Stream Computation of Shallow Water Equation Solver for FPGA-based 1D Tsunami Simulation

MOST (Method Of Splitting Tsunami) is widely used to solve shallow water equations (SWEs) for forecasting tsunami generated by an earthquake. Toward development of a power-efficient and high-performance computing system for 2D tsunami simulation, we conduct feasibility study on stream computation of 1D SWE solver with FPGA.We analyze an original code and design a stream algorithm with techniques of kernel fusion, shift buffering for streamed stencil-data access, and cascading processing elements for a longer pipeline. We implement a deep pipeline with at most 744 stages of 4 SPEs on 28 nm Stratix V FPGA, which achieves 82.4 GFlop/s at 200 MHz.