1/* strcpy/stpcpy implementation for x86-64.
2 Copyright (C) 2002-2024 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 <isa-level.h>
20
21#if ISA_SHOULD_BUILD (1)
22
23# include <sysdep.h>
24
25# ifndef STRCPY
26# define STRCPY __strcpy_sse2
27# endif
28
29 .text
30ENTRY (STRCPY)
31 movq %rsi, %rcx /* Source register. */
32 andl $7, %ecx /* mask alignment bits */
33 movq %rdi, %rdx /* Duplicate destination pointer. */
34
35 jz 5f /* aligned => start loop */
36
37 neg %ecx /* We need to align to 8 bytes. */
38 addl $8,%ecx
39 /* Search the first bytes directly. */
400:
41 movb (%rsi), %al /* Fetch a byte */
42 testb %al, %al /* Is it NUL? */
43 movb %al, (%rdx) /* Store it */
44 jz 4f /* If it was NUL, done! */
45 incq %rsi
46 incq %rdx
47 decl %ecx
48 jnz 0b
49
505:
51 movq $0xfefefefefefefeff,%r8
52
53 /* Now the sources is aligned. Unfortunately we cannot force
54 to have both source and destination aligned, so ignore the
55 alignment of the destination. */
56 .p2align 4
571:
58 /* 1st unroll. */
59 movq (%rsi), %rax /* Read double word (8 bytes). */
60 addq $8, %rsi /* Adjust pointer for next word. */
61 movq %rax, %r9 /* Save a copy for NUL finding. */
62 addq %r8, %r9 /* add the magic value to the word. We get
63 carry bits reported for each byte which
64 is *not* 0 */
65 jnc 3f /* highest byte is NUL => return pointer */
66 xorq %rax, %r9 /* (word+magic)^word */
67 orq %r8, %r9 /* set all non-carry bits */
68 incq %r9 /* add 1: if one carry bit was *not* set
69 the addition will not result in 0. */
70
71 jnz 3f /* found NUL => return pointer */
72
73 movq %rax, (%rdx) /* Write value to destination. */
74 addq $8, %rdx /* Adjust pointer. */
75
76 /* 2nd unroll. */
77 movq (%rsi), %rax /* Read double word (8 bytes). */
78 addq $8, %rsi /* Adjust pointer for next word. */
79 movq %rax, %r9 /* Save a copy for NUL finding. */
80 addq %r8, %r9 /* add the magic value to the word. We get
81 carry bits reported for each byte which
82 is *not* 0 */
83 jnc 3f /* highest byte is NUL => return pointer */
84 xorq %rax, %r9 /* (word+magic)^word */
85 orq %r8, %r9 /* set all non-carry bits */
86 incq %r9 /* add 1: if one carry bit was *not* set
87 the addition will not result in 0. */
88
89 jnz 3f /* found NUL => return pointer */
90
91 movq %rax, (%rdx) /* Write value to destination. */
92 addq $8, %rdx /* Adjust pointer. */
93
94 /* 3rd unroll. */
95 movq (%rsi), %rax /* Read double word (8 bytes). */
96 addq $8, %rsi /* Adjust pointer for next word. */
97 movq %rax, %r9 /* Save a copy for NUL finding. */
98 addq %r8, %r9 /* add the magic value to the word. We get
99 carry bits reported for each byte which
100 is *not* 0 */
101 jnc 3f /* highest byte is NUL => return pointer */
102 xorq %rax, %r9 /* (word+magic)^word */
103 orq %r8, %r9 /* set all non-carry bits */
104 incq %r9 /* add 1: if one carry bit was *not* set
105 the addition will not result in 0. */
106
107 jnz 3f /* found NUL => return pointer */
108
109 movq %rax, (%rdx) /* Write value to destination. */
110 addq $8, %rdx /* Adjust pointer. */
111
112 /* 4th unroll. */
113 movq (%rsi), %rax /* Read double word (8 bytes). */
114 addq $8, %rsi /* Adjust pointer for next word. */
115 movq %rax, %r9 /* Save a copy for NUL finding. */
116 addq %r8, %r9 /* add the magic value to the word. We get
117 carry bits reported for each byte which
118 is *not* 0 */
119 jnc 3f /* highest byte is NUL => return pointer */
120 xorq %rax, %r9 /* (word+magic)^word */
121 orq %r8, %r9 /* set all non-carry bits */
122 incq %r9 /* add 1: if one carry bit was *not* set
123 the addition will not result in 0. */
124
125 jnz 3f /* found NUL => return pointer */
126
127 movq %rax, (%rdx) /* Write value to destination. */
128 addq $8, %rdx /* Adjust pointer. */
129 jmp 1b /* Next iteration. */
130
131 /* Do the last few bytes. %rax contains the value to write.
132 The loop is unrolled twice. */
133 .p2align 4
1343:
135 /* Note that stpcpy needs to return with the value of the NUL
136 byte. */
137 movb %al, (%rdx) /* 1st byte. */
138 testb %al, %al /* Is it NUL. */
139 jz 4f /* yes, finish. */
140 incq %rdx /* Increment destination. */
141 movb %ah, (%rdx) /* 2nd byte. */
142 testb %ah, %ah /* Is it NUL?. */
143 jz 4f /* yes, finish. */
144 incq %rdx /* Increment destination. */
145 shrq $16, %rax /* Shift... */
146 jmp 3b /* and look at next two bytes in %rax. */
147
1484:
149# ifdef USE_AS_STPCPY
150 movq %rdx, %rax /* Destination is return value. */
151# else
152 movq %rdi, %rax /* Source is return value. */
153# endif
154 retq
155END (STRCPY)
156#endif
157

source code of glibc/sysdeps/x86_64/multiarch/strcpy-sse2.S