daal

What it is and what it does

oneDAL is Intel's oneAPI Data Analytics Library packaged for Python, providing optimized implementations of common machine learning algorithms (regression, clustering, forests) that run on CPUs, GPUs, and distributed clusters. It accelerates computation by leveraging SIMD instructions and cache optimization on CPUs, and SYCL/oneMKL on GPUs. The library is part of the UXL Foundation and is typically used either directly via its C++ interfaces or indirectly through the Extension for Scikit-learn, which patches scikit-learn to call oneDAL behind the scenes.

The Python package depends on tbb (Intel Threading Building Blocks) for parallelization and is distributed as platform-specific wheels for x86_64 Linux (manylinux_2_28) and Windows. Installation has medium friction due to these binary constraints. The library is actively maintained (last release 2026-06-10, repository last commit 2026-08-14) and is classified as Production/Stable, making it suitable for production workloads where tabular ML performance is critical.

Use it for:

• Accelerate scikit-learn pipelines by installing the Extension for Scikit-learn to call oneDAL routines transparently.
• Perform distributed K-means clustering and other algorithms across multi-node setups using oneDAL's MPI support.
• Speed up linear regression, random forest, and other tabular ML tasks on modern CPUs via SIMD and cache optimization.
• Leverage GPU acceleration for ML workloads on systems with SYCL-compatible GPUs using oneDAL's DPC++ interfaces.
• Integrate oneDAL C++ routines into Python data science applications for performance-critical tabular data processing.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

oneDAL is a C++ and DPC++ library that implements accelerated machine learning routines for tabular data (linear regression, K-means clustering, random forests, etc.) for CPUs, GPUs, and distributed setups, with a Python interface powered by tbb.

Yes, with conditions. oneDAL is actively maintained, production-stable, and offers genuine performance benefits for tabular machine learning on CPUs and GPUs. However, the unclear Intel Simplified Software License requires verification before commercial use, and installation is limited to x86_64 Windows or manylinux_2_28 Linux. Most users will benefit more from the Extension for Scikit-learn (which uses oneDAL transparently) than from direct use of this package.

Install

daal on PyPI

pip install daal

uv add daal

poetry add daal

Installing daal

Before you install

Medium install friction due to platform-specific wheels (manylinux_2_28_x86_64 and win_amd64) and a compiled runtime dependency on tbb. Active maintenance with recent releases and an engaged repository (651 stars, last commit 2026-08-14).

License in practice

Licensed under Intel Simplified Software License with unclear license treatment. Not an SPDX-recognized open-source license, so commercial or proprietary use may require explicit clarification with Intel before deployment.

Quickstart

pip install daal
import daal
# Access oneDAL routines via the daal module interface

Requires tbb runtime dependency; limited to x86_64 Windows or manylinux_2_28 Linux platforms.

Verify before relying

• Exact Python version support range (requires_python unspecified in metadata)
• Whether the PyPI daal package is the primary Python distribution or a wrapper around the C++ library
• Performance gains relative to standard scikit-learn on typical hardware
• Compatibility matrix with specific GPU vendors and SYCL implementations

Package facts

Evidence: daal-2026.1.0-py2.py3-none-manylinux_2_28_x86_64.whl; daal-2026.1.0-py2.py3-none-win_amd64.whl

Tags

More Libraries packages