ReleaseNVIDIANVIDIApublished Mar 12, 2026seen 15h

NVIDIA/cuDecomp v0.6.2

NVIDIA/cuDecomp

Open original ↗

Captured source

source ↗
published Mar 12, 2026seen 15hcaptured 15hhttp 200method plain

v0.6.2

Repository: NVIDIA/cuDecomp

Tag: v0.6.2

Published: 2026-03-12T21:30:30Z

Prerelease: no

Release notes:

What's Changed

This patch release includes a workaround for an out-of-bounds host write bug discovered in cuTENSOR v2.3.x - v2.5.x impacting cuDecomp when using large input sizes (see issue #112 for details). It also includes changes to better handle large inputs in general.

Breaking changes

None.

Deprecations

None.

PRs included in this release

  • Improve support for very large inputs. (https://github.com/NVIDIA/cuDecomp/pull/105)
  • Add workaround for cuTENSOR 2.3.x - 2.5.x OOB host write bug impacting large problem sizes (https://github.com/NVIDIA/cuDecomp/pull/113)

Full Changelog: https://github.com/NVIDIA/cuDecomp/compare/v0.6.1...v0.6.2