xref: /openbmc/linux/arch/x86/lib/memcpy_32.c (revision fb574682)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/string.h>
3 #include <linux/export.h>
4 
5 #undef memcpy
6 #undef memset
7 
8 __visible void *memcpy(void *to, const void *from, size_t n)
9 {
10 #if defined(CONFIG_X86_USE_3DNOW) && !defined(CONFIG_FORTIFY_SOURCE)
11 	return __memcpy3d(to, from, n);
12 #else
13 	return __memcpy(to, from, n);
14 #endif
15 }
16 EXPORT_SYMBOL(memcpy);
17 
18 __visible void *memset(void *s, int c, size_t count)
19 {
20 	return __memset(s, c, count);
21 }
22 EXPORT_SYMBOL(memset);
23 
24 __visible void *memmove(void *dest, const void *src, size_t n)
25 {
26 	int d0,d1,d2,d3,d4,d5;
27 	char *ret = dest;
28 
29 	__asm__ __volatile__(
30 		/* Handle more 16 bytes in loop */
31 		"cmp $0x10, %0\n\t"
32 		"jb	1f\n\t"
33 
34 		/* Decide forward/backward copy mode */
35 		"cmp %2, %1\n\t"
36 		"jb	2f\n\t"
37 
38 		/*
39 		 * movs instruction have many startup latency
40 		 * so we handle small size by general register.
41 		 */
42 		"cmp  $680, %0\n\t"
43 		"jb 3f\n\t"
44 		/*
45 		 * movs instruction is only good for aligned case.
46 		 */
47 		"mov %1, %3\n\t"
48 		"xor %2, %3\n\t"
49 		"and $0xff, %3\n\t"
50 		"jz 4f\n\t"
51 		"3:\n\t"
52 		"sub $0x10, %0\n\t"
53 
54 		/*
55 		 * We gobble 16 bytes forward in each loop.
56 		 */
57 		"3:\n\t"
58 		"sub $0x10, %0\n\t"
59 		"mov 0*4(%1), %3\n\t"
60 		"mov 1*4(%1), %4\n\t"
61 		"mov  %3, 0*4(%2)\n\t"
62 		"mov  %4, 1*4(%2)\n\t"
63 		"mov 2*4(%1), %3\n\t"
64 		"mov 3*4(%1), %4\n\t"
65 		"mov  %3, 2*4(%2)\n\t"
66 		"mov  %4, 3*4(%2)\n\t"
67 		"lea  0x10(%1), %1\n\t"
68 		"lea  0x10(%2), %2\n\t"
69 		"jae 3b\n\t"
70 		"add $0x10, %0\n\t"
71 		"jmp 1f\n\t"
72 
73 		/*
74 		 * Handle data forward by movs.
75 		 */
76 		".p2align 4\n\t"
77 		"4:\n\t"
78 		"mov -4(%1, %0), %3\n\t"
79 		"lea -4(%2, %0), %4\n\t"
80 		"shr $2, %0\n\t"
81 		"rep movsl\n\t"
82 		"mov %3, (%4)\n\t"
83 		"jmp 11f\n\t"
84 		/*
85 		 * Handle data backward by movs.
86 		 */
87 		".p2align 4\n\t"
88 		"6:\n\t"
89 		"mov (%1), %3\n\t"
90 		"mov %2, %4\n\t"
91 		"lea -4(%1, %0), %1\n\t"
92 		"lea -4(%2, %0), %2\n\t"
93 		"shr $2, %0\n\t"
94 		"std\n\t"
95 		"rep movsl\n\t"
96 		"mov %3,(%4)\n\t"
97 		"cld\n\t"
98 		"jmp 11f\n\t"
99 
100 		/*
101 		 * Start to prepare for backward copy.
102 		 */
103 		".p2align 4\n\t"
104 		"2:\n\t"
105 		"cmp  $680, %0\n\t"
106 		"jb 5f\n\t"
107 		"mov %1, %3\n\t"
108 		"xor %2, %3\n\t"
109 		"and $0xff, %3\n\t"
110 		"jz 6b\n\t"
111 
112 		/*
113 		 * Calculate copy position to tail.
114 		 */
115 		"5:\n\t"
116 		"add %0, %1\n\t"
117 		"add %0, %2\n\t"
118 		"sub $0x10, %0\n\t"
119 
120 		/*
121 		 * We gobble 16 bytes backward in each loop.
122 		 */
123 		"7:\n\t"
124 		"sub $0x10, %0\n\t"
125 
126 		"mov -1*4(%1), %3\n\t"
127 		"mov -2*4(%1), %4\n\t"
128 		"mov  %3, -1*4(%2)\n\t"
129 		"mov  %4, -2*4(%2)\n\t"
130 		"mov -3*4(%1), %3\n\t"
131 		"mov -4*4(%1), %4\n\t"
132 		"mov  %3, -3*4(%2)\n\t"
133 		"mov  %4, -4*4(%2)\n\t"
134 		"lea  -0x10(%1), %1\n\t"
135 		"lea  -0x10(%2), %2\n\t"
136 		"jae 7b\n\t"
137 		/*
138 		 * Calculate copy position to head.
139 		 */
140 		"add $0x10, %0\n\t"
141 		"sub %0, %1\n\t"
142 		"sub %0, %2\n\t"
143 
144 		/*
145 		 * Move data from 8 bytes to 15 bytes.
146 		 */
147 		".p2align 4\n\t"
148 		"1:\n\t"
149 		"cmp $8, %0\n\t"
150 		"jb 8f\n\t"
151 		"mov 0*4(%1), %3\n\t"
152 		"mov 1*4(%1), %4\n\t"
153 		"mov -2*4(%1, %0), %5\n\t"
154 		"mov -1*4(%1, %0), %1\n\t"
155 
156 		"mov  %3, 0*4(%2)\n\t"
157 		"mov  %4, 1*4(%2)\n\t"
158 		"mov  %5, -2*4(%2, %0)\n\t"
159 		"mov  %1, -1*4(%2, %0)\n\t"
160 		"jmp 11f\n\t"
161 
162 		/*
163 		 * Move data from 4 bytes to 7 bytes.
164 		 */
165 		".p2align 4\n\t"
166 		"8:\n\t"
167 		"cmp $4, %0\n\t"
168 		"jb 9f\n\t"
169 		"mov 0*4(%1), %3\n\t"
170 		"mov -1*4(%1, %0), %4\n\t"
171 		"mov  %3, 0*4(%2)\n\t"
172 		"mov  %4, -1*4(%2, %0)\n\t"
173 		"jmp 11f\n\t"
174 
175 		/*
176 		 * Move data from 2 bytes to 3 bytes.
177 		 */
178 		".p2align 4\n\t"
179 		"9:\n\t"
180 		"cmp $2, %0\n\t"
181 		"jb 10f\n\t"
182 		"movw 0*2(%1), %%dx\n\t"
183 		"movw -1*2(%1, %0), %%bx\n\t"
184 		"movw %%dx, 0*2(%2)\n\t"
185 		"movw %%bx, -1*2(%2, %0)\n\t"
186 		"jmp 11f\n\t"
187 
188 		/*
189 		 * Move data for 1 byte.
190 		 */
191 		".p2align 4\n\t"
192 		"10:\n\t"
193 		"cmp $1, %0\n\t"
194 		"jb 11f\n\t"
195 		"movb (%1), %%cl\n\t"
196 		"movb %%cl, (%2)\n\t"
197 		".p2align 4\n\t"
198 		"11:"
199 		: "=&c" (d0), "=&S" (d1), "=&D" (d2),
200 		  "=r" (d3),"=r" (d4), "=r"(d5)
201 		:"0" (n),
202 		 "1" (src),
203 		 "2" (dest)
204 		:"memory");
205 
206 	return ret;
207 
208 }
209 EXPORT_SYMBOL(memmove);
210