OpenBLAS is an optimized BLAS library derived from GotoBLAS2 1.13 BSD version, providing high-performance implementations of Basic Linear Algebra Subprograms for scientific computing and numerical algorithms. Written primarily in C, the library serves as a critical component in the computational mathematics ecosystem, with documented connections to major projects including Julia, SciPy, and Microsoft Visual Studio Code through overlapping contributor networks.
OpenBLAS includes LAPACK, the Linear Algebra Package, as an integrated component, offering both basic and higher-level linear algebra functionality. The library supports an extensive range of CPU architectures and operating systems, spanning x86/x86-64 processors from Intel Westmere through Sapphire Rapids and AMD architectures from Bobcat through Zen5, as well as ARM variants including ARMv6, ARMv7, and comprehensive ARM64 support covering Cortex processors, AWS Graviton variants, and Apple Vortex architectures. Additional platform support extends to MIPS32 and MIPS64, PPC/PPC64 including POWER8 through POWER10, IBM zEnterprise systems, RISC-V architectures, and LoongArch64.
The build system offers flexibility through both GNU Make and CMake, with automatic CPU detection and explicit target specification capabilities. Cross-compilation is supported through environment variable configuration, and the library can be compiled with optional MASS support on Power CPUs for mathematical acceleration. Debug builds are available, and installation can be directed to custom directories through PREFIX configuration.
Official binary packages are provided for Windows x86, x86-64, and ARM64 platforms, distributed through SourceForge and GitHub Releases. The library is additionally packaged across numerous package managers, with installation details documented in the official documentation. Building from source requires GNU Make or CMake, a C compiler such as GCC or Clang, and optionally a Fortran compiler for full LAPACK support, though machine-translated C versions of LAPACK are available when Fortran compilation is unavailable.
The repository maintains continuous integration testing across multiple platforms through Cirrus CI, OSUOSL POWERCI, and OSUOSL IBMZ-CI, ensuring reliability across the supported architecture spectrum. Documentation is hosted at openmathlib.org, with comprehensive build instructions and platform-specific guidance available for developers and users implementing OpenBLAS in their scientific computing workflows.
