koobas.hobune.stream
GPU Sorting
I'm trying to implement sorting using DirectCompute for my particle system. I found this post, which is great, but the sorting is incredibly slow. The C++ version can sort a million integers in a few milliseconds. The Unity/C# version is running at about 2000 integers per second. I'm getting better performance than that with CPU sorting right now.
Is this a limitation of Unity? Has anyone come across a fast sorting algorithm that can handle more than a few thousand elements?
Where exactly does the algorithm run slowly? Can you post your code/project? In my experience, computeBuffer.getData() runs EXTRE$$anonymous$$ELY slow (this is the part of the process when you get data back from your GPU into your CPU) and you need to watch out for that. It'll show up in the profiler as RenderTexture.SetActive or something like that.
