[vtkusers] VTK performance with 1000's of actors

[Nikhil Shetty]

Hi Andrew,

I definitely agree with you. I think VTK and OSG started from two different
perspectives. VTK focusing more on data and algorithm processing  and OSG
focusing more on efficient rendering. They started with two different
paradigms serving different initial needs and in someway has to merge.

Although there they can be merged easily but I think there is more work to
be done on the Mapper, Actors to get it working as efficiently as a
Geometry,Node of a scene graph.

-Nix




On Mon, Sep 27, 2010 at 12:46 PM, Andrew Cunningham <andrew at a-cunningham.com
> wrote:

> Hi Aashish,
> I am very familiar with both VTK and OSG as I did a 2 month evaluation of
> both for our needs, building a functioning prototype using both API's.
>
> VTK = great sci visualization algorithms, terrible as a scene graph
>
> OSG = great scene graph functionality, more modern C++ usage, but no
> packaged, integrated  sci-vis algorithms. Very good performance as it is
> targeted at OpenGL exclusively. Even ported to the iPhone..
>
> As you note, it is fairly easy to link many of the VTK sci-vis algorithms
> with OSG.
>
> Unfortunately, on the flip side,  not much can be done about VTK's
> limitations with scene graphs and it's rather poor performance with many
> actors if that is what is needed.
>
> Andrew
>
>
>
> On Mon, Sep 27, 2010 at 9:35 AM, Aashish Chaudhary <
> aashish.chaudhary at kitware.com> wrote:
>
>> Andrew,
>>
>> OpenSceneGraph uses entirely different framework and has optimizations
>> advantages because of its data structure and scene graph pipeline. That
>> pipeline works well for infovis but does not work as well for sci vis (I
>> have used OSG for 4 years).
>>
>> If you want to use OSG for rendering and VTK for algorithm what you can do
>> is construct a VTK pipeline,update it, get the data, convert it into OSG
>> structure. Update and get data is required only if your input parameters
>> changes.
>>
>> I have followed this approach primarily as I was doing mix of sci vis and
>> inforvis.
>>
>>
>>
>>
>> On Mon, Sep 27, 2010 at 12:18 PM, Andrew Cunningham <
>> andrew at a-cunningham.com> wrote:
>>
>>> HI Martin,
>>> I noted this problem in VTK 5.4, and did some diagnosis. The very poor
>>> performance with many actors , can be due, I would argue , to a n! ( n
>>> factorial) issue in the handling of sub-objects of a vtkAssembly when
>>> checking the modification time.
>>>
>>> See my post at
>>> http://public.kitware.com/pipermail/vtkusers/2009-May/100928.html
>>>
>>> Basically VTK 5.x is terrible at handling complex scene graphs. Try and
>>> flatten your scene.
>>>
>>> I switched to using OpenSceneGraph , which  handles extremely complex
>>> scene graphs very efficiently as well as proper handling of state
>>> inheritance, and can be interfaced to VTK algorithms
>>>
>>> Andrew
>>>
>>> _______________________________________________
>>> 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 VTK FAQ at:
>>> http://www.vtk.org/Wiki/VTK_FAQ
>>>
>>> Follow this link to subscribe/unsubscribe:
>>> http://www.vtk.org/mailman/listinfo/vtkusers
>>>
>>>
>>
>>
>> --
>> | Aashish Chaudhary
>> | R&D Engineer
>> | Kitware Inc.
>> | www.kitware.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 VTK FAQ at:
> http://www.vtk.org/Wiki/VTK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.vtk.org/mailman/listinfo/vtkusers
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20100927/cd80aa30/attachment.htm>