Fast and Adaptive List Intersections on the GPU

• Published in: 2018 IEEE High Performance extreme Computing Conference (HPEC) pp. 1 - 7
• Main Authors: Fox, James, Green, Oded, Gabert, Kasimir, An, Xiaojing, Bader, David A.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2018
• Subjects: Estimation; Graphics processing units; Instruction sets; Kernel; Partitioning algorithms; Prediction algorithms; Yarn
• DOI: 10.1109/HPEC.2018.8547759

List intersections are ubiquitous and can be found in a wide range of applications, including triangle counting and finding the maximal k-truss, both of which are part of the HPEC Static Graph Challenge. For many graph based problems it is necessary to find intersections for a very large number of lists-these lists tend to vary greatly in size and are difficult to efficiently load-balance. Numerous parallel algorithms on list intersections for triangle counting have been proposed, but load-balancing decisions are typically made at a global level. In this paper we present an efficient and adaptive approach to load-balancing at a finer granularity. Our approach assigns a different number of threads for different intersections in order to effectively utilize the resources of the GPU. We show the applicability of our load-balancing method to two different intersection methods, one search-based and one merge-based. Our algorithm outperforms several recent triangle counting algorithms, including recent HPEC Graph Challenge Champions.