CUDA streams are missing

Open ianmccul opened this issue 1 year ago • 1 comments

The GPU kernels need a mechanism to set and manage the cuda Stream. This is essential to enable parallelism on the GPU.

Nov 24 '24 11:11 ianmccul

This either can do using stream or just directly using threads. CUDA now give new default stream per thread instead of share same default stream.

Dec 12 '24 08:12 kaihsin