1 /* !! 1 /* SPDX-License-Identifier: GPL-2.0 */
2 * arch/xtensa/lib/memset.S !! 2 /* linux/arch/sparc/lib/memset.S: Sparc optimized memset, bzero and clear_user code
>> 3 * Copyright (C) 1991,1996 Free Software Foundation
>> 4 * Copyright (C) 1996,1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
>> 5 * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
3 * 6 *
4 * ANSI C standard library function memset !! 7 * Calls to memset returns initial %o0. Calls to bzero returns 0, if ok, and
5 * (Well, almost. .fixup code might return z !! 8 * number of bytes not yet set if exception occurs and we were called as
6 * !! 9 * clear_user.
7 * This file is subject to the terms and cond <<
8 * Public License. See the file "COPYING" in <<
9 * this archive for more details. <<
10 * <<
11 * Copyright (C) 2002 Tensilica Inc. <<
12 */ <<
13 <<
14 #include <linux/linkage.h> <<
15 #include <asm/asmmacro.h> <<
16 #include <asm/core.h> <<
17 <<
18 /* <<
19 * void *memset(void *dst, int c, size_t lengt <<
20 * <<
21 * The algorithm is as follows: <<
22 * Create a word with c in all byte position <<
23 * If the destination is aligned, <<
24 * do 16B chucks with a loop, and then fin <<
25 * 8B, 4B, 2B, and 1B stores conditional o <<
26 * If destination is unaligned, align it by <<
27 * setting 1B and 2B and then go to aligne <<
28 * This code tries to use fall-through branc <<
29 * case of an aligned destination (except <<
30 * the alignment labels). <<
31 */ <<
32 <<
33 .text <<
34 ENTRY(__memset) <<
35 WEAK(memset) <<
36 <<
37 abi_entry_default <<
38 # a2/ dst, a3/ c, a4/ length <<
39 extui a3, a3, 0, 8 # mask to just <<
40 slli a7, a3, 8 # duplicate ch <<
41 or a3, a3, a7 # ... <<
42 slli a7, a3, 16 # ... <<
43 or a3, a3, a7 # ... <<
44 mov a5, a2 # copy dst so <<
45 movi a6, 3 # for alignmen <<
46 bany a2, a6, .Ldstunaligned # if ds <<
47 .L0: # return here from .Ldstunaligned when <<
48 srli a7, a4, 4 # number of lo <<
49 # per iteratio <<
50 bnez a4, .Laligned <<
51 abi_ret_default <<
52 <<
53 /* <<
54 * Destination is word-aligned. <<
55 */ <<
56 # set 16 bytes per iteration for word- <<
57 .align 4 # 1 mod 4 alig <<
58 .byte 0 # (0 mod 4 ali <<
59 .Laligned: <<
60 #if XCHAL_HAVE_LOOPS <<
61 loopnez a7, .Loop1done <<
62 #else /* !XCHAL_HAVE_LOOPS */ <<
63 beqz a7, .Loop1done <<
64 slli a6, a7, 4 <<
65 add a6, a6, a5 # a6 = end of <<
66 #endif /* !XCHAL_HAVE_LOOPS */ <<
67 .Loop1: <<
68 EX(10f) s32i a3, a5, 0 <<
69 EX(10f) s32i a3, a5, 4 <<
70 EX(10f) s32i a3, a5, 8 <<
71 EX(10f) s32i a3, a5, 12 <<
72 addi a5, a5, 16 <<
73 #if !XCHAL_HAVE_LOOPS <<
74 blt a5, a6, .Loop1 <<
75 #endif /* !XCHAL_HAVE_LOOPS */ <<
76 .Loop1done: <<
77 bbci.l a4, 3, .L2 <<
78 # set 8 bytes <<
79 EX(10f) s32i a3, a5, 0 <<
80 EX(10f) s32i a3, a5, 4 <<
81 addi a5, a5, 8 <<
82 .L2: <<
83 bbci.l a4, 2, .L3 <<
84 # set 4 bytes <<
85 EX(10f) s32i a3, a5, 0 <<
86 addi a5, a5, 4 <<
87 .L3: <<
88 bbci.l a4, 1, .L4 <<
89 # set 2 bytes <<
90 EX(10f) s16i a3, a5, 0 <<
91 addi a5, a5, 2 <<
92 .L4: <<
93 bbci.l a4, 0, .L5 <<
94 # set 1 byte <<
95 EX(10f) s8i a3, a5, 0 <<
96 .L5: <<
97 .Lret1: <<
98 abi_ret_default <<
99 <<
100 /* <<
101 * Destination is unaligned <<
102 */ 10 */
103 11
104 .Ldstunaligned: !! 12 #include <linux/export.h>
105 bltui a4, 8, .Lbyteset # do s !! 13 #include <asm/ptrace.h>
106 bbci.l a5, 0, .L20 # bran <<
107 # dst is only byte aligned <<
108 # set 1 byte <<
109 EX(10f) s8i a3, a5, 0 <<
110 addi a5, a5, 1 <<
111 addi a4, a4, -1 <<
112 # now retest if dst aligned <<
113 bbci.l a5, 1, .L0 # if now align <<
114 .L20: <<
115 # dst half-aligned <<
116 # set 2 bytes <<
117 EX(10f) s16i a3, a5, 0 <<
118 addi a5, a5, 2 <<
119 addi a4, a4, -2 <<
120 j .L0 # dst is now a <<
121 14
122 /* !! 15 /* Work around cpp -rob */
123 * Byte by byte set !! 16 #define ALLOC #alloc
124 */ !! 17 #define EXECINSTR #execinstr
>> 18 #define EX(x,y,a,b) \
>> 19 98: x,y; \
>> 20 .section .fixup,ALLOC,EXECINSTR; \
>> 21 .align 4; \
>> 22 99: retl; \
>> 23 a, b, %o0; \
>> 24 .section __ex_table,ALLOC; \
>> 25 .align 4; \
>> 26 .word 98b, 99b; \
>> 27 .text; \
125 .align 4 28 .align 4
126 .byte 0 # 1 mod 4 alig <<
127 # (0 mod 4 ali <<
128 .Lbyteset: <<
129 #if XCHAL_HAVE_LOOPS <<
130 loopnez a4, .Lbytesetdone <<
131 #else /* !XCHAL_HAVE_LOOPS */ <<
132 beqz a4, .Lbytesetdone <<
133 add a6, a5, a4 # a6 = ending <<
134 #endif /* !XCHAL_HAVE_LOOPS */ <<
135 .Lbyteloop: <<
136 EX(10f) s8i a3, a5, 0 <<
137 addi a5, a5, 1 <<
138 #if !XCHAL_HAVE_LOOPS <<
139 blt a5, a6, .Lbyteloop <<
140 #endif /* !XCHAL_HAVE_LOOPS */ <<
141 .Lbytesetdone: <<
142 abi_ret_default <<
143 <<
144 ENDPROC(__memset) <<
145 EXPORT_SYMBOL(__memset) <<
146 EXPORT_SYMBOL(memset) <<
147 29
148 .section .fixup, "ax" !! 30 #define STORE(source, base, offset, n) \
149 .align 4 !! 31 98: std source, [base + offset + n]; \
>> 32 .section .fixup,ALLOC,EXECINSTR; \
>> 33 .align 4; \
>> 34 99: ba 30f; \
>> 35 sub %o3, n - offset, %o3; \
>> 36 .section __ex_table,ALLOC; \
>> 37 .align 4; \
>> 38 .word 98b, 99b; \
>> 39 .text; \
>> 40 .align 4;
>> 41
>> 42 #define STORE_LAST(source, base, offset, n) \
>> 43 EX(std source, [base - offset - n], \
>> 44 add %o1, offset + n);
>> 45
>> 46 /* Please don't change these macros, unless you change the logic
>> 47 * in the .fixup section below as well.
>> 48 * Store 64 bytes at (BASE + OFFSET) using value SOURCE. */
>> 49 #define ZERO_BIG_BLOCK(base, offset, source) \
>> 50 STORE(source, base, offset, 0x00); \
>> 51 STORE(source, base, offset, 0x08); \
>> 52 STORE(source, base, offset, 0x10); \
>> 53 STORE(source, base, offset, 0x18); \
>> 54 STORE(source, base, offset, 0x20); \
>> 55 STORE(source, base, offset, 0x28); \
>> 56 STORE(source, base, offset, 0x30); \
>> 57 STORE(source, base, offset, 0x38);
>> 58
>> 59 #define ZERO_LAST_BLOCKS(base, offset, source) \
>> 60 STORE_LAST(source, base, offset, 0x38); \
>> 61 STORE_LAST(source, base, offset, 0x30); \
>> 62 STORE_LAST(source, base, offset, 0x28); \
>> 63 STORE_LAST(source, base, offset, 0x20); \
>> 64 STORE_LAST(source, base, offset, 0x18); \
>> 65 STORE_LAST(source, base, offset, 0x10); \
>> 66 STORE_LAST(source, base, offset, 0x08); \
>> 67 STORE_LAST(source, base, offset, 0x00);
>> 68
>> 69 .text
>> 70 .align 4
>> 71
>> 72 .globl __bzero_begin
>> 73 __bzero_begin:
>> 74
>> 75 .globl __bzero
>> 76 .type __bzero,#function
>> 77 .globl memset
>> 78 EXPORT_SYMBOL(__bzero)
>> 79 EXPORT_SYMBOL(memset)
>> 80 memset:
>> 81 mov %o0, %g1
>> 82 mov 1, %g4
>> 83 and %o1, 0xff, %g3
>> 84 sll %g3, 8, %g2
>> 85 or %g3, %g2, %g3
>> 86 sll %g3, 16, %g2
>> 87 or %g3, %g2, %g3
>> 88 b 1f
>> 89 mov %o2, %o1
>> 90 3:
>> 91 cmp %o2, 3
>> 92 be 2f
>> 93 EX(stb %g3, [%o0], sub %o1, 0)
>> 94
>> 95 cmp %o2, 2
>> 96 be 2f
>> 97 EX(stb %g3, [%o0 + 0x01], sub %o1, 1)
>> 98
>> 99 EX(stb %g3, [%o0 + 0x02], sub %o1, 2)
>> 100 2:
>> 101 sub %o2, 4, %o2
>> 102 add %o1, %o2, %o1
>> 103 b 4f
>> 104 sub %o0, %o2, %o0
>> 105
>> 106 __bzero:
>> 107 clr %g4
>> 108 mov %g0, %g3
>> 109 1:
>> 110 cmp %o1, 7
>> 111 bleu 7f
>> 112 andcc %o0, 3, %o2
>> 113
>> 114 bne 3b
>> 115 4:
>> 116 andcc %o0, 4, %g0
>> 117
>> 118 be 2f
>> 119 mov %g3, %g2
>> 120
>> 121 EX(st %g3, [%o0], sub %o1, 0)
>> 122 sub %o1, 4, %o1
>> 123 add %o0, 4, %o0
>> 124 2:
>> 125 andcc %o1, 0xffffff80, %o3 ! Now everything is 8 aligned and o1 is len to run
>> 126 be 9f
>> 127 andcc %o1, 0x78, %o2
>> 128 10:
>> 129 ZERO_BIG_BLOCK(%o0, 0x00, %g2)
>> 130 subcc %o3, 128, %o3
>> 131 ZERO_BIG_BLOCK(%o0, 0x40, %g2)
>> 132 bne 10b
>> 133 add %o0, 128, %o0
>> 134
>> 135 orcc %o2, %g0, %g0
>> 136 9:
>> 137 be 13f
>> 138 andcc %o1, 7, %o1
>> 139
>> 140 srl %o2, 1, %o3
>> 141 set 13f, %o4
>> 142 sub %o4, %o3, %o4
>> 143 jmp %o4
>> 144 add %o0, %o2, %o0
>> 145
>> 146 ZERO_LAST_BLOCKS(%o0, 0x48, %g2)
>> 147 ZERO_LAST_BLOCKS(%o0, 0x08, %g2)
>> 148 13:
>> 149 be 8f
>> 150 andcc %o1, 4, %g0
>> 151
>> 152 be 1f
>> 153 andcc %o1, 2, %g0
>> 154
>> 155 EX(st %g3, [%o0], and %o1, 7)
>> 156 add %o0, 4, %o0
>> 157 1:
>> 158 be 1f
>> 159 andcc %o1, 1, %g0
>> 160
>> 161 EX(sth %g3, [%o0], and %o1, 3)
>> 162 add %o0, 2, %o0
>> 163 1:
>> 164 bne,a 8f
>> 165 EX(stb %g3, [%o0], and %o1, 1)
>> 166 8:
>> 167 b 0f
>> 168 nop
>> 169 7:
>> 170 be 13b
>> 171 orcc %o1, 0, %g0
>> 172
>> 173 be 0f
>> 174 8:
>> 175 add %o0, 1, %o0
>> 176 subcc %o1, 1, %o1
>> 177 bne 8b
>> 178 EX(stb %g3, [%o0 - 1], add %o1, 1)
>> 179 0:
>> 180 andcc %g4, 1, %g0
>> 181 be 5f
>> 182 nop
>> 183 retl
>> 184 mov %g1, %o0
>> 185 5:
>> 186 retl
>> 187 clr %o0
150 188
151 /* We return zero if a failure occurred. */ !! 189 .section .fixup,#alloc,#execinstr
>> 190 .align 4
>> 191 30:
>> 192 and %o1, 0x7f, %o1
>> 193 retl
>> 194 add %o3, %o1, %o0
152 195
153 10: !! 196 .globl __bzero_end
154 movi a2, 0 !! 197 __bzero_end:
155 abi_ret_default <<
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.