Acceleration of Turbomachinery Steady Simulations on GPU
Автор: von Karman Institute for Fluid Dynamics
Загружено: 2016-05-31
Просмотров: 1541
Описание:
Presentation by Mohamed Hassanine Aissa
PhD student at the von Karman Institute for Fluid Dynamics
Steady state CFD simulations, which rely on implicit time integration, are not experiencing great accelerations
on GPUs. Moreover, most of the reported acceleration effort concerns solving the linear system of equations
while neglecting the acceleration potential of running the system assembly also on the GPU. In this paper, we
present the software implementation of an implicit RANS CFD solver, which is fully running on GPU.We use the
flexible GMRES implementation of the Paralution package for the linear system solve along with the incomplete
LU factorization for the preconditioning. We propose also a control mechanism capable of accelerating the system
solve without altering the flow convergence by reducing the number of times an incomplete LU factorization is
performed. Speedups of 9.2x compared to a 1xcore CPU and 3.5x compared to a 4xcores CPU were measured for
3D flow predictions in turbine applications.
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
