Parallel Finite Volume Algorithm on Graphic Processing Units (GPU)
Mechanical and industrial Engineering Department, University of Toronto, Canada
22nd Annual Conference of the CFD Society of Canada, 2014
Capabilities of using Graphic Processing Units (GPU) as a computational tool in CFD have been investigated here. Several solvers for solving linear matrix equations have been benchmarked on GPU and is shown that Gauss-Seidle gives the best performance for the GPU architecture. Compared to CPU on a case of lid-driven cavity flow, speedups of up to 4 times were obtained by using GPU.
October 27, 2014 by hgpu
