Exercise 1.26

The square procedure takes a variable already computed and square this value. Using a multiplication instead of the square procedure tells the program to compute two times the expmod value. Thus we loose the advantage of divising by two the steps needed at each recusrsive call to expmod, transforming our $$\Theta(\log{n})$$ process in a $$\Theta(n)$$ process.