[package list]

rccl

User-installable modules (and EasyConfigs)

Install with the EasyBuild-user module:

eb <easyconfig> -r

To access module help after installation and get reminded for which stacks and partitions the module is installed, use module spider rccl/<version>.

EasyConfig:

• EasyConfig rccl-2.12.7-cpeGNU-22.08.eb, will build rccl/2.12.7-cpeGNU-22.08

Technical documentation

RCCL is a stand-alone library of standard collective communication routines for GPUs, implementing all-reduce, all-gather, reduce, broadcast, reduce-scatter, gather, scatter, and all-to-all. There is also initial support for direct GPU-to-GPU send and receive operations. It has been optimized to achieve high bandwidth on platforms using PCIe, xGMI as well as networking using InfiniBand Verbs or TCP/IP sockets. Plugins can be used for other network architectures so LUMI users should have a look at the aws-ofi-rccl plugin to extract the full performance out of the SlingShot 11 interconnect of LUMI.

As AMD sometimes uses upercase and sometimes lowercase letters, we had to make a choice and went with lowercase letters for the module name.

• RCCL on GitHub

  • GitHub releases

EasyBuild

The RCCL EasyConfigs are a development of the LUST team.