Full text loading...
-
Accelerating seismic parameter estimation with Adaptive Differential Evolution (JADE) and graphics processing units (GPUs)
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, Fourth EAGE Workshop on High Performance Computing for Upstream 2019, Oct 2019, Volume 2019, p.1 - 5
Abstract
Through the last decades, multiparametric traveltime has been used as an efficient technique for inversion and seismic imaging. Despite its efficiency, the required multiparameter estimations present itself as a challenge, since additional computational costs must be added. To mitigate this problem meta-heuristics, such as Adaptive Differential Evolution (JADE), can be applied. Despite the fast convergence of JADE, the runtime execution of parameter estimation can be suboptimal in many cases, mainly when the intrinsic embarrassingly parallel characteristic of the problem is not fully explored. In this paper, we propose a parallel implementation of parameter estimation with JADE in both GPU (through CUDA) and CPU (through OpenMP). We resort to cloud computing, specifically Amazon Web Service, to validate our implementation; executing our code on five different instances types. Experimental results show a considerable speedup obtained using GPU instances instead of CPU ones. Furthermore, qualitative results reveal similarity between both implementations. Thus, without any loss of quality, we are able to obtain an immense gain in execution time and a reduction of cloud computing costs.