At least Unums (1.0) and Arbs don't have that annoying asymmetry of IEEE 
floating point, you have exact zero, or inexact pos or neg "really small 
close to zero" in Unums,
or zero with a radius of 0, or a 0 with radius that can be *quite* small 
around zero (+/-), or something that can be as small as possible positive 
(but not zero), or small as possible negative (but not zero), with Arbs.
Nice and symmetric, which appeals to some of my OCD-like traits!

I've lately come to greatly appreciate Arbs compared to IEEE floats, thanks 
to Jeffrey Sarnoff's talk, and a number of people's work bringing Fredrijk 
Johansson's C Arb package https://github.com/fredrik-johansson/arb to Julia,
in WB Hart's et. al. Nemo.jl package https://github.com/wbhart/Nemo.jl, and 
Jeffrey's https://github.com/JuliaArbTypes/ArbFloats.jl.  

On Saturday, July 9, 2016 at 7:40:58 AM UTC-4, Tom Breloff wrote:
>
> Yes. They are different numbers. In a way, negative zero represents "a 
> really small negative number" that can't be represented exactly using 
> floating point.  

Reply via email to