Bullet physics and rendering in separate threads?

[ola]

Hi all,

I'm thinking about the possibility of separating the physics simulation and rendering for a new project, so that they run in two separate threads. The reason is to better utilize a dual-core processor. So, one thread should render the view as often as it can, the other should step Bullet at a as constant as possible 60 Herz. I'd need to create my own motion state handling for interpolating the graphics, then.

Have any of you tried this already? Does this approach make sense? Or are there better ways of spreading the cpu load onto several cores? The application will run on Linux, in double precision mode (beacuse of an earth-sized terrain). I am looking at BulletMultiThreaded, is that performing well on Linux?

Best regards,
Ola