Optimization of parallel computation in MOOSE

by Vivek Vidyasagaran for International Neuroinformatics Coordinating Facility

The Multiscale Object-Oriented Simulation Environment (MOOSE) has recently implemented a new MPI-based parallelization scheme that supports automatic setup-time load balancing and transparent object access across nodes. This project aims to improve parallelization in MOOSE using CUDA in order for better performance on GPGPUs.