1 | /* rawmemchr -- find a byte within an unsized memory block. |
2 | Copyright (C) 2013-2024 Free Software Foundation, Inc. |
3 | This file is part of the GNU C Library. |
4 | |
5 | The GNU C Library is free software; you can redistribute it and/or |
6 | modify it under the terms of the GNU Lesser General Public |
7 | License as published by the Free Software Foundation; either |
8 | version 2.1 of the License, or (at your option) any later version. |
9 | |
10 | The GNU C Library is distributed in the hope that it will be useful, |
11 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
12 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
13 | Lesser General Public License for more details. |
14 | |
15 | You should have received a copy of the GNU Lesser General Public |
16 | License along with the GNU C Library. If not, see |
17 | <https://www.gnu.org/licenses/>. */ |
18 | |
19 | #include <sysdep.h> |
20 | |
21 | .syntax unified |
22 | .text |
23 | |
24 | ENTRY (__rawmemchr) |
25 | @ r0 = start of string |
26 | @ r1 = character to match |
27 | @ returns a pointer to the match, which must be present. |
28 | ldrb r2, [r0] @ load first byte asap |
29 | |
30 | @ To cater to long strings, we want to search through a few |
31 | @ characters until we reach an aligned pointer. To cater to |
32 | @ small strings, we don't want to start doing word operations |
33 | @ immediately. The compromise is a maximum of 16 bytes less |
34 | @ whatever is required to end with an aligned pointer. |
35 | @ r3 = number of characters to search in alignment loop |
36 | and r3, r0, #7 |
37 | uxtb r1, r1 |
38 | rsb r3, r3, #15 @ 16 - 1 peeled loop iteration |
39 | cmp r2, r1 |
40 | it eq |
41 | bxeq lr |
42 | |
43 | @ Loop until we find ... |
44 | 1: ldrb r2, [r0, #1]! |
45 | subs r3, r3, #1 @ ... the alignment point |
46 | it ne |
47 | cmpne r2, r1 @ ... or C |
48 | bne 1b |
49 | |
50 | @ Disambiguate the exit possibilities above |
51 | cmp r2, r1 @ Found C |
52 | it eq |
53 | bxeq lr |
54 | add r0, r0, #1 |
55 | |
56 | @ So now we're aligned. |
57 | ldrd r2, r3, [r0], #8 |
58 | orr r1, r1, r1, lsl #8 @ Replicate C to all bytes |
59 | #ifdef ARCH_HAS_T2 |
60 | movw ip, #0x0101 |
61 | pld [r0, #64] |
62 | movt ip, #0x0101 |
63 | #else |
64 | ldr ip, =0x01010101 |
65 | pld [r0, #64] |
66 | #endif |
67 | orr r1, r1, r1, lsl #16 |
68 | |
69 | @ Loop searching for C, 8 bytes at a time. |
70 | @ Subtracting (unsigned saturating) from 1 means result of 1 for |
71 | @ any byte that was originally zero and 0 otherwise. Therefore |
72 | @ we consider the lsb of each byte the "found" bit. |
73 | 2: eor r2, r2, r1 @ Convert C bytes to 0 |
74 | eor r3, r3, r1 |
75 | uqsub8 r2, ip, r2 @ Find C |
76 | uqsub8 r3, ip, r3 |
77 | pld [r0, #128] |
78 | orrs r3, r3, r2 @ Test both words for found |
79 | it eq |
80 | ldrdeq r2, r3, [r0], #8 |
81 | beq 2b |
82 | |
83 | @ Found something. Disambiguate between first and second words. |
84 | @ Adjust r0 to point to the word containing the match. |
85 | @ Adjust r2 to the found bits for the word containing the match. |
86 | cmp r2, #0 |
87 | sub r0, r0, #4 |
88 | ite eq |
89 | moveq r2, r3 |
90 | subne r0, r0, #4 |
91 | |
92 | @ Find the bit-offset of the match within the word. Note that the |
93 | @ bit result from clz will be 7 higher than "true" , but we'll |
94 | @ immediately discard those bits converting to a byte offset. |
95 | #ifdef __ARMEL__ |
96 | rev r2, r2 @ For LE, count from the little end |
97 | #endif |
98 | clz r2, r2 |
99 | add r0, r0, r2, lsr #3 @ Adjust the pointer to the found byte |
100 | bx lr |
101 | |
102 | END (__rawmemchr) |
103 | |
104 | weak_alias (__rawmemchr, rawmemchr) |
105 | libc_hidden_def (__rawmemchr) |
106 | |