Re: How do I decrease run time?

From: Vittorio Boccone <boccone_at_cern.ch>
Date: Tue, 12 Jun 2012 20:09:30 +0200

Hi Vahen,

When you decide to do a simulation (and I'm not talking only about
Fluka) you treat the problem statistically and you don't need to
simulate the exact amount of particles you have in the real experiment.
For each primary particle you will have many secondaries and so on, then
depending on your problem (energy deposition, activation) you will reach
a point in which the statistical error of you simulation will be
sufficiently small. And this point (which depends on where and what you
are scoring) can be reached for sure with less particles.

In your case, I would start to simulate around 10^3 - 10^4 (even 10^5)
primaries and I would then check the statistical error of the relevant
resulting USRBIN/USRTRACK. Then check the time/primary and then scale
the number of primaries such a way you have reasonable running time.

You can multiply the result for the real number of impinging electrons,
but you have to make sure that the statistical error of the quantity you
want to score is small enough in your region of interest.

One might also want to analyze the simulation event by event (that's a
different problem... ), but still the number of primaries will be
(almost always) smaller that the real number of particles your have in
your problem.

cheers
Vittorio

On 12/06/2012 19:55, Vittorio Boccone wrote:
> Hi Vahen,
> placing a billion of primary blindly is the wrong approach.
>
> When you decide to do a simulation (and I'm not talking only about
> Fluka) you treat the problem statistically and you don't need to
> simulate the exact amount of particles you have in the real
> experiment. For each primary particle you will have many secondaries
> and so on, then depending on your problem (energy deposition,
> activation) you will reach a point in which the statistical error of
> you simulation will be sufficiently small. And this point (which
> depends on where and what you are scoring) can be reached for sure
> with less particles.
>
> In your case you will need to multiply the result for the real number
> of impinging electrons, and you have to make sure that the statistical
> error of the quantity you want to score is small enough in your region
> of interest.
>
> On 12/06/2012 14:20, Vahan Petrosyan wrote:
>> I want to calculate process in which electron beam which has 200pC
>> charge
>> and 20Mev energy hits target and 200pC = billion electrons.
>>
>> As I understood already fluka calculates quantities normalized to one
>> particle an to get real picture in case of 200pC charge I should
>> multiply
>> fluka output values (for example energy deposition) by beam current
>> instead of putting a lot primary particles in START card.
>> Am I right?
>>
>> Generaly my question was how to take into account in my
>> calculations the
>> total charge of a beam .
>>
>> Many thanks for attention.
>>
>
>

-- 
Dr. Vittorio Boccone - University of Geneva
o Address:
    UniGe: Département de physique nucléaire et corpusculaire
           24 Quai Ernest-Ansermet, CH-1211 Geneve 4, Switzerland
    CERN:  CERN, CH-1211 Geneve 23, Switzerland
o E-mail:
    vittorio.boccone_at_unige.ch
    vittorio.boccone_at_cern.ch
o Phone:
    Mobile: +41 76 487 5737
    UniGe:  +41 22 379 6357/6353
    CERN:   +41 22 767 5821
Received on Wed Jun 13 2012 - 09:50:52 CEST

This archive was generated by hypermail 2.2.0 : Wed Jun 13 2012 - 09:51:23 CEST