Description of the Xputer Paradigm
Many applications require intensively iterative data manipulations to be performed on a large amount of data, e.g. statement blocks in nested loops. The new architectural class of Xputers is especially designed to reduce the von Neumann bottleneck of repetetive decoding and address interpreting and data manipulations. This bottleneck contributes a significant amount ( up to 90% in image processing, 58% in digital signal processing) to the run time of previously mentionned type of algorithms. High performance improvements have been achieved for a large class of algorithms within important application areas such as e.g. digital signal processing, image processing, electronic design automatation and mathematical computation-intensive problems.
The Xputer is a non-von Neumann machine paradigm, since it does not have an instruction sequencer, but a data sequencer is used instead. The basic structure of an Xputer module consists of three major parts :
If you want to know more
about Xputers have a look on our FAQ-Pages
or on our Publications(HTML-Code!).
© Copyright 1996, University of Kaiserslautern, Kaiserslautern, GermanyWebmaster