[Insight-developers] itkMattesMutualInformationImageToImageMetricTest failures
Luis Ibanez
luis.ibanez at kitware.com
Mon Mar 24 17:11:04 EDT 2008
Hi Bill,
I'll do the runs with a single thread (I haven't done that yet). :-/
I was following the problems reported by valgrind, and I think
I fix them this time. However, the valgrind builds are using
the non-multi-threaded classes. (so there is still room for
errors in the multi-threaded ones.).
Luis
----------------------
Bill Lorensen wrote:
> Luis,
>
> You can run a test with any number of threads using the --withthreads
> n option. Have you tried this on your platform? I just updated itk and
> ran a failing test with --withthreads 1 and it passed. --withthreads 2
> and it still fails. I think this is true of the 6 tests failing on the
> borland build.
>
> Bill
>
> On Thu, Mar 20, 2008 at 6:00 PM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>
>>Hi Bill,
>>
>>Thanks for running the test with only one thread.
>>This is quite helpful.
>>
>>I believe the problem is that I haven't consolidated
>>the MetricDerivative contributions computed from each
>>one of the threads. I plan to put that together tomorrow.
>>
>>
>> Luis
>>
>>
>>--------------------
>>
>>Bill Lorensen wrote:
>>
>>>Luis,
>>>
>>>These failures (with optimized registration on) are related to
>>>threading. If I run the tests on my Borland build with number of
>>>thread = 1, they pass. Otherwise they fail as shown on my continuous
>>>build.
>>>
>>>Hope this helps find a solution to this.
>>>
>>>Bill
>>>
>>
>
More information about the Insight-developers
mailing list