Re: gcc 3.2.1 optimizer degradation (strlen, -O2)

["Marcel Cox" <marcel_cox at hotmail dot com>]

I think you forgot or lost the -O2 option. The 3.2.1 code you posted is what
I got when I don't use any -O? option at all. With -O2, optimized for
Pentium, I got the following code:

strlen.c:

_strlen:
 pushl %ebp
 xorl %eax, %eax
 movl %esp, %ebp
 movl 8(%ebp), %edx
 testl %edx, %edx
 je L1
 movl %edx, %eax
 cmpb $0, (%edx)
 je L9
 .p2align 4,,7
L7:
 incl %eax
 cmpb $0, (%eax)
 jne L7
L9:
 subl %edx, %eax
L1:
 popl %ebp
 ret


strlen2.c:

_strlen:
 pushl %ebp
 xorl %eax, %eax
 movl %esp, %ebp
 movl 8(%ebp), %ecx
 movl %ecx, %edx
 testl %ecx, %ecx
 je L1
 .p2align 4,,7
L3:
 movb (%edx), %al
 incl %edx
 testb %al, %al
 jne L3
 subl %ecx, %edx
 leal -1(%edx), %eax
L1:
 popl %ebp
 ret

Marcel

"Leonid Pauzner" <uue@pauzner.dnttm.ru> wrote in message
2.7.9.SBA.H8K8TU@pauzner.dnttm.ru">news:2.7.9.SBA.H8K8TU@pauzner.dnttm.ru...
>
> Sure, there is a "call _strlen" line in the assembler code,
> and I made the test program by calling
>   "gcc -o strlen.exe main.o strlen.o"
>
> This is a DJGPP environment.
>
> The key point in my original posting was the certain simple loop
> which happens to be optimized better with the old version of the compiler.
> Generated assembler attached.
>