Investigating Oil and Gas CSEM Application on Vector Architectures.

The importance of Controlled Source Electromagnetics (CSEM) has increased in the past decade. Along with this interest, its efficiency increased, data acquisition became easier and costs went down. For the Oil and Gas industry, modeling this data is necessary for exploration. The Modeling with Adaptively Refined Elements for 2D Electromagnetics (MARE2DEM), developed at Columbia University, is one of the tools used to model CSEM data. This paper will evaluate the performance observed during the investigation and implementation of the MARE2DEM's software using the vector architecture NEC SX-Aurora. MARE2DEM is a model for 2D electromagnetic geophysics, making it possible to model the presence of natural gas and petroleum on the depths of the ocean floor. Notably, we will show how the vector machine affects MARE2DEM's performance. It will be explained how the instrumentation of MARE2DEM works and the elaboration of the experiments to perform the investigation. Furthermore, it was necessary to elucidate the modifications done to the code source. The expected result is the runtime in seconds using different parallel decomposition settings. Lastly, we show results of this novel implementation with two workloads, a synthetic input and an input provided by Petrobras (oil & gas company), for which SX-Aurora provides performance improvements up to 27%.