[Robotics] Performance test with realistic robot model

[Florian Lier]

Hey,

nice work Benoit ... I'll test my robot modell too - as soon as I can 
build 2.5 on my machine.
Got some weird compile errors too ... I'll try the svn update.

I added the video!

Cheers, Florian

Benoit Bolsee wrote:
> Hi,
>
> I received from the KUL a nice robot model and I tested it with iTaSC.
> You can see the result here: http://vimeo.com/5857397
>
> It allowed me to test the performance of iTaSC compared to the old
> solver. The difference is amazing! iTaSC is able to operate this robot
> in real time at 25 fps using 10% of one CPU core (E8400 @ 3GHz). This
> percentage does not include the time spent on graphic (4%). In this test
> iTaSC computes 100 iterations per second.
>
> The exact same test but using the old solver uses 99% of one CPU core
> and runs at 4.4fps. This means that iTaSC is about 56 times faster than
> the old solver and produces much more coherent results. This is without
> SSE optimization (significant performance gain can be expected from
> Eigen2 template library when SSE optimizations are enabled).
>
> When I enable the auto step algorithm, the performance doubles but the
> movements are slightly less stable.  
>
> The iTaSC algorithm is available in the itasc branch (based on blender
> 2.5) and tests builds are available at graphicall.org
>
> Regards,
> Benoit
>
> P.S. Florian, can you add this video to the Blender and Robotics group
> in Vimeo. I couldn't find how to do it.
>
> _______________________________________________
> Robotics mailing list
> Robotics at blender.org
> http://lists.blender.org/mailman/listinfo/robotics
>
>