OpenMathLib/OpenBLAS

OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.

What it solves

OpenBLAS is an optimized library for Basic Linear Algebra Subprograms (BLAS), providing high-performance implementations of fundamental mathematical operations on vectors and matrices. It also includes the LAPACK (Linear Algebra Package) for higher-level linear algebra routines.

How it works

It is based on GotoBLAS2 and is highly optimized for a vast array of CPU architectures. It achieves performance by providing specialized kernels for specific CPU models (e.g., Intel Xeon, AMD Zen, ARM Cortex, RISC-V, POWER) and supporting dynamic architecture detection at runtime to ensure the most efficient instructions are used for the given hardware.

Who it’s for

It is designed for developers and researchers who need fast, reliable linear algebra operations for scientific computing, machine learning, and other data-intensive applications across various operating systems (Linux, Windows, macOS, Android, etc.) and hardware architectures.

Highlights

  • Broad Hardware Support: Optimized for x86, ARM, MIPS, PPC, RISC-V, LoongArch, and WebAssembly.
  • Dynamic Architecture Detection: Can be built with DYNAMIC_ARCH to automatically detect and use the best CPU-specific optimizations at runtime.
  • LAPACK Integration: Includes the Linear Algebra Package for more complex mathematical operations.
  • Multi-platform: Supports a wide range of OSs including GNU/Linux, Windows, macOS, and various BSDs.

Related

  • Project
    ROCm/rocBLAS

    rocBLAS is a Basic Linear Algebra Subprograms library implemented in HIP and optimized for AMD GPUs to provide high-performance linear algebra operations.

  • Project
    uncomplicate/neanderthal

    Neanderthal is a Clojure library for high-performance matrix and linear algebra computations utilizing optimized native BLAS and LAPACK routines on CPU and GPU.

  • Project
    vlang/vsl

    A high-performance scientific computing library for the V language providing linear algebra, numerical methods, and GPU acceleration for AI and scientific research.

  • Project
    uxlfoundation/oneDAL

    A C++ and DPC++ library that provides hardware-accelerated machine learning routines for tabular data across CPUs, GPUs, and distributed systems.

  • Project
    mlpack/ensmallen

    A high-quality C++ header-only library providing a wide range of non-linear numerical optimizers, including gradient-based and gradient-free methods.