Logo
PRESS RELEASE
Intel licenses technology from Oxford-based NAG

Intel's latest version of the high performance Math Kernel Library (MKL), Release 3.1, has added features and performance by adapting LAPACK software from the Numerical Algorithms Group (NAG), an Oxford-based software development company with almost thirty years of experience in scientific computing. These routines are both parallelised and optimised for best use of cache on Intel's new and future processor designs. The routines allow software developers to achieve excellent performance on standard PCs and an easy way to speed up their programs on multi-processor workstations and servers without rethinking and largely rewriting them.

The new routines solve medium to large size problems up to 3.8 times as fast on 4 processors as on a single processor. This same NAG software, running on other architectures has shown good scaling on up to 80 processors, designed to the LAPACK international standard makes it easy to use in financial, engineering and scientific programs.

Besides LAPACK, MKL contains optimized BLAS and FFTs. It comes with the cblas interface to ease use of the BLAS for the C/C++ programmer as well as a Fortran interface. It has full documentation for all the routines, with the LAPACK portion based on the highly regarded documentation for the NAG Fortran and Fortran SMP Libraries. MKL is available to all Intel software developers at:

http://developer.intel.com/vtune/perflibst/mkl/index.htm

"NAG looks forward to working with Intel to ensure that as many users as possible gain the full performance benefits of Intel's new processors with NAG technology" said Dr Brian Ford, Director of NAG.

Bruce Greer, manager for the Intel Math Kernel Library says "We wanted to offer MKL users the functionality of LAPACK with the high quality and additional performance available through NAG's implementation of this software."

Notes for Editors

These routines are available on other computing platforms as part of the NAG Fortran SMP Library which offers leading edge performance as well NAG's renowned reliability and accuracy. The NAG Fortran SMP Library contains all the functionality of the NAG Fortran Library so that many users only need to re-link to benefit from the better performance.

NAG has almost thirty years experience in developing, porting, distributing and supporting scientific software. NAG supplies a range of software libraries, compilers, tools and associated products for the professional programmer in business, industry and education. Based in Oxford, UK, NAG Ltd. has subsidiary companies in Chicago, Munich and Tokyo, as well as distributors worldwide.
 
 

For Further Information contact:

Liz Jupe
Marketing Communications Manager
NAG Ltd
Wilkinson House, Jordan Hill Road
Oxford, UK, OX2 8DR
Tel: +44 (0) 1865 511245
Fax: +44 (0) 1865 310139
WWW: www.nag.co.uk

15 March 1999

The Numerical Algorithms Group Ltd, Oxford UK, 1999