Efficient gather and scatter operations on graphics processors

Abstract

Gather and scatter are two fundamental data-parallel operations, where a large number of data items are read (gathered) from or are written (scattered) to given locations. In this paper, we study these two operations on graphics processing units (GPUs).