From: Szabolcs Nagy <nsz@port70.net>
To: musl@lists.openwall.com
Subject: Re: remquo - underlying logic
Date: Fri, 8 Dec 2017 01:42:03 +0100 [thread overview]
Message-ID: <20171208004203.GF15263@port70.net> (raw)
In-Reply-To: <alpine.LRH.2.02.1712071157010.3671@key0.esi.com.au>
* Damian McGuckin <damianm@esi.com.au> [2017-12-07 12:09:23 +1100]:
> That said, over that range, I am experimenting using a simplistic form of
> double-double arithmetic for that calculation.
>
> Would you agree that when
>
> (int) (x / y) < 2^52
>
> the computation (int) (x / y) is accurate to within epsilon, i.e. if it
> should be at most be incorrect by +/- 1.?
>
this part is not the issue:
if the exact mathematical (int)(x/y) is k, i.e.
k <= x/y < k+1
then the rounded double prec x/y is
k <= x/y <= k+1
because close to k+1 some values get rounded up,
so sometimes you would compute x-(k+1)*y instead
of x-k*y, but this is easy to correct: just add +y
in the end if the result is out of range.
(same is true when other round-to-int method is
used instead of trunc, assumes k and k+1 are
representable)
> If so, and using the same sort of logic that log.c uses to split the
> calculation of
>
> k * log(2.0)
>
> into a high and low component, or maybe into 4 components, would you agree
> that it is possible to come up with an accurate computation of
>
> x - y * (int) (x / y)
>
> It should be much quicker than long division.
the real problem is doing x-k*y exactly (it is
representable as double).
when evaluated without fma, there is a rounding
error on the mul (the sub is exact).
one possibility is if k < 2^26 and the bottom 26
bits of y are zeroed out in yz then
x - k*yz - k*(y-yz)
is exact, another way is to use veltkamp/dekker
exact multiplication algorithm.
prev parent reply other threads:[~2017-12-08 0:42 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-30 18:11 Damian McGuckin
2017-11-30 18:59 ` Szabolcs Nagy
2017-11-30 20:29 ` Damian McGuckin
2017-11-30 21:16 ` John Reiser
2017-11-30 21:17 ` Szabolcs Nagy
2017-12-04 8:10 ` Damian McGuckin
2017-12-06 1:17 ` Damian McGuckin
2017-12-06 10:37 ` Szabolcs Nagy
2017-12-07 1:09 ` Damian McGuckin
2017-12-08 0:42 ` Szabolcs Nagy [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=20171208004203.GF15263@port70.net \
--to=nsz@port70.net \
--cc=musl@lists.openwall.com \
/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.
Code repositories for project(s) associated with this public inbox
https://git.vuxu.org/mirror/musl/
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 NNTP newsgroup(s).