This paper outlines the strategy and methodology used to create a portable coding environment for the commercial package MADYMO. The objective is to design a global data structure that efficiently utilises the memory and cache of computers, so that one source code can be used for serial, vector and parallel computer architectures. At this stage porting is confined to shared memory computers, however, the techniques used to parallelise MADYMO can also be applied to distributed memory architectures.

A description is given of the techniques that were used and the problems that were encountered when attempting to port MADYMO to a variety of computer platforms. By using techniques such as array index swapping, array padding, array blocking and unrolling, a substantial reduction in run-time has been achieved on serial, vector and parallel shared memory computers. Performance results are presented on various computer platforms, including Cray C90 and Silicon Graphics Power Challenge.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £82 +P&P)