Strategies to Improve the Performance of a Geophysics Model for Different Manycore Systems

Abstract

Abstract-Many software mechanisms for geophysics exploration in Oil & Gas industries are based on wave propagation simulation. To perform such simulations, state-of-art HPC architectures are employed, generating results faster and with more accuracy at each generation. The software must evolve to support the new features of each design to keep performance scaling. Furthermore, it is important to understand the impact of each change applied to the software, in order to improve the performance as most as possible. In this paper, we propose several optimization strategies for a wave propagation model for five architectures: Intel Haswell, Intel Knights Corner, Intel Knights Landing, NVIDIA Kepler and NVIDIA Maxwell. We focus on improving the cache memory usage, vectorization, and locality in the memory hierarchy. We analyze the hardware impact of the optimizations, providing insights of how each strategy can improve the performance. The results show that NVIDIA Maxwell improves over Intel Haswell, Intel Knights Corner, Intel Knights Landing and NVIDIA Kepler performance by up to 17.9x.

Optimization strategies for geophysics models on manycore systems

“…This article is an extension to our previous work (Serpa et al, 2017). In this previous research work, we applied several optimization techniques on a largely used seismic imaging simulator running on new architectures.…”

A Data-Centric Directive-Based Framework to Accelerate Out-of-Core Stencil Computation on a GPU

“…In this section, we use the acoustic wave propagator [1] as an example of a reputed stencil application that is widely used for geophysics exploration in the petroleum industry. Acoustic wave propagation is governed by the following equation:…”

“…In this way, we can discretize (1); that is, we can describe it in the form of a 25-point stencil computation ( Fig. 1) in which calculation of an element relies on 24 surrounding elements (i.e., halo regions).…”

“…Stencil computation is one of the most important classes in scientific computing with a key principle of iteratively updating an input array by applying a fixed calculation pattern (i.e., stencil) to all the elements of the array. Stencil applications appear in a wide range of fields, including geophysics simulations [1], [2], computational electromagnetics [3], and image processing [4], [5]. Because computation time and memory consumption grow linearly with the size of input arrays, parallel implementations of stencil computation are of great importance [6].…”

Programação Paralela em Memória Compartilhada e Avaliação de Desempenho com Contadores de Hardware

“…A versão final, que combina todas otimizações, teve o melhor desempenho, de 75,5⇥ em relação a versão sequencial. Essas e outras propostas de otimização para essa aplicação de geofísica podem ser vistas em diversos trabalhos (SERPA et al, 2017(SERPA et al, , 2018b(SERPA et al, , 2018a(SERPA et al, , 2019. 3.7.…”