NCCLNet Plugin
NCCLNet provides a set of plugins designed to optimize NVIDIA's NCCL library for Spectrum-X and Quantum platforms. It can be used to achieve more resilient and performant communications for these platforms. In HPC-X, the plugin is located at $HPCX_DIR/ncclnet_plugin/lib
.
For NCCL to detect the network plugin, add the plugin path to the library search path environment variable.
The plugin can be loaded by explicitly setting the library search path using LD_LIBRARY_PATH:
$ export LD_LIBRARY_PATH=$HPCX_DIR/ncclnet_plugin/lib:$LD_LIBRARY_PATH$ <run command>
With HPCX, the plugin can also be loaded by NCCL's environment variable NCCL_NET_PLUGIN=ncclnet