In this paper we propose a dense instruction set architecture which offers some advantages over existing architectures. We approach the problem from different points of view: Entropy measurements show high redundancy in instruction streams, especially when treated as higher order Markov sources; the memory fetch time for instructions has been a major bottleneck for many years, especially for multiprocessors with shared memory systems; ever larger programs are in use which require more and more memory; networks often wasting their time transfering redundant information, and loading times, and bus traffic increase with larger programs. We suggest a computer architecture which can eliminate the problems mentioned above using dense instruction sets.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados