From: ams AT ludd DOT luth DOT se (Martin Str|mberg)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: gcc optimizes divisons!
Date: 23 Sep 1999 19:37:54 GMT
Organization: University of Lulea, Sweden
Lines: 32
Message-ID: <7sdvii$ipb$1@news.luth.se>
References: <37e37095 AT news DOT xenologics DOT com> <7scukd$rm6$1 AT antares DOT lu DOT erisoft DOT se> <37ea6cda AT news DOT xenologics DOT com>
NNTP-Posting-Host: queeg.ludd.luth.se
X-Newsreader: TIN [UNIX 1.3 950824BETA PL0]
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Oliver Roese (oroese AT edina DOT xnc DOT com) wrote:
: 
: Martin Stromberg <eplmst AT lu DOT erisoft DOT se> schrieb in im Newsbeitrag:
: 7scukd$rm6$1 AT antares DOT lu DOT erisoft DOT se...
: (...)
: >
: > Well... Obviously(?) x/3 == 1431655766*x + (x<0?1:0) (mod 2^32)
: > Or perhaps (mod 2^32) should be (mod 2^31)...
: >
: > It's very probably mathematically provable using rings and fields and
: such.
: >
: 
: The above equation cannot hold, since x/3 is fractional most of the
: time.

You are missing the fact that we are operating on the ring Z(2^32), so
the mapping is Z(2^32) -> Z(2^32) on the division operation, where /3
is a special case.

: Even if this was a mistake and you meant floor(x/3), the equation cannot
: hold, since
: floor(x/n) is a staircase function.

floor() is something which operates on R or (in computers) floating
point. We are dealing with integers here (and in this case, a subset
of them, Z(2^32)).


Right,

							MartinS