Efficient Parallel Implementation of Coherent Stacking Algorithms in Seismic Data Processing

Abstract

We discuss efficient parallel implementation of coherent stacking algorithms which form basis for a class of seismic processing procedures. In detail we address the problem of processing data of microseismic monitoring for localizing seismic events in space and time. Continuous data recording by seismic array quickly generates terabytes of data to be processed in a timely manner, including real-time analysis in some cases. Thus processing requires efficient parallel implementation with a special attention to data partitioning between nodes, and using computations to mask data reading from disk. Efforts were taken to minimize cache misses and vectorize loops. Sequential version of the code demonstrates 8x speed up compared to a naive implementation of the algorithm; parallel code scales almost linearly.