[Insight-users] Re sampleImageFilter and ITK_USE_OPTIMIZED_REGISTRATION_METHODS

[Luis Ibanez]

Hi Pete79,

Your finding of narrowing down the problem to the ROI filter is quite
interesting.

You may have identified a bug in this filter...

Could you please clarify the following two circumstances ?:

A) Are you using an independent ROI filter for each one of your filters ?
    or where you sharing the output of a single ROI filter with all the N
    subsequent filters that you run in separate threads ?

B) If you are using independent ROI filters, could you give us
     more details on how you were computing the ROI regions
     for each one of the instances  ?


BTW:
It will be great if we could get a minimal piece of code
in which we can replicate the problem....    :-)



   Thanks


         Luis


-----------------------------------------------------------------------
On Thu, Oct 22, 2009 at 5:29 PM, Pete79 <pkoniusz at hotmail.com> wrote:
>
> <qoute>It sounds like you are doing your own threading, and you have multiple
> threads trying to execute the pipeline. ITK expects only one thread to be
> executing or updating the pipeline. The pipeline is not thread safe in the
> way you appear to be assuming. This would explain the sporadic crashing.
> </qoute>
>
> Bradely, hi.
>
> I guess ITK can host several pipelines though. Alas, on the outset I had
> only made sure images fed to the thread were already updated - that's where
> one would hope the main thread pipeline can be queried by multiple threads
> without any harm being done. Though (!!!) ...
>
> I can imagine that when I relied on ROI filter, it was reaching source image
> pipeline while being updated itself. Source image was fed to many threads
> and that could potentially confuse behaviour of this pipeline being queried
> by a bunch of ROIs at the same time. I thought only updating (the actually
> triggered process) changes status of a pipeline. Though, even if up-to-date,
> calling Update must write to status status structures or so. I believe
> that's where it gets multithread unsafe.
>
> Well done then, coz copying image ROI with an iterator (instead of ROI
> filter) separates pipelines and this seems like the right way forward then.
>
> Cheers a lot,
> Peter
> --
> View this message in context: http://www.nabble.com/ResampleImageFilter-and-ITK_USE_OPTIMIZED_REGISTRATION_METHODS-tp25967879p26017542.html
> Sent from the ITK - Users mailing list archive at Nabble.com.
>
> _____________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the ITK FAQ at: http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-users
>