期刊名称：Bulletin of the Technical Committee on Data Engineering
出版年度：2017
卷号：40
期号：3
页码：30
出版社：IEEE Computer Society
摘要：Developing parallel graph algorithms with correctness guarantees is nontrivial even for experiencedprogrammers. Is it possible to parallelize existing sequential graph algorithms, without recasting thealgorithms into a parallel model? Better yet, can the parallelization guarantee to converge at correctanswers as long as the sequential algorithms provided are correct? GRAPE tackles these questions, tomake parallel graph computations accessible to a large group of users. This paper presents (a) the parallelmodel of GRAPE, based on partial evaluation and incremental computation, and (b) a performancestudy, showing that GRAPE achieves performance comparable to the state-of-the-art systems.
