A summation algorithm with error correction for parallel computers

Abstract

This paper proposes an algorithm to accurately compute the sum of floating-point numbers on parallel computers. This algorithm is an extension of the wellknown recursive doubling technique which computes the sum of n floating-point number in log2n parallel steps. The time complexity of the present algorithm also is O(log n), and the space complexity is O(n… (More)
DOI: 10.1002/scj.4690240706

Topics

4 Figures and Tables

Slides referencing similar topics