Hi Nick,
Thanks for the reply. I will look into the memory limit issue some more (we
recently instituted per job memory limits), but I have done a test run with a
16GB memory limit (running one job per node).
Meanwhile, thanks for the novtk suggestion. Our fallback option is to just
disable the ribbon generation. I don't think we ever use them for anything.
Don
> Subject: Re: [Freesurfer] vtkMultiThreader error
> From: ni...@nmr.mgh.harvard.edu
> To: martin.ka...@gmail.com; dhagle...@hotmail.com
> Date: Mon, 8 Jun 2009 11:15:51 -0400
> CC: freesurfer@nmr.mgh.harvard.edu
>
> Don and Martin,
>
> I dont have a good solution for you to this problem. The error code
> indicates an 'out-of-memory' error returned by pthread_create, and I do
> know the mris_vol_mask uses a lot of memory (at least 1GB).
>
> I have not been able to find any clues to the source of this problem on
> the web either. The only curious thing to it is that in the source
> code, that section is conditionally compiled to ignore the return code
> if 'VTK_HP_PTHREADS' is defined, which implies some sort of odd behavior
> on some special HP platform build.
>
> mris_vol_mask has been a real thorn in the side of the recon-all stream,
> but its not easy to replace because it does a subjectively better job of
> finding the voxels which intersect the pial and white surfaces. For
> comparison, or even to solve your problem, you can use
> mris_vol_mask_novtk to create the ribbon.mgz files. The output is not
> the same as that produced by mris_vol_mask though. Perhaps you can run
> it, and compare the results side-by-side and give us your opinion on
> which is the better ribbon (as there is no objective measure on what a
> 'correct' ribbon.mgz should like like, ie, should it include voxels that
> intersect more than half way through a surface, or not at all?).
>
> Nick
>
>
> On Mon, 2009-06-08 at 09:16 +0200, Martin Kavec wrote:
> > Hi Don,
> >
> > I am in a similar situation. I am on a IBM BladeCenter cluster (though with
> > one blade only) with 2 Quad core Intels, and I experience intermittent
> > crashes of mri_volmask as well. Similarly as you, when I run the
> > mri_volmask
> > on the crashed subject on a desktop, this passes perfectly fine and
> > sometimes
> > also on the cluster as well.
> >
> > I am wonder whether how do you obtain the debug information about the cause
> > of
> > the crash (being in the vtkMultiThreader)?
> >
> > I also wonder, whether the crashes could be due to different compilation
> > environments of the VTK libraries used in teh FS distribution (centos) and
> > your redhat. I went on the bug-report site of VTK and there really isn't
> > anythink major reported on this class.
> >
> > In my experience the mri_volmask runs fine for quite a while and then
> > crashes.
> > >From the debug messages you show it would thus could be that the process
> > >may
> > hit some limit of the threads which could be allocated to a process.
> >
> > What do you think?
> >
> > Martin
> >
> > On Saturday 06 June 2009 04:19:51 Don Hagler wrote:
> > > When we run v430 on our cluster, running rocks redhat 4.3, we get the
> > > error
> > > message below when mris_volmask is run. From a google search, I see that
> > > someone had the same error message back in Oct 2007. Was a solution found
> > > for this? It runs fine on our desktop machines as well as when we log in
> > > directly to a node and run the job interactively.
> > >
> > >
> > >
> > > mris_volmask --label_left_white 2 --label_left_ribbon 3
> > > --label_right_white 41 --label_right_ribbon 42 --save_ribbon
> > > --save_distance FREESURFERRECON_NTNU_PRE_095v2_20071126.173245.968000_1
> > >
> > > SUBJECTS_DIR is /home/dhagler/data_test
> > > ERROR: In /usr/pubsw/packages/vtk/5.2.0/src/Common/vtkMultiThreader.cxx,
> > > line 372 vtkMultiThreader (0x1306ec80): Unable to create a thread.
> > > pthread_create() returned 12
> > >
> > > ERROR: In /usr/pubsw/packages/vtk/5.2.0/src/Common/vtkMultiThreader.cxx,
> > > line 372 vtkMultiThreader (0x1306ec80): Unable to create a thread.
> > > pthread_create() returned 12
> > >
> > > ERROR: In /usr/pubsw/packages/vtk/5.2.0/src/Common/vtkMultiThreader.cxx,
> > > line 372 vtkMultiThreader (0x1306ec80): Unable to create a thread.
> > > pthread_create() returned 12
> > >
> > > ERROR: In /usr/pubsw/packages/vtk/5.2.0/src/Common/vtkMultiThreader.cxx,
> > > line 372 vtkMultiThreader (0x1306ec80): Unable to create a thread.
> > > pthread_create() returned 12
> > >
> > > ERROR: In /usr/pubsw/packages/vtk/5.2.0/src/Common/vtkMultiThreader.cxx,
> > > line 372 vtkMultiThreader (0x1306ec80): Unable to create a thread.
> > > pthread_create() returned 12
> > >
> > > ERROR: In /usr/pubsw/packages/vtk/5.2.0/src/Common/vtkMultiThreader.cxx,
> > > line 372 vtkMultiThreader (0x1306ec80): Unable to create a thread.
> > > pthread_create() returned 12
> > >
> > > ERROR: In /usr/pubsw/packages/vtk/5.2.0/src/Common/vtkMultiThreader.cxx,
> > > line 372 vtkMultiThreader (0x1306ec80): Unable to create a thread.
> > > pthread_create() returned 12
> > >
> > > Segmentation fault
> > > Linux compute-0-12.local 2.6.9-55.0.2.ELsmp #1 SMP Tue Jun 26 14:14:47 EDT
> > > 2007 x86_64 x86_64 x86_64 GNU/Linux
> > >
> > > recon-all exited with ERRORS at Fri Jun 5 18:58:09 PDT 2009
> > >
> > >
> > >
> > >
> > >
> > > _________________________________________________________________
> > > Hotmail® has ever-growing storage! Don’t worry about storage limits.
> > > http://windowslive.com/Tutorial/Hotmail/Storage?ocid=TXT_TAGLM_WL_HM_Tutori
> > >al_Storage_062009
> >
> >
> > _______________________________________________
> > Freesurfer mailing list
> > Freesurfer@nmr.mgh.harvard.edu
> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> >
> >
>
> _______________________________________________
> Freesurfer mailing list
> Freesurfer@nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_________________________________________________________________
Insert movie times and more without leaving Hotmail®.
http://windowslive.com/Tutorial/Hotmail/QuickAdd?ocid=TXT_TAGLM_WL_HM_Tutorial_QuickAdd_062009
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
