UCCL

Pinned Loading

1. uccl uccl Public

   UCCL is an efficient communication library for GPUs, covering collectives, P2P (e.g., KV cache transfer, RL weight transfer), and EP (e.g., GPU-driven)

   C++ 1.5k 172

2. mKernel mKernel Public

   mKernel: fast multi-node, multi-GPU fused kernels

   Cuda 270 25

3. CommBench CommBench Public

   Can LLMs Write Correct and Efficient GPU Communication Code?

   Python 63 2

4. rdmatop rdmatop Public

   htop-like TUI for real-time RDMA network monitoring.

   Rust 90 9