Floating point number operation are not guaranteed to hold correct results. A certain range of errors can occur silently (overflow, underflow, imprecision, etc). This library aims to provide a replacement for floating-point number that warns on error.

The library is already in a relatively advanced state. The goal of this project is to bring it in a review-ready state. In this application, I suggest exploring a new design that hopefully should end up being more performant than the current one. I also want to implement the missing features and to complete tests and documentation.



Thomas Ferrand


  • Damian A Vicino