2.7.0

Overview

This release modernizes mkl-service with updated Intel MKL instruction set support, improved dependency management, and enhanced code quality. Python 3.9 support has been dropped, and several legacy CBWR branches have been removed in alignment with the latest Intel® oneAPI Math Kernel Library (oneMKL).

Highlights

🎯 New Intel Architecture Support

• Added support for AVX10 and AVX512_E5 instruction set architecture (ISA) constants
• Introduced AVX10 conditional numerical reproducibility (CNR) constants with strict variants

📦 Improved Dependency Management

• MKL is now explicitly declared as a runtime dependency
• Build process simplified with mkl-devel as a formal build dependency

🔧 Code Quality Improvements

• Fixed Cython compilation warnings for cleaner builds
• Removed star imports from package initialization, preventing namespace pollution
• Improved module import hygiene

⚠️ Breaking Changes

• Python 3.9 support dropped - Minimum supported version is now Python 3.10+
• Legacy CBWR branches removed: ssse3, sse4_1, avx, avx512_mic, avx512_mic,strict, and avx512_mic_e1 are no longer supported

Python Version Support

• ✅ Python 3.10, 3.11, 3.12, 3.13, 3.14
• ❌ Python 3.9 and earlier

Detailed Changes

Added

• Added support for ISA constants "avx10" and "avx512_e5" and CNR constants "avx10" and "avx10,strict" gh-175

Changed

• Added mkl as a runtime dependency and mkl-devel as build dependency in pyproject.toml gh-177

Removed

• Dropped support for Python 3.9 gh-118
• Dropped support for "ssse3", "sse4_1", "avx", "avx512_mic", "avx512_mic,strict", and "avx512_mic_e1" cbwr branches gh-173

Fixed

• Fixed Cython warnings when building mkl-service gh-183
• Removed use of star import in mkl-service initialization, which removes imported module pollution from the namespace gh-178

Contributors

• Anton Volkov (@antonwolfy)
• Evseniia Komarova (@ekomarova)
• Nikita Grigorian (@ndgrigorian)

New Contributors

• Nikolay Petrov (@napetrov)

Full Changelog: 2.6.1...2.7.0

v2.6.1

This is a bug-fix release, including resolving:

• an issue with the run-time dependencies of mkl-service package to explicitly depend on a non–free-threaded (GIL-enabled) Python gh-111

Full Changelog: 2.6.0...2.6.1

v2.6.0

Added

• Enabled support of Python 3.14 gh-100

Changed

• Used GIT_DESCRIBE_TAG and GIT_DESCRIBE_NUMBER in meta.yaml instead of manual stepping the numbers gh-98
• Updated meta.yaml recipe and documentation with approved oneMKL name gh-87

Contributors

• @antonwolfy
• @ekomarova

Full Changelog: v2.5.2...2.6.0

v2.5.2

This is a bug-fix release, including:

• Updated meta.yaml with proper license description to pass the validation rules gh-87

Contributors

• @antonwolfy

Full Changelog: v2.5.1...v2.5.2

v2.5.1

This is a bug-fix release, including resolving:

• import issue in the virtual environment which breaks loading of MKL libs gh-85

New Contributors

• @antonwolfy

Full Changelog: v2.5.0...v2.5.1

v2.5.0

Added support of Python 3.13.

Transition from setup.py to pyproject.toml.
Added out of the box support in virtual environment.

v2.4.2

Incremental release.

Tests checking library version moved to the end of the test suite, as after it is run, the state of the library is finalized, and tests that modify that state may fail.

Updated installation instructions.

v2.4.1

Transition from nose to unittest and then to pytest to enable support for Python 3.12. Added Github Actions CI. Removed six as a dependency.

v2.4.0.post1

• Update description for Pypi package installation

v2.4.0

Fixed issue #14.

Added mkl.set_num_stripes and mkl.get_num_stripes

Also expanded support isa keyword argument values in mkl.enable_instructions(isa=isa) function per recent Intel(R) oneMKL support.