NetBSD/lib/libc/arch/i386/string/strcmp.S

/*
 * Written by J.T. Conklin <jtc@netbsd.org>.
 * Public domain.
 */

#include <machine/asm.h>

#if defined(LIBC_SCCS)
	RCSID("$NetBSD: strcmp.S,v 1.11 1995/04/28 22:58:10 jtc Exp $")
#endif

/*
 * NOTE: I've unrolled the loop eight times: large enough to make a
 * significant difference, and small enough not to totally trash the
 * cache.
 */

ENTRY(strcmp)
	movl	0x04(%esp),%eax
	movl	0x08(%esp),%edx
	jmp	L2			/* Jump into the loop! */

	.align	2,0x90
L1:	incl	%eax
	incl	%edx
L2:	movb	(%eax),%cl
	testb	%cl,%cl			/* null terminator??? */
	jz	L3
	cmpb	%cl,(%edx)		/* chars match??? */
	jne	L3
	incl	%eax
	incl	%edx
	movb	(%eax),%cl
	testb	%cl,%cl
	jz	L3
	cmpb	%cl,(%edx)
	jne	L3
	incl	%eax
	incl	%edx
	movb	(%eax),%cl
	testb	%cl,%cl
	jz	L3
	cmpb	%cl,(%edx)
	jne	L3
	incl	%eax
	incl	%edx
	movb	(%eax),%cl
	testb	%cl,%cl
	jz	L3
	cmpb	%cl,(%edx)
	jne	L3
	incl	%eax
	incl	%edx
	movb	(%eax),%cl
	testb	%cl,%cl
	jz	L3
	cmpb	%cl,(%edx)
	jne	L3
	incl	%eax
	incl	%edx
	movb	(%eax),%cl
	testb	%cl,%cl
	jz	L3
	cmpb	%cl,(%edx)
	jne	L3
	incl	%eax
	incl	%edx
	movb	(%eax),%cl
	testb	%cl,%cl
	jz	L3
	cmpb	%cl,(%edx)
	jne	L3
	incl	%eax
	incl	%edx
	movb	(%eax),%cl
	testb	%cl,%cl
	jz	L3
	cmpb	%cl,(%edx)
	je	L1
	.align 2, 0x90
L3:	movzbl	(%eax),%eax		/* unsigned comparison */
	movzbl	(%edx),%edx
	subl	%edx,%eax
	ret
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`/*`
Winning Strategies has placed this code into the Public Domain 1995-04-29 02:53:59 +04:00			`* Written by J.T. Conklin <jtc@netbsd.org>.`
			`* Public domain.`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`*/`

Use RCSID macro for RCS ID's. 1994-03-12 04:39:55 +03:00			`#include <machine/asm.h>`

Change instances of LIBC_RCS back to LIBC_SCCS. If we are going to change th LIBC_RCS, the entire source tree should be updated at once. Remove && !defined(lint) from conditionals around rcs id's. No one is ever going to pass a *.s file through lint. 1993-10-21 04:39:55 +03:00			`#if defined(LIBC_SCCS)`
$Id: $ -> $NetBSD: $ 1995-04-29 02:57:54 +04:00			`RCSID("$NetBSD: strcmp.S,v 1.11 1995/04/28 22:58:10 jtc Exp $")`
Change instances of LIBC_RCS back to LIBC_SCCS. If we are going to change th LIBC_RCS, the entire source tree should be updated at once. Remove && !defined(lint) from conditionals around rcs id's. No one is ever going to pass a *.s file through lint. 1993-10-21 04:39:55 +03:00			`#endif`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00
			`/*`
Winning Strategies has placed this code into the Public Domain 1995-04-29 02:53:59 +04:00			`* NOTE: I've unrolled the loop eight times: large enough to make a`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`* significant difference, and small enough not to totally trash the`
Fix spelling mistake: cashe -> cache. 1994-02-14 20:45:16 +03:00			`* cache.`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`*/`

			`ENTRY(strcmp)`
So many files in arch/i386 have been changed I might as well run my remove unecessary trailing whitespace script. 1993-10-21 04:48:14 +03:00			`movl 0x04(%esp),%eax`
			`movl 0x08(%esp),%edx`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`jmp L2 /* Jump into the loop! */`

			`.align 2,0x90`
			`L1: incl %eax`
So many files in arch/i386 have been changed I might as well run my remove unecessary trailing whitespace script. 1993-10-21 04:48:14 +03:00			`incl %edx`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`L2: movb (%eax),%cl`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`testb %cl,%cl /* null terminator??? */`
Style police. Use `je' and `jne' only on the result of an equality test; use `jz' and `jnz' in all other cases. 1994-02-14 23:05:52 +03:00			`jz L3`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`cmpb %cl,(%edx) /* chars match??? */`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`jne L3`
			`incl %eax`
			`incl %edx`
			`movb (%eax),%cl`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`testb %cl,%cl`
Style police. Use `je' and `jne' only on the result of an equality test; use `jz' and `jnz' in all other cases. 1994-02-14 23:05:52 +03:00			`jz L3`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`cmpb %cl,(%edx)`
			`jne L3`
			`incl %eax`
			`incl %edx`
			`movb (%eax),%cl`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`testb %cl,%cl`
Style police. Use `je' and `jne' only on the result of an equality test; use `jz' and `jnz' in all other cases. 1994-02-14 23:05:52 +03:00			`jz L3`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`cmpb %cl,(%edx)`
			`jne L3`
			`incl %eax`
			`incl %edx`
			`movb (%eax),%cl`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`testb %cl,%cl`
Style police. Use `je' and `jne' only on the result of an equality test; use `jz' and `jnz' in all other cases. 1994-02-14 23:05:52 +03:00			`jz L3`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`cmpb %cl,(%edx)`
			`jne L3`
			`incl %eax`
			`incl %edx`
			`movb (%eax),%cl`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`testb %cl,%cl`
Style police. Use `je' and `jne' only on the result of an equality test; use `jz' and `jnz' in all other cases. 1994-02-14 23:05:52 +03:00			`jz L3`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`cmpb %cl,(%edx)`
			`jne L3`
			`incl %eax`
			`incl %edx`
			`movb (%eax),%cl`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`testb %cl,%cl`
Style police. Use `je' and `jne' only on the result of an equality test; use `jz' and `jnz' in all other cases. 1994-02-14 23:05:52 +03:00			`jz L3`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`cmpb %cl,(%edx)`
			`jne L3`
			`incl %eax`
			`incl %edx`
			`movb (%eax),%cl`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`testb %cl,%cl`
Style police. Use `je' and `jne' only on the result of an equality test; use `jz' and `jnz' in all other cases. 1994-02-14 23:05:52 +03:00			`jz L3`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`cmpb %cl,(%edx)`
			`jne L3`
			`incl %eax`
			`incl %edx`
			`movb (%eax),%cl`
Use "testb %cl,%cl" instead of "cmpb $0,%cl". This saves three cycles per character, and reduces the size of the function by eight bytes (the loop is unrolled eight times). Thanks to davidg for pointing this out. 1993-10-08 04:25:51 +03:00			`testb %cl,%cl`
Style police. Use `je' and `jne' only on the result of an equality test; use `jz' and `jnz' in all other cases. 1994-02-14 23:05:52 +03:00			`jz L3`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`cmpb %cl,(%edx)`
			`je L1`
			`.align 2, 0x90`
Use movzbl, not movsbl, to zero extend for unsigned comparison. Fix from Bruce Evans. 1994-02-07 23:57:05 +03:00			`L3: movzbl (%eax),%eax /* unsigned comparison */`
			`movzbl (%edx),%edx`
make an arch directory in libc, and move all machine dirs into it 1993-10-07 03:17:24 +03:00			`subl %edx,%eax`
			`ret`