| OCR Text |
Show CHAPTER 8 CONCLUSIONS It is widely accepted that the organization of next generation computers will reflect a radical departure from traditional Von Neumann architectures. Recent advancement in applications of computers suggests that the processing of symbols rather than numbers will be the basis for the next generation of computers [91,192, 197,199]. Concurrency will have to be exploited at all levels of the computer system to meet the boundless computing demands of artificial intelligence applications. In this thesis, we have presented a parallel processing framework for exploring parallelism in a sequential CLP backtracking search. By making use of high-level control parallelism and low-level data parallelism, several AI computer architectures for constraint-based symbolic processing have been developed that offer many orders of magnitude of performance improvement over sequential and parallel algorithms running on presently existing computer machines. Modern VLSI and WSI technologies make possible the user-oriented exploration of new computer architectures which provides the opportunity to freely make design tradeoffs without being 'locked in' to conventional styles or basing the design on the availability of standard parts or existing computer machines. Structured CAD tools permit the implementation of parallel CLP machines within relatively short engineering cycles with superior performance/ cost figures. 1 1 Interestingly enough, for example, the search space prune engine (the key component) can be implemented in about n man-hours (n is problem size) using PPL [66]; it takes as little as a |
