08:43
2026-05-29
marktechpost.com
ai-infrastructure
Meet mKernel: A Multi-GPU, Multi-Node Fused Kernel Library for GPU-Driven Communication
Researchers from UC Berkeley's UCCL project released mKernel, a library of persistent CUDA kernels that fuse intra-node NVLink communication, inter-node RDMA, and compute into a single kernel to addre…