GPU Enhanced Smoothed Particle Hydrodynamics Simulation

Smoothed Particle Hydrodynamics is a powerful tool for simulating fluid dynamics. Moreover, it is easily parallelizable, as the interaction between two particles is independent of the others. However, with a large number of particles there would be a significant amount of computation involved for calculating the interaction between the particles. So its necessity of an algorithm that is suitable for such parallelization using GPUs. An analysis of the implementation of smoothed particle hydrodynamics (SPH) simulation in a parallelized manner is presented here. In normal implementation, there is very much data transfer overhead. It is because, after performing physical computation every time it copies all data to the main memory for displaying them. But constantly sending this data makes this computation extremely slow. So to overcome this problem, a parallel implementation of SPH simulation using shared memory is used in proposed work. Speaking simply, proposed work performs all the physical computations at GPU and updated data is sent to the main memory only when it needs to be displayed. This will help to minimize the CPU-GPU data transfer overhead and speeding-up the performance.