xref: /illumos-gate/usr/src/lib/libc/i386/gen/strcat.S (revision 5d9d9091f564c198a760790b0bfa72c44e17912b)
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
23 * Use is subject to license terms.
24 */
25
26	.file	"strcat.s"
27
28/
29/ strcat(s1, s2)
30/
31/ Concatenates s2 on the end of s1.  s1's space must be large enough.
32/ Returns s1.
33/
34/ Fast assembly language version of the following C-program strcat
35/ which represents the `standard' for the C-library.
36/
37/	char *
38/	strcat(char *s1, const char *s2)
39/	{
40/		char	*os1 = s1;
41/
42/		while (*s1++)
43/			;
44/		--s1;
45/		while (*s1++ = *s2++)
46/			;
47/		return (os1);
48/	}
49/
50/ In this assembly language version, the following expression is used
51/ to check if a 32-bit word data contains a null byte or not:
52/	(((A & 0x7f7f7f7f) + 0x7f7f7f7f) | A) & 0x80808080
53/ If the above expression geneates a value other than 0x80808080,
54/ that means the 32-bit word data contains a null byte.
55/
56
57#include "SYS.h"
58
59	ENTRY(strcat)
60	pushl	%edi			/ save register variable
61	/ find a null byte in destination string
62	movl	8(%esp), %edi		/ %edi = destination string address
63	testl	$3, %edi		/ if %edi not word aligned
64	jnz	.L1			/ goto .L1
65	.align	4
66.L2:
67	movl	(%edi), %edx		/ move 1 word from (%edi) to %edx
68	movl	$0x7f7f7f7f, %ecx
69	andl	%edx, %ecx		/ %ecx = %edx & 0x7f7f7f7f
70	addl	$4, %edi		/ next word
71	addl	$0x7f7f7f7f, %ecx	/ %ecx += 0x7f7f7f7f
72	orl	%edx, %ecx		/ %ecx |= %edx
73	andl	$0x80808080, %ecx	/ %ecx &= 0x80808080
74	cmpl	$0x80808080, %ecx	/ if no null byte in this word
75	je	.L2			/ goto .L2
76	subl	$4, %edi		/ post-incremented
77.L1:
78	cmpb	$0, (%edi)		/ if a byte in (%edi) is null
79	je	.L3			/ goto .L3
80	incl	%edi			/ next byte
81	testl	$3, %edi		/ if %edi not word aligned
82	jnz	.L1			/ goto .L1
83	jmp	.L2			/ goto .L2 (%edi word aligned)
84	.align	4
85.L3:
86	/ %edi points to a null byte in destination string
87	movl	12(%esp), %eax		/ %eax = source string address
88	testl	$3, %eax		/ if %eax not word aligned
89	jnz	.L4			/ goto .L4
90	.align	4
91.L5:
92	movl	(%eax), %edx		/ move 1 word from (%eax) to %edx
93	movl	$0x7f7f7f7f, %ecx
94	andl	%edx, %ecx		/ %ecx = %edx & 0x7f7f7f7f
95	addl	$4, %eax		/ next word
96	addl	$0x7f7f7f7f, %ecx	/ %ecx += 0x7f7f7f7f
97	orl	%edx, %ecx		/ %ecx |= %edx
98	andl	$0x80808080, %ecx	/ %ecx &= 0x80808080
99	cmpl	$0x80808080, %ecx	/ if null byte in this word
100	jne	.L7			/ goto .L7
101	movl	%edx, (%edi)		/ copy this word to (%edi)
102	addl	$4, %edi		/ next word
103	jmp	.L5			/ goto .L5
104.L7:
105	subl	$4, %eax		/ post-incremented
106	.align	4
107.L4:
108	movb	(%eax), %dl		/ %dl = a byte in (%eax)
109	cmpb	$0, %dl			/ compare %dl with a null byte
110	movb	%dl, (%edi)		/ copy %dl to (%edi)
111	je	.L6			/ if %dl is a null, goto .L6
112	incl	%eax			/ next byte
113	incl	%edi			/ next byte
114	testl	$3, %eax		/ if %eax not word aligned
115	jnz	.L4			/ goto .L4
116	jmp	.L5			/ goto .L5 (%eax word aligned)
117	.align	4
118.L6:
119	movl	8(%esp), %eax		/ return the destination address
120	popl	%edi			/ restore register variable
121	ret
122	SET_SIZE(strcat)
123