02-04-2023, 05:58 PM
Unfortunately NVidia's CUDA implementation does not necessarily produce identical results to their CPU implementation, and the implementation itself is closed-source, so I can't dig into the code to find out why.
When I move tyFlow's PhysX version up to 5 I hope to explore this a little more in-depth and also contact the PhysX devs about it to get it fixed.
When I move tyFlow's PhysX version up to 5 I hope to explore this a little more in-depth and also contact the PhysX devs about it to get it fixed.