Fluid simulation algorithms are typically accelerated via Graphics Processing Unit(s), or GPUs, since most of these algorithms are extremely data parallel. However, GPUs may still have comparatively higher latency than FPGAs when it comes to highly irregular computational tasks. This weakness is especially apparent in Fixed-Radius-Near-Neighbors (FRNN) searches. Our project aims to use an FPGA to accelerate the fluid simulation algorithm detailed in a recent paper, “Position Based Fluids” by Miles Macklin and Matthias Muller, which implements FRNN searches. The project will consist of an interface where users can customize a scene for input, and will perform simulation and generate a video filer for output. We will have everything running on an Ultra96 FPGA board.
Team E1: FPGA Accelerated Fluid Simulation
Carnegie Mellon University Capstone, Spring 2022 – Jeremy Dropkin, Alice Lai, Ziyi Zuo