PiFie: A Platform-independent Fuzzy Instruction Set Extension

One of the main obstacles in using the modern fuzzy control theory in engineering applications is its very high computational cost with respect to common traditional approaches. This paper focuses on improving the execution time of software implementation of different fuzzy applications on various processors. A platform-independent fuzzy instruction set extension (PiFie) is introduced in this paper. It contains sixteen different instructions which can significantly accelerate the run times of nearly all fuzzy as well as some non-fuzzy applications. These instructions do not demand specific hardware blocks and so can be augmented into almost any existing processor at a low cost. Also, we show that adding these instructions to an existing datapath does not cause any change in system critical path and so, the clock frequency will not be affected significantly. To show the efficiency of the PiFie, the instruction counts and also cycle counts of executing different fuzzy operations are computed both for a common instruction set and also for an extended instruction set which includes the PiFie