Aplicação de estêncils compactos e reordenamento de dados na modelagem de ondas acústicas

Abstract

This work presents the motivations, implementation details and tests of two acoustic wave modeling algorithms: the Reordered Acoustic Modeling and the Compact Reordered Acoustic Modeling. Seismic wave modeling is an utmost important process in several areas of geophysics and seismic imaging and processing. Due to temporal and spatial cache locality matters, these algorithms were developed to try to improve the performance of wave modeling in both CPUs and GPUs alike. They were implemented using Mamute as a base, which is a seismic imaging and processing software, and both showed performance improvements for larger problem when compared to other algorithms, justified by the reduction of the cache-miss rate.