nullandkale

To display the closeups I'm using super splat in the browser and in the training tool I'm using my own hand written renderer. Mostly so it's easier to swap out splats during training

We have thought of supporting the tensor cores in cuda devices as well, we could probably use the same abstractions we need for that for the amx support. Unfortunately we mainly focus on cuda support because most cases people are using cuda for gpu compute purposes.

Currently ILGPU already supports arm64 linux if you have a cuda or opencl device, so supporting apple silicon is not out of the question. It has been floated that we would support Metal via Vulkan support and MoltenVK but not much work has been done on that front.

Currently SIMD support for a fast CPU accelerator is the main focus for the next big accelerator type.

I use ILGPU all the time! You would be amazed the performance you can get out of c# if you are careful about how you write your code. I recently wrote a super fast gaussian splat renderer in c# and used ILGPU to sort the splats on the gpu. You can even do weird cuda stuff like pass buffers between opengl and cuda, all in c#! https://github.com/NullandKale/OpenTKSplat/blob/main/OpenTKS...

I use ILGPU all the time for prototyping different image processing and rendering things for work and for my personal projects. I use cpp for a few things where the extra efficiency is worth it, but for most tasks c# is plenty fast enough.