Selective nesting of circular convolution algorithms

The problem of the proper use of nesting techniques applied to circular convolution computation is highlighted. The well known method of transforming the convolution nesting problem into a problem of nesting polynomial products named split-nesting is considered. It is then shown that for receiving optimum results polynomial products should be nested only if an inequality named the rule of number 2 is satisfied. Examples show that the algorithms obtained are much better than the split-radix ones. Using the nesting technique for multidimensional data is also discussed