[vtk-developers] Shared Memory Parallelism in VTK
jianwei li
ljianweiyx at gmail.com
Tue Mar 17 10:19:46 EDT 2015
Hi
There is no the <vtkSMPTools.h> in the VTK source file.Is this a unused
class,and we have to complete it in the future.
Best Regard
Jianwei Li
2015-03-17 21:54 GMT+08:00 jianwei li <ljianweiyx at gmail.com>:
> Hi
> Now,in my proposal,I want use the implemented paralle methods to
> implemente the cutting algorithm.But,do you have a guide line for the
> implemented paralle methods.
> Thank you
> Cheers
> Jianwei
>
> 2015-03-17 11:17 GMT+08:00 David Gobbi <david.gobbi at gmail.com>:
>
>> Hi Jianwei,
>>
>> You wouldn't directly use TBB or Kaapi, because VTK provides an abstract
>> SMP interface. This doxygen page provides a brief synopsis. The "Functor"
>> described on the page is a function-like object with the signature "void
>> func(vtkIdType first, vtkIdType last)" where vtkIdType is an integer type.
>> Basically, the SMP For() function is given a (first,last) range that covers
>> the entire data set, and divides it into smaller (first,last) sub-ranges
>> that the Functor operates on from within the worker threads.
>> http://www.vtk.org/doc/nightly/html/classvtkSMPTools.html
>>
>> To complete the picture, you can take a look at the Kaapi back end and
>> the TBB back end:
>>
>> https://gitlab.kitware.com/vtk/vtk/blob/master/Common/Core/SMP/Kaapi/vtkSMPToolsInternal.h.in
>>
>> https://gitlab.kitware.com/vtk/vtk/blob/master/Common/Core/SMP/TBB/vtkSMPToolsInternal.h.in
>>
>> So to answer your question, the proposal would be to write (and then
>> evaluate) an efficient parallel VTK algorithm that uses the abstract
>> interface described on the first web page.
>>
>> Cheers,
>> - David
>>
>>
>> On Mon, Mar 16, 2015 at 7:56 PM, jianwei li <ljianweiyx at gmail.com> wrote:
>>
>>> So,do I need to write a proposal to implement anyone algorithm using TBB
>>> or XKaapi
>>>
>>> 2015-03-17 3:59 GMT+08:00 Berk Geveci <berk.geveci at kitware.com>:
>>>
>>>> Maybe a brief Wiki page describing the overall volume splitting and
>>>> static thread assignment would be good. We can put a link from the project
>>>> page. Beyond that, when/if we pick a student for the project and Google
>>>> funds it, we can elaborate.
>>>>
>>>> -berk
>>>>
>>>> On Mon, Mar 16, 2015 at 3:17 PM, David Gobbi <david.gobbi at gmail.com>
>>>> wrote:
>>>>
>>>>> On Mon, Mar 16, 2015 at 12:52 PM, Berk Geveci <berk.geveci at kitware.com
>>>>> > wrote:
>>>>>>
>>>>>>
>>>>>> David: Is there a reference describing the VTK threaded image
>>>>>> algorithm infrastructure or do we just point to the code?
>>>>>>
>>>>>
>>>>> There's no reference for vtkThreadedImageAlgorithm that I'm aware of,
>>>>> though I can write something up on the wiki. Even though the code itself
>>>>> isn't very complicated, it would be hard to follow for anyone who isn't
>>>>> familiar with a) the streaming pipeline and b) the vtkMultiThreader.
>>>>>
>>>>> - David
>>>>>
>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20150317/4a9cfdf8/attachment.html>
More information about the vtk-developers
mailing list