FICHA · MANJARO

nccl

Library for NVIDIA multi-GPU and multi-node collective communication primitives

  • Library
  • LIBRARY
  • GPU-COMPUTE
  • HPC
  • Dependency only
official+codex · reviewed · May 28, 2026 description in en

Description

Provides NVIDIA's collective communication library for multi-GPU and multi-node computing. Machine learning and HPC frameworks use it to move data efficiently between GPUs during distributed training or numerical workloads.

Most users install it as a dependency of GPU compute software. It requires compatible NVIDIA drivers and hardware, and direct use is mainly for developers or researchers.

Permissions

Permissions not analysed for this source yet.