initial commit
[glibc.git] / sysdeps / aarch64 / memrchr.S
1 /* memrchr - find the last occurrence of a byte in a memory block
2
3 Copyright (C) 2015-2022 Free Software Foundation, Inc.
4
5 This file is part of the GNU C Library.
6
7 The GNU C Library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Lesser General Public
9 License as published by the Free Software Foundation; either
10 version 2.1 of the License, or (at your option) any later version.
11
12 The GNU C Library is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Lesser General Public License for more details.
16
17 You should have received a copy of the GNU Lesser General Public
18 License along with the GNU C Library. If not, see
19 <https://www.gnu.org/licenses/>. */
20
21 #include <sysdep.h>
22
23 /* Assumptions:
24 *
25 * ARMv8-a, AArch64, Advanced SIMD.
26 * MTE compatible.
27 */
28
29 /* Arguments and results. */
30 #define srcin x0
31 #define chrin w1
32 #define cntin x2
33 #define result x0
34
35 #define src x3
36 #define cntrem x4
37 #define synd x5
38 #define shift x6
39 #define tmp x7
40 #define end x8
41 #define endm1 x9
42
43 #define vrepchr v0
44 #define qdata q1
45 #define vdata v1
46 #define vhas_chr v2
47 #define vend v3
48 #define dend d3
49
50 /*
51 Core algorithm:
52 For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
53 per byte. We take 4 bits of every comparison byte with shift right and narrow
54 by 4 instruction. Since the bits in the nibble mask reflect the order in
55 which things occur in the original string, counting leading zeros identifies
56 exactly which byte matched. */
57
58 ENTRY (__memrchr)
59 PTR_ARG (0)
60 SIZE_ARG (2)
61 add end, srcin, cntin
62 sub endm1, end, 1
63 bic src, endm1, 15
64 cbz cntin, L(nomatch)
65 ld1 {vdata.16b}, [src]
66 dup vrepchr.16b, chrin
67 cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
68 neg shift, end, lsl 2
69 shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
70 fmov synd, dend
71 lsl synd, synd, shift
72 cbz synd, L(start_loop)
73
74 clz synd, synd
75 sub result, endm1, synd, lsr 2
76 cmp cntin, synd, lsr 2
77 csel result, result, xzr, hi
78 ret
79
80 L(start_loop):
81 sub tmp, end, src
82 subs cntrem, cntin, tmp
83 b.ls L(nomatch)
84
85 /* Make sure that it won't overread by a 16-byte chunk */
86 add tmp, cntrem, 15
87 tbnz tmp, 4, L(loop32_2)
88
89 .p2align 4
90 L(loop32):
91 ldr qdata, [src, -16]!
92 cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
93 umaxp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
94 fmov synd, dend
95 cbnz synd, L(end)
96
97 L(loop32_2):
98 ldr qdata, [src, -16]!
99 subs cntrem, cntrem, 32
100 cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
101 b.ls L(end)
102 umaxp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
103 fmov synd, dend
104 cbz synd, L(loop32)
105 L(end):
106 shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
107 fmov synd, dend
108
109 add tmp, src, 15
110 #ifdef __AARCH64EB__
111 rbit synd, synd
112 #endif
113 clz synd, synd
114 sub tmp, tmp, synd, lsr 2
115 cmp tmp, srcin
116 csel result, tmp, xzr, hs
117 ret
118
119 L(nomatch):
120 mov result, 0
121 ret
122
123 END (__memrchr)
124 weak_alias (__memrchr, memrchr)
125 libc_hidden_builtin_def (memrchr)