ORNL / ORNL/HeCBench

[pingpong-hip] main-mpi requires GPU-aware MPI

Open
#170 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
306
Forks
119
Avg merge
2d 6h
Merged PRs (30d)
18

Description

Although the general README mentions that certain benchmarks require GPU-aware MPI, it's not explicitly mentioned anywhere that pingpong-hip is one of those benchmarks. I have two suggestions:

  1. Add a README.md file to pingpong-hip (and probably pingpong-cuda too) explicitly mentioning that it requires GPU-aware MPI.
  2. Add a fallback implementation for main-mpi that doesn't require GPU-aware MPI. If it's available, we use GPU-aware MPI, if it's not, we explicitly do mem copies and use host pointers for MPI calls instead of device pointers. If this is of interest, I do have a patch for this that I can share.

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

Inspect the pingpong-hip benchmark, its proposed README.md, the likely pingpong-cuda README.md, and the main-mpi entry point. First determine how GPU-aware MPI is currently required, then clarify whether the work is documentation only or also the fallback path. Done means the requirement is explicit and, if implemented, main-mpi works with host-pointer copies when GPU-aware MPI is unavailable.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
distributed-systems, performance
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.