Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!uwm.edu!spool.mu.edu!uunet!karln!karln!karln
From: karln@uunet.uu.net
Newsgroups: comp.lang.c
Subject: Re: Divide and C
Message-ID: <1991Mar29.163422.14672@uunet.uu.net>
Date: 29 Mar 91 16:34:22 GMT
References: <1991Mar27.185804.7221@uunet.uu.net> <639@taumet.com>
Reply-To: karln!karln@uunet.uu.net
Organization: Sam76 - Pennington NJ
Lines: 40

>
>If your compiler is not this smart and you know that the time spent in
>the extra division is making your program unacceptably slow, you can
>try the Standard C library function div(), which takes a numerator
>and denominator and returns a structure containing the quotient and
>remainder.  Use of this function might turn out to be slower than
>the extra divide.
>
>Steve Clamage, TauMetric Corp, steve@taumet.com

	This shows a good question someone pointed out
about the speed of an assembly divide plus a multiply
verses doing the real binary divide in C with subtracts and
shifts. Here is the posting that did not make into the
summary.


int twodivide(val,divisor,&result,&remain)
{
        int r;
        r=val/divisor;
        *result=r;
        *remain=val-divisor*r;
}
 
that's with a multiply (divisor*r) - but that's probably a lot faster 
than looping, subtracting divisor each time, adding 1 onto r, you know.

Russell Schulz             ersys!rschulz@nro.cs.athabascau.ca
Edmonton Remote Systems:  Serving Northern Alberta since 1982


  Do you suppose that is true? I do have that 'Binary Divide' code around 
somewhere. I think I will find it and do a benchmark. Stay tuned. 

	Karl Nicholas
	karln!karln@uunet.uu.net

PS. I wonder if the optimizer will recoqnize the Binary Divide
routine and just do an Assembly Divide :-)