Executing narrow phase in parallel with broad phase

[console]

I am researching collision detection and I've implemented the broad phase on an FPGA, which is essentially a programmable chip. I then overrode Bullet's broad phase to use the FPGA. My FPGA algorithm generates the list of colliding objects in "bunches".

Currently, I have to wait until all of these bunches arrive before beginning narrow phase and physics computation. Is there some way I can begin processing some bunches (in a second thread) while I am still waiting for other bunches to arrive?

Thanks.

[Erwin Coumans]

This is possible, but it requires a custom collision/dynamics world and dispatcher. If you are interested, can you please file a request in the Bullet issue tracker?
(to make sure your request gets attention)

We are currently working on Bullet 3.x, and this should provide better support for parallelism in general.
Thanks,
Erwin

[console]

Thanks for your suggestion.

I've posted it to the issue tracker at http://code.google.com/p/bullet/issues/detail?id=327