NVIDIA Inference Xfer Library (NIXL) GitHub repository

NIXL 'is targeted for accelerating point to point communications in AI inference frameworks such as NVIDIA Dynamo, while providing an abstraction over various types of memory (e.g., CPU and GPU) and storage (e.g., file, block and object store) through a modular plug-in architecture'. Platform constraint stated: 'NIXL is supported on a Linux environment only... macOS and Windows are not currently supported'. Stack: tested with UCX 1.23.x, optional GDRCopy ('necessary for maximum performance, but UCX and NIXL will work without it'), optional etcd for cross-node metadata, and bundled NVIDIA ucx IB modules (libuct_ib_mlx5_ext, libuct_ib_mlx5_gda, libuct_ib_mlx5_gdp) under the NVIDIA Proprietary License. Ships nixlbench and kvbench benchmark harnesses.