Fast Fluid Dynamics is a computationally cheap way to evolve stable fluids. The seminal paper in this topic was written by Jos Stam. In this project, the lid driven cavity problem which is a benchmark CFD problem is simulated using Fast Fluid Dynamics. The implementation has been written in Arrayfire a general purpose GPU library as well as Octave/Matlab. The code is a modified version of another code available on github. The code is completely vectorized and the difference can be seen in the execution time. Here's a gif obtained by evolving for a grid size of 200x200 cells using the octave/matlab implementation.

The values obtained using the code have been compared to verified values from the paper by Ghia, Ghia, Shin. The particular comparison of the velocity of the fluid in the x direction for a lid driven cavity can be seen here for a Reynolds number of 100 and 400. These trends show that the value is converging to the verified values as the grid size used for the same lid decreases.

The aim of the thesis which was to develop a code which implements Fast Fluid Dynamics and verifies the claim that FFD is indeed faster than CFD methods has been achieved. The code is completely vectorized and the improvements made on the available code is starkly visible. The pdf of the thesis is also uploaded, which details the work done, specifically the improvements made in speed of the simulation.

Thanks.