xref: /linux/arch/arm/lib/memset.S (revision f3a8b6645dc2e60d11f20c1c23afd964ff4e55ae)
1/*
2 *  linux/arch/arm/lib/memset.S
3 *
4 *  Copyright (C) 1995-2000 Russell King
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 *
10 *  ASM optimised string functions
11 */
12#include <linux/linkage.h>
13#include <asm/assembler.h>
14#include <asm/unwind.h>
15#include <asm/export.h>
16
17	.text
18	.align	5
19
20ENTRY(mmioset)
21ENTRY(memset)
22UNWIND( .fnstart         )
23	ands	r3, r0, #3		@ 1 unaligned?
24	mov	ip, r0			@ preserve r0 as return value
25	bne	6f			@ 1
26/*
27 * we know that the pointer in ip is aligned to a word boundary.
28 */
291:	orr	r1, r1, r1, lsl #8
30	orr	r1, r1, r1, lsl #16
31	mov	r3, r1
32	cmp	r2, #16
33	blt	4f
34
35#if ! CALGN(1)+0
36
37/*
38 * We need 2 extra registers for this loop - use r8 and the LR
39 */
40	stmfd	sp!, {r8, lr}
41UNWIND( .fnend              )
42UNWIND( .fnstart            )
43UNWIND( .save {r8, lr}      )
44	mov	r8, r1
45	mov	lr, r1
46
472:	subs	r2, r2, #64
48	stmgeia	ip!, {r1, r3, r8, lr}	@ 64 bytes at a time.
49	stmgeia	ip!, {r1, r3, r8, lr}
50	stmgeia	ip!, {r1, r3, r8, lr}
51	stmgeia	ip!, {r1, r3, r8, lr}
52	bgt	2b
53	ldmeqfd	sp!, {r8, pc}		@ Now <64 bytes to go.
54/*
55 * No need to correct the count; we're only testing bits from now on
56 */
57	tst	r2, #32
58	stmneia	ip!, {r1, r3, r8, lr}
59	stmneia	ip!, {r1, r3, r8, lr}
60	tst	r2, #16
61	stmneia	ip!, {r1, r3, r8, lr}
62	ldmfd	sp!, {r8, lr}
63UNWIND( .fnend              )
64
65#else
66
67/*
68 * This version aligns the destination pointer in order to write
69 * whole cache lines at once.
70 */
71
72	stmfd	sp!, {r4-r8, lr}
73UNWIND( .fnend                 )
74UNWIND( .fnstart               )
75UNWIND( .save {r4-r8, lr}      )
76	mov	r4, r1
77	mov	r5, r1
78	mov	r6, r1
79	mov	r7, r1
80	mov	r8, r1
81	mov	lr, r1
82
83	cmp	r2, #96
84	tstgt	ip, #31
85	ble	3f
86
87	and	r8, ip, #31
88	rsb	r8, r8, #32
89	sub	r2, r2, r8
90	movs	r8, r8, lsl #(32 - 4)
91	stmcsia	ip!, {r4, r5, r6, r7}
92	stmmiia	ip!, {r4, r5}
93	tst	r8, #(1 << 30)
94	mov	r8, r1
95	strne	r1, [ip], #4
96
973:	subs	r2, r2, #64
98	stmgeia	ip!, {r1, r3-r8, lr}
99	stmgeia	ip!, {r1, r3-r8, lr}
100	bgt	3b
101	ldmeqfd	sp!, {r4-r8, pc}
102
103	tst	r2, #32
104	stmneia	ip!, {r1, r3-r8, lr}
105	tst	r2, #16
106	stmneia	ip!, {r4-r7}
107	ldmfd	sp!, {r4-r8, lr}
108UNWIND( .fnend                 )
109
110#endif
111
112UNWIND( .fnstart            )
1134:	tst	r2, #8
114	stmneia	ip!, {r1, r3}
115	tst	r2, #4
116	strne	r1, [ip], #4
117/*
118 * When we get here, we've got less than 4 bytes to zero.  We
119 * may have an unaligned pointer as well.
120 */
1215:	tst	r2, #2
122	strneb	r1, [ip], #1
123	strneb	r1, [ip], #1
124	tst	r2, #1
125	strneb	r1, [ip], #1
126	ret	lr
127
1286:	subs	r2, r2, #4		@ 1 do we have enough
129	blt	5b			@ 1 bytes to align with?
130	cmp	r3, #2			@ 1
131	strltb	r1, [ip], #1		@ 1
132	strleb	r1, [ip], #1		@ 1
133	strb	r1, [ip], #1		@ 1
134	add	r2, r2, r3		@ 1 (r2 = r2 - (4 - r3))
135	b	1b
136UNWIND( .fnend   )
137ENDPROC(memset)
138ENDPROC(mmioset)
139EXPORT_SYMBOL(memset)
140EXPORT_SYMBOL(mmioset)
141