summaryrefslogtreecommitdiffstats
path: root/arch/x86/lib/memmove_64.S
blob: 0ecb8433e5a8e2f62c0a678da69ccd16f233b304 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
/*
 * Normally compiler builtins are used, but sometimes the compiler calls out
 * of line code. Based on asm-i386/string.h.
 *
 * This assembly file is re-written from memmove_64.c file.
 *	- Copyright 2011 Fenghua Yu <fenghua.yu@intel.com>
 */
#define _STRING_C
#include <linux/linkage.h>
#include <asm/dwarf2.h>

#undef memmove

/*
 * Implement memmove(). This can handle overlap between src and dst.
 *
 * Input:
 * rdi: dest
 * rsi: src
 * rdx: count
 *
 * Output:
 * rax: dest
 */
ENTRY(memmove)
	CFI_STARTPROC
	/* Handle more 32bytes in loop */
	mov %rdi, %rax
	cmp $0x20, %rdx
	jb	1f

	/* Decide forward/backward copy mode */
	cmp %rdi, %rsi
	jb	2f

	/*
	 * movsq instruction have many startup latency
	 * so we handle small size by general register.
	 */
	cmp  $680, %rdx
	jb	3f
	/*
	 * movsq instruction is only good for aligned case.
	 */

	cmpb %dil, %sil
	je 4f
3:
	sub $0x20, %rdx
	/*
	 * We gobble 32byts forward in each loop.
	 */
5:
	sub $0x20, %rdx
	movq 0*8(%rsi), %r11
	movq 1*8(%rsi), %r10
	movq 2*8(%rsi), %r9
	movq 3*8(%rsi), %r8
	leaq 4*8(%rsi), %rsi

	movq %r11, 0*8(%rdi)
	movq %r10, 1*8(%rdi)
	movq %r9, 2*8(%rdi)
	movq %r8, 3*8(%rdi)
	leaq 4*8(%rdi), %rdi
	jae 5b
	addq $0x20, %rdx
	jmp 1f
	/*
	 * Handle data forward by movsq.
	 */
	.p2align 4
4:
	movq %rdx, %rcx
	movq -8(%rsi, %rdx), %r11
	lea -8(%rdi, %rdx), %r10
	shrq $3, %rcx
	rep movsq
	movq %r11, (%r10)
	jmp 13f
	/*
	 * Handle data backward by movsq.
	 */
	.p2align 4
7:
	movq %rdx, %rcx
	movq (%rsi), %r11
	movq %rdi, %r10
	leaq -8(%rsi, %rdx), %rsi
	leaq -8(%rdi, %rdx), %rdi
	shrq $3, %rcx
	std
	rep movsq
	cld
	movq %r11, (%r10)
	jmp 13f

	/*
	 * Start to prepare for backward copy.
	 */
	.p2align 4
2:
	cmp $680, %rdx
	jb 6f
	cmp %dil, %sil
	je 7b
6:
	/*
	 * Calculate copy position to tail.
	 */
	addq %rdx, %rsi
	addq %rdx, %rdi
	subq $0x20, %rdx
	/*
	 * We gobble 32byts backward in each loop.
	 */
8:
	subq $0x20, %rdx
	movq -1*8(%rsi), %r11
	movq -2*8(%rsi), %r10
	movq -3*8(%rsi), %r9
	movq -4*8(%rsi), %r8
	leaq -4*8(%rsi), %rsi

	movq %r11, -1*8(%rdi)
	movq %r10, -2*8(%rdi)
	movq %r9, -3*8(%rdi)
	movq %r8, -4*8(%rdi)
	leaq -4*8(%rdi), %rdi
	jae 8b
	/*
	 * Calculate copy position to head.
	 */
	addq $0x20, %rdx
	subq %rdx, %rsi
	subq %rdx, %rdi
1:
	cmpq $16, %rdx
	jb 9f
	/*
	 * Move data from 16 bytes to 31 bytes.
	 */
	movq 0*8(%rsi), %r11
	movq 1*8(%rsi), %r10
	movq -2*8(%rsi, %rdx), %r9
	movq -1*8(%rsi, %rdx), %r8
	movq %r11, 0*8(%rdi)
	movq %r10, 1*8(%rdi)
	movq %r9, -2*8(%rdi, %rdx)
	movq %r8, -1*8(%rdi, %rdx)
	jmp 13f
	.p2align 4
9:
	cmpq $8, %rdx
	jb 10f
	/*
	 * Move data from 8 bytes to 15 bytes.
	 */
	movq 0*8(%rsi), %r11
	movq -1*8(%rsi, %rdx), %r10
	movq %r11, 0*8(%rdi)
	movq %r10, -1*8(%rdi, %rdx)
	jmp 13f
10:
	cmpq $4, %rdx
	jb 11f
	/*
	 * Move data from 4 bytes to 7 bytes.
	 */
	movl (%rsi), %r11d
	movl -4(%rsi, %rdx), %r10d
	movl %r11d, (%rdi)
	movl %r10d, -4(%rdi, %rdx)
	jmp 13f
11:
	cmp $2, %rdx
	jb 12f
	/*
	 * Move data from 2 bytes to 3 bytes.
	 */
	movw (%rsi), %r11w
	movw -2(%rsi, %rdx), %r10w
	movw %r11w, (%rdi)
	movw %r10w, -2(%rdi, %rdx)
	jmp 13f
12:
	cmp $1, %rdx
	jb 13f
	/*
	 * Move data for 1 byte.
	 */
	movb (%rsi), %r11b
	movb %r11b, (%rdi)
13:
	retq
	CFI_ENDPROC
ENDPROC(memmove)
OpenPOWER on IntegriCloud