Loading...
1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 * linux/arch/arm/lib/memset.S
4 *
5 * Copyright (C) 1995-2000 Russell King
6 *
7 * ASM optimised string functions
8 */
9#include <linux/linkage.h>
10#include <asm/assembler.h>
11#include <asm/unwind.h>
12
13 .text
14 .align 5
15
16ENTRY(__memset)
17ENTRY(mmioset)
18WEAK(memset)
19UNWIND( .fnstart )
20 and r1, r1, #255 @ cast to unsigned char
21 ands r3, r0, #3 @ 1 unaligned?
22 mov ip, r0 @ preserve r0 as return value
23 bne 6f @ 1
24/*
25 * we know that the pointer in ip is aligned to a word boundary.
26 */
271: orr r1, r1, r1, lsl #8
28 orr r1, r1, r1, lsl #16
29 mov r3, r1
307: cmp r2, #16
31 blt 4f
32UNWIND( .fnend )
33
34#if ! CALGN(1)+0
35
36/*
37 * We need 2 extra registers for this loop - use r8 and the LR
38 */
39UNWIND( .fnstart )
40UNWIND( .save {r8, lr} )
41 stmfd sp!, {r8, lr}
42 mov r8, r1
43 mov lr, r3
44
452: subs r2, r2, #64
46 stmiage ip!, {r1, r3, r8, lr} @ 64 bytes at a time.
47 stmiage ip!, {r1, r3, r8, lr}
48 stmiage ip!, {r1, r3, r8, lr}
49 stmiage ip!, {r1, r3, r8, lr}
50 bgt 2b
51 ldmfdeq sp!, {r8, pc} @ Now <64 bytes to go.
52/*
53 * No need to correct the count; we're only testing bits from now on
54 */
55 tst r2, #32
56 stmiane ip!, {r1, r3, r8, lr}
57 stmiane ip!, {r1, r3, r8, lr}
58 tst r2, #16
59 stmiane ip!, {r1, r3, r8, lr}
60 ldmfd sp!, {r8, lr}
61UNWIND( .fnend )
62
63#else
64
65/*
66 * This version aligns the destination pointer in order to write
67 * whole cache lines at once.
68 */
69
70UNWIND( .fnstart )
71UNWIND( .save {r4-r8, lr} )
72 stmfd sp!, {r4-r8, lr}
73 mov r4, r1
74 mov r5, r3
75 mov r6, r1
76 mov r7, r3
77 mov r8, r1
78 mov lr, r3
79
80 cmp r2, #96
81 tstgt ip, #31
82 ble 3f
83
84 and r8, ip, #31
85 rsb r8, r8, #32
86 sub r2, r2, r8
87 movs r8, r8, lsl #(32 - 4)
88 stmiacs ip!, {r4, r5, r6, r7}
89 stmiami ip!, {r4, r5}
90 tst r8, #(1 << 30)
91 mov r8, r1
92 strne r1, [ip], #4
93
943: subs r2, r2, #64
95 stmiage ip!, {r1, r3-r8, lr}
96 stmiage ip!, {r1, r3-r8, lr}
97 bgt 3b
98 ldmfdeq sp!, {r4-r8, pc}
99
100 tst r2, #32
101 stmiane ip!, {r1, r3-r8, lr}
102 tst r2, #16
103 stmiane ip!, {r4-r7}
104 ldmfd sp!, {r4-r8, lr}
105UNWIND( .fnend )
106
107#endif
108
109UNWIND( .fnstart )
1104: tst r2, #8
111 stmiane ip!, {r1, r3}
112 tst r2, #4
113 strne r1, [ip], #4
114/*
115 * When we get here, we've got less than 4 bytes to set. We
116 * may have an unaligned pointer as well.
117 */
1185: tst r2, #2
119 strbne r1, [ip], #1
120 strbne r1, [ip], #1
121 tst r2, #1
122 strbne r1, [ip], #1
123 ret lr
124
1256: subs r2, r2, #4 @ 1 do we have enough
126 blt 5b @ 1 bytes to align with?
127 cmp r3, #2 @ 1
128 strblt r1, [ip], #1 @ 1
129 strble r1, [ip], #1 @ 1
130 strb r1, [ip], #1 @ 1
131 add r2, r2, r3 @ 1 (r2 = r2 - (4 - r3))
132 b 1b
133UNWIND( .fnend )
134ENDPROC(memset)
135ENDPROC(mmioset)
136ENDPROC(__memset)
137
138ENTRY(__memset32)
139UNWIND( .fnstart )
140 mov r3, r1 @ copy r1 to r3 and fall into memset64
141UNWIND( .fnend )
142ENDPROC(__memset32)
143ENTRY(__memset64)
144UNWIND( .fnstart )
145 mov ip, r0 @ preserve r0 as return value
146 b 7b @ jump into the middle of memset
147UNWIND( .fnend )
148ENDPROC(__memset64)
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, [r0], #1 @ 1
23 strleb r1, [r0], #1 @ 1
24 strb r1, [r0], #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 ands r3, r0, #3 @ 1 unaligned?
33 bne 1b @ 1
34/*
35 * we know that the pointer in r0 is aligned to a word boundary.
36 */
37 orr r1, r1, r1, lsl #8
38 orr r1, r1, r1, lsl #16
39 mov r3, r1
40 cmp r2, #16
41 blt 4f
42
43#if ! CALGN(1)+0
44
45/*
46 * We need an extra register for this loop - save the return address and
47 * use the LR
48 */
49 str lr, [sp, #-4]!
50 mov ip, r1
51 mov lr, r1
52
532: subs r2, r2, #64
54 stmgeia r0!, {r1, r3, ip, lr} @ 64 bytes at a time.
55 stmgeia r0!, {r1, r3, ip, lr}
56 stmgeia r0!, {r1, r3, ip, lr}
57 stmgeia r0!, {r1, r3, ip, lr}
58 bgt 2b
59 ldmeqfd sp!, {pc} @ Now <64 bytes to go.
60/*
61 * No need to correct the count; we're only testing bits from now on
62 */
63 tst r2, #32
64 stmneia r0!, {r1, r3, ip, lr}
65 stmneia r0!, {r1, r3, ip, lr}
66 tst r2, #16
67 stmneia r0!, {r1, r3, ip, lr}
68 ldr lr, [sp], #4
69
70#else
71
72/*
73 * This version aligns the destination pointer in order to write
74 * whole cache lines at once.
75 */
76
77 stmfd sp!, {r4-r7, lr}
78 mov r4, r1
79 mov r5, r1
80 mov r6, r1
81 mov r7, r1
82 mov ip, r1
83 mov lr, r1
84
85 cmp r2, #96
86 tstgt r0, #31
87 ble 3f
88
89 and ip, r0, #31
90 rsb ip, ip, #32
91 sub r2, r2, ip
92 movs ip, ip, lsl #(32 - 4)
93 stmcsia r0!, {r4, r5, r6, r7}
94 stmmiia r0!, {r4, r5}
95 tst ip, #(1 << 30)
96 mov ip, r1
97 strne r1, [r0], #4
98
993: subs r2, r2, #64
100 stmgeia r0!, {r1, r3-r7, ip, lr}
101 stmgeia r0!, {r1, r3-r7, ip, lr}
102 bgt 3b
103 ldmeqfd sp!, {r4-r7, pc}
104
105 tst r2, #32
106 stmneia r0!, {r1, r3-r7, ip, lr}
107 tst r2, #16
108 stmneia r0!, {r4-r7}
109 ldmfd sp!, {r4-r7, lr}
110
111#endif
112
1134: tst r2, #8
114 stmneia r0!, {r1, r3}
115 tst r2, #4
116 strne r1, [r0], #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, [r0], #1
123 strneb r1, [r0], #1
124 tst r2, #1
125 strneb r1, [r0], #1
126 mov pc, lr
127ENDPROC(memset)