Alan Jorgensen – True North Floating Point
For years I have been concerned that floating point operations can yield undetectably incorrect results. I have patented a solution to that problem. The paper presented here describes the problems with floating point and testing floating point and describes how bounded floating point can be used, not just to provide reliable results, but provide a means of testing floating point applications.
Alan Jorgensen, 2020 Technical Presentation, Paper, Slides