--> --> ABSTRACT: Meeting Exploration Challenges with High Performance Computing Applications, by Alhashim, Fadhel; #90141 (2012)

Datapages, Inc.Print this page

Meeting Exploration Challenges with High Performance Computing Applications

Alhashim, Fadhel *1
(1) Saudi Aramco, Dhahran, Saudi Arabia.

Successful and safe hydrocarbon drilling operations require creation of a detailed and accurate estimate of the earth’s physical properties before the well location is planned. Creating such a reservoir model typically involves the interpretation of seismic data to predict macro- and micro-scale rock properties and to guide the interpolation of reservoir properties between existing well locations.

Unfortunately, the resolution of seismic data is typically less than desired for accurate reservoir description, therefore significant effort is made to extract earth properties near the limit of seismic resolution. This resolution-pushing requirement drives our development of more accurate and efficient algorithms for improving the apparent resolution of seismic data. In addition to the safety issues, inaccurate earth models can cost oil companies tens or hundreds of millions of dollars to drill a single dry hole.

We start this paper with a brief review of common spatial interpolation algorithms that have historically been used for traditional image processing. Bi-cubic interpolation is found to have the most favorable features for crossover applications in seismic processing. Unfortunately bi-cubic interpolation is computationally inefficient when applied to large seismic data volumes.

New multi-core CPUs and GPUs allow scientists to take advantage of powerful, yet computationally expensive algorithms on individual user's desktops. Through the application of parallel programming techniques we adapt the bi-cubic interpolation algorithm using both a CPU-based shared memory architecture (OpenMP) and a graphics processing (GPU) architecture. Tradeoffs between these two approaches are discussed and performance comparisons are made. During this research we discovered that bi-cubic interpolation can be performed efficiently on GPU hardware, and there are bottlenecks, like memory bandwidth, that limit the application of GPU technology in the seismic domain.

 

AAPG Search and Discovery Article #90141©2012, GEO-2012, 10th Middle East Geosciences Conference and Exhibition, 4-7 March 2012, Manama, Bahrain