GPU-Based In Situ Visualization for Large-Scale Discrete Element Simulations

Abstract

The result of discrete element simulations has tens of thousands of time frames and hundreds of billions of interacting particles in each frame, which may lead to very large data sets. The existing postprocessing visualization technology cannot deal with such large-scale data, which will bring a negative impact to the simulation, affect the simulation speed, and make it difficult to realize the runtime steering. In this work, an attempt has been made to optimize visualization methods by directly accessing DEM simulation results, data exchanging straightly between GPUs, and accelerating pixel composition by GPU. The key feature of this work is that the entire process from DEM simulation to graphical rendering and image composition is done inside the same GPU.