StreamSim on massively parallel multi-core architecture

Three fold speed up reported with OpenMP-based code parallelization—bests cellular approach.

StreamSim Technologies and the Fraunhofer Institute have adapted StreamSim’s commercial reservoir fluid flow simulator to run on parallel multi-core architectures leveraging the OpenMP programming model. The work was presented at the Society of Petroleum Engineers Reservoir Simulation Symposium held in The Woodlands, Texas last month.

Parallelization was facilitated by the fact that the bulk of the serial program’s run time involved computing independent solutions for each streamline. Fraunhofer’s in-house software was used to identify ‘thread-safe’ variables that would migrate to the multi-core architecture with minimal rewrite of the existing code base. The test code was run on an AMD Opteron 8218-based machine with eight, 2.6 GHz dual core processors—for up to 16 OpenMP threads on RedHat Linux.

The parallel simulator was tested on several models including a Forties field model and a Middle East dual-porosity model. Speed up of between 2.5 to 3.5 fold was seen for 8-threads—reducing run time for these large models from around 12 hours to under 4 hours. Beyond 8-threads the serial portion of code became the limiting factor. The authors believe that the inherently parallelizable nature of streamline simulation makes it better suited to several modern reservoir evaluation workflows than traditional finite difference, cellular simulators. The full paper can be obtained from www.spe.org.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.