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
297
298
299
300
301
302
303
304
305
|
/*
* Copyright (C) 2013, 2014-2015, 2017 Synopsys, Inc. (www.synopsys.com)
* Copyright (C) 2007 ARC International (UK) LTD
*
* Licensed under the LGPL v2.1 or later, see the file COPYING.LIB in this tarball.
*/
#include <sysdep.h>
#if !defined(__ARC700__) && !defined(__ARCHS__)
#error "Neither ARC700 nor ARCHS is defined!"
#endif
ENTRY(memcpy)
#ifdef __ARC700__
/* This memcpy implementation does not support objects of 1GB or larger -
the check for alignment does not work then. */
/* We assume that most sources and destinations are aligned, and
that also lengths are mostly a multiple of four, although to a lesser
extent. */
or r3,r0,r1
asl_s r3,r3,30
mov_s r5,r0
brls.d r2,r3,.Lcopy_bytewise
sub.f r3,r2,1
ld_s r12,[r1,0]
asr.f lp_count,r3,3
bbit0.d r3,2,.Lnox4
bmsk_s r2,r2,1
st.ab r12,[r5,4]
ld.a r12,[r1,4]
.Lnox4:
lppnz .Lendloop
ld_s r3,[r1,4]
st.ab r12,[r5,4]
ld.a r12,[r1,8]
st.ab r3,[r5,4]
.Lendloop:
breq r2,0,.Last_store
ld r3,[r5,0]
#ifdef __LITTLE_ENDIAN__
add3 r2,-1,r2
; uses long immediate
xor_s r12,r12,r3
bmsk r12,r12,r2
xor_s r12,r12,r3
#else /* BIG ENDIAN */
sub3 r2,31,r2
; uses long immediate
xor_s r3,r3,r12
bmsk r3,r3,r2
xor_s r12,r12,r3
#endif /* ENDIAN */
.Last_store:
j_s.d [blink]
st r12,[r5,0]
.balign 4
.Lcopy_bytewise:
jcs [blink]
ldb_s r12,[r1,0]
lsr.f lp_count,r3
bhs_s .Lnox1
stb.ab r12,[r5,1]
ldb.a r12,[r1,1]
.Lnox1:
lppnz .Lendbloop
ldb_s r3,[r1,1]
stb.ab r12,[r5,1]
ldb.a r12,[r1,2]
stb.ab r3,[r5,1]
.Lendbloop:
j_s.d [blink]
stb r12,[r5,0]
#endif /* __ARC700__ */
#ifdef __ARCHS__
#ifdef __LITTLE_ENDIAN__
# define SHIFT_1(RX,RY,IMM) asl RX, RY, IMM ; <<
# define SHIFT_2(RX,RY,IMM) lsr RX, RY, IMM ; >>
# define MERGE_1(RX,RY,IMM) asl RX, RY, IMM
# define MERGE_2(RX,RY,IMM)
# define EXTRACT_1(RX,RY,IMM) and RX, RY, 0xFFFF
# define EXTRACT_2(RX,RY,IMM) lsr RX, RY, IMM
#else
# define SHIFT_1(RX,RY,IMM) lsr RX, RY, IMM ; >>
# define SHIFT_2(RX,RY,IMM) asl RX, RY, IMM ; <<
# define MERGE_1(RX,RY,IMM) asl RX, RY, IMM ; <<
# define MERGE_2(RX,RY,IMM) asl RX, RY, IMM ; <<
# define EXTRACT_1(RX,RY,IMM) lsr RX, RY, IMM
# define EXTRACT_2(RX,RY,IMM) lsr RX, RY, 0x08
#endif
#if defined(__LL64__) || defined(__ARC_LL64__)
# define PREFETCH_READ(RX) prefetch [RX, 56]
# define PREFETCH_WRITE(RX) prefetchw [RX, 64]
# define LOADX(DST,RX) ldd.ab DST, [RX, 8]
# define STOREX(SRC,RX) std.ab SRC, [RX, 8]
# define ZOLSHFT 5
# define ZOLAND 0x1F
#else
# define PREFETCH_READ(RX) prefetch [RX, 28]
# define PREFETCH_WRITE(RX) prefetchw [RX, 32]
# define LOADX(DST,RX) ld.ab DST, [RX, 4]
# define STOREX(SRC,RX) st.ab SRC, [RX, 4]
# define ZOLSHFT 4
# define ZOLAND 0xF
#endif
prefetch [r1] ; Prefetch the read location
prefetchw [r0] ; Prefetch the write location
mov.f 0, r2
;;; if size is zero
jz.d [blink]
mov r3, r0 ; don't clobber ret val
;;; if size <= 8
cmp r2, 8
bls.d @.Lsmallchunk
mov.f lp_count, r2
and.f r4, r0, 0x03
rsub lp_count, r4, 4
lpnz @.Laligndestination
;; LOOP BEGIN
ldb.ab r5, [r1,1]
sub r2, r2, 1
stb.ab r5, [r3,1]
.Laligndestination:
;;; Check the alignment of the source
and.f r4, r1, 0x03
bnz.d @.Lsourceunaligned
;;; CASE 0: Both source and destination are 32bit aligned
;;; Convert len to Dwords, unfold x4
lsr.f lp_count, r2, ZOLSHFT
lpnz @.Lcopy32_64bytes
;; LOOP START
LOADX (r6, r1)
PREFETCH_READ (r1)
PREFETCH_WRITE (r3)
LOADX (r8, r1)
LOADX (r10, r1)
LOADX (r4, r1)
STOREX (r6, r3)
STOREX (r8, r3)
STOREX (r10, r3)
STOREX (r4, r3)
.Lcopy32_64bytes:
and.f lp_count, r2, ZOLAND ;Last remaining 31 bytes
.Lsmallchunk:
lpnz @.Lcopyremainingbytes
;; LOOP START
ldb.ab r5, [r1,1]
stb.ab r5, [r3,1]
.Lcopyremainingbytes:
j [blink]
;;; END CASE 0
.Lsourceunaligned:
cmp r4, 2
beq.d @.LunalignedOffby2
sub r2, r2, 1
bhi.d @.LunalignedOffby3
ldb.ab r5, [r1, 1]
;;; CASE 1: The source is unaligned, off by 1
;; Hence I need to read 1 byte for a 16bit alignment
;; and 2bytes to reach 32bit alignment
ldh.ab r6, [r1, 2]
sub r2, r2, 2
;; Convert to words, unfold x2
lsr.f lp_count, r2, 3
MERGE_1 (r6, r6, 8)
MERGE_2 (r5, r5, 24)
or r5, r5, r6
;; Both src and dst are aligned
lpnz @.Lcopy8bytes_1
;; LOOP START
ld.ab r6, [r1, 4]
prefetch [r1, 28] ;Prefetch the next read location
ld.ab r8, [r1,4]
prefetchw [r3, 32] ;Prefetch the next write location
SHIFT_1 (r7, r6, 24)
or r7, r7, r5
SHIFT_2 (r5, r6, 8)
SHIFT_1 (r9, r8, 24)
or r9, r9, r5
SHIFT_2 (r5, r8, 8)
st.ab r7, [r3, 4]
st.ab r9, [r3, 4]
.Lcopy8bytes_1:
;; Write back the remaining 16bits
EXTRACT_1 (r6, r5, 16)
sth.ab r6, [r3, 2]
;; Write back the remaining 8bits
EXTRACT_2 (r5, r5, 16)
stb.ab r5, [r3, 1]
and.f lp_count, r2, 0x07 ;Last 8bytes
lpnz @.Lcopybytewise_1
;; LOOP START
ldb.ab r6, [r1,1]
stb.ab r6, [r3,1]
.Lcopybytewise_1:
j [blink]
.LunalignedOffby2:
;;; CASE 2: The source is unaligned, off by 2
ldh.ab r5, [r1, 2]
sub r2, r2, 1
;; Both src and dst are aligned
;; Convert to words, unfold x2
lsr.f lp_count, r2, 3
#ifdef __BIG_ENDIAN__
asl.nz r5, r5, 16
#endif
lpnz @.Lcopy8bytes_2
;; LOOP START
ld.ab r6, [r1, 4]
prefetch [r1, 28] ;Prefetch the next read location
ld.ab r8, [r1,4]
prefetchw [r3, 32] ;Prefetch the next write location
SHIFT_1 (r7, r6, 16)
or r7, r7, r5
SHIFT_2 (r5, r6, 16)
SHIFT_1 (r9, r8, 16)
or r9, r9, r5
SHIFT_2 (r5, r8, 16)
st.ab r7, [r3, 4]
st.ab r9, [r3, 4]
.Lcopy8bytes_2:
#ifdef __BIG_ENDIAN__
lsr.nz r5, r5, 16
#endif
sth.ab r5, [r3, 2]
and.f lp_count, r2, 0x07 ;Last 8bytes
lpnz @.Lcopybytewise_2
;; LOOP START
ldb.ab r6, [r1,1]
stb.ab r6, [r3,1]
.Lcopybytewise_2:
j [blink]
.LunalignedOffby3:
;;; CASE 3: The source is unaligned, off by 3
;;; Hence, I need to read 1byte for achieve the 32bit alignment
;; Both src and dst are aligned
;; Convert to words, unfold x2
lsr.f lp_count, r2, 3
#ifdef __BIG_ENDIAN__
asl.ne r5, r5, 24
#endif
lpnz @.Lcopy8bytes_3
;; LOOP START
ld.ab r6, [r1, 4]
prefetch [r1, 28] ;Prefetch the next read location
ld.ab r8, [r1,4]
prefetchw [r3, 32] ;Prefetch the next write location
SHIFT_1 (r7, r6, 8)
or r7, r7, r5
SHIFT_2 (r5, r6, 24)
SHIFT_1 (r9, r8, 8)
or r9, r9, r5
SHIFT_2 (r5, r8, 24)
st.ab r7, [r3, 4]
st.ab r9, [r3, 4]
.Lcopy8bytes_3:
#ifdef __BIG_ENDIAN__
lsr.nz r5, r5, 24
#endif
stb.ab r5, [r3, 1]
and.f lp_count, r2, 0x07 ;Last 8bytes
lpnz @.Lcopybytewise_3
;; LOOP START
ldb.ab r6, [r1,1]
stb.ab r6, [r3,1]
.Lcopybytewise_3:
j [blink]
#endif /* __ARCHS__ */
END(memcpy)
libc_hidden_def(memcpy)
|
