[Gate-users] Non-independent results when using the GATE job splitter

[Simon Stute]

Hi,

You are wright, the problem was already mentioned and is caused by the new
random engine.
In fact this new engine takes different seeds than the last one but the gjs
program was not updated to produce these new seeds ...

To fix this out you have to change the definition of the random engine at
the beginning of the main program to set the old one instead of the new one.
I send you with this mail the file Gate.cc where I did the modification.
Just replace the original Gate.cc file by this one and recompile, then it
will use the old random engine with the old seed files generated by the gjs
program and it will work fine.

A new patch is coming soon with which you can manage all the random engines
and seeds aspects with macro commands.

Sorry for your CPU time lost !
Cheers,
Simon

On Wed, Dec 10, 2008 at 4:01 PM, KB <kd8had at gmail.com> wrote:

> Hello All,
>
> GATE version name: gate_v4.0.0  (20-August-2008). I am running SGE.
>
> Using gjs, I split the main macro into 25 parts and submit them as a array
> job using qsub:
>
> "qsub -t 1-25 -V /home/data/.Gate/Hits/Hits.sh"
>
> This produced 25 root files which I am supposed to merge with gjm. I did so
> and noticed that the resultant root file has some strange behavior when some
> parameters were histogrammed. There was a periodicity of 25. So I randomly
> picked 2 of the 25 root files  and using root printed out the events. They
> are unfortunately identical. This makes the culster version of GATE almost
> useless (at least for me). Does anyone else notice the same thing?
>
> Here is some output from running root:
> I open the first file:
> root [0] TFile f2("CylinderHitsSingleGamma3.root")
> root [1] Hits->Show(1)
> ======> EVENT:1
>  PDGEncoding     = 22
>  trackID         = 4
>  parentID        = 1
>  time            = 0.400002
>  edep            = 0.0018285
>  stepLength      = 0.00627867
>  posX            = 455.453
>  posY            = 75.4412
>  posZ            = -30.2955
>  localPosX       = -1.67338
>  localPosY       = -0.606797
>  localPosZ       = 0.267013
>  gantryID        = 0
>  blockID         = 0
>  crystalID       = 339
>  unused3ID       = -1
>  unused4ID       = -1
>  unused5ID       = -1
>  photonID        = 1
>  nPhantomCompton = 6
>  nCrystalCompton = 0
>  nPhantomRayleigh = 0
>  nCrystalRayleigh = 0
>  primaryID       = 1
>  sourcePosX      = 17.0792
>  sourcePosY      = -100.416
>  sourcePosZ      = -49.8924
>  sourceID        = 0
>  eventID         = 3
>  runID           = 0
>  axialPos        = 0
>  rotationAngle   = 0
>  volumeID        = 0,
>                   0, 0, 339, -1, -1, -1, -1, -1, -1
>  processName     = LowEnPhotoElec
>  comptVolName    = wholecylinder_P
>  RayleighVolName = NULL
>
>
> Here is the 1st event  from another root file:
> root [0] TFile f1("CylinderHitsSingleGamma10.root")
> root [1] Hits->Show(1)
> ======> EVENT:1
>  PDGEncoding     = 22
>  trackID         = 4
>  parentID        = 1
>  time            = 1.8
>  edep            = 0.0018285
>  stepLength      = 0.00627867
>  posX            = 455.453
>  posY            = 75.4412
>  posZ            = -30.2955
>  localPosX       = -1.67338
>  localPosY       = -0.606797
>  localPosZ       = 0.267013
>  gantryID        = 0
>  blockID         = 0
>  crystalID       = 339
>  unused3ID       = -1
>  unused4ID       = -1
>  unused5ID       = -1
>  photonID        = 1
>  nPhantomCompton = 6
>  nCrystalCompton = 0
>  nPhantomRayleigh = 0
>  nCrystalRayleigh = 0
>  primaryID       = 1
>  sourcePosX      = 17.0792
>  sourcePosY      = -100.416
>  sourcePosZ      = -49.8924
>  sourceID        = 0
>  eventID         = 3
>  runID           = 1
>  axialPos        = 0
>  rotationAngle   = 0
>  volumeID        = 0,
>                   0, 0, 339, -1, -1, -1, -1, -1, -1
>  processName     = LowEnPhotoElec
>  comptVolName    = wholecylinder_P
>  RayleighVolName = NULL
>
> The only thing that changes is the time. Everything else is identical.
> Printint out the other events shows the same similarities.
>
> Karthik.
>
> _______________________________________________
> Gate-users mailing list
> Gate-users at lists.healthgrid.org
> http://lists.healthgrid.org/mailman/listinfo/gate-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengatecollaboration.org/mailman/private/gate-users/attachments/20081210/ad87f791/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Gate.cc
Type: text/x-c++src
Size: 9768 bytes
Desc: not available
URL: <http://lists.opengatecollaboration.org/mailman/private/gate-users/attachments/20081210/ad87f791/attachment.cc>