high performance computing on graphics processing units: hgpu.org

SpGEMM (General Sparse Matrix-Matrix Multiplication) has attracted much attention from researchers in fields of multigrid methods and graph analysis. Many optimization techniques have been developed for certain application fields and computing architecture over the decades. The objective of this paper is to provide a structured and comprehensive overview of the research on SpGEMM. Existing optimization techniques have been grouped into different categories based on their target problems and architectures. Covered topics include SpGEMM applications, size prediction of result matrix, matrix partitioning and load balancing, result accumulating, and target architecture-oriented optimization. The rationales of different algorithms in each category are analyzed, and a wide range of SpGEMM algorithms are summarized. This survey sufficiently reveals the latest progress and research status of SpGEMM optimization from 1977 to 2019. More specifically, an experimentally comparative study of existing implementations on CPU and GPU is presented. Based on our findings, we highlight future research directions and how future studies can leverage our findings to encourage better design and implementation.