Skip to content

MNNVL fix#604

Merged
chhwang merged 2 commits intomainfrom
chhwang/mnnvl
Aug 8, 2025
Merged

MNNVL fix#604
chhwang merged 2 commits intomainfrom
chhwang/mnnvl

Conversation

@chhwang
Copy link
Contributor

@chhwang chhwang commented Aug 8, 2025

No description provided.

@chhwang chhwang requested a review from Copilot August 8, 2025 04:05

This comment was marked as outdated.

@chhwang chhwang requested a review from Copilot August 8, 2025 04:09

This comment was marked as resolved.

@chhwang chhwang requested a review from Copilot August 8, 2025 04:12

This comment was marked as outdated.

@chhwang chhwang requested a review from Copilot August 8, 2025 04:18
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull Request Overview

This PR addresses an MNNVL (Multi-Node NVLink) fix by implementing conditional CUDA NVLS support in semaphore allocation and updating CUDA IPC transport logic.

  • Adds conditional NVLS-based allocation for GPU tokens when NVLS is supported
  • Modifies CUDA IPC transport condition to be more permissive
  • Includes necessary header for GPU utilities

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File Description
src/semaphore.cc Adds NVLS support check and conditional allocation path for GPU tokens
src/registered_memory.cc Removes host hash requirement from CUDA IPC transport condition

@chhwang chhwang enabled auto-merge (squash) August 8, 2025 19:20
@chhwang chhwang merged commit 1cc1b82 into main Aug 8, 2025
14 checks passed
@chhwang chhwang deleted the chhwang/mnnvl branch August 8, 2025 19:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants