This is the mail archive of the egcs@egcs.cygnus.com mailing list for the EGCS project. See the EGCS home page for more information.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Small slowdown...


  In message <19980522112932.39503@atrey.karlin.mff.cuni.cz>you write:
  > Hi
  > Just for fun I've tires to compile following small program with
  > both old gcc and egcs-1.0.1 (sorry I don't have latest snapshot here,
  > but I will try it at home later).
  > 
  > Program is:
  > 
  > #include <stdio.h>
  > main()
  > {
  >   int i,a=1,b=0;
  >   for(i=0;i<99999999;i++) 
  >    a=a+b,b=2*a;
  >   printf("%i\n",a);
  > }
  > 
  > It is compiled with -O3 -m486
  > Output from the gcc2.7.2 is:
  > 	pushl %ebp
  > 	movl %esp,%ebp
  > 	movl $1,%eax
  > 	xorl %ecx,%ecx
  > 	movl $9998,%edx
  > 	.align 4
  > .L11:
  > 	addl %ecx,%eax
  > 	leal (%eax,%eax),%ecx
  > 	decl %edx
  > 	jns .L11
  > 
  > (just the internal loop)
  > And egcs:
  > 
  > 	pushl %ebp
  > 	movl $1,%edx
  > 	movl %esp,%ebp
  > 	xorl %ecx,%ecx
  > 	movl $9998,%eax
  > 	.align 4
  > .L5:
  > 	addl %ecx,%edx
  > 	movl %edx,%ecx
  > 	addl %edx,%ecx
  > 	decl %eax
  > 	jns .L5
  > Looks IMO worse... Why it didn't used leal anymore? I think it is also
  > slower (at least here at K5).
I just checked this with the current snapshot and the loop looks like this now:

.L6:
        addl %edx,%eax
        leal 0(,%eax,2),%edx
        decl %ecx
        jns .L6

jeff