Re: A book on floating-point

On Jun 7, 2018, at 2:43 AM, Dirk Laurie <dirk.laurie@gmail.com> wrote:

In view of the recurrent posts on floating-point issues, I pass on an
announcement from this week's NA Digest:

2nd edition of "Handbook of Floating-Point Arithmetic" (Birkhauser,
2018), Authors: Muller, Brunie, de Dinechin, Jeannerod, Joldes,
Lefevre, Melquiond, Revol, and Torres.

Thanks for your last post on floating-point issue.

Yes, unspecified rounding is worse than extended precision rounding.

I keep saying "just pick one", but the Windows example probably kill it.

One of the author you mentioned, Vincent Lefèvre, also wrote

an article on issues with extended precision rounding [1]

Lefèvre is also a main developer in MPFR, an arbitrary precision C library

that that do correct rounding in arbitrary bits. [2]

[1] https://www.vinc17.net/research/extended.en.html

[2] https://www.mpfr.org/credit.html