The authors describe a practical technique for improving the performan
ce of square-and-multiply exponentiation. A family of linear time algo
rithms, denoted by SS(l) where l determines the maximum length of prec
omputed exponents, is presented. Analysis on n-bit exponents shows tha
t the average number of multiplications required tends to nl(l + 1) fo
r large n.