NVIDIA / NVIDIA/cccl

[cuda.compute] Add support for bf16 types with callable ops

Open
#10,941 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
C++
Stars
2.5k
Forks
486
Avg merge
2d 6h
Merged PRs (30d)
295

Description

https://github.com/NVIDIA/cccl/pull/10933/ skips support for bf16 with Python callable ops because of missing support in `numba-cuda` for `bf16`. Since `numba-cuda-mlir` _does_ support bf16, enable it here.

Contributor guide

Open the contributing guide

Research direction

Start by reading pull request #10933 and the existing CUDA compute support for Python callable ops. Compare the skipped bf16 path with the bf16 support available through numba-cuda-mlir; done means callable operations support bf16 without relying on missing numba-cuda support.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp, python
Domain
hpc
Issue type
Feature
Difficulty
3/5
Estimated time
1-2 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
64/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.