4 Pull requests merged by 2 people

• Multi-stream CUDA IPC

  #326 merged Jun 4, 2025

• Use a stream pool for gpuCalloc*()

  #509 merged Jun 4, 2025

• Improved documentation & minor interface revision

  #541 merged Jun 3, 2025

• Initial Steps on Channel Based Syntax

  #530 merged May 29, 2025

3 Pull requests opened by 2 people

• Implementing Operations Channel Based DSL

  #539 opened May 29, 2025

• Use the ibverbs library found during build

  #543 opened Jun 3, 2025

• Do not use tail replica by default

  #544 opened Jun 4, 2025

1 Issue closed by 1 person

• [Bug] About Performance drop in Cpp API

  #531 closed Jun 3, 2025

2 Issues opened by 2 people

• [Feature] Automatically Find the best NIC for each GPU

  #542 opened Jun 3, 2025

• [Feature] ProxyService avoid keeping ownership of the RegisteredMemory

  #540 opened Jun 2, 2025

7 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [Draft] Change executor to load new execution plan

  #538 commented on Jun 4, 2025 • 1 new comment

• MSCCL++ v0.7.0 Release Plan

  #494 commented on Jun 2, 2025 • 0 new comments

• [Bug] `cudaMemcpyAsync` blocks proxy thread making program hang

  #516 commented on Jun 4, 2025 • 0 new comments

• Revise NVLS interface

  #458 commented on Jun 2, 2025 • 0 new comments

• [Draft] DSL design for next version

  #519 commented on Jun 4, 2025 • 0 new comments

• Enable channel creation within a single device

  #532 commented on May 30, 2025 • 0 new comments

• Fix memory leak bug

  #534 commented on Jun 3, 2025 • 0 new comments