[ODE] hardware physics chip
ed.jones at oracle.com
Tue Mar 8 14:18:14 MST 2005
I imagine this would necessitate some form of "OpenP[hysics]L" or
"OpenD[ynamics]L" abstraction layer so that you can write your physics
code independent of the acceleration hardware being used?
Isn't the biggest benefit of video acceleration that you can just chuck
a load of vertices and textures onto the card, leave them there, and
periodically tell the card where to draw them without having to pull and
push stuff back and forth across the bus? If you had a physics
accelerator you could send the whole scene definition to the card, but
then each simulation step you'd have to pull the positional data off
that card and squirt it at the graphics card. To get full benefit
wouldn't you need the physics accelerator on-board the graphics card?
Sounds interesting though.
Jeffrey Smith wrote:
>Vrej Melkonian [mailto:vmelkon at yahoo.com] wrote:
>>Is it possible to create a special chip for physics? It's not like
>>graphics that can really benefit from a specially designed chip.
>Yes it is possible, and yes physics can benefit from a specially designed
>chip. It is possible to parallelize much of the physics computation
>required for medium-to-large scenes in addition to having specialized
>hardware for extremely fast 4x4 matrix computations
>Limor Schweitzer <limor666 at hotmail.com> wrote:
>>Does anyone know if these guys base their stuff on ODE ?
>They are using Novodex (www.novodex.com), which has been written by some
>ex-MathEngine employees (among others), and thus shares some architectural
>similarities with ODE.
>It's an extremely robust physics engine, you should check out the demos.
>ODE mailing list
>ODE at q12.org
More information about the ODE