STTR: Utilizing GPUs as an Efficient Chemical Kinetics Co-processor

The goal of the Phase I effort is demonstrating viability of the approach to utilizing GPUs as an efficient "chemical kinetics co-processor" for improving simulation runtimes of missile plume simulations. Work in the several keys areas is highlighted below and will serve to accomplish three principal milestones: 1. Modify and successfully test the plume simulation software for thread-based programming and thermochemistry hand-off, necessary for efficient GPU I/O interaction; 2. Develop preliminary GPU-based algorithms for accurately and efficiently computing chemical kinetic source terms; and 3. Demonstrate chemical kinetics coprocessor approach transferring relevant data between CPU and GPU using a simplified plume simulation scenario. The product delivered at the end of Phase I will consist of a technical report describing work accomplished and findings in each of the areas discussed below. During Phase II, implementation of more detailed chemistry models, performance benefit using real plume simulations of current interest, improvements to GPU coprocessor utilization, evaluation of streaming compilers to facilitate code porting, and extension of learned methodologies to other simulation bottlenecking areas are envisioned.