Dual Logic Based Techniques For Area And Power Optimization

With the rapid development of integrated circuit technology and process, power, area, speedand verification has already been four main challenges that IC technique facing. During IC design,logic level plays an important role that joining RTL level and physical level and its EDA tools arethe most mature. Currently, almost all of the IC design methods are developed from theAND/OR/NOT operator based traditional boolean(TB) logic. But in fact, logic functions can alsobe implemented by the AND/XOR based Reed-Muller(RM) logic. Both TB logic and RM logicare complete set and have their own superiority. Studies have shown that most of the digitalsystems can get the greatest performance optimization using AND/OR/NOT/XOR operator baseddual logic to implement. This subject is a study of dual logic based techniques for area and poweroptimization from thi point. Before logic synthesis, logic detection and partition should be done todivide the logic function into two parts, which is suitable to TB logic and RM logic implementationrespectively. As TB logic is very mature, the area and power optimization of RM logic is studiednext. The main contents of this paper consist of three parts shown as follows:1. First of all, the disjointed cubes generation algorithm is proposed to make the products ofthe logic function be disjointed. Combining the existing logic detection and partition methods andthe characteristics of XOR logic, two rules that is suitable to RM logic implementation is proposed.Through the ruels, logic function is divided into two parts.2. As disjointed cubes have the characteristic that “OR” operation is equivalent to “XOR”operation, which is the theoretical basis of tabular techniques, a disjoint cubes based tabulartechnique for mix polarity conversion is proposed. The proposed algorithm avoids transforminglogic function into minterms, the number of products during conversion and time complexity isgreatly reduced compared to that of tranditional tabular technique.3. Analysis shows that the relationship between the signal probability and switching activity isuncertain. Combining with the Boolean difference operation and the definition of AND and XORlogic, we get the power estimation model of AND and XOR. At last the optimized power is obtained through polarity conversion using proposed low power decomposition method.