Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!att!alberta!ubc-cs!fs1!coho.ee.ubc.ca!mikeb
From: mikeb@coho.ee.ubc.ca (Mike Bolotski)
Newsgroups: comp.lang.c
Subject: Re: Optimal for loop on the 68020.
Message-ID: <304@fs1.ee.ubc.ca>
Date: 7 Jun 89 01:27:44 GMT
References: <11993@well.UUCP> <5260014@hplsla.HP.COM>
Sender: news@fs1.ee.ubc.ca
Reply-To: mikeb@ee.ubc.ca (Mike Bolotski)
Organization: Dept. of Electrical Engineering University of B.C.
Lines: 39

In article <5260014@hplsla.HP.COM> jima@hplsla.HP.COM (Jim Adcock) writes:
>Below is your examples, redone with a null loop body, plus *my* favorite
>loop construct: for(i=COUNT; i--;).  

This is my favorite loop, too.  But if i is defined to be short, instead
of int, the following results (Sun3/260), gcc 1.35 -O -S.  

Note: no -fstrength_reduce required.

main()
{
  volatile int x;
  short j;

  for (j=5; j--; )
    x = 0;
}

#NO_APP
gcc_compiled.:
.text
        .even
.globl _main
_main:
        link a6,#-4
        moveq #5,d0
        jra L2
L5:
        clrl a6@(-4)
L2:
        dbra d0,L5
        unlk a6
        rts


Mike Bolotski, Department of Electrical Engineering,
               University of British Columbia, Vancouver, Canada 
mikeb@ee.ubc.ca                    | mikeb%ee.ubc.ca@relay.ubc.ca
ee.ubc.ca!mikeb@uunet.uu.net       | ...!ubc-vision!ee.ubc.ca!mikeb