Unnecessary local mempcy in postAllReduce

Open
#5,567 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
42/100
Issue type
Refactor
Clarity
Mostly clear
Activity status
Stale
Tech stack
cpp
Domain
distributed-systems, performance

Research direction

Start in csrc/multidevice/communication.cpp around line 493 and trace postAllReduce's host IR generation. Check how the current local memcpy is introduced and how the input and output buffers are represented for allreduce. Done means the generated host IR reuses the same input and output buffer and no longer creates the unnecessary local copy.

Written by the indexing model from the issue text.

Description

Multi-GPU

https://github.com/NVIDIA/Fuser/blob/cce887595dc86b099506b70f88d653880fde5116/csrc/multidevice/communication.cpp#L493

This can be fixed by generating host IR that reuses the same input and output buffer for allreduce.

Dominant language
C++
Stars
404
Forks
84
No merged PRs in 30d

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.

More from NVIDIA/Fuser

All issues in NVIDIA/Fuser

Similar issues

More C++ issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.