[Gate-users] Parallel simulations without cluster-aware version of Gate

[Luis Ammour]

Hello Sébastien,

1. I didn't check in the 7.0 but the activity should be corrected from the
time decay if start time is not 0. You should try with examples provided in
GATE.
2. Random numbers generator ! The seed have to change in each simulation.
3. I don't know, but normal version of GATE works fine on cluster, as far
as I know.

Hope this helps,

Luis.


2014-12-30 20:32 GMT+01:00 Sébastien Laberge <sebastien.laberge at polymtl.ca>:

>
> Hi,
>
> I am trying to split my Gate jobs into many processes to speed up my
> simulations. I have access to a computer cluster but not to a cluster-aware
> version of Gate. To get things going nonetheless, I decided not to use the
> jobsplitter and to use on the cluster a build of the regular version of
> Gate 7.0 available on the website. I am now wondering whether my approach
> is correct.
>
> Right now, I am manually dividing my simulation macro file into several
> versions, which differ only by the name of the output file and the initial
> and final time of simulation, as set using the macro functions
> /gate/application/setTimeStart and /gate/application/setTimeStop. I set
> these times so that the whole acquisition time is divided into regular
> intervals, one for each process.
>
> I am only simulating models of tumors (in voxelized volumes) that fit into
> the scanner's field of view. I neither need to modify the geometry during
> the simulation nor to divide the simulation into more than one run.
>
> To keep the simulations realistic, my only worry is that the activity in
> the voxelized volume be adjusted at the beginning of each process to take
> into account the attenuation of activity due to radioactive decay during
> the time interval separating the initial time of the entire acquisition and
> the beginning of the time interval being simulated in a given process.
>
> I guess my question could be divided into three parts :
>
> 1-When using setTimeStart to set the initial time to something other than
> zero, is the activity in the voxelized volume adjusted automatically by
> Gate to take into account activity attenuation or should I do that manually
> on the input of each process I call?
>
> 2-Apart maybe from simulation time optimization, does there seem to be
> anything I am overlooking that would render my approach fundamentally
> flawed?
>
> 3-In case I have to resort to using a cluster-aware version of Gate, how
> can I have access to it?
>
> Thank you
>
> _______________________________________________
> Gate-users mailing list
> Gate-users at lists.opengatecollaboration.org
> http://lists.opengatecollaboration.org/mailman/listinfo/gate-users
>
-------------- section suivante --------------
Une pi?ce jointe HTML a ?t? nettoy?e...
URL: <http://lists.opengatecollaboration.org/mailman/private/gate-users/attachments/20141230/548daaed/attachment.html>