Global Memory Access Modelling for Efficient Implementation of the Lattice Boltzmann Method on Graphics Processing Units
Christian Obrecht (Centre de Thermique de Lyon (CETHIL))
Frédéric Kuznik (Centre de Thermique de Lyon (CETHIL))
Bernard Tourancheau (Laboratoire de l’Informatique du Parallélisme (LIP))
Jean-Jacques Roux (Centre de Thermique de Lyon (CETHIL))
In this work, we investigate the global memory access mechanism on recent GPUs. For the purpose of this study, we created specific benchmark programs, which allowed us to explore the scheduling of global memory transactions. Thus, we formulate a model capable of estimating the execution time for a large class of applications. Our main goal is to facilitate optimisation of regular data-parallel applications on GPUs. As an example, we finally describe our CUDA implementations of LBM flow solvers on which our model was able to estimate performance with less than 5% relative error.
Parallel and Distributed Computing, Multiscale and Multiphysics Problems, Performance Analysis