From: Paul Zimmermann <Paul.Zimmermann@inria.fr>
To: Carlos O'Donell <carlos@redhat.com>
Cc: libc-alpha@sourceware.org, sibid@uvic.ca
Subject: Re: correctly rounded mathematical functions
Date: Thu, 13 Jan 2022 06:00:48 +0100 [thread overview]
Message-ID: <mwfspsxm9r.fsf@tomate.loria.fr> (raw)
In-Reply-To: <f19641ca-b47f-3e97-3c23-ef1b2d9b6d6d@redhat.com> (message from Carlos O'Donell on Wed, 12 Jan 2022 17:12:40 -0500)
Dear Carlos,
> For a very long time we've considered that there are really 3 sets of users:
>
> (a) Loosely correct answers provided quickly (high ULPs errors allowed, >9 ULPs)
> (b) Correct answers (no errors >9 ULPs)
> (c) Correctly rounded answers (0 ULPs)
>
> Note: Rounding modes matter.
I would say that rounding modes only matter for (c), since for (a) and (b)
most users will use the default rounding-to-nearest-ties-to-even mode, and
if not, another rounding mode will maybe add a few ulps to the error, which
should not matter for them.
However for (c), users who want to perform interval arithmetic for example
will want correct rounding for all rounding modes.
Progress of the CORE-MATH project can be followed on
https://homepages.loria.fr/PZimmermann/CORE-MATH/.
Take a look at the tanf code and the associated performance graph:
our correctly rounded code is up to 3 times faster than GNU libc.
(and also outperforms the Intel Math Library).
We plan to release our code under the MIT license, I believe this is
compatible for inclusion into GNU libc.
Best regards,
Paul
prev parent reply other threads:[~2022-01-13 5:00 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-01-03 12:48 Paul Zimmermann
2022-01-04 19:15 ` Joseph Myers
2022-01-05 16:03 ` Paul Zimmermann
2022-01-05 18:44 ` Joseph Myers
2022-01-06 9:55 ` Paul Zimmermann
2022-01-12 22:17 ` Carlos O'Donell
2022-01-12 22:12 ` Carlos O'Donell
2022-01-13 5:00 ` Paul Zimmermann [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=mwfspsxm9r.fsf@tomate.loria.fr \
--to=paul.zimmermann@inria.fr \
--cc=carlos@redhat.com \
--cc=libc-alpha@sourceware.org \
--cc=sibid@uvic.ca \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).