In this paper the operation of a superscalar processor is studied. An analytical model is developed for computing the throughput and the speedup of the superscalar processors. The throughput is computed as a function of the reoccurrence period of the instructions in the program. It is shown that, when the number of the instructions is sufficiently large, the probability of referring to the different pipes and the number of pipes are more important than the length of the pipes.Also the results show that maximum throughput of a p-issues superscalar processor is equal to p instructions per cycle. This upper bound is reachable if and only if some conditions for reoccurrence period of the instructions in the program are fulfilled. At the end of the paper some remarks related to the structural designing and the number of pipes are offered.
/lp/association-for-computing-machinery/achieving-the-best-performance-on-superscalar-processors-0XdeU0uywU