Ubuntu
nvidia-nccl package

nvidia-nccl 2.18.5-1-2 source package in Ubuntu

Changelog

nvidia-nccl (2.18.5-1-2) unstable; urgency=medium

  * Team upload.
  * ci: Let blhc ignore some nonverbose lines.
  * Add upstream metadata.
  * With debhelper-compat 13 there is no more need for dh-exec.
  * Bump Standards-Version to 4.6.2, no changes needed.
  * Fix or override some issues found by Lintian.
  * Enable more hardening flags.

 -- Andreas Beckmann <email address hidden>  Sun, 12 Nov 2023 02:15:42 +0100

Upload details

Uploaded by:
Debian NVIDIA Maintainers
Uploaded to:
Sid
Original maintainer:
Debian NVIDIA Maintainers
Architectures:
amd64 arm64 ppc64el
Section:
misc
Urgency:
Medium Urgency

See full publishing history Publishing

Series Pocket Published Component Section
Noble release multiverse misc

Downloads

File Size SHA-256 Checksum
nvidia-nccl_2.18.5-1-2.dsc 2.1 KiB 31616a0c0b45e726acdfb553e949ec174503eb0469c4c23676cf974b36976787
nvidia-nccl_2.18.5-1.orig.tar.gz 354.1 KiB b4f5d7d9eea2c12e32e7a06fe138b2cfc75969c6d5c473aa6f819a792db2fc96
nvidia-nccl_2.18.5-1-2.debian.tar.xz 4.9 KiB 2ae0d151024c52faac933b01bbd779e0eb079cf1d4be3fea9f4abe9f62da3983

No changes file available.

Binary packages built by this source

libnccl-dev: NVIDIA Optimized primitives for inter-GPU communication (development)

 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.
 .
 This package contains the development files.

libnccl2: NVIDIA Optimized primitives for inter-GPU communication

 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.
 .
 This package contains the shared objects.

libnccl2-dbgsym: debug symbols for libnccl2