Thanks Jeffrey. Starred and followed. For those following from a distance, ArbFloats (IMO) are a *much* faster alternative to "fixed-precision Unums 1.0", and could nicely form the basis of interval arithmetic with open/closed endpoints, paving/uboxing algorithms, and more.
On Sat, Jun 25, 2016 at 4:29 AM, Jeffrey Sarnoff <[email protected]> wrote: > If the first ArbFloats.jl was proof of concept, this one is concept of > proof. > > This is a pre-announcement in response to some requests for access and my > preference. > > These quantities are information-richest floating point representations of > using arithmetic, elementary and some special functions. > ArbFloats presents digits that neither lead nor mislead the perception of > value, showing more or fewer digits would be less informative. > > When announced (ANN), I will offer more explanation and examples and any > overlooked details, > no worries -- if you do not need this now, wait for me to have fuller > tests written, etc. > . > If you do need it or just want to play with it, please do and share your > experience with me. > > https://github.com/jsarnoff-juliacon/ArbFloats.jl > >
