This paper presents a paradigm of real-time processing on the lowest level of computing systems: the arithmetic unit. The arithmetic unit based on this principle containing addition, subtraction, multiplication and division operations is described. The development of the computation model is based on the Soft Computing and the Imprecise Computation paradigms, combined with the MSB-First and the Interval Arithmetic techniques. Those paradigms and techniques give the arithmetic unit design the ability to compute with precisions as a function of time available or accuracy needed. The predictability of processing time and result?s accuracy are obtained by means of processing granularity of k-bits and by using look-up tables. We present an evaluation of the operation in time delay and computation accuracy that shows significant performance improvement over conventional arithmetic unit architecture, that is, the ability to produce intermediate-result during execution time, to give certainty in computation accuracy even before the process finish time by providing two intermediate-results, which act as the lower and upper bound of the real and complete computation result, and finally, gain high computation accuracy from the early time of the execution process.
展开▼
