Add CUDA IPC transport implementation #120
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Implements a new transport layer using CUDA IPC (Inter-Process Communication) for direct GPU-to-GPU memory sharing within a single node. This enables zero-copy transfers between processes by sharing GPU memory handles instead of copying data through CPU memory. Key features: - Direct GPU memory access using CUDA IPC handles - Eliminates GPU->CPU->GPU copies for intra-node transfers - Leverages NVLink/PCIe P2P when available - Automatic fallback for non-CUDA tensors Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>
meta-cla
bot
added
the
CLA Signed
This was referenced Feb 7, 2026
Member
|
a general question is when do we find cuda IPC useful? my understanding is it would only be single host synchronous weight sync cases. do you see this as a common usage case? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Implements a new transport layer using CUDA IPC (Inter-Process Communication) for direct GPU-to-GPU memory sharing within a single node.
This enables zero-copy transfers between processes by sharing GPU memory handles instead of copying data through CPU memory.
Key features: