# Re: Trackr common is full!!, increase the Mxtrck parameter

From: Vasilis Vlachoudis <Vasilis.Vlachoudis_at_cern.ch>
Date: Tue, 26 Oct 2010 17:02:51 +0200

Hi Maki,

please look in the slides of the FLUKA course

even though you are in vacuum FLUKA is performing the tracking by
iterations (linear sub-steps). The size of each sub-step is a calculated
from the user provided parameters, the magnetic field and the particle
momentum. Therefore using excessive number of steps can have a penalty
in cpu time.

The reason of performing the tracking with steps is that calculating the
distance to the next boundary inside a magnetic field would require the
solution of a high order non-linear equation (calculation of the
intersection of a curved path with a quadratic surface, with also
variable magnetic field, and dE/dx if you are in matter).

Regards
Vasilis

On 10/25/10 16:40, Chrysostomos Valderanis wrote:
> Dear Vasili,
>
> In fact I was planning the other way. I want to force Fluka transporting
> through the magnetic field in one step, while keeping the accuracy.
> Since I am asking for transport in vacuum it shouldn't matter. Is this
> possible?
> I was under the impression that transporting in vacuum shouldn't cost in
> time since the solution is known analytically. Is this the case?
>
> Thank you again,
> Makis
>
> ________________________________
>
> From: Vasilis Vlachoudis
> Sent: Mon 25/10/2010 16:18
> To: Chrysostomos Valderanis=20
> Cc: fluka-discuss_at_fluka.org
> Subject: Re: Trackr common is full!!, increase the Mxtrck parameter
>
>
> Hi Maki,
>
> then you have to find a suitable combination with the STEPSIZE and
> MGNFIELD cards for the Minimum, Maximum step, Boundary accuracy and Max
> angle.
> In your case you were requesting a maximum angle (default) 57deg,
> boundary accuracy of 1e-10cm and Maximum step of 700cm. It was ending,
> by breaking the 57deg into too many substeps more than the maximum limit
> in the trackr common of 2500.
> To go to 1e-10 accuracy (which will exploded your CPU tracking time and
> it wont make any sense either for the calculation) reduce the maximum
> step in STEPSIZE down to 0.1cm
>
> Cheers
> Vasilis
>
> On 10/25/10 10:55, Chrysostomos Valderanis wrote:
>
> Dear Vasili,
>
> Thank you very much for the explanation. But putting 1e-4 to WHAT(2) is
> something I would prefer to avoid. I really would like to be able to
> have double precision accuracy in my tracking under the magnetic field
> and asking for a boundary accuracy of 1e-4 is not helping me. So, is
> there another way to achive double precision in tracking?
>
> Thank you again,
> Makis
>
> ________________________________
>
> From: Vasilis Vlachoudis
> Sent: Mon 25/10/2010 10:44
> To: Chrysostomos Valderanis
> Cc: fluka-discuss_at_fluka.org
> Subject: Re: Trackr common is full!!, increase the Mxtrck parameter
>
> Dear Maki,
>
> the problem was coming from the too small value set in WHAT(2) of
> MGNFIELD. You have requested as upper limit for the boundary accuracy
> 1e-10cm during the tracking in the mangetic field. Therefore it was
> forcing the code to make small steps to check for the boundary leading
> to and overflow in the trackr common. Please increase the WHAT(2) e.g.
> 1e-4 and you will see that the message will disappear.
> FLUKA was displaying only the first ten messages in order not to flood
> your output with such messages.
>
> Regards
> Vasilis
>
> On 10/18/10 08:31, Chrysostomos Valderanis wrote:
>
> Hello,
>
> I am trying to run a simple example and I get this message in the
> first event of the err file. My questions
> 1. Is it an error or a warning. It doesn't seem to influence the
> result, but the fact that is written to the error file is worrying.
> 2. Since only tracking is involved, shouldn't it be also produced in
> the other events? Is it suppressed?
> 3. How to avoid it? I thought that enlarging the maximum step in the
> magnetic field was going to remove it, since the magnetic field region
> was going to be passed in one step, but it isn't the case.
> 4. changing the mxtrck parameter in the Trackr include module. It
> gives an error in the linking.
>
> Thank you,
> Makis
>
>
Received on Tue Oct 26 2010 - 18:47:16 CEST

This archive was generated by hypermail 2.2.0 : Tue Oct 26 2010 - 18:47:16 CEST