Ubuntu
gloo-cuda package

gloo-cuda 0.0~git20230519.597accf-3 source package in Ubuntu

Changelog

gloo-cuda (0.0~git20230519.597accf-3) unstable; urgency=medium

  * Rebuild src:gloo with CUDA (non-free) enabled. (Closes: #1063743)
    Due to non-free B-D, section is changed from main to contrib.
  * Automatically generated by src:gloo :: debian/cudabuild.sh
    Please always merge the changes to the cuda variant back to
    src:gloo, thanks.

 -- Mo Zhou <email address hidden>  Wed, 24 Apr 2024 22:18:04 -0400

Upload details

Uploaded by:
Debian Deep Learning Team
Uploaded to:
Sid
Original maintainer:
Debian Deep Learning Team
Architectures:
any
Section:
misc
Urgency:
Medium Urgency

See full publishing history Publishing

Series Pocket Published Component Section
Oracular release multiverse misc

Downloads

File Size SHA-256 Checksum
gloo-cuda_0.0~git20230519.597accf-3.dsc 2.3 KiB 136c8cdaa51a777b03282074e375007d13610398bb3b56e066d46cb9f7f933d0
gloo-cuda_0.0~git20230519.597accf.orig.tar.xz 180.8 KiB 68f7bb91c706808d653cb4c9f81537f7c8a732544f76f992bf4e8d7be29c803a
gloo-cuda_0.0~git20230519.597accf-3.debian.tar.xz 5.2 KiB c66c6060d5fd77752693116d624a2707e92a054d7b699997a2c687666e862923

No changes file available.

Binary packages built by this source

libgloo-cuda-0: Collective communications library (shared object)

 Gloo is a collective communications library. It comes with a number of
 collective algorithms useful for machine learning applications. These
 include a barrier, broadcast, and allreduce.
 .
 Transport of data between participating machines is abstracted so that
 IP can be used at all times, or InifiniBand (or RoCE) when available.
 In the latter case, if the InfiniBand transport is used, GPUDirect can
 be used to accelerate cross machine GPU-to-GPU memory transfers.
 .
 Where applicable, algorithms have an implementation that works with system
 memory buffers, and one that works with NVIDIA GPU memory buffers. In the
 latter case, it is not necessary to copy memory between host and device;
 this is taken care of by the algorithm implementations.
 .
 This package ships the shared object for Gloo.

libgloo-cuda-0-dbgsym: debug symbols for libgloo-cuda-0
libgloo-cuda-dev: Collective communications library (development files)

 Gloo is a collective communications library. It comes with a number of
 collective algorithms useful for machine learning applications. These
 include a barrier, broadcast, and allreduce.
 .
 Transport of data between participating machines is abstracted so that
 IP can be used at all times, or InifiniBand (or RoCE) when available.
 In the latter case, if the InfiniBand transport is used, GPUDirect can
 be used to accelerate cross machine GPU-to-GPU memory transfers.
 .
 Where applicable, algorithms have an implementation that works with system
 memory buffers, and one that works with NVIDIA GPU memory buffers. In the
 latter case, it is not necessary to copy memory between host and device;
 this is taken care of by the algorithm implementations.
 .
 This package ships the development files.