pytorch / pytorch/gloo

Compile USE_IBVERBS & work with RoCEv2

Open
#304 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
1.5k
Forks
368
Avg merge
19h 32m
Merged PRs (30d)
3

Description

Hi,
we are trying to enable the ibverbs & make it work with RoCEv2 (e.g., using both librdma_cm & libibverbs). However, we cannot successfully compile the source code.

Our question is whether Gloo can work with RDMA (RoCEv2 instead of IPoIB or RoCEv1) now? Do we have any documentation for transport part (it shows 'TBD' in doc folder)? Thanks.

Extra Info:
When we tried to build Gloo with USE_IBVERBs option, following error shows up:
/home/tian/gloo/gloo/transport/ibverbs/device.h:22:2: error: #error "Expected GLOO_HAVE_TRANSPORT_IBVERBS to be defined"

Contributor guide

Open the contributing guide

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 with gloo/transport/ibverbs/device.h and the build configuration for USE_IBVERBS, focusing on why GLOO_HAVE_TRANSPORT_IBVERBS is not defined. Review the transport documentation in doc/, which is marked TBD, and establish whether RoCEv2 is supported. Done means the build behavior and RoCEv2 support are verified and documented.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
build-system, documentation, networking
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.