Path: csiph.com!eternal-september.org!feeder.eternal-september.org!mx02.eternal-september.org!.POSTED!not-for-mail
From: Keith Thompson <kst-u@mib.org>
Newsgroups: comp.lang.c
Subject: Re: Learning functions; roots of the quadratic; two return values
Date: Wed, 07 Oct 2015 16:23:28 -0700
Organization: None to speak of
Lines: 26
Message-ID: <lnwpuyi7pr.fsf@kst-u.example.com>
References: <cb70be98-6d3b-4814-bd1d-37289fa8d0b7@googlegroups.com> <623fef72-989a-4973-9d04-b1a867a00703@googlegroups.com> <pnr21b1287fff5arfericujjih1vk7to5b@4ax.com> <5611ccfd$0$3060$426a34cc@news.free.fr> <mut2g2$6m1$1@speranza.aioe.org> <MPG.307c4d1b6913286e74@news.eternal-september.org> <muutbr$mdg$1@speranza.aioe.org> <5613071A.2070609@verizon.net> <muvg2s$ncq$1@speranza.aioe.org> <mv0n91$v9l$1@dont-email.me> <mv1ksu$q4n$1@speranza.aioe.org> <20151007051930.3973122b@maxa-pc> <lnmvvuk68e.fsf@kst-u.example.com> <20151008005326.47e5ac1f@maxa-pc>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Injection-Info: mx02.eternal-september.org; posting-host="945944de09706c9b4e29b53c9d2efdc2"; logging-data="31158"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX18UBs/x9ZJoah3MnBILLY9d"
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.1 (gnu/linux)
Cancel-Lock: sha1:gTfHXV8dkQH96wgwvvzWHn5hbDA= sha1:k1bREvEKmY6kirSn1+ATZJ5/cHQ=
Xref: csiph.com comp.lang.c:72163

Melzzzzz <mel@zzzzz.com> writes:
> On Wed, 07 Oct 2015 09:12:33 -0700
> Keith Thompson <kst-u@mib.org> wrote:
[...]
>> The behavior of the sqrt() function is defined by the standard.  A
>> compiler can take advantage of the fact that calling sqrt() twice with
>> the same argument will yield the same result both times, and will not
>> have side effects.
>
> Yes, I see that compiler (gcc) takes advantage of knowing contents of
> sqrt, but there is some weird constraint that make it generate calls
> for several sqrts anyway. That is it does not calls function first,
> rather it executes sqrt instruction directly, but then it checks
> result with itself (ucomisd) and if parity flag is set (unordered) it
> calls sqrt then checks result again and if parity calls sqrt again.
> This code is not present if -ffast-math or -Ofast flag is present.

I have no idea.  The standard requires a call to the standard sqrt()
function to behave in a certain way.  A compiler is free to do whatever
it likes to achieve that effect.

-- 
Keith Thompson (The_Other_Keith) kst-u@mib.org  <http://www.ghoti.net/~kst>
Working, but not speaking, for JetHead Development, Inc.
"We must do something.  This is something.  Therefore, we must do this."
    -- Antony Jay and Jonathan Lynn, "Yes Minister"