Improving the Efficiency of the Preconditioning of Iterative Solutions to the Kinetic Equation
To achieve the reality of fusion, a greater understanding of plasma is required. The kinetic equation can be evolved simultaneously alongside the fluid equations to solve for kinetic closures. NIMROD performs this with numerical solvers where the General Minimum Residual (GMRES) solver becomes more efficient with a preconditioning matrix as input. Using a GPU-enabled library, the efficiency of GPU offloading to the preconditioning step was tested. A significant decrease in the factoring time of preconditioning matrix was observed. This suggests that the allocation of GPUs is worth investigating for NIMROD’s own benefit, but also anyone seeking to improve the efficiency of their scientific program.
Price, D. Caleb, "Improving the Efficiency of the Preconditioning of Iterative Solutions to the Kinetic Equation" (2022). Physics Capstone Projects. Paper 102.