Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
freebsd
GitHub Repository: freebsd/freebsd-src
Path: blob/main/contrib/arm-optimized-routines/string/aarch64/strchrnul-mte.S
39498 views
1
/*
2
* strchrnul - find a character or nul in a string
3
*
4
* Copyright (c) 2020-2022, Arm Limited.
5
* SPDX-License-Identifier: MIT OR Apache-2.0 WITH LLVM-exception
6
*/
7
8
/* Assumptions:
9
*
10
* ARMv8-a, AArch64, Advanced SIMD.
11
* MTE compatible.
12
*/
13
14
#include "asmdefs.h"
15
16
#define srcin x0
17
#define chrin w1
18
#define result x0
19
20
#define src x2
21
#define tmp1 x1
22
#define tmp2 x3
23
24
#define vrepchr v0
25
#define vdata v1
26
#define qdata q1
27
#define vhas_nul v2
28
#define vhas_chr v3
29
#define vend v4
30
#define dend d4
31
32
/*
33
Core algorithm:
34
For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
35
per byte. We take 4 bits of every comparison byte with shift right and narrow
36
by 4 instruction. Since the bits in the nibble mask reflect the order in
37
which things occur in the original string, counting leading zeros identifies
38
exactly which byte matched. */
39
40
ENTRY (__strchrnul_aarch64_mte)
41
bic src, srcin, 15
42
dup vrepchr.16b, chrin
43
ld1 {vdata.16b}, [src]
44
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
45
cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b
46
lsl tmp2, srcin, 2
47
shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
48
fmov tmp1, dend
49
lsr tmp1, tmp1, tmp2 /* Mask padding bits. */
50
cbz tmp1, L(loop)
51
52
rbit tmp1, tmp1
53
clz tmp1, tmp1
54
add result, srcin, tmp1, lsr 2
55
ret
56
57
.p2align 4
58
L(loop):
59
ldr qdata, [src, 16]
60
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
61
cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b
62
umaxp vend.16b, vhas_chr.16b, vhas_chr.16b
63
fmov tmp1, dend
64
cbnz tmp1, L(end)
65
ldr qdata, [src, 32]!
66
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
67
cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b
68
umaxp vend.16b, vhas_chr.16b, vhas_chr.16b
69
fmov tmp1, dend
70
cbz tmp1, L(loop)
71
sub src, src, 16
72
L(end):
73
shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
74
add src, src, 16
75
fmov tmp1, dend
76
#ifndef __AARCH64EB__
77
rbit tmp1, tmp1
78
#endif
79
clz tmp1, tmp1
80
add result, src, tmp1, lsr 2
81
ret
82
83
END (__strchrnul_aarch64_mte)
84
85
86