[vtkusers] How to visualize thousands of actors with good performance

[Jim Peterson]

I had started to do some testing using Java as the host language, and 
Start and End event callbacks to the renderer to get
render times while interacting with the display. I also found the number 
of actors did not appear to be the significant factor, perhaps if Bill 
can run similar test with 100 and 1000 spheres will give some additional 
points for comparison across the versions. It may be the render times 
for increasing objects or points in not linear across the VTK versions.

FWIW, this is beginning to sound like an available memory or paging 
effect on the timings rather than a simple CPU time issue. perhaps the 
newer versions require a larger virtual storage load and hit the thrash 
boundary for paging earlier than older releases. This might be a reason 
for a knee in the graph of frame rate performance at some number of 
objects. The knee will likely be at a different point per version 
depending on the relative storage usage. 

HTH,
Jim

Bill Lorensen wrote:
> I just ran a small test on VTK4.4, VTK5.0, VTK5.2 and VTK Head. All
> built Release
>
> Rendered 10000 randomly distributed spheres. Ignore the absolute times
> (I have an old hand-me-down). All reported in frames per second (FPS):
>
> Version  MultipleActors  Single Actor
> 4.4.2             .51 FPS         .6 FPS
> 5.0.4           1.24 FPS       5.9   FPS
> 5.2.1             .92 FPS       6.0   FPS
> Head             .88 FPS       5.9   FPS
>
> So, in this test, I don't see degradation in the 5.x releases. It is
> interesting that 4.4.2 does not benefit from a single actor. But,
> also, the performance has improved from 4.4 to present. Affolter's
> pipeline must be different I think.
>
> I've attached my test code. Sorry for it being somewhat ugly (it has
> memory leaks), but I wanted something that would compile on all of the
> releases.
>
> Bill
>
> On Fri, Oct 1, 2010 at 11:19 AM, Pascal Augustin
> <pascal at rogue-research.com> wrote:
>   
>> Hi,
>>
>>     
>>> I'd just like to know the alternatives (if any), before I consider rewriting our selection and highlighting mechanism... It would really be great, if we could find a solution to this.
>>>       
>>  In your particular situation, would it be possible to use a single mapper for all your objects? Perhaps it could be a viable solution. For example, if you wanted to show 10000 spheres, you could do:
>>
>> 1) 10000 actors, 10000 mappers -> very slow
>> 2) 1 actor, 1 mapper -> very fast
>> 3) 10000 actors, 1 mapper -> acceptable speed
>>
>> Pascal
>>
>>
>> _______________________________________________
>> 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
>>
>>     
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> 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
>>