R2F2: a Run-time Reconfigurable Flexible Floating Point Multiplier

(paper) Exploring and Exploiting Runtime Reconfigurable Floating Point Precision in Scientific Computing: a Case Study for Solving PDEs

Cong Hao (Georgia Tech), ASP-DAC 2025

Files:

• templated_floating_point.cpp
  • Arbitrary precision floating point multiplier, whose decision is decided at compile time
  • Acronym in code: TFP
• flexible_floating_point.cpp
  • The proposed R2F2, run-time reconfigurable flexible floating point multiplier
  • Acronym in code: FFP
• tb.cpp
  • The testbench for validating the accuracy of TFP and FFP
• heat_equation.cpp + plot_heat_equation.ipynb
  • C++ implementation of 1D heat equation and its visualization
• shallow_water_equation_python.ipynb
  • Original python implementation of shallow water equation (SWE)
• shallow_water_equation.cpp + plot_shallow_water_equation.ipynb
  • C++ implementation of SWE and its visualization
• script.tcl
  • Synthesis the designs into FPGA implementation