xref: /titanic_41/usr/src/lib/libc/sparcv9/gen/strcmp.s (revision b7d3956b92a285d8dac2c7f5f7e28d2ef5347ef8)
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/*
23 * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
24 * Use is subject to license terms.
25 */
26
27#pragma ident	"%Z%%M%	%I%	%E% SMI"
28
29	.file	"%M%"
30
31/* strcmp(s1, s2)
32 *
33 * Compare strings:  s1>s2: >0  s1==s2: 0  s1<s2: <0
34 *
35 * Fast assembler language version of the following C-program for strcmp
36 * which represents the `standard' for the C-library.
37 *
38 *	int
39 *	strcmp(s1, s2)
40 *	register const char *s1;
41 *	register const char *s2;
42 *	{
43 *
44 *		if(s1 == s2)
45 *			return(0);
46 *		while(*s1 == *s2++)
47 *			if(*s1++ == '\0')
48 *				return(0);
49 *		return(*s1 - s2[-1]);
50 *	}
51 */
52
53#include <sys/asm_linkage.h>
54
55	! This strcmp implementation first determines whether s1 is aligned.
56	! If it is not, it attempts to align it and then checks the
57	! alignment of the destination string.  If it is possible to
58	! align s2, this also happens and then the compare begins.  Otherwise,
59	! a different compare for non-aligned strings is used.
60
61	ENTRY(strcmp)
62
63	.align 32
64
65	subcc	%o0, %o1, %o2		! s1 == s2 ?
66	bz,pn	%xcc, .stringsequal	! yup, same string, done
67 	sethi	%hi(0x01010101), %o5	! magic2<31:13>
68	andcc	%o0, 7, %o3		! s1 sword-aligned ?
69	or	%o5, %lo(0x01010101),%o5! magic2<31:0>
70	bz,pn	%xcc, .s1aligned	! yup
71	sllx	%o5, 32, %o4		! magic2<63:32>
72	sub 	%o3, 8, %o3		! number of bytes till s1 aligned
73
74.aligns1:
75	ldub	[%o1 + %o2], %o0	! s1[]
76	ldub	[%o1], %g1		! s2[]
77	subcc	%o0, %g1, %o0		! s1[] != s2[] ?
78	bne,pn	%xcc, .done		! yup, done
79	addcc	%o0, %g1, %g0		! s1[] == 0 ?
80	bz,pn	%xcc, .done		! yup, done
81	inccc	%o3			! s1 aligned yet?
82	bnz,pt	%xcc, .aligns1		! nope, compare another pair of bytes
83	inc	%o1			! s1++, s2++
84
85.s1aligned:
86	andcc	%o1, 7, %o3		! s2 dword aligned ?
87	or	%o5, %o4, %o5		! magic2<63:0>
88	bz,pn	%xcc, .s2aligned	! yup
89	sllx	%o5, 7, %o4		! magic1
90	sllx	%o3, 3, %g5		! leftshift = 8*ofs
91	orn	%g0, %g0, %g1		! all ones
92	and	%o1, -8, %o1		! round s1 down to next aligned dword
93	srlx	%g1, %g5, %g1		! mask for fixing up bytes
94	ldx	[%o1], %o0		! new lower dword in s2
95	orn	%o0, %g1, %o0 		! force start bytes to non-zero
96	sub	%g0, %g5, %g4		! rightshift = -(8*ofs) mod 64
97	sllx	%o0, %g5, %g1		! partial unaligned word from s2
98	add	%o2, %o3, %o2		! adjust pointers
99	nop				! align loop to 16-byte boundary
100	nop				! align loop to 16-byte boundary
101
102.cmp:
103	andn	%o4, %o0, %o3		! ~word & 0x80808080
104	sub	%o0, %o5, %o0		! word - 0x01010101
105	andcc	%o0, %o3, %g0		! (word - 0x01010101) & ~word & 0x80808080
106	bz,a,pt	%xcc, .doload		! no null byte in previous word from s2
107 	ldx	[%o1+8], %o0		! next aligned word in s2
108.doload:
109	srlx	%o0, %g4, %o3		! bytes from aligned word from s2
110	or	%g1, %o3, %g1		! merge to get unaligned word from s2
111	ldx	[%o1 + %o2], %o3	! word from s1
112	cmp	%o3, %g1		! *s1 != *s2 ?
113	bne,pn	%xcc, .wordsdiffer	! yup, find the byte that is different
114	add	%o1, 8, %o1		! s1+=8, s2+=8
115	andn	%o4, %o3, %g1		! ~word & 0x80808080
116	sub	%o3, %o5, %o3		! word - 0x01010101
117	andcc	%o3, %g1, %g0		! (word - 0x01010101) & ~word & 0x80808080
118	bz,pt	%xcc, .cmp		! no null-byte in s1 yet
119	sllx	%o0, %g5, %g1		! partial unaligned word from s2
120
121	! words are equal but the end of s1 has been reached
122	! this means the strings must be equal
123.strequal:
124	retl				! return from leaf function
125	mov	%g0, %o0		! return 0, i.e. strings are equal
126	nop
127
128.s2aligned:
129	ldx	[%o1 + %o2], %o3	! load word from s1
130
131.cmpaligned:
132	ldx	[%o1], %g1		! load word from s2
133	cmp	%o3, %g1		! *scr1 == *src2 ?
134	bne,pn	%xcc, .wordsdiffer	! nope, find mismatching character
135	add	%o1, 8, %o1		! src1 += 8, src2 += 8
136	andn	%o4, %o3, %o0		! ~word & 0x80808080
137	sub	%o3, %o5, %o3		! word - 0x01010101
138	andcc	%o3, %o0, %g0		! (word - 0x01010101) & ~word & 0x80808080
139	bz,a,pt	%xcc, .cmpaligned	! no null-byte in s1 yet
140	ldx	[%o1 + %o2], %o3	! load word from s1
141
142	! words are equal but the end of s1 has been reached
143	! this means the strings must be equal
144
145.stringsequal:
146	retl				! return from leaf function
147	mov	%g0, %o0		! return 0, i.e. strings are equal
148	nop				! align loop on 16-byte boundary
149	nop				! align loop on 16-byte boundary
150	nop				! align loop on 16-byte boundary
151
152.wordsdiffer:
153	mov	56, %o4			! initial shift count
154	srlx	%g1, %o4, %o2		! first byte of mismatching word in s2
155.cmpbytes:
156	srlx	%o3, %o4, %o1		! first byte of mismatching word in s1
157	subcc	%o1, %o2, %o0		! *s1-*s2
158	bnz,pn	%xcc, .done		! bytes differ, return difference
159	nop
160	andcc	%o1, 0xff, %o0		! *s1 == 0 ?
161	bz,pn	%xcc, .done		! yup, strings match
162	subcc	%o4, 8, %o4		! shift_count -= 8
163	bpos,pt	%xcc, .cmpbytes		! until all bytes processed
164	srlx	%g1, %o4, %o2		! first byte of mismatching word in s2
165
166.done:
167	retl				! return from leaf routine
168	nop				! padding
169
170	SET_SIZE(strcmp)
171