NVIDIA / NVIDIA/cuda-python

Rely more on the driver for input checks to reduce latency

Open
#685 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

cuda.core enhancement P1
Dominant language
Cython
Stars
3.4k
Forks
329
Avg merge
1d 23h
Merged PRs (30d)
116

Description

Instead of doing all of these checks in Python, if we just set the options as passed and try the launch, does CUDA error gracefully where we can post process the error code to make it meaningful to the user? Doing all of these checks in Python feels expensive and will hurt kernel launch latency.

Originally posted by @kkraus14 in https://github.com/NVIDIA/cuda-python/pull/676#discussion_r2126787942

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

The issue names no files, tests, or entry points. Review the Python-side input checks and the CUDA launch and error-handling path, then determine which checks can be delegated safely and how driver errors should be made meaningful to users. Done means reducing launch latency without losing useful input errors.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
performance
Issue type
Refactor
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.