From: Richard Biener <richard.guenther@gmail.com>
To: Jeff Law <law@redhat.com>
Cc: Giuliano Augusto Faulin Belinassi <giuliano.belinassi@usp.br>,
Wilco Dijkstra <Wilco.Dijkstra@arm.com>,
Segher Boessenkool <segher@kernel.crashing.org>,
GCC Patches <gcc-patches@gcc.gnu.org>,
giuliano.belinassi@gmail.com, nd <nd@arm.com>
Subject: Re: [PATCH] Add sinh(tanh(x)) and cosh(tanh(x)) rules
Date: Tue, 23 Oct 2018 09:52:00 -0000 [thread overview]
Message-ID: <CAFiYyc1kZKLRzAkM+daZer+ory-gJy1af7SFxnuj8qeMdmt4gA@mail.gmail.com> (raw)
In-Reply-To: <5b717a08-f994-0de8-5092-d11f5d9f8bd0@redhat.com>
On Mon, Oct 22, 2018 at 10:09 PM Jeff Law <law@redhat.com> wrote:
>
> On 10/20/18 9:47 AM, Giuliano Augusto Faulin Belinassi wrote:
> > So I did some further investigation comparing the ULP error.
> >
> > With the formula that Wilco Dijkstra provided, there are cases where
> > the substitution is super precise.
> > With floats:
> > with input : = 9.99999940395355224609375000000000000000000000000000e-01
> > sinh: before: = 2.89631005859375000000000000000000000000000000000000e+03
> > sinh: after : = 2.89630932617187500000000000000000000000000000000000e+03
> > sinh: mpfr : = 2.89630924626497842670468162463283783344599446025119e+03
> > ulp err befr: = 3
> > ulp err aftr: = 0
> >
> > With doubles:
> > with input : = 9.99999999999999888977697537484345957636833190917969e-01
> > sinh: before: = 6.71088640000000298023223876953125000000000000000000e+07
> > sinh: after : = 6.71088639999999925494194030761718750000000000000000e+07
> > sinh: mpfr : = 6.71088639999999944120645523071287770030292885894208e+07
> > ulp err befr: = 3
> > ulp err aftr: = 0
> >
> > *However*, there are cases where some error shows up. The biggest ULP
> > error that I could find was 2.
> >
> > With floats:
> > with input : = 9.99968349933624267578125000000000000000000000000000e-01
> > sinh: before: = 1.25686134338378906250000000000000000000000000000000e+02
> > sinh: after : = 1.25686149597167968750000000000000000000000000000000e+02
> > sinh: mpfr : = 1.25686137592274042266452526368087062890399889097864e+02
> > ulp err befr: = 0
> > ulp err aftr: = 2
> >
> > With doubles:
> > with input : = 9.99999999999463651256803586875321343541145324707031e-01
> > sinh: before: = 9.65520209507428342476487159729003906250000000000000e+05
> > sinh: after : = 9.65520209507428109645843505859375000000000000000000e+05
> > sinh: mpfr : = 9.65520209507428288553227922831618987450806468855883e+05
> > ulp err befr: = 0
> > ulp err aftr: = 2
> >
> > And with FMA we have the same results showed above. (super precise
> > cases, and maximum ULP error equal 2).
> >
> > So maybe update the patch with the following rules?
> > * If FMA is available, then compute 1 - x*x with it.
> > * If FMA is not available, then do the dijkstra substitution when |x| > 0.5.
> So I think the runtime math libraries shoot for .5 ULP (yes, they don't
> always make it, but that's their goal). We should probably have the
> same goal. Going from 0 to 2 ULPs would be considered bad.
But we do that everywhere (with -funsafe-math-optimizations or
-fassociative-math).
Richard.
> So ideally we'd have some way to distinguish between the cases where we
> actually improve things (such as in your example). I don't know if
> that's possible.
>
> jeff
next prev parent reply other threads:[~2018-10-23 9:17 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-10-18 23:22 Wilco Dijkstra
2018-10-19 13:39 ` Giuliano Augusto Faulin Belinassi
2018-10-19 13:46 ` Wilco Dijkstra
2018-10-19 14:49 ` Jakub Jelinek
2018-10-19 15:35 ` Wilco Dijkstra
2018-10-19 14:03 ` Segher Boessenkool
2018-10-19 15:10 ` Giuliano Augusto Faulin Belinassi
2018-10-19 17:28 ` Wilco Dijkstra
2018-10-20 19:06 ` Giuliano Augusto Faulin Belinassi
2018-10-22 20:58 ` Jeff Law
2018-10-23 9:52 ` Richard Biener [this message]
2018-10-23 11:31 ` Wilco Dijkstra
2018-10-23 11:35 ` Jakub Jelinek
2018-10-23 12:02 ` Wilco Dijkstra
2018-11-07 22:21 ` Jeff Law
2018-11-07 22:34 ` Wilco Dijkstra
2018-11-08 1:44 ` Segher Boessenkool
2018-11-08 13:33 ` Wilco Dijkstra
2018-11-09 20:04 ` Jeff Law
2018-11-09 22:04 ` Giuliano Augusto Faulin Belinassi
2018-11-10 5:36 ` Segher Boessenkool
2018-11-12 14:17 ` Richard Biener
2018-12-02 14:37 ` Segher Boessenkool
-- strict thread matches above, loose matches on Subject: below --
2018-08-07 20:00 Giuliano Augusto Faulin Belinassi
2018-08-07 20:29 ` Paul Koning
2018-08-07 20:42 ` Giuliano Augusto Faulin Belinassi
2018-08-08 13:59 ` Paul Koning
2018-08-08 18:57 ` Giuliano Augusto Faulin Belinassi
2018-08-08 19:55 ` Paul Koning
2018-10-12 4:55 ` Jeff Law
2018-10-12 14:55 ` Giuliano Augusto Faulin Belinassi
2018-10-16 21:16 ` Jeff Law
2018-10-17 21:43 ` Giuliano Augusto Faulin Belinassi
2018-10-17 22:45 ` Jeff Law
2018-10-18 5:38 ` Giuliano Augusto Faulin Belinassi
2018-10-18 17:10 ` Jeff Law
2018-10-18 22:46 ` Giuliano Belinassi
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=CAFiYyc1kZKLRzAkM+daZer+ory-gJy1af7SFxnuj8qeMdmt4gA@mail.gmail.com \
--to=richard.guenther@gmail.com \
--cc=Wilco.Dijkstra@arm.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=giuliano.belinassi@gmail.com \
--cc=giuliano.belinassi@usp.br \
--cc=law@redhat.com \
--cc=nd@arm.com \
--cc=segher@kernel.crashing.org \
/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).