Hi guys,
I hit an assert when building the static mesh shape for a relatively huge world (precision issue), so I might switch to the double precision mode.
I have observed that using double precision would disable SSE usage - mostly in the broadphase implementation - so I am affraid that performance will suffer noticeably..
Two questions :
1 - Has anyone benchmarked Bullet when using double precision mode, and compared the results against a single precision build ?
2 - Couldn't one imagine using double precision SSE2 instructions as well ?
Thanks a lot,
Cheers,
Gregory