Intel® Math Kernel Library 10.1 for Windows*, Linux*, and Mac OS* X

Product Brief The Flagship for High-Performance Computing Intel® Math Kernel Library 10.1 Math Software for Windows*, Linux*, and Mac OS* X Intel® Math Kernel Library (Intel® MKL) is a library of highly optimized, extensively threaded math routines for science, engineering, and financial applications that require maximum performance.

Availability • Intel® C++ Compiler Professional Editions (Windows, Linux, Mac OS X)

• Intel® Fortran Compiler Professional Editions (Windows, Linux, Mac OS X)

• Intel® Cluster Toolkit Compiler Edition (Windows, Linux)

• Intel® Math Kernel Library 10.1 (Windows, Linux, Mac OS X)

Functionality “By adopting the • Linear Algebra—BLAS and LAPACK • Fast Fourier Transforms Intel MKL DGEMM • Linear Algebra—ScaLAPACK • Vector Math Library libraries, our standard • Linear Algebra—Sparse Solvers • Vector Random Number Generators benchmarks timing DGEMM Threaded Performance Intel® Xeon® Quad-Core Processor E5472 3.0GHZ, 8MB L2 Cache,16GB Memory Intel® Xeon® Quad-Core Processor improved between Redhat 5 Server Intel® MKL 10.1; ATLAS 3.8.0 DGEMM Function

43 percent and 71 Intel MKL - 8 Threads Intel MKL - 1 Thread ATLAS - 8 Threads 100 percent…” ATLAS - 1 Thread 90

Matt Dunbar 80

Software Developer, 70 ABAQUS, Inc. s 60 p

GFlo 50

40

30

20

10

0 4 6 8 2 4 0 8 2 8 4 6 0 4 2 4 8 6 6 8 9 10 11 12 13 14 16 18 19 20 22 25 32 38 51

Matrix Size (M=20000, N=4000, K=64, ..., 512) Features and Benefits Vector Random Number Generators • Outstanding performance Intel MKL Vector Statistical Library (VSL) is a collection of 9 random number generators and 22 probability distributions • Multicore and multiprocessor ready that deliver significant performance improvements in physics, • Extensive parallelism and scaling chemistry, and financial analysis. • Royalty free redistribution

• Standard APIs in C and Fortran Random-Number Generators Probability Distributions Pseudo-random Continuous Discrete • World-class technical support Multiplicative Congruential 59-bit Uniform Uniform BLAS and LAPACK Multiplicative Congruential 31-bit Gaussian UniformBits Multiple Recursive GaussianMV Bernoulli Intel MKL provides extremely well-tuned BLAS and LAPACK Feedback shift register Exponential Geometric implementations that deliver significant performance leadership Wichman-Hill Laplace Binomial over alternative math libraries. Mersenne Twister 19937 Weibull Hypergeometric

Mersenne Twister 2203 Cauchy Poisson PTPE

ScaLAPACK Quasi-random Rayleigh Poisson Norm

Intel MKL includes a highly optimized version of ScaLAPACK Sobol Lognormal Poisson V regardless of block size and delivers significant performance Negative Niederreiter Gumbel improvements over the NETLIB* implementation. Binomial Gamma —

Beta — Fast Fourier Transforms Intel MKL Fast Fourier Transforms are highly optimized and provide Sparse Solvers significant performance gains over alternative libraries for medium The library includes both direct and iterative sparse solvers: and large transform sizes. Direct—PARDISO: A threaded, high-performance, memory Features efficient solver for large sparse linear systems of equations. • Outstanding multiprocessor scaling Version 10.0 introduced support for out-of-core memory! • Modern easy-to-use interface Iterative—FGMRES and Conjugate Gradient Solvers: FMGRES • FFTW interface wrappers for current FFTW users adds the capability to solve general sparse systems of linear • Support for distributed memory systems (clusters) equations while the Conjugate Gradient solver solves symmetric positive-definite systems

2D FFT Intel® Xeon® Quad-Core Processor E5472 3.0 GHz, 12MB L2 Cache, 16GB Memory Single-precision complex Redhat 5.0 Server Out of Place Intel® MKL 10.1, FFTW 3.1.2 Vector Math Library Intel MKL - 8T 35 Intel MKL - 1T FFTW - 8T Intel MKL provides vector implementations of computationally FFTW - 1T 30 intensive core mathematical functions. These include: 25 s p

lo 20 F

G Power, Exp, Log 15 Math Root Trig Hyper Tounding Special 10 Add Pow Cos Cosh Floor Exp 5 Sub Powx Sin Sinh Ceil Expm1 0

4 8 8 4 6 8 8 6 4 2 8 2 2 8 2 4 6 4 6 2 Div Pow2o3 SinCos Tanh Round Ln 6 4

4 4 x x 4 8 16x1 64x6 256x3 64x12 16x51 128x12 512x6 128x25 1024x6 256x25 1024x12 256x51 1024x25 256x102 1024x51 256x204 1024x102 4096x51 1024x204 2048x204 8192x102 2048x409 Sqr Pow3o2 Cis Asinh Trunc Log10 Transform Size Mul Sqrt Tan Acosh Rint Log1p

Conj Cbrt Acos Atanh NearbyInt Erf

MulByConj InvSqrt Asin - Modf Erfc

Abs InvCbrt Atan ErfInv

Inv Hypot Atan Performance Achieve outstanding performance from a math library that is highly optimized for Intel® Xeon®, Intel® Core™ i7, Intel® Itanium®, Intel® Pentium® and Intel® Core® processor-based systems. Intel MKL strives for competitive performance on Intel architecture compatible processors, which makes it the best choice for developers across all x86 platforms.

Compatibility Intel MKL runs on a variety of workstations, servers, and personal computers running Linux*, Windows*, and Mac OS* X operating systems. For details on hardware and software requirements please refer to www.intel. com/software/products/mkl.

Support Every purchase of Intel MKL includes one year of free upgrades, Intel® Premier Support, MKL user forum, and Intel knowledge base access.

Share experiences with other users of Intel MKL at the Intel moderated Intel MKL Discussion Forum at: http://softwarecommunity.intel.com/isn/Community/ en-US/forums/1273/ShowForum.aspx

Download a trial version today. www.intel.com/software/products/mkl

© 2009, Intel Corporation. All rights reserved. Intel and the Intel logo are trademarks of Intel Corporation in the U.S. and other countries. *Other names and brands may be claimed as the property of others. 0209/BLA/CMD/PDF 321488-001