NVIDIA/cuDecomp v0.6.0

v0.6.0

Repository: NVIDIA/cuDecomp

Tag: v0.6.0

Published: 2025-11-18T15:50:54Z

Prerelease: no

Release notes:

What's Changed

This release includes new features in cuDecomp, relicenses the project to Apache 2.0, and adds support for CUDA 13.0. This release adds the ability for users to transpose the 2D process rank assignment from the row-major to column-major (enabled via new environment variable CUDECOMP_USE_COL_MAJOR_RANK_ORDER, see https://github.com/NVIDIA/cuDecomp/pull/85 for more details).

Breaking changes

None.

Deprecations

None.

PRs included in this release

• Add user option for column-major rank assignment. (https://github.com/NVIDIA/cuDecomp/pull/85)
• Update to Apache 2.0 license. (https://github.com/NVIDIA/cuDecomp/pull/91)
• Adding function for MPI communicators in Fortran integer format to simplify Fortran interfaces. (https://github.com/NVIDIA/cuDecomp/pull/92)
• Making required changes for CUDA 13 support. (https://github.com/NVIDIA/cuDecomp/pull/86)
• Fix minor issue in Poisson example. (https://github.com/NVIDIA/cuDecomp/pull/94)
• Correctly propagate C++17 standard flags to NVCC compilation with older CMake versions. (https://github.com/NVIDIA/cuDecomp/pull/93)

Full Changelog: https://github.com/NVIDIA/cuDecomp/compare/v0.5.1...v0.6.0