Eventos Anais de eventos
ENCIT 2022
19th Brazilian Congress of Thermal Sciences and Engineering
A GPU ACCELERATED ALGORITHM FOR SOLVING NAVIER-STOKES EQUATIONS
Submission Author:
Daniel Botezelli , SP
Co-Authors:
Daniel Botezelli, Elisan dos Santos Magalhães, Davi Antônio dos Santos
Presenter: Daniel Botezelli
doi://10.26678/ABCM.ENCIT2022.CIT22-0610
Abstract
Modern parallel programming technology has granted a recent speed enhancement in the system of equations solutions. The present paper proposes a CUDA-C-based numeric algorithm to analyze GPU devices' acceleration in solving thermo-fluid problems through the heat transfer and Navier-Stokes equations. The thermal study is approached through the enthalpy function of non-linear thermal conductivity. In order to study this algorithm's performance, this paper solves a two-dimensional lid-driven cavity benchmark with heat conduction in various Reynolds Number (Re), linear system of equations solvers, and mesh refinement. One of the objectives is to develop a high accelerated graphics processing units (GPU) based algorithm. The problems are numerically solved using the semi-implicit finite volume method in a GPU device. The results show that the parallel solver achieved 400-450 times faster than CPU-based codes under double variables precision.
Keywords
Parallel GPU Solver, Lid-driven cavity, thermal-fluid analysis
