Class

Article

College

College of Science

Department

English Department

Faculty Mentor

J. Andrew Spencer

Presentation Type

Poster Presentation

Abstract

Solving the plasma kinetic equation provides higher-order moments to better approximate an accurate time evolution of plasma behavior contained in a tokamak. NIMROD seeks to produce code to simulate and model these macroscopic dynamics. Essential to this code’s efficiency is the preconditioning of matrices. I studied how compiling a GPU-enabled third-party library, SuperLU, would improve this preconditioning time for a test version of NIMROD's code. We found that the time of preconditioning decreased significantly when GPUs were allocated versus when only CPUs were, which provides understanding for the future utilization of GPUs by the NIMROD code directly.

Location

Logan, UT

Start Date

4-8-2022 12:00 AM

Included in

Physics Commons

Share

COinS
 
Apr 8th, 12:00 AM

GPU Accelerated Preconditioning of the Plasma Kinetic Equation

Logan, UT

Solving the plasma kinetic equation provides higher-order moments to better approximate an accurate time evolution of plasma behavior contained in a tokamak. NIMROD seeks to produce code to simulate and model these macroscopic dynamics. Essential to this code’s efficiency is the preconditioning of matrices. I studied how compiling a GPU-enabled third-party library, SuperLU, would improve this preconditioning time for a test version of NIMROD's code. We found that the time of preconditioning decreased significantly when GPUs were allocated versus when only CPUs were, which provides understanding for the future utilization of GPUs by the NIMROD code directly.