Miles Macklin

Hi @kmarchais, agreed this is very useful for a number of applications - the `wp.hashgrid` right now is tiling periodically, i.e.: if a particle lies on a boundary cell (as...

Yes, we plan to support tf32, bf16, fp8, fp4 in roughly that order. Please add any additional information on your workload / GEMM requirements if you have them. Thanks!

Hi @BernardoCovas, this looks like an interesting use case, and I can see that having the ability to construct an array at kernel time from ptr would be useful indeed....

Just want to ask a clarifying question - in your example, would `image_trgts` be an array of e.g.: uint64 addresses representing the array memory for an image?

Thanks a lot for the repro @steinraf, I've not seen this before - we will investigate and get back to you.

Thanks @nithinsomu, unfortunately this is expected at present. I will add a task for us to address this for the next version. Best, Miles

Hi @yuhan-zh , we've just shipped a new cloth self-collision method in the VBDIntegrator, please try it out and let us know if you have any feedback and open another...

Hi @nithinsomu, I'm not particularly familiar with Torch Geometric - can you expand on which particular features would be the most useful? Or is it mostly a matter of data...

Hi @BernardoCovas, thanks for the report - indeed our CPU execution does not currently support multi-threaded execution. The main reason for this is the use of a static variable (https://github.com/NVIDIA/warp/blob/main/warp/codegen.py#L2401)...

Hi @marcelroed, good question, I think this should be possible. Let me ask our JAX experts to see if they can offer some advice. Best, Miles