SATPC0038370 Tab 04 4 SOW.pdf

This Statement of Work (SOW) details a research and development effort for GPGPU Acceleration for Linear Solvers and Modern Compute Architecture Enhancements in the Loci Framework, sponsored by NASA. The primary objective is to incorporate a GPU-enabled version of the Portable, Extensible Toolkit for Scientific Computation (PETSc) library into the Loci framework to investigate performance benefits for linear solver systems, with a focus on computational fluid dynamics (CFD) tools.

The SOW outlines three key tasks: 1) Enhance the Loci PETSc interface to offload linear system solving to GPGPU, testing two mapping strategies for CPU to GPGPU performance, 2) Fine-tune the Loci PETSc interface by addressing issues encountered during testing with various solvers, and 3) Explore optimizations for new GPGPU computing architectures like NVIDIA Grace-Hopper and latest ARM-based CPUs. Deliverables include an initial and optimized Loci PETSc GPGPU interface and a final report, with delivery dates specified relative to the award to proceed (ATP) date, ranging from 2 to 9 months after project initiation.