xref: /titanic_50/usr/src/lib/libc/sparc/gen/strchr.s (revision 9acbbeaf2a1ffe5c14b244867d427714fab43c5c)
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, Version 1.0 only
6 * (the "License").  You may not use this file except in compliance
7 * with the License.
8 *
9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10 * or http://www.opensolaris.org/os/licensing.
11 * See the License for the specific language governing permissions
12 * and limitations under the License.
13 *
14 * When distributing Covered Code, include this CDDL HEADER in each
15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16 * If applicable, add the following below this CDDL HEADER, with the
17 * fields enclosed by brackets "[]" replaced with your own identifying
18 * information: Portions Copyright [yyyy] [name of copyright owner]
19 *
20 * CDDL HEADER END
21 */
22/*
23 * Copyright 2003 Sun Microsystems, Inc.  All rights reserved.
24 * Use is subject to license terms.
25 */
26
27.ident	"%Z%%M%	%I%	%E% SMI"
28
29	.file	"%M%"
30
31/*
32 * The strchr() function returns a pointer to the first occurrence of c
33 * (converted to a char) in string s, or a null pointer if c does not occur
34 * in the string.
35 */
36
37#include <sys/asm_linkage.h>
38#include "synonyms.h"
39
40	! Here, we start by checking to see if we're searching the dest
41	! string for a null byte.  We have fast code for this, so it's
42	! an important special case.  Otherwise, if the string is not
43	! word aligned, we check a for the search char a byte at a time
44	! until we've reached a word boundary.  Once this has happened
45	! some zero-byte finding values are initialized and the string
46	! is checked a word at a time
47
48	ENTRY(strchr)
49
50	.align 32
51
52	andcc	%o1, 0xff, %o1		! search only for this one byte
53	bz	.searchnullbyte		! faster code for searching null
54	andcc	%o0, 3, %o4		! str word aligned ?
55	bz,a	.prepword2		! yup, prepare for word-wise search
56	sll	%o1, 8, %g1		! start spreading findchar across word
57
58	ldub	[%o0], %o2		! str[0]
59	cmp	%o2, %o1		! str[0] == findchar ?
60	be	.done			! yup, done
61	tst	%o2			! str[0] == 0 ?
62	bz	.notfound		! yup, return null pointer
63	cmp	%o4, 3			! only one byte needed to align?
64	bz	.prepword		! yup, prepare for word-wise search
65	inc	%o0			! str++
66	ldub	[%o0], %o2		! str[1]
67	cmp	%o2, %o1		! str[1] == findchar ?
68	be	.done			! yup, done
69	tst	%o2			! str[1] == 0 ?
70	bz	.notfound		! yup, return null pointer
71	cmp	%o4, 2			! only two bytes needed to align?
72	bz	.prepword		! yup, prepare for word-wise search
73	inc	%o0			! str++
74	ldub	[%o0], %o2		! str[2]
75	cmp	%o2, %o1		! str[2] == findchar ?
76	be	.done			! yup, done
77	tst	%o2			! str[2] == 0 ?
78	bz	.notfound		! yup, return null pointer
79	inc	%o0			! str++
80
81.prepword:
82	sll	%o1, 8, %g1		! spread findchar ------+
83.prepword2:							!
84	sethi	%hi(0x01010101), %o4	! Alan Mycroft's magic1 !
85	or	%o1, %g1, %o1		!  across all <---------+
86	sethi	%hi(0x80808080), %o5	! Alan Mycroft's magic2	!
87	sll	%o1, 16, %g1		!   four bytes <--------+
88	or	%o4, %lo(0x01010101), %o4			!
89	or	%o1, %g1, %o1		!    of a word <--------+
90	or	%o5, %lo(0x80808080), %o5
91
92.searchchar:
93	lduw	[%o0], %o2		! src word
94	andn	%o5, %o2, %o3		! ~word & 0x80808080
95	sub	%o2, %o4, %g1		! word = (word - 0x01010101)
96	andcc	%o3, %g1, %g0		! ((word - 0x01010101) & ~word & 0x80808080)
97	bnz	.haszerobyte		! zero byte if magic expression != 0
98	xor	%o2, %o1, %g1		! tword = word ^ findchar
99	andn	%o5, %g1, %o3		! ~tword & 0x80808080
100	sub	%g1, %o4, %o2		! (tword - 0x01010101)
101	andcc	%o3, %o2, %g0		! ((tword - 0x01010101) & ~tword & 0x80808080)
102	bz,a	.searchchar		! no findchar if magic expression == 0
103	add	%o0, 4, %o0		! str += 4
104
105	! here we know "word" contains the searched character, but no null
106	! byte. if there was a null byte, we would have gone to .haszerobyte
107	! "tword" has null bytes where "word" had findchar. Examine "tword"
108
109.foundchar:
110	set	0xff000000, %o4		! mask for 1st byte
111	andcc	%g1, %o4, %g0		! first byte zero (= found search char) ?
112	bz	.done			! yup, done
113	set	0x00ff0000, %o5		! mask for 2nd byte
114	inc	%o0			! str++
115	andcc	%g1, %o5, %g0		! second byte zero (= found search char) ?
116	bz	.done			! yup, done
117	srl	%o4, 16, %o4		! 0x0000ff00 = mask for 3rd byte
118	inc	%o0			! str++
119	andcc	%g1, %o4, %g0		! third byte zero (= found search char) ?
120	bnz,a	.done			! nope, increment in delay slot
121	inc	%o0			! str++
122
123.done:
124	retl				! done with leaf function
125	nop				! padding
126
127	! Here we know that "word" contains a null byte indicating the
128	! end of the string. However, "word" might also contain findchar
129	! "tword" (in %g1) has null bytes where "word" had findchar. So
130	! check both "tword" and "word"
131
132.haszerobyte:
133	set	0xff000000, %o4		! mask for 1st byte
134	andcc	%g1, %o4, %g0		! first byte == findchar ?
135	bz	.done			! yup, done
136	andcc	%o2, %o4, %g0		! first byte == 0 ?
137	bz	.notfound		! yup, return null pointer
138	set	0x00ff0000, %o4		! mask for 2nd byte
139	inc	%o0			! str++
140	andcc	%g1, %o4, %g0		! second byte == findchar ?
141	bz	.done			! yup, done
142	andcc	%o2, %o4, %g0		! second byte == 0 ?
143	bz	.notfound		! yup, return null pointer
144	srl	%o4, 8, %o4		! mask for 3rd byte = 0x0000ff00
145	inc	%o0			! str++
146	andcc	%g1, %o4, %g0		! third byte == findchar ?
147	bz	.done			! yup, done
148	andcc	%o2, %o4, %g0		! third byte == 0 ?
149	bz	.notfound		! yup, return null pointer
150	andcc	%g1, 0xff, %g0		! fourth byte == findchar ?
151	bz	.done			! yup, done
152	inc	%o0			! str++
153
154.notfound:
155	retl				! done with leaf function
156	xor	%o0, %o0, %o0		! return null pointer
157
158	! since findchar == 0, we only have to do one test per item
159	! instead of two. This makes the search much faster.
160
161.searchnullbyte:
162	bz	.straligned		! str is word aligned
163	nop				! padding
164
165	cmp	%o4, 2			! str halfword aligned ?
166	be	.s2aligned		! yup
167	ldub	[%o0], %o1		! str[0]
168	tst	%o1			! byte zero?
169	bz	.done2			! yup, done
170	cmp	%o4, 3			! only one byte needed to align?
171	bz	.straligned		! yup
172	inc	%o0			! str++
173
174	! check to see if we're half word aligned, which it better than
175	! not being aligned at all.  Search the first half of the word
176	! if we are, and then search by whole word.
177
178.s2aligned:
179	lduh	[%o0], %o1		! str[]
180	srl	%o1, 8, %o4		! %o4<7:0> = first byte
181	tst	%o4			! first byte zero ?
182	bz	.done2			! yup, done
183	andcc	%o1, 0xff, %g0		! second byte zero ?
184	bz,a	.done2			! yup, done
185	inc	%o0			! str++
186	add	%o0, 2, %o0		! str+=2
187
188.straligned:
189	sethi	%hi(0x01010101), %o4	! Alan Mycroft's magic1
190	sethi	%hi(0x80808080), %o5	! Alan Mycroft's magic2
191	or	%o4, %lo(0x01010101), %o4
192	or	%o5, %lo(0x80808080), %o5
193
194.searchword:
195	lduw	[%o0], %o1		! src word
196	andn	%o5, %o1, %o3		! ~word & 0x80808080
197	sub	%o1, %o4, %g1		! word = (word - 0x01010101)
198	andcc	%o3, %g1, %g0		! ((word - 0x01010101) & ~word & 0x80808080)
199	bz,a	.searchword		! no zero byte if magic expression == 0
200	add	%o0, 4, %o0		! str += 4
201
202.zerobyte:
203	set	0xff000000, %o4		! mask for 1st byte
204	andcc	%o1, %o4, %g0		! first byte zero?
205	bz	.done2			! yup, done
206	set	0x00ff0000, %o5		! mask for 2nd byte
207	inc	%o0			! str++
208	andcc	%o1, %o5, %g0		! second byte zero?
209	bz	.done2			! yup, done
210	srl	%o4, 16, %o4		! 0x0000ff00 = mask for 3rd byte
211	inc	%o0			! str++
212	andcc	%o1, %o4, %g0		! third byte zero?
213	bnz,a	.done2			! nope, increment in delay slot
214	inc	%o0			! str++
215.done2:
216    	retl				! return from leaf function
217	nop				! padding
218
219	SET_SIZE(strchr)
220