1/* strcpy/stpcpy implementation for x86-64.
2 Copyright (C) 2002-2020 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Andreas Jaeger <aj@suse.de>, 2002.
5
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
10
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <https://www.gnu.org/licenses/>. */
19
20#include <sysdep.h>
21#include "asm-syntax.h"
22
23#ifndef USE_AS_STPCPY
24# define STRCPY strcpy
25#endif
26
27 .text
28ENTRY (STRCPY)
29 movq %rsi, %rcx /* Source register. */
30 andl $7, %ecx /* mask alignment bits */
31 movq %rdi, %rdx /* Duplicate destination pointer. */
32
33 jz 5f /* aligned => start loop */
34
35 neg %ecx /* We need to align to 8 bytes. */
36 addl $8,%ecx
37 /* Search the first bytes directly. */
380:
39 movb (%rsi), %al /* Fetch a byte */
40 testb %al, %al /* Is it NUL? */
41 movb %al, (%rdx) /* Store it */
42 jz 4f /* If it was NUL, done! */
43 incq %rsi
44 incq %rdx
45 decl %ecx
46 jnz 0b
47
485:
49 movq $0xfefefefefefefeff,%r8
50
51 /* Now the sources is aligned. Unfortunatly we cannot force
52 to have both source and destination aligned, so ignore the
53 alignment of the destination. */
54 .p2align 4
551:
56 /* 1st unroll. */
57 movq (%rsi), %rax /* Read double word (8 bytes). */
58 addq $8, %rsi /* Adjust pointer for next word. */
59 movq %rax, %r9 /* Save a copy for NUL finding. */
60 addq %r8, %r9 /* add the magic value to the word. We get
61 carry bits reported for each byte which
62 is *not* 0 */
63 jnc 3f /* highest byte is NUL => return pointer */
64 xorq %rax, %r9 /* (word+magic)^word */
65 orq %r8, %r9 /* set all non-carry bits */
66 incq %r9 /* add 1: if one carry bit was *not* set
67 the addition will not result in 0. */
68
69 jnz 3f /* found NUL => return pointer */
70
71 movq %rax, (%rdx) /* Write value to destination. */
72 addq $8, %rdx /* Adjust pointer. */
73
74 /* 2nd unroll. */
75 movq (%rsi), %rax /* Read double word (8 bytes). */
76 addq $8, %rsi /* Adjust pointer for next word. */
77 movq %rax, %r9 /* Save a copy for NUL finding. */
78 addq %r8, %r9 /* add the magic value to the word. We get
79 carry bits reported for each byte which
80 is *not* 0 */
81 jnc 3f /* highest byte is NUL => return pointer */
82 xorq %rax, %r9 /* (word+magic)^word */
83 orq %r8, %r9 /* set all non-carry bits */
84 incq %r9 /* add 1: if one carry bit was *not* set
85 the addition will not result in 0. */
86
87 jnz 3f /* found NUL => return pointer */
88
89 movq %rax, (%rdx) /* Write value to destination. */
90 addq $8, %rdx /* Adjust pointer. */
91
92 /* 3rd unroll. */
93 movq (%rsi), %rax /* Read double word (8 bytes). */
94 addq $8, %rsi /* Adjust pointer for next word. */
95 movq %rax, %r9 /* Save a copy for NUL finding. */
96 addq %r8, %r9 /* add the magic value to the word. We get
97 carry bits reported for each byte which
98 is *not* 0 */
99 jnc 3f /* highest byte is NUL => return pointer */
100 xorq %rax, %r9 /* (word+magic)^word */
101 orq %r8, %r9 /* set all non-carry bits */
102 incq %r9 /* add 1: if one carry bit was *not* set
103 the addition will not result in 0. */
104
105 jnz 3f /* found NUL => return pointer */
106
107 movq %rax, (%rdx) /* Write value to destination. */
108 addq $8, %rdx /* Adjust pointer. */
109
110 /* 4th unroll. */
111 movq (%rsi), %rax /* Read double word (8 bytes). */
112 addq $8, %rsi /* Adjust pointer for next word. */
113 movq %rax, %r9 /* Save a copy for NUL finding. */
114 addq %r8, %r9 /* add the magic value to the word. We get
115 carry bits reported for each byte which
116 is *not* 0 */
117 jnc 3f /* highest byte is NUL => return pointer */
118 xorq %rax, %r9 /* (word+magic)^word */
119 orq %r8, %r9 /* set all non-carry bits */
120 incq %r9 /* add 1: if one carry bit was *not* set
121 the addition will not result in 0. */
122
123 jnz 3f /* found NUL => return pointer */
124
125 movq %rax, (%rdx) /* Write value to destination. */
126 addq $8, %rdx /* Adjust pointer. */
127 jmp 1b /* Next iteration. */
128
129 /* Do the last few bytes. %rax contains the value to write.
130 The loop is unrolled twice. */
131 .p2align 4
1323:
133 /* Note that stpcpy needs to return with the value of the NUL
134 byte. */
135 movb %al, (%rdx) /* 1st byte. */
136 testb %al, %al /* Is it NUL. */
137 jz 4f /* yes, finish. */
138 incq %rdx /* Increment destination. */
139 movb %ah, (%rdx) /* 2nd byte. */
140 testb %ah, %ah /* Is it NUL?. */
141 jz 4f /* yes, finish. */
142 incq %rdx /* Increment destination. */
143 shrq $16, %rax /* Shift... */
144 jmp 3b /* and look at next two bytes in %rax. */
145
1464:
147#ifdef USE_AS_STPCPY
148 movq %rdx, %rax /* Destination is return value. */
149#else
150 movq %rdi, %rax /* Source is return value. */
151#endif
152 retq
153END (STRCPY)
154#ifndef USE_AS_STPCPY
155libc_hidden_builtin_def (strcpy)
156#endif
157