Re: Re: [fluka-discuss]: How to optimize input files?

From: <yyc2011_at_mail.ustc.edu.cn>
Date: Sat, 14 Dec 2013 12:49:42 +0800 (CST)

Dear Santana,

  Thanks for your suggestions.In your reply,you said "In particular, for your proton case you could save a lot of CPU with EMF-OFF." Why using EMF-OFF will save a lot of CPU for protons? On the other hand,will the result be changed when use EMF-OFF comparing with not using EMF-OFF?

Best Regards
Chen Yuan


> -----Original E-mail-----
> From: "Santana, Mario" <msantana_at_slac.stanford.edu>
> Sent Time: 2013-12-14 3:28:40
> To: "fluka-discuss_at_fluka.org" <fluka-discuss_at_fluka.org>
> Cc:
> Subject: Re: [fluka-discuss]: How to optimize input files?
>
> Chen,
>
> I believe the purpose of this list is to provide guidance of specific
> FLUKA questions. With that spirit, I will give you some tips and
> recommendations, but I cannot go in-depth and perform your work.
>
> The first thing to say, is that speed of computation is not an absolute
> number to quantify the efficiency of your input file. It is all relative
> to the statistical uncertainty of the observables that you are scoring for
> that amount of CPU. For example, the CPU/history may be high when you have
> intense region biasing, but this does not necessarily mean you are wasting
> CPU, on the contrary, if region biasing is well tuned, you may be
> following many split trajectories towards your region of interest for a
> single history.
>
> Now, the question is whether you are wasting CPU time in processes that
> you don't need. In that regard I have recommend the following:
> 1) Reduce the number of particles per cycle and run several cycles.
> Analyze your results and evaluate how many particles you need in total to
> converge to the desired result
>
> 2) You use PRECISIO as your DEFAULT. This cards includes many options and
> precise settings which you may not need. I copy them at the end of the
> email. You should decide whether this is the DEFAULT that you want (maybe
> you could survive with NEW-DEFA?), or switch off those cards that are not
> needed for your problem. In particular, for your proton case you could
> save a lot of CPU with EMF-OFF. For your ELECTRON case you could play with
> the cut-off energies (the lower the energy the longer the simulation).
> 3) You are using MULSOPT. This may impact your processing speed.
> 4) Your USRBIN has a 2000x2000 grid. The more you write to files the
> slower the simulation can get. Try having a coarser grid if CPU is a
> concern to you. Also, the smaller the bin, the more individual statistical
> error you will get for a given CPU time.
>
> I hope this helps
>
>
> DEFAULTS=PRECISIO
>
>
> * EMF <http://www.fluka.org/fluka.php?id=man_onl&sub=25> on
>
> * Rayleigh scattering and inelastic form factor corrections to Compton
> scattering and Compton profiles activated
>
> * Detailed photoelectric edge treatment and fluorescence photons activated
>
> * Low energy neutron transport on down to thermal energies included, (high
> energy neutron threshold at 20 MeV)
>
> * Fully analogue absorption for low-energy neutrons
>
> * Particle transport threshold set at 100 keV, except neutrons (1E-5 eV),
> and (anti)neutrinos (0, but they are discarded by default anyway)
>
> * Multiple scattering threshold at minimum allowed energy, for both
> primary and secondary charged particles
>
> * Delta ray production on with threshold 100 keV (see option DELTARAY
> <http://www.fluka.org/fluka.php?id=man_onl&sub=21>)
>
> * Restricted ionisation fluctuations on, for both hadrons/muons and EM
> particles (see option IONFLUCT
> <http://www.fluka.org/fluka.php?id=man_onl&sub=42>)
>
> * Tabulation ratio for hadron/muon dp/dx set at 1.04, fraction of the
> kinetic energy to be lost in a step set at 0.05, number of dp/dx
> tabulation points set at 80 (see options DELTARAY
> <http://www.fluka.org/fluka.php?id=man_onl&sub=21>, EMFFIX
> <http://www.fluka.org/fluka.php?id=man_onl&sub=28>, FLUKAFIX
> <http://www.fluka.org/fluka.php?id=man_onl&sub=35>)
>
> * Heavy particle e+/e- pair production activated with full explicit
> production (with the minimum threshold = 2m_e)
>
> * Heavy particle bremsstrahlung activated with explicit photon production
> above 300 keV
>
> * Muon photonuclear interactions activated with explicit generation of
> secondaries
>
> * Heavy fragment transport activated
>
>
>
>
>
> On 12/13/13 3:38 AM, "yyc2011_at_mail.ustc.edu.cn" <yyc2011_at_mail.ustc.edu.cn>
> wrote:
>
> >Dear all,
> >
> > The attachmens are my input files.The first one is the input file for
> >protons.since I have run it for eleven days,but my computer have not
> >finished the first cycle.The calculation is too slow.So,I wish you could
> >see my input file,and optimize the input file so as to improve the
> >calculation speed.The second input file is about electrons.Likewise,I
> >wish you could see my input file and find whether there are something
> >wrong in my input file.I am not sure about this.So I wish you could help
> >me.
> > Thanks for any suggestion.Thanks for your help.
> >
> >Best Regards
> >Chen Yuan
> >
> >
> >> -----Original E-mail-----
> >> From: "Giuseppe Battistoni" <giuseppe.battistoni_at_mi.infn.it>
> >> Sent Time: 2013-11-25 22:56:18
> >> To: yyc2011_at_mail.ustc.edu.cn
> >> Cc: fluka-discuss_at_fluka.org
> >> Subject: Re: [fluka-discuss]: some questions about charged particle
> >>transport
> >>
> >> Dear Chen Yuan
> >> you are asking something very difficult to be summarized by e-mail:
> >> you have to refer to text books on the argument.
> >> From your questions it seems that you have to
> >> improve your background knowledge in the studies of
> >> interaction of radiation with matter: this is essential for
> >> any user of general purpose Monte Carlo codes as FLUKA.
> >>
> >> If I may give and advice, I strongly suggest to use always multiple
> >>scattering
> >> for all standard problems of realistic detectors and situation. Avoid
> >>single scattering
> >> which is to be considered only something exceptional for very very thin
> >>layers.
> >> (by the way, it's very CPU time consuming).
> >>
> >> For your study on Molière's Theory of Multiple Scattering
> >> let me also advice a historical paper by H. Bethe in Phys. Rev. 89,
> >>1256­1266 (1953)
> >>
> >> Best regards
> >> Giuseppe Battistoni
> >>
> >> On 11/25/2013 11:54 AM, yyc2011_at_mail.ustc.edu.cn wrote:
> >> > Dear Giuseppe,
> >> >
> >> > First,Thanks for your reply.But,I still have some questions.
> >> > (1)I wil study the paper so as to solve the problem.
> >> > (2)In the Fluka course,you refer to the validity conditions of
> >>Moliere theory more than once. But, what are the validity conditions of
> >>Moliere theory? Could you tell me in detail?
> >> >
> >> > Any multiple scattering therory applies when in a given step you have
> >>several coulomb scattering.Several is a qualitative word. In practice,
> >>as a rule of thumb, you may be safe if you expect many tens of
> >>scatterings.
> >> >
> >> > So,According to your view,when refer to scattering,I can only
> >>estimate scattering numbers to determine whether it is appropriate to
> >>use the mutiple scattering.And in general,how could you estimate
> >>scattering numbers?
> >> > (3)We advice single scattering only when you have thickness of
> >>materials so small that the above conditions risks to be not respected.A
> >>couple of practical examples:
> >> > - layers of nuclear photographic emulsions.where you aim to measure
> >>charged tracks with few micron accuracy
> >> > - very thin layers of silicon trackers (order of 100 micron or less)
> >> >
> >> > According to your idea,There are no specific conditions to tell me
> >>when should I use single scattering.I must judge whether I should use
> >>single sattering as specific questions change.
> >> >
> >> > I don't know whether my understanding are right.Wish to receive
> >>from you.
> >> >
> >> > Best regards
> >> > Chen Yuan
> >> >
> >> >
> >> >
> >> >> -----Original E-mail-----
> >> >> From: "Giuseppe Battistoni" <giuseppe.battistoni_at_mi.infn.it>
> >> >> Sent Time: 2013-11-25 16:59:43
> >> >> To: yyc2011_at_mail.ustc.edu.cn, fluka-discuss_at_fluka.org
> >> >> Cc:
> >> >> Subject: Re: [fluka-discuss]: some questions about charged particle
> >>transport
> >> >>
> >> >> Dear Chen Yuan
> >> >>>
> >> >>> When I learn charged particle transport,There are a few
> >>problems I cannot understand.The questions are as followings.
> >> >>> (1)The Fluka course said "Accurate PLC (not the average value
> >>but sampled from a distribution), giving a complete independence from
> >>step size",I cannot understand this sentence at all. What kind of the
> >>distribution is? Could you give me a simple interpretations?
> >> >>
> >> >> I advice to study the paper (quoted in FLUKA references):
> >> >> A. Ferrari et al., Nucl. Instr. Meth. in Phys. Res. B71, 412-426
> >>(1992)
> >> >> and references therein
> >> >>
> >> >>> (2)In the Fluka course,you refer to the validity conditions of
> >>Moliere theory more
> >> >> than once. But, what are the validity conditions of Moliere
> >>theory? Could you tell me in detail?
> >> >>
> >> >> Any multiple scattering therory applies when in a given step you
> >>have several coulomb scattering.
> >> >> Several is a qualitative word. In practice, as a rule of thumb, you
> >>may be safe
> >> >> if you expect many tens of scatterings.
> >> >>
> >> >>> (3)In the Fluka course,you mentioned "full single scattering
> >>are used to handle very low energy electron problems and very thin
> >>layers problems. But,how much low is the electron energy? How much thin
> >>are the layers?
> >> >>> In a word,I want to know when I should use mutiple scattering
> >>and when I should use the full single scattering.
> >> >>> I'm looking forward to receive from you.Thanks for your
> >>time.Thanks for your reply.
> >> >>
> >> >> We advice single scattering only when you have thickness of
> >>materials so small that the above
> >> >> conditions risks to be not respected.
> >> >> A couple of practical examples:
> >> >> - layers of nuclear photographic emulsions.
> >> >> where you aim to measure charged tracks with few micron accuracy
> >> >> - very thin layers of silicon trackers (order of 100 micron or less)
> >> >>
> >> >> Best regards
> >> >> Giuseppe Battistoni
> >> >>
> >> >> --
> >> >> INFN Milano
> >> >> via Celoria 16, 20133 Milano, Italy
> >> >> tel: +39 02 50317307
> >> >> fax: +39 02 50317617
> >> >>
> >>
>
Received on Sat Dec 14 2013 - 06:49:35 CET

This archive was generated by hypermail 2.3.0 : Sat Dec 14 2013 - 06:49:37 CET