SATPC0039290 TAB 04 4 SOW.pdf

This Statement of Work (SOW) details GPGPU Programming Support for Loci Applications, focusing on enhancing the Loci framework's parallel computing capabilities. The project aims to develop advanced GPGPU (General Purpose Graphics Processing Unit) compatibility for the Loci framework, with specific tasks including: developing GPGPU-compatible container types (storeVec, multiMap, MapVec, multiStore), implementing memory scheduler enhancements, creating thread block partitioning support, developing global and local reduction scheduling APIs, extending the Loci preprocessor, and creating a GPU/flowPsi technology demonstrator.

The key deliverables include implementing new Loci container interfaces, memory scheduler improvements, runtime scheduling system enhancements, global reduction scheduling APIs, local reduction scheduling with sub-domain decomposition, Loci preprocessor extensions to support GPGPU execution, and a technology demonstrator. The project is part of a NASA/NSSC sole-source contract with Loci Consulting Services LLC, with performance located at NASA Marshall Space Flight Center, aimed at advancing high-performance computational modeling capabilities for complex scientific simulations.