MatX icon indicating copy to clipboard operation
MatX copied to clipboard

Published 20 hours ago • verified publisher icon NVIDIA

Refactor to use thrust::reduce on any.

Open ZelboK opened this issue 7 months ago • 11 comments

Saw the issue open for this and did this on a whim. I'm not too comfortable deleting code from this repository so for brevity I'll leave things as is. All this does is just use thrust::reduce instead of the defined kernels

Some clarifying points:

  1. This PR is likely not complete. I figured it would be better to get feedback earlier on incase I am doing something wrong.
  2. Tests are seemingly passing on my 3080 WSL2 build but some of them segfault on main so I can't replicate a perfect test scenario.
  3. I am leaving the dispatching to thrust. The Executors I didn't spend too much time studying.

ZelboK avatar Jul 27 '24 20:07 ZelboK