A novel greedy approach to harmonic summing using GPUs

One-dimensional incoherent harmonic summing is a technique used to improve the sensitivity of Fourier domain search methods. In time-domain radio astronomy, the harmonic sum is a part of the Fourier domain periodicity search, which aims to detect isolated single pulsars. We present a new harmonic sum algorithm based on the greedy approach and an implementation of this on NVIDIA GPUs using the CUDA programming language. The new algorithm determines which samples to add according to short-term gains. We present an evaluation of the sensitivity of the new approach and its performance against the PRESTO harmonic sum a standard used in radio astronomy. The new Greedy harmonic sum has, on average, 20% higher sensitivity whilst it has the same performance as the standard algorithm. This work forms part of the AstroAccelerate project which is a GPU accelerated software package for processing time-domain radio astronomy data