Math Kernel Library

from Wikipedia, the free encyclopedia
Math Kernel Library
Basic data

Maintainer Intel
developer Intel
Publishing year 2003
Current  version MKL 2020
(December 11, 2019)
operating system Unixoide , mac OS X , Microsoft Windows
programming language C , C ++ , Fortran
category Program library for mathematics
License proprietary
software.intel.com/en-us/intel-mkl

The Intel Math Kernel Library (Intel MKL) is a program library for mathematical calculations from Intel . The main components are Basic Linear Algebra Subprograms (BLAS), LAPACK , ScaLAPACK , solver for sparse matrices , fast Fourier transformations and vector mathematics.

The Math Kernel Library was released by Intel on May 9, 2003. It is particularly suitable for Intel microprocessors and use in multithreading . Processors not from Intel are at a disadvantage, as the program code runs much slower than necessary, slowdowns by a factor of 3–10 have been observed in some cases, and performance-enhancing instruction sets such as AVX are not used, although they may be available. There are now hotfixes to temporarily restore the lost performance.

A license is required for use, and there is a free community license. The library supports C , C ++ and Fortran natively and is compatible with Java , C # , Python and other languages.

In the area of numerical linear algebra , BLAS and LAPACK are included and have been expanded to include high-performance calculations. These include LR , Cholesky and QR decomposition , eigenvalue problems and the least squares method . The Math Kernel Library contains fast Fourier transforms (FFT) for any number of dimensions and different data types. Like the FFTs, LAPACK is also included as a cluster version. Vector math functions are implemented in single and double precision , and are available for complex and floating point numbers . The statistics area includes random number generators and probability distributions . Different splines are included for interpolation . The Math Kernel Library has functions for neural networks (e.g. backpropagation ).

Web links

Individual evidence

  1. ^ Indiana University: What is the Intel Math Kernel Library, and where is it installed at IU and on XSEDE? In: Indiana University. Indiana University System , August 20, 2015, accessed March 25, 2017 .
  2. ^ Richard Friedman: Intel MKL and Intel TBB Working Together for Performance. In: Inside HPC. insideHPC Media, March 23, 2017, accessed on March 25, 2017 .
  3. ↑ AMD Ryzen 3900X vs Intel Xeon 2175W Python numpy - MKL vs OpenBLAS. Retrieved December 11, 2019 .
  4. ↑ How To Use MKL with AMD Ryzen and Threadripper CPU's (Effectively) for Python Numpy (And Other Applications). Retrieved December 11, 2019 .
  5. ↑ Alexander Neumann: Intel assembles distribution for Python developers. In: heise Developer . Heise Medien GmbH & Co. KG, March 21, 2016, accessed on March 25, 2017 .
  6. ↑ Intel: Intel Math Kernel Library. In: Intel. Intel Corp., accessed March 25, 2017 .