ROCm / ROCm/AMDMIGraphX

Measure when MLIR is faster or slower to update the heuristic

Open
#4,231 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
333
Forks
150
Avg merge
4d 19h
Merged PRs (30d)
54

Description

Gather mxr/pythons files from all the models including for NCHW/NHWC/fp16/bf16/fp32 and rewrite_dot/on/off. Measure the performance for each file with and without MLIR on all the different hardware. Make a list for when MLIR is slower and use that list to update the heuristic for when we pick MLIR.

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

Begin by locating the mxr/pythons files for the models and the heuristic that chooses MLIR. Benchmark each file with and without MLIR across NCHW, NHWC, fp16, bf16, fp32, rewrite_dot on/off, and the available hardware. Done means documenting slower cases and updating the heuristic based on those measurements.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp, python
Domain
compilers, machine-learning, performance
Issue type
Refactor
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.