作者: G.A. Jullien , V.S. Dimitrov , B. Li , W.C. Miller , A. Lee
DOI: 10.1109/ISCAS.1999.777792
关键词: Algorithm 、 Numerical digit 、 Redundancy (engineering) 、 Computation 、 Arbitrary-precision arithmetic 、 Cellular neural network 、 Very-large-scale integration 、 Computer science 、 Reduction (complexity) 、 Real number 、 Artificial neural network 、 Representation (mathematics)
摘要: This paper introduces a modification to an index calculus representation for the double-base number system (DBNS). The DBNS uses bases 2 and 3; it is redundant (very sparse) has simple two-dimensional representation. An extremely sparse form of single non-zero digit represent any real with arbitrary precision. In this case can be identified by its coordinates (indices) in space. proposed paper, targeted DSP inner product computations, coefficient vector 2-digit data vector. We show that reduction over 80% hardware cost possible using hybrid compared original single-digit technique.