Optimal group distribution in carry-skip adders

The carry-skip adder, because of its greater topological regularity and layout simplicity, is considered a good compromise in terms of area and performance. Some general rules have been suggested for its design, but they tend to overlook many important implementation details and cannot be applied to carry-skip adders with more than two levels of carry-skip or with different delays in the carry paths. The result is a nonoptimal distribution of groups and subgroups of the carry-skip circuits, degrading the worst-case delay of the adder. A new algorithm for determining the optimal distribution with no restriction on the number of skip levels is presented. Some results and conclusions are presented regularly in the realization of such an adder in bipolar ECL technology.<<ETX>>