DAILY NEWS AND INFORMATION FOR THE GLOBAL GRID COMMUNITY / SEPTEMBER 1, 2003; VOL. 2 NO. 35    ( Table of Contents )

Applications:

GRID COMPUTING GETS A SPEED BOOST FROM MATHEMATICA 5

Wolfram Research has significantly increased the performance of its grid computing application with the release of gridMathematica 1.1. gridMathematica now includes Mathematica 5, the new high-performance version of Wolfram's leading technical computing software. The update delivers large performance gains in dense numerical linear algebra, support for sparse numerical linear algebra, large-scale linear programming, arbitrary precision computation and interprocess communication speed.

Originally unveiled at the Supercomputing 2002 conference, gridMathematica is a complete parallel computing solution for dedicated grids or clusters. It offers a highly cost-effective way of deploying Mathematica for parallel computations. gridMathematica can run on any cluster of machines, including Unix, Linux, Windows and Mac OS X, and requires only TCP/IP connectivity. Version 1.1 capitalizes on a new TCP/IP protocol in Mathematica 5 that allows it to communicate at the speed of the underlying network. On most standard networks this means that users will see bandwidth improvements of a factor of 10 and latency improvements of a factor of 200; on faster networks the gains are even higher.

With gridMathematica, simple commands such as "ParallelEvaluate," "ParallelTable" and "ParallelMap" can often be used to quickly parallelize sequential codes or develop parallel applications from scratch. gridMathematica supports all common parallel computing constructs such as virtual shared memory, distributed memory, automatic or explicit process scheduling, explicit concurrency primitives and failure recovery of stranded processes.

"Mathematica 5 now offers the fastest technical computing on a single computer. gridMathematica 1.1 multiplies that by the grid to provide the best high-performance computing environment available," says Roger Germundsson, director of research and development at Wolfram Research. Both individual users and large-scale organizations can optimize their computations with this sophisticated, easy-to-use alternative for high-level programming applications.