Primary applications for SPACE are AI algorithms that require fast searching and processing within large, rapidly changing data structures. The PADMAVATI prototype has a peak performance of 136 billion 32-bit comparisons per second when the CAPPs are driven by transputers. The existing CAPP modules could provide a throughput of over 1 Tera comparisons per second with a dedicated microcontroller design. Primitive parallel search and write instructions can be composed into arbitrarily complex arithmetic and logical operations, allowing SPACE to be used as a powerful SIMD processor.
A paper presented at the 3rd International Workshop on VLSI for Neural Networks and Artificial Intelligence, Oxford, 1992, describes the machine in detail. Here are PostScript and PDF files without figures, and this is HTML generated automatically from the TeX source.