xref: /linux/arch/arm/lib/memset.S (revision c3d6b628395fe6ec3442a83ddf02334c54867d43)
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
15	.text
16	.align	5
17	.word	0
18
191:	subs	r2, r2, #4		@ 1 do we have enough
20	blt	5f			@ 1 bytes to align with?
21	cmp	r3, #2			@ 1
22	strltb	r1, [ip], #1		@ 1
23	strleb	r1, [ip], #1		@ 1
24	strb	r1, [ip], #1		@ 1
25	add	r2, r2, r3		@ 1 (r2 = r2 - (4 - r3))
26/*
27 * The pointer is now aligned and the length is adjusted.  Try doing the
28 * memset again.
29 */
30
31ENTRY(memset)
32/*
33 * Preserve the contents of r0 for the return value.
34 */
35	mov	ip, r0
36	ands	r3, ip, #3		@ 1 unaligned?
37	bne	1b			@ 1
38/*
39 * we know that the pointer in ip is aligned to a word boundary.
40 */
41	orr	r1, r1, r1, lsl #8
42	orr	r1, r1, r1, lsl #16
43	mov	r3, r1
44	cmp	r2, #16
45	blt	4f
46
47#if ! CALGN(1)+0
48
49/*
50 * We need 2 extra registers for this loop - use r8 and the LR
51 */
52	stmfd	sp!, {r8, lr}
53	mov	r8, r1
54	mov	lr, r1
55
562:	subs	r2, r2, #64
57	stmgeia	ip!, {r1, r3, r8, lr}	@ 64 bytes at a time.
58	stmgeia	ip!, {r1, r3, r8, lr}
59	stmgeia	ip!, {r1, r3, r8, lr}
60	stmgeia	ip!, {r1, r3, r8, lr}
61	bgt	2b
62	ldmeqfd	sp!, {r8, pc}		@ Now <64 bytes to go.
63/*
64 * No need to correct the count; we're only testing bits from now on
65 */
66	tst	r2, #32
67	stmneia	ip!, {r1, r3, r8, lr}
68	stmneia	ip!, {r1, r3, r8, lr}
69	tst	r2, #16
70	stmneia	ip!, {r1, r3, r8, lr}
71	ldmfd	sp!, {r8, lr}
72
73#else
74
75/*
76 * This version aligns the destination pointer in order to write
77 * whole cache lines at once.
78 */
79
80	stmfd	sp!, {r4-r8, lr}
81	mov	r4, r1
82	mov	r5, r1
83	mov	r6, r1
84	mov	r7, r1
85	mov	r8, r1
86	mov	lr, r1
87
88	cmp	r2, #96
89	tstgt	ip, #31
90	ble	3f
91
92	and	r8, ip, #31
93	rsb	r8, r8, #32
94	sub	r2, r2, r8
95	movs	r8, r8, lsl #(32 - 4)
96	stmcsia	ip!, {r4, r5, r6, r7}
97	stmmiia	ip!, {r4, r5}
98	tst	r8, #(1 << 30)
99	mov	r8, r1
100	strne	r1, [ip], #4
101
1023:	subs	r2, r2, #64
103	stmgeia	ip!, {r1, r3-r8, lr}
104	stmgeia	ip!, {r1, r3-r8, lr}
105	bgt	3b
106	ldmeqfd	sp!, {r4-r8, pc}
107
108	tst	r2, #32
109	stmneia	ip!, {r1, r3-r8, lr}
110	tst	r2, #16
111	stmneia	ip!, {r4-r7}
112	ldmfd	sp!, {r4-r8, lr}
113
114#endif
115
1164:	tst	r2, #8
117	stmneia	ip!, {r1, r3}
118	tst	r2, #4
119	strne	r1, [ip], #4
120/*
121 * When we get here, we've got less than 4 bytes to zero.  We
122 * may have an unaligned pointer as well.
123 */
1245:	tst	r2, #2
125	strneb	r1, [ip], #1
126	strneb	r1, [ip], #1
127	tst	r2, #1
128	strneb	r1, [ip], #1
129	mov	pc, lr
130ENDPROC(memset)
131