JuliaGPU / JuliaGPU/KernelAbstractions.jl

Pre-launch workgroupsize auto-tuning

Open
#216 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Julia
Stars
523
Forks
88
Avg merge
1d 11h
Merged PRs (30d)
25

Description

If the caller (host-side code) of a kernel needs to pre-allocate buffer that depends on workgroupsize and the workgroupsize is not specified, the caller needs to run the auto-tuning of workgroupsize before launching the kernel. For example, I used it for implementing "mapreduce" kernel in FoldsCUDA.jl. Can we have an API for invoking workgroupsize auto-tuning before launching the kernel?

Contributor guide

No contributing guide indexed for this repository

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

Start by reading the linked FoldsCUDA.jl example in src/kernels.jl, especially lines 87-98, and trace KernelAbstractions.jl's existing workgroupsize auto-tuning path. Done means a documented API can invoke that tuning before kernel launch so host code can use the selected workgroupsize for pre-allocation; validate it against the linked mapreduce use case.

Written by the indexing model from the issue text.

Assessment

Tech stack
julia
Domain
hpc
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.