SlackBuilds Repository

15.0 > Development > nccl (2.30.4.1)

NCCL (pronounced "Nickel") is a stand-alone library of standard
communication routines for GPUs, implementing all-reduce, all-gather,
reduce, broadcast, reduce-scatter, as well as any send/receive based
communication pattern. It has been optimized to achieve high bandwidth
on platforms using PCIe, NVLink, NVswitch, as well as networking using
InfiniBand Verbs or TCP/IP sockets. NCCL supports an arbitrary number
of GPUs installed in a single node or across multiple nodes, and can
be used in either single- or multi-process (e.g., MPI) applications.

Although this SlackBuild REQUIRES="cudatoolkit_13", cudatoolkit_12 has
also been successfully tested and may be substituted if neccessary.
Other CUDA toolkit versions may also be suitable but have not been
tested and are not supported.

Building nccl requires one of the cudatoolkits to be installed. Since
none of the cudatoolkits support anything other than x86_64 systems,
only x86_64 system architecture is supported by this SlackBuild.

This requires: cudatoolkit_13

Maintained by: Christoph Willing
Keywords: nccl,nvidia,multi,gpu,cudnn,ollama
ChangeLog: nccl