JuliaGPU/CUDA.jl

Consider maximizing grid utilization

Open

#1,321 opened on Jan 13, 2022

View on GitHub
 (1 comment) (0 reactions) (0 assignees)Julia (274 forks)batch import
good first issueperformance

Repository metrics

Stars
 (1,408 stars)
PR merge metrics
 (Avg merge 5d 5h) (16 merged PRs in 30d)

Description

We currently maximize block utilization (taking the max threads), which may leave SMs underutilized. We should consider first selecting an optimal amount of blocks, before maximizing the thread could:

    config = launch_configuration(kernel.fun)
    threads = min(length(ps), config.threads)
    # XXX: this kernel performs much better with all blocks active
    blocks = max(cld(length(ps), threads), config.blocks)
    threads = cld(length(ps), blocks)

I'm sure this will lead to some kernels performing worse, though, but it's probably a good thing to test.

Contributor guide

Research direction
Implement the proposed change to maximize grid utilization by first selecting an optimal number of blocks, then adjust threads accordingly, and run benchmarks on various kernels to measure performance impact.
Tech stack
None
Domain
performancebackend
Issue type
Performance
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
Julia
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
45