Abstract | ||
---|---|---|
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. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1007/978-3-319-49583-5_34 | ALGORITHMS AND ARCHITECTURES FOR PARALLEL PROCESSING, ICA3PP 2016 |
Keywords | Field | DocType |
Optimization,Computation locality,Stencil | Kernel (linear algebra),Locality,Locality of reference,Computer science,Stencil,Parallel computing,SIMD,Stencil code,Compiler,Computation | Conference |
Volume | ISSN | Citations |
10048 | 0302-9743 | 1 |
PageRank | References | Authors |
0.35 | 7 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Lufeng Yuan | 1 | 2 | 0.71 |
Junhong Liu | 2 | 18 | 1.94 |
Yulong Luo | 3 | 18 | 1.63 |
Guangming Tan | 4 | 436 | 48.90 |