Stencil computation is a performance critical kernel that is widely used in scientific and engineering applications. In this paper we develop a redundant computation elimination (RCE) algorithm to exploit temporal locality. We implement the RCE optimization strategy using ROSE compiler infrastructure. The experiments with a benchmark of eleven stencil applications show that temporal locality of RCE averagely improves performance by 15.4% and 10.1% for benchmark without or with SIMD optimization.
CITATION STYLE
Yuan, L., Liu, J., Luo, Y., & Tan, G. (2016). Locality of computation for stencil optimization. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 10048 LNCS, pp. 449–456). Springer Verlag. https://doi.org/10.1007/978-3-319-49583-5_34
Mendeley helps you to discover research relevant for your work.