[Gate-users] Possibly a bug in GATE geometry check?

[Ida Häggström]

Dear John,

Thank you so much for your reply! I find it very strange too that the
visualization enabled or disabled gives different results. The reason I
started with this geometry check is that my GATE run is aborted and I can't
figure out why. As I found errors in the geometry test, I figured that might
be the root to by problems (I'm not sure about this however, but it's the
best guess I have so far).

I've looked through the GATE users manual, but the only type of test I could
find there was the one I'm using, i.e. "/geometry/test/recursive_test". I
have built up my geometries step by step with the visaulization on, and
everything looks OK in my eyes, i.e. no overlapping crystals, blocks, foils
etc. Is there a better way you think to check the geometry?

I tried running a new voxelized phantom today, with visualization disabled.
I got the error:
-------------------------------------------
GeomTest Error: Overshooting daughter volume
    The volume xcatPhantomVoxel_PVP[-1] appears to extend outside the mother
volume xcatPhantom_P[1]
    at the following points in global coordinates: (list truncated)
.
.
.

970000 singles stored in LMF fileTime set to (s) 1
1950000 singles stored in LMF fileTime set to (s) 2
2930000 singles stored in LMF fileTime set to (s) 3
ERROR - G4PhantomParameterisation::GetReplicaNo()
        localPoint - (1.837206641,1.061885918,-0.07676457566) - is outside
container solid: xcat_S

*** G4Exception : InvalidSetup
      issued by : G4PhantomParameterisation::GetReplicaNo()
Point outside voxels!
*** Fatal Error In Argument *** core dump ***

*** G4Exception: Aborting execution ***
3530000 singles stored in LMF fileGatedied with signal 6
-----------------------------------------
 I don't know how the voxels can extend outside the mother xcatPhantom. I
only create the regular matrix xcatPhantom, and the voxels are created by
the program. I get the same geometry error for the voxelized phantom I
attached in the zip-file in the previous post. Anway, I figured the latter
error that aborted the run is associated to the geometry check error.
With visualization enabled, I don't get any geometry check errors, but the
computer hangs after a while, before I arrive at the error point! I tried
four times, same result. I guess the CPU gets overloaded or something with
visualization enabled, I don't know. I really don't know how to proceed,
cause I need to run these voxelized phantoms.

Sorry for making this post looonger and looonger, but I really have no clue
what to do, or if I'm trying to solve the right problem. Thank you for your
help!
Ida

Den 5 oktober 2009 15.25 skrev John Apostolakis <john.apostolakis at cern.ch>:

>  Dear Ida,
>
> I write from the perspective of a Geant4 developer - but the observations
> should apply equally to any GATE application.
>
>  - It is unexpected that using visualisation should change the result of
> any geometry check.
>  - The results of the recursive_test should be reasonably accurate; that is
> at the limit of high statistics you should get a correct answer, and using
> default number of trial lines you should get an answer which does not vary
> from run to run.
>
> There is another method to check a Geant4 geometry, which involves checking
> every volume which is positioned to see that it does not overlap with a
> volume at the same level or protrude from its parent volume.  I am not
> certain whether this capability is exploited already in Gate, in all or
> recent releases.
>
> Best regards,
> John A.
>
> Ida Häggström wrote:
>
> Dear all,
>
> I've recently posted some questions regarding segmentation violations,
> geometry test errors and voxelized phantoms. I've experienced a lot of
> problems with getting a proper run from GATE, and I've spent many hours
> (piuh!) troubleshooting! =) Anyway, I discovered something that I'm not sure
> is a bug in GATE or if it's my code that is erroneous...
>
> I run a geometry check as I initialize my run, to check that there are no
> geometry overlaps etc since I believe this can cause segmentation violations
> which aborts the entire run.
> -------------
> /run/initialize
> /geometry/test/recursive_test
> -------------
> I usually keep the visualization disabled since this saves a lot of time
> for longer runs:
> -------------
> /vis/disable
> -------------
>  I noticed that I get errors regarding my voxelized phantom (overshooting
> daughter volume) which I don't think should be there (emphasis on THINK!).
> When I turn the visualization ON however, the errors disappear and I only
> get results like:
> ----------
> GeomTest: no daughter volume extending outside mother detected.
> GeomTest: no overlapping daughters detected.
> ---------
> from the geometry check. My visualization macro is very simple:
> ----------
> # Visualization.mac
> /vis/open OGLSX
> /vis/viewer/reset
> /vis/drawView 45 45
> /vis/viewer/zoom 2
> /vis/viewer/set/style surface
> /vis/drawVolume
> /tracking/storeTrajectory 1
> /vis/scene/endOfEventAction accumulate
> /vis/viewer/update
> ------------
> and I don't think the visualization should have anything to do with the
> geometry setup at all?(!) Can this be a GATE bug? Note that the only thing I
> change in my main macro is changing "/vis/disable" to
> ----------
> #/vis/disable
> /control/execute Visualization.mac
> ----------
> and I get no errors. I've also tried to disable all visualization of the
> phantom (in the range.dat file, setting visualization to "false") but this
> has no affect on the run results (no matter if I have the main visualization
> on or off).
>
> I have attached my main macro, GATE run output log, voxelized phantom macro
> with related .dat files and a screenshot of my setup with camera and
> phantom.
>
> Please, if anyone has any idea what could cause these problems, please let
> me know! I'm really running out of ideas... =) It's not a good option to
> always have to run with visualization on for longer runs...
> Thanks!
> Ida
>
>
> ------------------------------
>
>  ------------------------------
>
> _______________________________________________
> Gate-users mailing listGate-users at lists.healthgrid.orghttp://lists.healthgrid.org/mailman/listinfo/gate-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengatecollaboration.org/mailman/private/gate-users/attachments/20091006/8449e9cc/attachment.htm>