xref: /linux/arch/arm/lib/memmove.S (revision 6fdcba32711044c35c0e1b094cbd8f3f0b4472c9)
1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 *  linux/arch/arm/lib/memmove.S
4 *
5 *  Author:	Nicolas Pitre
6 *  Created:	Sep 28, 2005
7 *  Copyright:	(C) MontaVista Software Inc.
8 */
9
10#include <linux/linkage.h>
11#include <asm/assembler.h>
12#include <asm/unwind.h>
13
14		.text
15
16/*
17 * Prototype: void *memmove(void *dest, const void *src, size_t n);
18 *
19 * Note:
20 *
21 * If the memory regions don't overlap, we simply branch to memcpy which is
22 * normally a bit faster. Otherwise the copy is done going downwards.  This
23 * is a transposition of the code from copy_template.S but with the copy
24 * occurring in the opposite direction.
25 */
26
27ENTRY(memmove)
28	UNWIND(	.fnstart			)
29
30		subs	ip, r0, r1
31		cmphi	r2, ip
32		bls	memcpy
33
34		stmfd	sp!, {r0, r4, lr}
35	UNWIND(	.fnend				)
36
37	UNWIND(	.fnstart			)
38	UNWIND(	.save	{r0, r4, lr}		) @ in first stmfd block
39		add	r1, r1, r2
40		add	r0, r0, r2
41		subs	r2, r2, #4
42		blt	8f
43		ands	ip, r0, #3
44	PLD(	pld	[r1, #-4]		)
45		bne	9f
46		ands	ip, r1, #3
47		bne	10f
48
491:		subs	r2, r2, #(28)
50		stmfd	sp!, {r5 - r8}
51	UNWIND(	.fnend				)
52
53	UNWIND(	.fnstart			)
54	UNWIND(	.save	{r0, r4, lr}		)
55	UNWIND(	.save	{r5 - r8}		) @ in second stmfd block
56		blt	5f
57
58	CALGN(	ands	ip, r0, #31		)
59	CALGN(	sbcsne	r4, ip, r2		)  @ C is always set here
60	CALGN(	bcs	2f			)
61	CALGN(	adr	r4, 6f			)
62	CALGN(	subs	r2, r2, ip		)  @ C is set here
63	CALGN(	rsb	ip, ip, #32		)
64	CALGN(	add	pc, r4, ip		)
65
66	PLD(	pld	[r1, #-4]		)
672:	PLD(	subs	r2, r2, #96		)
68	PLD(	pld	[r1, #-32]		)
69	PLD(	blt	4f			)
70	PLD(	pld	[r1, #-64]		)
71	PLD(	pld	[r1, #-96]		)
72
733:	PLD(	pld	[r1, #-128]		)
744:		ldmdb	r1!, {r3, r4, r5, r6, r7, r8, ip, lr}
75		subs	r2, r2, #32
76		stmdb	r0!, {r3, r4, r5, r6, r7, r8, ip, lr}
77		bge	3b
78	PLD(	cmn	r2, #96			)
79	PLD(	bge	4b			)
80
815:		ands	ip, r2, #28
82		rsb	ip, ip, #32
83		addne	pc, pc, ip		@ C is always clear here
84		b	7f
856:		W(nop)
86		W(ldr)	r3, [r1, #-4]!
87		W(ldr)	r4, [r1, #-4]!
88		W(ldr)	r5, [r1, #-4]!
89		W(ldr)	r6, [r1, #-4]!
90		W(ldr)	r7, [r1, #-4]!
91		W(ldr)	r8, [r1, #-4]!
92		W(ldr)	lr, [r1, #-4]!
93
94		add	pc, pc, ip
95		nop
96		W(nop)
97		W(str)	r3, [r0, #-4]!
98		W(str)	r4, [r0, #-4]!
99		W(str)	r5, [r0, #-4]!
100		W(str)	r6, [r0, #-4]!
101		W(str)	r7, [r0, #-4]!
102		W(str)	r8, [r0, #-4]!
103		W(str)	lr, [r0, #-4]!
104
105	CALGN(	bcs	2b			)
106
1077:		ldmfd	sp!, {r5 - r8}
108	UNWIND(	.fnend				) @ end of second stmfd block
109
110	UNWIND(	.fnstart			)
111	UNWIND(	.save	{r0, r4, lr}		) @ still in first stmfd block
112
1138:		movs	r2, r2, lsl #31
114		ldrbne	r3, [r1, #-1]!
115		ldrbcs	r4, [r1, #-1]!
116		ldrbcs	ip, [r1, #-1]
117		strbne	r3, [r0, #-1]!
118		strbcs	r4, [r0, #-1]!
119		strbcs	ip, [r0, #-1]
120		ldmfd	sp!, {r0, r4, pc}
121
1229:		cmp	ip, #2
123		ldrbgt	r3, [r1, #-1]!
124		ldrbge	r4, [r1, #-1]!
125		ldrb	lr, [r1, #-1]!
126		strbgt	r3, [r0, #-1]!
127		strbge	r4, [r0, #-1]!
128		subs	r2, r2, ip
129		strb	lr, [r0, #-1]!
130		blt	8b
131		ands	ip, r1, #3
132		beq	1b
133
13410:		bic	r1, r1, #3
135		cmp	ip, #2
136		ldr	r3, [r1, #0]
137		beq	17f
138		blt	18f
139	UNWIND(	.fnend				)
140
141
142		.macro	backward_copy_shift push pull
143
144	UNWIND(	.fnstart			)
145	UNWIND(	.save	{r0, r4, lr}		) @ still in first stmfd block
146		subs	r2, r2, #28
147		blt	14f
148
149	CALGN(	ands	ip, r0, #31		)
150	CALGN(	sbcsne	r4, ip, r2		)  @ C is always set here
151	CALGN(	subcc	r2, r2, ip		)
152	CALGN(	bcc	15f			)
153
15411:		stmfd	sp!, {r5 - r9}
155	UNWIND(	.fnend				)
156
157	UNWIND(	.fnstart			)
158	UNWIND(	.save	{r0, r4, lr}		)
159	UNWIND(	.save	{r5 - r9}		) @ in new second stmfd block
160
161	PLD(	pld	[r1, #-4]		)
162	PLD(	subs	r2, r2, #96		)
163	PLD(	pld	[r1, #-32]		)
164	PLD(	blt	13f			)
165	PLD(	pld	[r1, #-64]		)
166	PLD(	pld	[r1, #-96]		)
167
16812:	PLD(	pld	[r1, #-128]		)
16913:		ldmdb   r1!, {r7, r8, r9, ip}
170		mov     lr, r3, lspush #\push
171		subs    r2, r2, #32
172		ldmdb   r1!, {r3, r4, r5, r6}
173		orr     lr, lr, ip, lspull #\pull
174		mov     ip, ip, lspush #\push
175		orr     ip, ip, r9, lspull #\pull
176		mov     r9, r9, lspush #\push
177		orr     r9, r9, r8, lspull #\pull
178		mov     r8, r8, lspush #\push
179		orr     r8, r8, r7, lspull #\pull
180		mov     r7, r7, lspush #\push
181		orr     r7, r7, r6, lspull #\pull
182		mov     r6, r6, lspush #\push
183		orr     r6, r6, r5, lspull #\pull
184		mov     r5, r5, lspush #\push
185		orr     r5, r5, r4, lspull #\pull
186		mov     r4, r4, lspush #\push
187		orr     r4, r4, r3, lspull #\pull
188		stmdb   r0!, {r4 - r9, ip, lr}
189		bge	12b
190	PLD(	cmn	r2, #96			)
191	PLD(	bge	13b			)
192
193		ldmfd	sp!, {r5 - r9}
194	UNWIND(	.fnend				) @ end of the second stmfd block
195
196	UNWIND(	.fnstart			)
197	UNWIND(	.save {r0, r4, lr}		) @ still in first stmfd block
198
19914:		ands	ip, r2, #28
200		beq	16f
201
20215:		mov     lr, r3, lspush #\push
203		ldr	r3, [r1, #-4]!
204		subs	ip, ip, #4
205		orr	lr, lr, r3, lspull #\pull
206		str	lr, [r0, #-4]!
207		bgt	15b
208	CALGN(	cmp	r2, #0			)
209	CALGN(	bge	11b			)
210
21116:		add	r1, r1, #(\pull / 8)
212		b	8b
213	UNWIND(	.fnend				)
214
215		.endm
216
217
218		backward_copy_shift	push=8	pull=24
219
22017:		backward_copy_shift	push=16	pull=16
221
22218:		backward_copy_shift	push=24	pull=8
223
224ENDPROC(memmove)
225