The Application of Simplified Strassen Algorithm to Snow Simulation with Material Point Method

Abstract

As a beautiful yet complicated material, snow and its simulations are significantly challenging in the computer graphics community. Since the introduction of the material point method, abbreviated as MPM, the performance of snow simulations has been augmented in many aspects, including accuracy, stability, and efficiency. Nowadays, researchers work on developing simpler mathematical models or incorporating better hardware design to accelerate the snow simulations, but the overall running speed remains a problem. Instead of improving on the simulation methods, this paper presents a novel approach to optimize the efficiency of snow simulations by applying the simplified Strassen algorithm, an abridgment of a famous fast matrix multiplication algorithm, to snow simulations implemented with MPM. Rather than straightforwardly showing the triumph of the simplified Strassen algorithm over the standard matrix multiplication algorithm in runtime, the results present the evidence that leads to the inference that for simulations with a certain level of disorder, the simplified Strassen algorithm will surpass the standard matrix multiplication algorithm as the simulations progress.

Mingrui Wang

CS PhD Student

My research interests include computer graphics, physics-based simulation and deep Learning.