by Matt Gillespie, technical author and editor. Intel Corp.

The Intel® Math Kernel Library is a highly optimized, thread-safe, multithreading-capable set of libraries of functions for linear algebra, discrete Fourier transforms, vector transcendental math, and statistics. Intel MKL 7.0 offers substantial improvements over previous versions of the libraries, while requiring very little expenditure on the part of developers in terms of money or effort. This latest version incorporates new optimizations to existing functional areas, as well as the addition of the following new functionality:

• ScaLAPACK routines extend linear algebra functionality to distributed-memory computing architectures.
• PARDISO direct sparse solver supports highly optimized computations on systems of linear equations with many null values.
• New Vector Statistical Library Random Number Generator Functions improve the flexibility and performance of implementations based on pseudorandom numbers, including Monte Carlo simulations.

Intel MKL 7.0 provides significant new optimizations to many of the functions that were included in previous versions of the libraries. These improvements build on previous Intel MKL functionality, including LAPACK solvers and eigensolvers, BLAS (level 1, 2, and 3) functions, Discrete Fourier Transforms, the Vector Math Library, and the Vector Statistical Library. Like previous versions, Intel MKL 7.0 is royalty-free and optimized for Intel mobile, desktop, and server architectures. The libraries are thread-safe and multi-threading-capable using OpenMP, and they include interfaces to popular workstation compilers for convenience.

This article provides independent software vendors, decision makers and programmers with the background they need to make an informed decision about adopting Intel MKL 7.0.