System#: High-level synthesis of physical simulations for FPGA-based real-time execution

FPGA-accelerated real-time simulations of physical phenomena gain importance in the area of Hardware-in-the-Loop tests for electrified powertrain components. Especially the challenging timing requirements demand dedicated hardware solutions. We propose System#, which is a highly modular, flexible and extensible open source environment for synthesizing physical computations to FPGAs. We demonstrate its capabilities by using an existing model of a squirrel-cage induction machine. We oppose several design variants, using fixed point and floating point arithmetic, and various settings for schedule length and pipelining depth of arithmetic operators. We conclude that the synthesized designs are on the par with a manually optimized implementation. Moreover, our approach scales well by allowing a time/area tradeoff over a wide range.