Does Mathematica take advantage of multiple processors?

Mathematica® is able to take direct advantage of multiple processors in particular areas through the use of optimized libraries that support threading. In Mathematica 5.0 and 5.1, threading support was focused on the area where it made the most sense and would have the most impact, which is linear algebra on machine-precision real numbers. Expanding support for threading in Mathematica is an ongoing area of research and is a particular area of focus as hardware vendors move toward multi-core processors.

Linear algebra is accelerated through machine-optimized BLAS and LAPACK libraries. As important as it is to increase performance, the primary focus is on obtaining correct results and ensuring the stability of Mathematica. On some systems a non-threaded BLAS library provides the best performance while meeting quality expectations. However, for every release of Mathematica, the available libraries are re-evaluated on all platforms. So it is possible that a future version of Mathematica will be able to take advantage of multiple processors on systems that do not yet support this. Currently, Windows, Macintosh, 32-bit Linux x86, and HP-UX are able to take direct advantage of multiple processors.

On systems that are not able to take direct advantage of multiple processors, there are indirect performance gains based on the ability of the operating system to schedule processes across multiple processors. Since the Mathematica front end and kernel execute as separate processes, Mathematica is particularly suited to take advantage of these indirect gains. Additionally, many computations can benefit directly from multiple processors by linking multiple Mathematica kernels with each other using MathLink. Wolfram Research has an application package entitled Parallel Computing Toolkit that provides tools for accomplishing this. Wolfram Research also has a product entitled gridMathematica that leverages the tools in the Parallel Computing Toolkit for use on clusters, grids, and larger SMP systems. For more information on gridMathematica and Parallel Computing Toolkit, please visit the following URLs.

http://www.wolfram.com/products/gridmathematica/

http://www.wolfram.com/products/applications/parallel/

Download this FAQ as a Mathematica 6.0 Notebook

Questions or comments? Send email to support@wolfram.com.