xref: /kernel/linux/linux-5.10/arch/x86/lib/memcpy_32.c (revision 8c2ecf20)
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}
16EXPORT_SYMBOL(memcpy);
17
18__visible void *memset(void *s, int c, size_t count)
19{
20	return __memset(s, c, count);
21}
22EXPORT_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}
209EXPORT_SYMBOL(memmove);
210