Highly-Scalable Computational-Based Engineering Algorithms for Emerging Parallel Machine Architectures
ABSTRACT: In order to significantly advance the fidelity and capability of current computational engineering tools, algorithms, techniques, and software which map effectively to emerging massively parallel hardware architectures must be developed. The objective of this project is to investigate and demonstrate specific techniques for enhancing the scalability of computational fluid dynamics (CFD), computational structural dynamics (CSD), and coupled CFD/CSD aeroelastic simulation tools. These include improved strategies for large-scale parallel partitioning, novel techniques for extracting additional parallelism from the time dimension for time dependent problems, and a flexible software development strategy that enables hybrid programming models, encapsulation of hardware specific tasks, and simple reconfiguration as new parallel architectures evolve. In addition to enhancing the scalability of individual engineering simulation disciplines such as CFD and CSD, novel approaches for ensuring optimal scalability of the fully coupled problem will also be considered. In the Phase 1 project, various key techniques for achieving increased scalability of standalone and coupled CFD/CSD simulations on diverse HPC hardware configurations were demonstrated at moderate core counts. In Phase 2, these techniques will be refined and productionalized and benchmarked on hardware configurations with over 50,000 cores. At the same time, additional techniques will be developed to further enhance the scalability of multidisciplinary simulations on massively parallel and heterogeneous hardware. BENEFIT: Although the ability to run effectively on massively parallel architectures as typified by large government installations is not currently an important consideration for most aerodynamic engineering departments, the explosive growth of parallelism means that even low and mid-range future computer hardware installations will contain many more cores than is currently the case. This project aims to meet this demand head on as it develops over the next few years and to build, in the process, a suite of highly scalable and flexible simulation tools which will enable commercial and government users to take full advantage of emerging hardware capabilities. At the same time, this project will result in the development of a high fidelity coupled aeroelastic simulation tool for use in the fixed wing and rotary wing aircraft communities. Finally, adoption of specific isolated software components developed within this project will be facilitated by conforming to a set of well defined software interfaces used within current DoD programs.
Small Business Information at Submission:
Research Institution Information:
Scientific Simulations LLC
1582 Inca Laramie, WY -
Number of Employees:
University of Wyoming
Research Office, Dept 3355
1000 E, University Ave.
Laramie, WY 82071-