xref: /openbmc/linux/arch/x86/lib/memcpy_32.c (revision 75016ca3)
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 	return __memcpy(to, from, n);
11 }
12 EXPORT_SYMBOL(memcpy);
13 
14 __visible void *memset(void *s, int c, size_t count)
15 {
16 	return __memset(s, c, count);
17 }
18 EXPORT_SYMBOL(memset);
19 
20 __visible void *memmove(void *dest, const void *src, size_t n)
21 {
22 	int d0,d1,d2,d3,d4,d5;
23 	char *ret = dest;
24 
25 	__asm__ __volatile__(
26 		/* Handle more 16 bytes in loop */
27 		"cmp $0x10, %0\n\t"
28 		"jb	1f\n\t"
29 
30 		/* Decide forward/backward copy mode */
31 		"cmp %2, %1\n\t"
32 		"jb	2f\n\t"
33 
34 		/*
35 		 * movs instruction have many startup latency
36 		 * so we handle small size by general register.
37 		 */
38 		"cmp  $680, %0\n\t"
39 		"jb 3f\n\t"
40 		/*
41 		 * movs instruction is only good for aligned case.
42 		 */
43 		"mov %1, %3\n\t"
44 		"xor %2, %3\n\t"
45 		"and $0xff, %3\n\t"
46 		"jz 4f\n\t"
47 		"3:\n\t"
48 		"sub $0x10, %0\n\t"
49 
50 		/*
51 		 * We gobble 16 bytes forward in each loop.
52 		 */
53 		"3:\n\t"
54 		"sub $0x10, %0\n\t"
55 		"mov 0*4(%1), %3\n\t"
56 		"mov 1*4(%1), %4\n\t"
57 		"mov  %3, 0*4(%2)\n\t"
58 		"mov  %4, 1*4(%2)\n\t"
59 		"mov 2*4(%1), %3\n\t"
60 		"mov 3*4(%1), %4\n\t"
61 		"mov  %3, 2*4(%2)\n\t"
62 		"mov  %4, 3*4(%2)\n\t"
63 		"lea  0x10(%1), %1\n\t"
64 		"lea  0x10(%2), %2\n\t"
65 		"jae 3b\n\t"
66 		"add $0x10, %0\n\t"
67 		"jmp 1f\n\t"
68 
69 		/*
70 		 * Handle data forward by movs.
71 		 */
72 		".p2align 4\n\t"
73 		"4:\n\t"
74 		"mov -4(%1, %0), %3\n\t"
75 		"lea -4(%2, %0), %4\n\t"
76 		"shr $2, %0\n\t"
77 		"rep movsl\n\t"
78 		"mov %3, (%4)\n\t"
79 		"jmp 11f\n\t"
80 		/*
81 		 * Handle data backward by movs.
82 		 */
83 		".p2align 4\n\t"
84 		"6:\n\t"
85 		"mov (%1), %3\n\t"
86 		"mov %2, %4\n\t"
87 		"lea -4(%1, %0), %1\n\t"
88 		"lea -4(%2, %0), %2\n\t"
89 		"shr $2, %0\n\t"
90 		"std\n\t"
91 		"rep movsl\n\t"
92 		"mov %3,(%4)\n\t"
93 		"cld\n\t"
94 		"jmp 11f\n\t"
95 
96 		/*
97 		 * Start to prepare for backward copy.
98 		 */
99 		".p2align 4\n\t"
100 		"2:\n\t"
101 		"cmp  $680, %0\n\t"
102 		"jb 5f\n\t"
103 		"mov %1, %3\n\t"
104 		"xor %2, %3\n\t"
105 		"and $0xff, %3\n\t"
106 		"jz 6b\n\t"
107 
108 		/*
109 		 * Calculate copy position to tail.
110 		 */
111 		"5:\n\t"
112 		"add %0, %1\n\t"
113 		"add %0, %2\n\t"
114 		"sub $0x10, %0\n\t"
115 
116 		/*
117 		 * We gobble 16 bytes backward in each loop.
118 		 */
119 		"7:\n\t"
120 		"sub $0x10, %0\n\t"
121 
122 		"mov -1*4(%1), %3\n\t"
123 		"mov -2*4(%1), %4\n\t"
124 		"mov  %3, -1*4(%2)\n\t"
125 		"mov  %4, -2*4(%2)\n\t"
126 		"mov -3*4(%1), %3\n\t"
127 		"mov -4*4(%1), %4\n\t"
128 		"mov  %3, -3*4(%2)\n\t"
129 		"mov  %4, -4*4(%2)\n\t"
130 		"lea  -0x10(%1), %1\n\t"
131 		"lea  -0x10(%2), %2\n\t"
132 		"jae 7b\n\t"
133 		/*
134 		 * Calculate copy position to head.
135 		 */
136 		"add $0x10, %0\n\t"
137 		"sub %0, %1\n\t"
138 		"sub %0, %2\n\t"
139 
140 		/*
141 		 * Move data from 8 bytes to 15 bytes.
142 		 */
143 		".p2align 4\n\t"
144 		"1:\n\t"
145 		"cmp $8, %0\n\t"
146 		"jb 8f\n\t"
147 		"mov 0*4(%1), %3\n\t"
148 		"mov 1*4(%1), %4\n\t"
149 		"mov -2*4(%1, %0), %5\n\t"
150 		"mov -1*4(%1, %0), %1\n\t"
151 
152 		"mov  %3, 0*4(%2)\n\t"
153 		"mov  %4, 1*4(%2)\n\t"
154 		"mov  %5, -2*4(%2, %0)\n\t"
155 		"mov  %1, -1*4(%2, %0)\n\t"
156 		"jmp 11f\n\t"
157 
158 		/*
159 		 * Move data from 4 bytes to 7 bytes.
160 		 */
161 		".p2align 4\n\t"
162 		"8:\n\t"
163 		"cmp $4, %0\n\t"
164 		"jb 9f\n\t"
165 		"mov 0*4(%1), %3\n\t"
166 		"mov -1*4(%1, %0), %4\n\t"
167 		"mov  %3, 0*4(%2)\n\t"
168 		"mov  %4, -1*4(%2, %0)\n\t"
169 		"jmp 11f\n\t"
170 
171 		/*
172 		 * Move data from 2 bytes to 3 bytes.
173 		 */
174 		".p2align 4\n\t"
175 		"9:\n\t"
176 		"cmp $2, %0\n\t"
177 		"jb 10f\n\t"
178 		"movw 0*2(%1), %%dx\n\t"
179 		"movw -1*2(%1, %0), %%bx\n\t"
180 		"movw %%dx, 0*2(%2)\n\t"
181 		"movw %%bx, -1*2(%2, %0)\n\t"
182 		"jmp 11f\n\t"
183 
184 		/*
185 		 * Move data for 1 byte.
186 		 */
187 		".p2align 4\n\t"
188 		"10:\n\t"
189 		"cmp $1, %0\n\t"
190 		"jb 11f\n\t"
191 		"movb (%1), %%cl\n\t"
192 		"movb %%cl, (%2)\n\t"
193 		".p2align 4\n\t"
194 		"11:"
195 		: "=&c" (d0), "=&S" (d1), "=&D" (d2),
196 		  "=r" (d3),"=r" (d4), "=r"(d5)
197 		:"0" (n),
198 		 "1" (src),
199 		 "2" (dest)
200 		:"memory");
201 
202 	return ret;
203 
204 }
205 EXPORT_SYMBOL(memmove);
206