A stochastic roundoff error analysis for the convolution

Abstract

We study the accuracy of an algorithm which computes the convolution via Radix-2 fast Fourier transforms. Upper bounds are derived for the expected value and the variance of the accompanying linear forms in terms of the expected value and variance of the relative roundoff errors for the elementary operations of addition and multiplication. These results are compared with the corresponding ones for two algorithms computing the convolution directly, via Homer's sums and using cascade summation, respectively.