Tom Breloff

Related? https://arxiv.org/abs/1604.07070v2 On Sunday, October 9, 2016, Alex Williams [email protected] wrote: > This paper has a maybe slightly better summary of it: > https://arxiv.org/pdf/1603.06160v2.pdf (see Algorithm 2). > > I...

Thanks for this @SimonDanisch. This is heavily related to what I try to do in Plots, which is creating compressed, equivalent definitions of generic visualization parameters. As an example, consider...

The extra bits don't actually add information... they summarize the contents. Many of the Inf/NaN/negative/0 exceptional states require semi-complex logic involving **all** of the bits of the number. The proposal...

Thinking about this type: ``` abstract AbstractFixedUnum

What do you think about: ``` typealias Unum{ESZ, UINT} FixedUnum{2, ESZ, UINT} typealias DecimalUnum{ESZ, UINT} FixedUnum{10, ESZ, UINT} typealias Unum16 Unum{3, UInt16} typealias Unum32 Unum{7, UInt32} typealias Unum64 Unum{16, UInt64}...

No those were thoughtless, arbitrary choices. I am open to better defaults. On Thu, Jul 30, 2015 at 1:54 PM, Scott P. Jones [email protected] wrote: > Are the 3, 7,...

A quick note... I'm going to remove the summary bits for now. I'm building it in a flexible enough way that luckily that's pretty easy.

John: can you expand on your "block floating point" comment? I don't quite understand the operation(s) that they are necessary for, but I certainly believe that I'm missing something. The...

This is really interesting, and can be a great extension to integers! I'll probably add a new set of: typealias IntUnum{NBITS} Unum{1,NBITS} Which represent the full range of integers (minus...

I could see allowing direct conversion to Integers when the IntUnum is exact, finite, (and positive for Unsigned), but requiring a call to round otherwise. (convert would throw an InexactError)...