Fluid instability plays a fundamental role in the research of astrophysics, energy power, chemical industry and new materials. The Smoothed Particle Hydrodynamics (SPH) method is a useful tool for simulating interfacial flows such as multiphase flow, high-velocity impact, explosion phenomenon. However, SPH method harnesses an enormous amount of particles for accuracy, which consumes a lot of computing power. In this paper, we present a massively parallel SPH scheme on the new Sunway supercomputer, SWSPH. In order to take full advantage of large-scale heterogeneous many-core computing system, we propose a series of parallel strategies and optimization methods. Experiments show that SWSPH has the capability of handling hundred-billion-particles simulations of fluid instability phenomenon on 39 million cores with a performance of 76% parallel efficiency.
