1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
|
/* Optimized memcpy for Xtensa.
Copyright (C) 2001, 2007 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
#include <sysdep.h>
#include <bits/xtensa-config.h>
.macro src_b r, w0, w1
#ifdef __XTENSA_EB__
src \r, \w0, \w1
#else
src \r, \w1, \w0
#endif
.endm
.macro ssa8 r
#ifdef __XTENSA_EB__
ssa8b \r
#else
ssa8l \r
#endif
.endm
/* If the Xtensa Unaligned Load Exception option is not used, this
code can run a few cycles faster by relying on the low address bits
being ignored. However, if the code is then run with an Xtensa ISS
client that checks for unaligned accesses, it will produce a lot of
warning messages. Set this flag to disable the use of unaligned
accesses and keep the ISS happy. */
#define UNALIGNED_ADDRESSES_CHECKED 1
/* Do not use .literal_position in the ENTRY macro. */
#undef LITERAL_POSITION
#define LITERAL_POSITION
/* void *memcpy (void *dst, const void *src, size_t len)
The algorithm is as follows:
If the destination is unaligned, align it by conditionally
copying 1- and/or 2-byte pieces.
If the source is aligned, copy 16 bytes with a loop, and then finish up
with 8, 4, 2, and 1-byte copies conditional on the length.
Else (if source is unaligned), do the same, but use SRC to align the
source data.
This code tries to use fall-through branches for the common
case of aligned source and destination and multiple of 4 (or 8) length. */
/* Byte by byte copy. */
.text
.align 4
.literal_position
__memcpy_aux:
/* Skip a byte to get 1 mod 4 alignment for LOOPNEZ
(0 mod 4 alignment for LBEG). */
.byte 0
.Lbytecopy:
#if XCHAL_HAVE_LOOPS
loopnez a4, 2f
#else
beqz a4, 2f
add a7, a3, a4 /* a7 = end address for source */
#endif
1: l8ui a6, a3, 0
addi a3, a3, 1
s8i a6, a5, 0
addi a5, a5, 1
#if !XCHAL_HAVE_LOOPS
blt a3, a7, 1b
#endif
2: retw
/* Destination is unaligned. */
.align 4
.Ldst1mod2: /* dst is only byte aligned */
/* Do short copies byte-by-byte. */
_bltui a4, 7, .Lbytecopy
/* Copy 1 byte. */
l8ui a6, a3, 0
addi a3, a3, 1
addi a4, a4, -1
s8i a6, a5, 0
addi a5, a5, 1
/* Return to main algorithm if dst is now aligned. */
_bbci.l a5, 1, .Ldstaligned
.Ldst2mod4: /* dst has 16-bit alignment */
/* Do short copies byte-by-byte. */
_bltui a4, 6, .Lbytecopy
/* Copy 2 bytes. */
l8ui a6, a3, 0
l8ui a7, a3, 1
addi a3, a3, 2
addi a4, a4, -2
s8i a6, a5, 0
s8i a7, a5, 1
addi a5, a5, 2
/* dst is now aligned; return to main algorithm. */
j .Ldstaligned
ENTRY (memcpy)
/* a2 = dst, a3 = src, a4 = len */
mov a5, a2 /* copy dst so that a2 is return value */
_bbsi.l a2, 0, .Ldst1mod2
_bbsi.l a2, 1, .Ldst2mod4
.Ldstaligned:
/* Get number of loop iterations with 16B per iteration. */
srli a7, a4, 4
/* Check if source is aligned. */
movi a8, 3
_bany a3, a8, .Lsrcunaligned
/* Destination and source are word-aligned, use word copy. */
#if XCHAL_HAVE_LOOPS
loopnez a7, 2f
#else
beqz a7, 2f
slli a8, a7, 4
add a8, a8, a3 /* a8 = end of last 16B source chunk */
#endif
1: l32i a6, a3, 0
l32i a7, a3, 4
s32i a6, a5, 0
l32i a6, a3, 8
s32i a7, a5, 4
l32i a7, a3, 12
s32i a6, a5, 8
addi a3, a3, 16
s32i a7, a5, 12
addi a5, a5, 16
#if !XCHAL_HAVE_LOOPS
blt a3, a8, 1b
#endif
/* Copy any leftover pieces smaller than 16B. */
2: bbci.l a4, 3, 3f
/* Copy 8 bytes. */
l32i a6, a3, 0
l32i a7, a3, 4
addi a3, a3, 8
s32i a6, a5, 0
s32i a7, a5, 4
addi a5, a5, 8
3: bbsi.l a4, 2, 4f
bbsi.l a4, 1, 5f
bbsi.l a4, 0, 6f
retw
/* Copy 4 bytes. */
4: l32i a6, a3, 0
addi a3, a3, 4
s32i a6, a5, 0
addi a5, a5, 4
bbsi.l a4, 1, 5f
bbsi.l a4, 0, 6f
retw
/* Copy 2 bytes. */
5: l16ui a6, a3, 0
addi a3, a3, 2
s16i a6, a5, 0
addi a5, a5, 2
bbsi.l a4, 0, 6f
retw
/* Copy 1 byte. */
6: l8ui a6, a3, 0
s8i a6, a5, 0
.Ldone:
retw
/* Destination is aligned; source is unaligned. */
.align 4
.Lsrcunaligned:
/* Avoid loading anything for zero-length copies. */
_beqz a4, .Ldone
/* Copy 16 bytes per iteration for word-aligned dst and
unaligned src. */
ssa8 a3 /* set shift amount from byte offset */
#if UNALIGNED_ADDRESSES_CHECKED
and a11, a3, a8 /* save unalignment offset for below */
sub a3, a3, a11 /* align a3 */
#endif
l32i a6, a3, 0 /* load first word */
#if XCHAL_HAVE_LOOPS
loopnez a7, 2f
#else
beqz a7, 2f
slli a10, a7, 4
add a10, a10, a3 /* a10 = end of last 16B source chunk */
#endif
1: l32i a7, a3, 4
l32i a8, a3, 8
src_b a6, a6, a7
s32i a6, a5, 0
l32i a9, a3, 12
src_b a7, a7, a8
s32i a7, a5, 4
l32i a6, a3, 16
src_b a8, a8, a9
s32i a8, a5, 8
addi a3, a3, 16
src_b a9, a9, a6
s32i a9, a5, 12
addi a5, a5, 16
#if !XCHAL_HAVE_LOOPS
blt a3, a10, 1b
#endif
2: bbci.l a4, 3, 3f
/* Copy 8 bytes. */
l32i a7, a3, 4
l32i a8, a3, 8
src_b a6, a6, a7
s32i a6, a5, 0
addi a3, a3, 8
src_b a7, a7, a8
s32i a7, a5, 4
addi a5, a5, 8
mov a6, a8
3: bbci.l a4, 2, 4f
/* Copy 4 bytes. */
l32i a7, a3, 4
addi a3, a3, 4
src_b a6, a6, a7
s32i a6, a5, 0
addi a5, a5, 4
mov a6, a7
4:
#if UNALIGNED_ADDRESSES_CHECKED
add a3, a3, a11 /* readjust a3 with correct misalignment */
#endif
bbsi.l a4, 1, 5f
bbsi.l a4, 0, 6f
retw
/* Copy 2 bytes. */
5: l8ui a6, a3, 0
l8ui a7, a3, 1
addi a3, a3, 2
s8i a6, a5, 0
s8i a7, a5, 1
addi a5, a5, 2
bbsi.l a4, 0, 6f
retw
/* Copy 1 byte. */
6: l8ui a6, a3, 0
s8i a6, a5, 0
retw
libc_hidden_def (memcpy)
|
