Parallel Integration Using OpenMP and GPU to Solve Engineering Problems

Abstract

The paper presents the results of a comparative analysis of parallel integration methods to solve engineering problems. We examine a problem of velocity and displacement data reconstruction from measured acceleration signal. Many engineering tasks such as vibration data acquisition from underground traffic, dwellings vibrations due to wind pressure, earthquake data acquisition, etc. involve field measurements that are mostly performed using accelerometers. For engineering analysis purposes these records are to be integrated to get velocity and displacement. Since some of the measurements are quite long, moreover number of acquisition channels often exceeds dozen, we get a large amount of data to be analyzed fast. The best solution for the problem is parallelization of the integration algorithm, which involves using some modern techniques like OpenMP or GPU computing. This article shows the performance acceleration on different processors for OpenMP technology as well as performance acceleration for different GPU supported devices. Some aspects of algorithm implementations are discussed and future work suggested.