>
    > >
    > > --
    > >
    > > 
    > +---------------------------------------------------------------------
    > > -------+
    > > |  Alfredo Ferrari                ||  Tel.:  
    > > +41.22.767.6119                  |
    > > |  CERN-AB                        ||  Fax.:  
    > > +41.22.767.7555                  |
    > > |  1211 Geneva 23                 ||  e-mail:  
    > > Alfredo.Ferrari@cern.ch        |
    > > |  Switzerland                    ||           
    > > Alfredo.Ferrari@mi.infn.it     |
    > > 
    > +---------------------------------------------------------------------
    > > -------+
    > 
    > <><><><><><><><><><><><><><><><><><><><><><>
    > Dr Silvia Pani
    > University College London
    > Department of Medical Physics and Bioengineering
    > Malet Place Engineering Building
    > London WC1E 6BT
    > Tel +44 20 76790259
    > e-mail spani@medphys.ucl.ac.uk
    > <><><><><><><><><><><><><><><><><><><><><><>
    > 
    > 
    > 
    

  • Next message: Giuseppe Battistoni: "Re: Issue with USRYIELD results for e+/e-"

    This archive was generated by hypermail 2.1.6 : Mon Jan 22 2007 - 10:40:16 CET