Release 4.0 of ACML reaches new heights in mathematical functionality
December 2007. Experts from the Numerical Algorithms Group (NAG), the world's leading provider of mathematical and statistical software components, have once again teamed up with AMD to co-develop the next version of the AMD Core Math Library (ACML).
ACML is a highly tuned mathematical and statistical components library developed for high performance computing, particularly for users of 64-bit AMD chip systems. The library provides a suite of Basic Linear Algebra (BLAS) and LAPACK routines with many being specifically tuned for high performance computing. It also features comprehensive sets of Fast Fourier Transforms and Random Number Generators, again, both with highly tuned code.
New features in version 4.0 of ACML include:
- Support for AMD Family 10h (Barcelona) processors – double and single precision BLAS and LAPACK numerical routines have been tuned to deliver enhanced performance on quad core Barcelona systems
- Upgraded to include the latest LAPACK routines (version 3.1.1) and all LAPACK 3.1.1 bug fixes and enhancements applied
Release 4.0 of ACML further extends NAG and AMD’s longstanding collaborative relationship, which began with NAG and AMD developers working jointly on mathematical functionality even before the first version of ACML. Since then the two organizations have continuously developed ACML providing users with high quality, flexible and robust numerical functionality.
“AMD’s new Barcelona chip, with its enhanced floating-point performance, is a fantastic platform for technical computing”, said Mike Dewar, Chief Technical Officer at NAG. “NAG and AMD developers have re-written many of the assembler-code kernels from previous versions of ACML to take advantage of the characteristics of this new architecture, ensuring that whatever platform the software is running on it delivers the best possible performance to the user “
Marketing Communications Manager
The Numerical Algorithms Group (NAG) is dedicated to making world-class cross-platform mathematical, statistical, data mining components and tools for developers as well as 3D visualization application development environments. NAG serves it customers from offices in Oxford, Chicago and Tokyo supporting over 10,000 customer sites worldwide in finance, engineering, and scientific research. NAG software is the choice of over 25 independent software vendors including Oracle, IBM, DemandTec and many others.