(Yilin Fang, Steve Yabusaki, Bruce Palmer) Enlarged View
Advanced Scalability for STOMP: Subsurface Simulation and Characterization at Extreme Resolution
Lead Investigator: Dr. Steve Yabusaki
Researchers have developed a highly scalable, next-generation version of PNNL's subsurface simulator, STOMP. The new implementation is able to deliver performance and still allow those familiar with the serial STOMP code to easily modify and enhance code capabilities without learning new languages, data structures, or programming styles. The project focused on the re-engineering of STOMP operational modes, which addressed Department of Energy missions in environmental restoration of nuclear production legacy waste sites, geologic sequestration of CO2 and the development of alternative energy resources.
Comparison between conceptual model (top) and a two-dimensional slice through the Ritzi numerical model (bottom). A 30x1 m section through one unit bar deposit is shown (cm resolution). Voxel colors indicate sedimentary architecture types
(e.g., facies) Enlarged View
- The STOMP routines, excluding third-party libraries, have exhibited linear speedup for large processor core counts.
- The new code, re-engineered with PNNL's GA Toolkit and a new grid component, has demonstrated very good scalability.
- Led to eSTOMP scaling studies using up to 131,040 processors.