Physics computing on GPU

I’m currently writing a rendering engine based on Vulkan.
For physics simulations, I plan on doing most/all of the calculations on the GPU.

So my question is, would it be better to use a shader running on Vulkan Compute for this, or an OpenCL kernel?

If it is tightly coupled with vis, I would start with Vulkan. OpenCL interop does not seem to get much traction: https://github.com/KhronosGroup/Vulkan-Ecosystem/issues/42. Meanwhile scientific compute features are rolled out in Vulkan (like controling rounding modes).

1 Like

This topic was automatically closed 183 days after the last reply. New replies are no longer allowed.