[vtk-developers] TestFixedPointRayCasterLinearCropped failure

[David Gobbi]

To complicate things further, the mapper uses a multithreaded helper
filter internally: vtkVolumeRayCastSpaceLeapingImageFilter.  And there
isn't any way for a user to control the number of threads used by the
helper other than by setting the GlobalDefaultNumberOfThreads.

 - David

On Sun, Feb 24, 2013 at 12:36 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
> This is a complex test.
>
> II had to set the number of threads in the mappers in the included file.
> Then it fails when I explicitly set the number of threads for the mappers.
>
> On Sun, Feb 24, 2013 at 1:51 PM, Bill Lorensen <bill.lorensen at gmail.com>
> wrote:
>>
>> Ok, now it fails.
>>
>> But, this test has a lot of filters other than volume mappers. If I
>> explicitly set the number of threads to 1 for each volume mapper, it still
>> fails.
>>
>>
>>
>> On Sun, Feb 24, 2013 at 1:38 PM, David Gobbi <david.gobbi at gmail.com>
>> wrote:
>>>
>>> I see the failure on my MacBook Pro when I add the following
>>> lines to the top of the test:
>>>
>>>   vtkMultiThreader multithread
>>>   multithread SetGlobalDefaultNumberOfThreads 64
>>>
>>> The lines have to be added to the top (i.e. before the mappers
>>> are instantiated) in order to take effect.
>>>
>>>  - David
>>>
>>> On Sun, Feb 24, 2013 at 11:25 AM, Bill Lorensen <bill.lorensen at gmail.com>
>>> wrote:
>>> > I can't get to to fail on my fedora system. I get a 0 error no matter
>>> > how
>>> > many threads I specify. I'll try a couple of other systems I have
>>> > access to.
>>> >
>>> >
>>> > On Sun, Feb 24, 2013 at 1:17 PM, David Gobbi <david.gobbi at gmail.com>
>>> > wrote:
>>> >>
>>> >> I haven't been able to find the old mailing list discussions about
>>> >> this topic, but here is the original bug report that I made in 2011
>>> >> when I noticed the failing regression test.  It gives instructions
>>> >> that any developer can use to reproduce the bug.
>>> >> http://www.vtk.org/Bug/view.php?id=12115
>>> >>
>>> >> A more recent report from a user who saw the same bug.
>>> >> http://www.vtk.org/Bug/view.php?id=13420
>>> >>
>>> >> I've already fixed a few issues in this volume mapper related to
>>> >> 32-bit overflow, but this threading bug should probably get the
>>> >> attention of someone who is more familiar with the code.
>>> >>
>>> >>  - David
>>> >>
>>> >>
>>> >> On Sun, Feb 24, 2013 at 10:34 AM, David Gobbi <david.gobbi at gmail.com>
>>> >> wrote:
>>> >> > Bill,
>>> >> >
>>> >> > I can reproduce the failure on any computer by increasing the
>>> >> > DefaultNumberOfThreads in the multithreader.
>>> >> >
>>> >> > Also, this is the "fixed point" mapper, it uses integer arithmetic.
>>> >> > It should produce _identical_ results on all platforms.
>>> >> >
>>> >> > I'll see if I can dig up some of my earlier list postings on it.
>>> >> > I've
>>> >> > spent many hours investigating this issue, both by myself and with
>>> >> > other VTK users.
>>> >> >
>>> >> >  - David
>>> >> >
>>> >> > On Sun, Feb 24, 2013 at 9:57 AM, Bill Lorensen
>>> >> > <bill.lorensen at gmail.com>
>>> >> > wrote:
>>> >> >> David,
>>> >> >>
>>> >> >> What makes you think this is a threading issue? The failng error is
>>> >> >> pretty
>>> >> >> low, <20 on all failing systems and the images look visually
>>> >> >> similar.
>>> >> >> http://open.cdash.org/testDetails.php?test=178462199&build=2824975
>>> >> >>
>>> >> >> Bill
>>> >> >>
>>> >> >> --
>>> >> >> Unpaid intern in BillsBasement at noware dot com
>>> >
>>> >
>>> >
>>> >
>>> > --
>>> > Unpaid intern in BillsBasement at noware dot com
>>
>>
>>
>>
>> --
>> Unpaid intern in BillsBasement at noware dot com
>
>
>
>
> --
> Unpaid intern in BillsBasement at noware dot com