Optimizing stencil code via locality of computation

Stencil computation is a performance critical kernel used in scientific and engineering applications. We define a term of locality of computation to guide stencil optimization by either architecture or compiler. Being analogous to locality of reference, computational behavior is also classified into spatial locality and temporal locality. This paper develops equivalent computation elimination (ECE) approach in multi-level loop for exploiting temporal locality of computation. The strength of ECE lies on an intermediate-based searching algorithm to eliminate inter-iteration computational redundancies of all possible combination and a multiple dimensions replacement algorithm to replace redundant computation across loops of multiple dimensions. We implemented ECE in ROSE compiler infrastructure. The experiment shows that ECE improves performance by 20% on average due to the consciousness of temporal locality.