meta Feb 24, 2026

RCCLX: Innovating GPU communications on AMD platforms (opens in new tab)

gpu-acceleration distributed-computing llm-inference rcclx amd-rocm torchcomms tensor-parallelism low-precision-collectives