NVIDIA Linux open GPU kernel module source

Infiniband Verbs Performance Tests

ROCm Communication Collectives Library (RCCL)

AWS Libfabric

Samples for CUDA Developers which demonstrates features in CUDA Toolkit

This is a plugin which lets EC2 developers use libfabric as network provider while running NCCL applications.

Open Fabric Interfaces

NCCL Tests

ludicrously awesome [w]rapper for nvidia-docker

Source code examples from the Parallel Forall Blog

Optimized primitives for collective multi-GPU communication

Open MPI main development repository