MKL

Intel Math Kernel Library (MKL) is a proprietary, high-performance collection of mathematical routines for scientific computing and data analysis. It provides optimized implementations of the basic linear algebra subprograms (BLAS), the linear algebra package (LAPACK), fast Fourier transforms (FFT), random number generation, and a vector math library (VML). Designed to accelerate numerical workloads on Intel processors, MKL emphasizes architecture-specific optimizations, threaded execution, and scalable performance across cores and vector units.

Key components include dense and sparse linear algebra routines (BLAS/LAPACK), eigenvalue problems, FFTs, and a sparse solver suite such as PARDISO. The library also offers the Vector Math Library (VML) for vector operations and high-quality random number generators, providing a broad set of capabilities used in scientific and engineering applications.

APIs are provided for C, C++, and Fortran, with compatibility to common BLAS/LAPACK interfaces. MKL supports Windows, Linux, and macOS, and works with a range of compilers. It enables multi-threading through OpenMP and internal threading, and can automatically select optimized kernels based on hardware features such as AVX, AVX2, and AVX-512 to maximize performance on supported CPUs.

Distribution and evolution: MKL is distributed by Intel as part of its development tools and HPC product lines. Since the early 2020s, Intel has positioned MKL within the oneAPI initiative, consolidating it under the oneMKL umbrella to enable cross-architecture performance on CPUs and accelerators while preserving API compatibility with existing MKL code.