A theory of plausibility for computer architecture designs
U. Aguero · 1987
Every design should have certain merits in order to be considered worthy of being implemented. This requirement is especially critical when the design is to be physically implemented, and the implementation cost is too high to allow multiple iterations of the design-implementation cycle. Under such circumstances, the merits of a design need to be determined prior to implementation such that, hopefully, the implementation will only confirm the design's merits. Hence, the merits of a design determine its plausibility. In this dissertation, we present a general theory of plausibility for computer architecture designs. The theory is based on the premise that an architecture design should be viewed as a specification of constraints that must be satisfied by some (possibly abstract) implementation. The theory consists of principles that (a) characterize the nature of plausible constraints; (b) define a way in which plausibility statements can be developed to make claims about a design's merits or demerits; (c) define the main evolutionary properties of plausible designs; and (d) serve as reasoning guidelines for plausibility-driven design methods.