1 | /* strchrnul (str, chr) -- Return pointer to first occurrence of CHR in STR
|
---|
2 | or the final NUL byte.
|
---|
3 | For Intel 80x86, x>=3.
|
---|
4 | Copyright (C) 1994-1997, 1999, 2000, 2005 Free Software Foundation, Inc.
|
---|
5 | This file is part of the GNU C Library.
|
---|
6 | Contributed by Ulrich Drepper <drepper@gnu.org>
|
---|
7 | Some optimisations by Alan Modra <Alan@SPRI.Levels.UniSA.Edu.Au>
|
---|
8 |
|
---|
9 | The GNU C Library is free software; you can redistribute it and/or
|
---|
10 | modify it under the terms of the GNU Lesser General Public
|
---|
11 | License as published by the Free Software Foundation; either
|
---|
12 | version 2.1 of the License, or (at your option) any later version.
|
---|
13 |
|
---|
14 | The GNU C Library is distributed in the hope that it will be useful,
|
---|
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
---|
17 | Lesser General Public License for more details.
|
---|
18 |
|
---|
19 | You should have received a copy of the GNU Lesser General Public
|
---|
20 | License along with the GNU C Library; if not, write to the Free
|
---|
21 | Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
|
---|
22 | 02111-1307 USA. */
|
---|
23 |
|
---|
24 | #include <sysdep.h>
|
---|
25 | #include "asm-syntax.h"
|
---|
26 | #include "bp-sym.h"
|
---|
27 | #include "bp-asm.h"
|
---|
28 |
|
---|
29 | #define PARMS LINKAGE+4 /* space for 1 saved reg */
|
---|
30 | #define RTN PARMS
|
---|
31 | #define STR RTN+RTN_SIZE
|
---|
32 | #define CHR STR+PTR_SIZE
|
---|
33 |
|
---|
34 | .text
|
---|
35 | ENTRY (BP_SYM (__strchrnul))
|
---|
36 | ENTER
|
---|
37 |
|
---|
38 | pushl %edi /* Save callee-safe registers used here. */
|
---|
39 | cfi_adjust_cfa_offset (4)
|
---|
40 | cfi_rel_offset (edi, 0)
|
---|
41 |
|
---|
42 | movl STR(%esp), %eax
|
---|
43 | movl CHR(%esp), %edx
|
---|
44 | CHECK_BOUNDS_LOW (%eax, STR(%esp))
|
---|
45 |
|
---|
46 | /* At the moment %edx contains CHR. What we need for the
|
---|
47 | algorithm is CHR in all bytes of the dword. Avoid
|
---|
48 | operations on 16 bit words because these require an
|
---|
49 | prefix byte (and one more cycle). */
|
---|
50 | movb %dl, %dh /* now it is 0|0|c|c */
|
---|
51 | movl %edx, %ecx
|
---|
52 | shll $16, %edx /* now it is c|c|0|0 */
|
---|
53 | movw %cx, %dx /* and finally c|c|c|c */
|
---|
54 |
|
---|
55 | /* Before we start with the main loop we process single bytes
|
---|
56 | until the source pointer is aligned. This has two reasons:
|
---|
57 | 1. aligned 32-bit memory access is faster
|
---|
58 | and (more important)
|
---|
59 | 2. we process in the main loop 32 bit in one step although
|
---|
60 | we don't know the end of the string. But accessing at
|
---|
61 | 4-byte alignment guarantees that we never access illegal
|
---|
62 | memory if this would not also be done by the trivial
|
---|
63 | implementation (this is because all processor inherent
|
---|
64 | boundaries are multiples of 4. */
|
---|
65 |
|
---|
66 | testb $3, %al /* correctly aligned ? */
|
---|
67 | jz L(11) /* yes => begin loop */
|
---|
68 | movb (%eax), %cl /* load byte in question (we need it twice) */
|
---|
69 | cmpb %cl, %dl /* compare byte */
|
---|
70 | je L(6) /* target found => return */
|
---|
71 | testb %cl, %cl /* is NUL? */
|
---|
72 | jz L(6) /* yes => return NULL */
|
---|
73 | incl %eax /* increment pointer */
|
---|
74 |
|
---|
75 | testb $3, %al /* correctly aligned ? */
|
---|
76 | jz L(11) /* yes => begin loop */
|
---|
77 | movb (%eax), %cl /* load byte in question (we need it twice) */
|
---|
78 | cmpb %cl, %dl /* compare byte */
|
---|
79 | je L(6) /* target found => return */
|
---|
80 | testb %cl, %cl /* is NUL? */
|
---|
81 | jz L(6) /* yes => return NULL */
|
---|
82 | incl %eax /* increment pointer */
|
---|
83 |
|
---|
84 | testb $3, %al /* correctly aligned ? */
|
---|
85 | jz L(11) /* yes => begin loop */
|
---|
86 | movb (%eax), %cl /* load byte in question (we need it twice) */
|
---|
87 | cmpb %cl, %dl /* compare byte */
|
---|
88 | je L(6) /* target found => return */
|
---|
89 | testb %cl, %cl /* is NUL? */
|
---|
90 | jz L(6) /* yes => return NULL */
|
---|
91 | incl %eax /* increment pointer */
|
---|
92 |
|
---|
93 | /* No we have reached alignment. */
|
---|
94 | jmp L(11) /* begin loop */
|
---|
95 |
|
---|
96 | /* We exit the loop if adding MAGIC_BITS to LONGWORD fails to
|
---|
97 | change any of the hole bits of LONGWORD.
|
---|
98 |
|
---|
99 | 1) Is this safe? Will it catch all the zero bytes?
|
---|
100 | Suppose there is a byte with all zeros. Any carry bits
|
---|
101 | propagating from its left will fall into the hole at its
|
---|
102 | least significant bit and stop. Since there will be no
|
---|
103 | carry from its most significant bit, the LSB of the
|
---|
104 | byte to the left will be unchanged, and the zero will be
|
---|
105 | detected.
|
---|
106 |
|
---|
107 | 2) Is this worthwhile? Will it ignore everything except
|
---|
108 | zero bytes? Suppose every byte of LONGWORD has a bit set
|
---|
109 | somewhere. There will be a carry into bit 8. If bit 8
|
---|
110 | is set, this will carry into bit 16. If bit 8 is clear,
|
---|
111 | one of bits 9-15 must be set, so there will be a carry
|
---|
112 | into bit 16. Similarly, there will be a carry into bit
|
---|
113 | 24. If one of bits 24-31 is set, there will be a carry
|
---|
114 | into bit 32 (=carry flag), so all of the hole bits will
|
---|
115 | be changed.
|
---|
116 |
|
---|
117 | 3) But wait! Aren't we looking for CHR, not zero?
|
---|
118 | Good point. So what we do is XOR LONGWORD with a longword,
|
---|
119 | each of whose bytes is CHR. This turns each byte that is CHR
|
---|
120 | into a zero. */
|
---|
121 |
|
---|
122 | /* Each round the main loop processes 16 bytes. */
|
---|
123 |
|
---|
124 | ALIGN(4)
|
---|
125 |
|
---|
126 | L(1): addl $16, %eax /* adjust pointer for whole round */
|
---|
127 |
|
---|
128 | L(11): movl (%eax), %ecx /* get word (= 4 bytes) in question */
|
---|
129 | xorl %edx, %ecx /* XOR with word c|c|c|c => bytes of str == c
|
---|
130 | are now 0 */
|
---|
131 | movl $0xfefefeff, %edi /* magic value */
|
---|
132 | addl %ecx, %edi /* add the magic value to the word. We get
|
---|
133 | carry bits reported for each byte which
|
---|
134 | is *not* CHR */
|
---|
135 |
|
---|
136 | /* According to the algorithm we had to reverse the effect of the
|
---|
137 | XOR first and then test the overflow bits. But because the
|
---|
138 | following XOR would destroy the carry flag and it would (in a
|
---|
139 | representation with more than 32 bits) not alter then last
|
---|
140 | overflow, we can now test this condition. If no carry is signaled
|
---|
141 | no overflow must have occurred in the last byte => it was 0. */
|
---|
142 | jnc L(7)
|
---|
143 |
|
---|
144 | /* We are only interested in carry bits that change due to the
|
---|
145 | previous add, so remove original bits */
|
---|
146 | xorl %ecx, %edi /* ((word^charmask)+magic)^(word^charmask) */
|
---|
147 |
|
---|
148 | /* Now test for the other three overflow bits. */
|
---|
149 | orl $0xfefefeff, %edi /* set all non-carry bits */
|
---|
150 | incl %edi /* add 1: if one carry bit was *not* set
|
---|
151 | the addition will not result in 0. */
|
---|
152 |
|
---|
153 | /* If at least one byte of the word is CHR we don't get 0 in %edi. */
|
---|
154 | jnz L(7) /* found it => return pointer */
|
---|
155 |
|
---|
156 | /* Now we made sure the dword does not contain the character we are
|
---|
157 | looking for. But because we deal with strings we have to check
|
---|
158 | for the end of string before testing the next dword. */
|
---|
159 |
|
---|
160 | xorl %edx, %ecx /* restore original dword without reload */
|
---|
161 | movl $0xfefefeff, %edi /* magic value */
|
---|
162 | addl %ecx, %edi /* add the magic value to the word. We get
|
---|
163 | carry bits reported for each byte which
|
---|
164 | is *not* 0 */
|
---|
165 | jnc L(7) /* highest byte is NUL => return NULL */
|
---|
166 | xorl %ecx, %edi /* (word+magic)^word */
|
---|
167 | orl $0xfefefeff, %edi /* set all non-carry bits */
|
---|
168 | incl %edi /* add 1: if one carry bit was *not* set
|
---|
169 | the addition will not result in 0. */
|
---|
170 | jnz L(7) /* found NUL => return NULL */
|
---|
171 |
|
---|
172 | movl 4(%eax), %ecx /* get word (= 4 bytes) in question */
|
---|
173 | xorl %edx, %ecx /* XOR with word c|c|c|c => bytes of str == c
|
---|
174 | are now 0 */
|
---|
175 | movl $0xfefefeff, %edi /* magic value */
|
---|
176 | addl %ecx, %edi /* add the magic value to the word. We get
|
---|
177 | carry bits reported for each byte which
|
---|
178 | is *not* CHR */
|
---|
179 | jnc L(71) /* highest byte is CHR => return pointer */
|
---|
180 | xorl %ecx, %edi /* ((word^charmask)+magic)^(word^charmask) */
|
---|
181 | orl $0xfefefeff, %edi /* set all non-carry bits */
|
---|
182 | incl %edi /* add 1: if one carry bit was *not* set
|
---|
183 | the addition will not result in 0. */
|
---|
184 | jnz L(71) /* found it => return pointer */
|
---|
185 | xorl %edx, %ecx /* restore original dword without reload */
|
---|
186 | movl $0xfefefeff, %edi /* magic value */
|
---|
187 | addl %ecx, %edi /* add the magic value to the word. We get
|
---|
188 | carry bits reported for each byte which
|
---|
189 | is *not* 0 */
|
---|
190 | jnc L(71) /* highest byte is NUL => return NULL */
|
---|
191 | xorl %ecx, %edi /* (word+magic)^word */
|
---|
192 | orl $0xfefefeff, %edi /* set all non-carry bits */
|
---|
193 | incl %edi /* add 1: if one carry bit was *not* set
|
---|
194 | the addition will not result in 0. */
|
---|
195 | jnz L(71) /* found NUL => return NULL */
|
---|
196 |
|
---|
197 | movl 8(%eax), %ecx /* get word (= 4 bytes) in question */
|
---|
198 | xorl %edx, %ecx /* XOR with word c|c|c|c => bytes of str == c
|
---|
199 | are now 0 */
|
---|
200 | movl $0xfefefeff, %edi /* magic value */
|
---|
201 | addl %ecx, %edi /* add the magic value to the word. We get
|
---|
202 | carry bits reported for each byte which
|
---|
203 | is *not* CHR */
|
---|
204 | jnc L(72) /* highest byte is CHR => return pointer */
|
---|
205 | xorl %ecx, %edi /* ((word^charmask)+magic)^(word^charmask) */
|
---|
206 | orl $0xfefefeff, %edi /* set all non-carry bits */
|
---|
207 | incl %edi /* add 1: if one carry bit was *not* set
|
---|
208 | the addition will not result in 0. */
|
---|
209 | jnz L(72) /* found it => return pointer */
|
---|
210 | xorl %edx, %ecx /* restore original dword without reload */
|
---|
211 | movl $0xfefefeff, %edi /* magic value */
|
---|
212 | addl %ecx, %edi /* add the magic value to the word. We get
|
---|
213 | carry bits reported for each byte which
|
---|
214 | is *not* 0 */
|
---|
215 | jnc L(72) /* highest byte is NUL => return NULL */
|
---|
216 | xorl %ecx, %edi /* (word+magic)^word */
|
---|
217 | orl $0xfefefeff, %edi /* set all non-carry bits */
|
---|
218 | incl %edi /* add 1: if one carry bit was *not* set
|
---|
219 | the addition will not result in 0. */
|
---|
220 | jnz L(72) /* found NUL => return NULL */
|
---|
221 |
|
---|
222 | movl 12(%eax), %ecx /* get word (= 4 bytes) in question */
|
---|
223 | xorl %edx, %ecx /* XOR with word c|c|c|c => bytes of str == c
|
---|
224 | are now 0 */
|
---|
225 | movl $0xfefefeff, %edi /* magic value */
|
---|
226 | addl %ecx, %edi /* add the magic value to the word. We get
|
---|
227 | carry bits reported for each byte which
|
---|
228 | is *not* CHR */
|
---|
229 | jnc L(73) /* highest byte is CHR => return pointer */
|
---|
230 | xorl %ecx, %edi /* ((word^charmask)+magic)^(word^charmask) */
|
---|
231 | orl $0xfefefeff, %edi /* set all non-carry bits */
|
---|
232 | incl %edi /* add 1: if one carry bit was *not* set
|
---|
233 | the addition will not result in 0. */
|
---|
234 | jnz L(73) /* found it => return pointer */
|
---|
235 | xorl %edx, %ecx /* restore original dword without reload */
|
---|
236 | movl $0xfefefeff, %edi /* magic value */
|
---|
237 | addl %ecx, %edi /* add the magic value to the word. We get
|
---|
238 | carry bits reported for each byte which
|
---|
239 | is *not* 0 */
|
---|
240 | jnc L(73) /* highest byte is NUL => return NULL */
|
---|
241 | xorl %ecx, %edi /* (word+magic)^word */
|
---|
242 | orl $0xfefefeff, %edi /* set all non-carry bits */
|
---|
243 | incl %edi /* add 1: if one carry bit was *not* set
|
---|
244 | the addition will not result in 0. */
|
---|
245 | jz L(1) /* no NUL found => restart loop */
|
---|
246 |
|
---|
247 | L(73): addl $4, %eax /* adjust pointer */
|
---|
248 | L(72): addl $4, %eax
|
---|
249 | L(71): addl $4, %eax
|
---|
250 |
|
---|
251 | /* We now scan for the byte in which the character was matched.
|
---|
252 | But we have to take care of the case that a NUL char is
|
---|
253 | found before this in the dword. */
|
---|
254 |
|
---|
255 | L(7): testb %cl, %cl /* is first byte CHR? */
|
---|
256 | jz L(6) /* yes => return pointer */
|
---|
257 | cmpb %dl, %cl /* is first byte NUL? */
|
---|
258 | je L(6) /* yes => return NULL */
|
---|
259 | incl %eax /* it's not in the first byte */
|
---|
260 |
|
---|
261 | testb %ch, %ch /* is second byte CHR? */
|
---|
262 | jz L(6) /* yes => return pointer */
|
---|
263 | cmpb %dl, %ch /* is second byte NUL? */
|
---|
264 | je L(6) /* yes => return NULL? */
|
---|
265 | incl %eax /* it's not in the second byte */
|
---|
266 |
|
---|
267 | shrl $16, %ecx /* make upper byte accessible */
|
---|
268 | testb %cl, %cl /* is third byte CHR? */
|
---|
269 | jz L(6) /* yes => return pointer */
|
---|
270 | cmpb %dl, %cl /* is third byte NUL? */
|
---|
271 | je L(6) /* yes => return NULL */
|
---|
272 |
|
---|
273 | /* It must be in the fourth byte and it cannot be NUL. */
|
---|
274 | incl %eax
|
---|
275 |
|
---|
276 | L(6): CHECK_BOUNDS_HIGH (%eax, STR(%esp), jb)
|
---|
277 | RETURN_BOUNDED_POINTER (STR(%esp))
|
---|
278 | popl %edi /* restore saved register content */
|
---|
279 | cfi_adjust_cfa_offset (-4)
|
---|
280 | cfi_restore (edi)
|
---|
281 |
|
---|
282 | LEAVE
|
---|
283 | RET_PTR
|
---|
284 | END (BP_SYM (__strchrnul))
|
---|
285 |
|
---|
286 | weak_alias (BP_SYM (__strchrnul), BP_SYM (strchrnul))
|
---|