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
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
|
/*
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#include "libyuv/compare_row.h"
#include "libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// This module is for GCC x86 and x64.
#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
#if defined(__x86_64__)
uint32_t HammingDistance_SSE42(const uint8_t* src_a,
const uint8_t* src_b,
int count) {
uint64_t diff = 0u;
asm volatile(
"xor %3,%3 \n"
"xor %%r8,%%r8 \n"
"xor %%r9,%%r9 \n"
"xor %%r10,%%r10 \n"
// Process 32 bytes per loop.
LABELALIGN
"1: \n"
"mov (%0),%%rcx \n"
"mov 0x8(%0),%%rdx \n"
"xor (%1),%%rcx \n"
"xor 0x8(%1),%%rdx \n"
"popcnt %%rcx,%%rcx \n"
"popcnt %%rdx,%%rdx \n"
"mov 0x10(%0),%%rsi \n"
"mov 0x18(%0),%%rdi \n"
"xor 0x10(%1),%%rsi \n"
"xor 0x18(%1),%%rdi \n"
"popcnt %%rsi,%%rsi \n"
"popcnt %%rdi,%%rdi \n"
"add $0x20,%0 \n"
"add $0x20,%1 \n"
"add %%rcx,%3 \n"
"add %%rdx,%%r8 \n"
"add %%rsi,%%r9 \n"
"add %%rdi,%%r10 \n"
"sub $0x20,%2 \n"
"jg 1b \n"
"add %%r8, %3 \n"
"add %%r9, %3 \n"
"add %%r10, %3 \n"
: "+r"(src_a), // %0
"+r"(src_b), // %1
"+r"(count), // %2
"=r"(diff) // %3
:
: "memory", "cc", "rcx", "rdx", "rsi", "rdi", "r8", "r9", "r10");
return (uint32_t)(diff);
}
#else
uint32_t HammingDistance_SSE42(const uint8_t* src_a,
const uint8_t* src_b,
int count) {
uint32_t diff = 0u;
asm volatile(
// Process 16 bytes per loop.
LABELALIGN
"1: \n"
"mov (%0),%%ecx \n"
"mov 0x4(%0),%%edx \n"
"xor (%1),%%ecx \n"
"xor 0x4(%1),%%edx \n"
"popcnt %%ecx,%%ecx \n"
"add %%ecx,%3 \n"
"popcnt %%edx,%%edx \n"
"add %%edx,%3 \n"
"mov 0x8(%0),%%ecx \n"
"mov 0xc(%0),%%edx \n"
"xor 0x8(%1),%%ecx \n"
"xor 0xc(%1),%%edx \n"
"popcnt %%ecx,%%ecx \n"
"add %%ecx,%3 \n"
"popcnt %%edx,%%edx \n"
"add %%edx,%3 \n"
"add $0x10,%0 \n"
"add $0x10,%1 \n"
"sub $0x10,%2 \n"
"jg 1b \n"
: "+r"(src_a), // %0
"+r"(src_b), // %1
"+r"(count), // %2
"+r"(diff) // %3
:
: "memory", "cc", "ecx", "edx");
return diff;
}
#endif
static const vec8 kNibbleMask = {15, 15, 15, 15, 15, 15, 15, 15,
15, 15, 15, 15, 15, 15, 15, 15};
static const vec8 kBitCount = {0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
uint32_t HammingDistance_SSSE3(const uint8_t* src_a,
const uint8_t* src_b,
int count) {
uint32_t diff = 0u;
asm volatile(
"movdqa %4,%%xmm2 \n"
"movdqa %5,%%xmm3 \n"
"pxor %%xmm0,%%xmm0 \n"
"pxor %%xmm1,%%xmm1 \n"
"sub %0,%1 \n"
LABELALIGN
"1: \n"
"movdqa (%0),%%xmm4 \n"
"movdqa 0x10(%0), %%xmm5 \n"
"pxor (%0,%1), %%xmm4 \n"
"movdqa %%xmm4,%%xmm6 \n"
"pand %%xmm2,%%xmm6 \n"
"psrlw $0x4,%%xmm4 \n"
"movdqa %%xmm3,%%xmm7 \n"
"pshufb %%xmm6,%%xmm7 \n"
"pand %%xmm2,%%xmm4 \n"
"movdqa %%xmm3,%%xmm6 \n"
"pshufb %%xmm4,%%xmm6 \n"
"paddb %%xmm7,%%xmm6 \n"
"pxor 0x10(%0,%1),%%xmm5 \n"
"add $0x20,%0 \n"
"movdqa %%xmm5,%%xmm4 \n"
"pand %%xmm2,%%xmm5 \n"
"psrlw $0x4,%%xmm4 \n"
"movdqa %%xmm3,%%xmm7 \n"
"pshufb %%xmm5,%%xmm7 \n"
"pand %%xmm2,%%xmm4 \n"
"movdqa %%xmm3,%%xmm5 \n"
"pshufb %%xmm4,%%xmm5 \n"
"paddb %%xmm7,%%xmm5 \n"
"paddb %%xmm5,%%xmm6 \n"
"psadbw %%xmm1,%%xmm6 \n"
"paddd %%xmm6,%%xmm0 \n"
"sub $0x20,%2 \n"
"jg 1b \n"
"pshufd $0xaa,%%xmm0,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"movd %%xmm0, %3 \n"
: "+r"(src_a), // %0
"+r"(src_b), // %1
"+r"(count), // %2
"=r"(diff) // %3
: "m"(kNibbleMask), // %4
"m"(kBitCount) // %5
: "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
"xmm7");
return diff;
}
#ifdef HAS_HAMMINGDISTANCE_AVX2
uint32_t HammingDistance_AVX2(const uint8_t* src_a,
const uint8_t* src_b,
int count) {
uint32_t diff = 0u;
asm volatile(
"vbroadcastf128 %4,%%ymm2 \n"
"vbroadcastf128 %5,%%ymm3 \n"
"vpxor %%ymm0,%%ymm0,%%ymm0 \n"
"vpxor %%ymm1,%%ymm1,%%ymm1 \n"
"sub %0,%1 \n"
LABELALIGN
"1: \n"
"vmovdqa (%0),%%ymm4 \n"
"vmovdqa 0x20(%0), %%ymm5 \n"
"vpxor (%0,%1), %%ymm4, %%ymm4 \n"
"vpand %%ymm2,%%ymm4,%%ymm6 \n"
"vpsrlw $0x4,%%ymm4,%%ymm4 \n"
"vpshufb %%ymm6,%%ymm3,%%ymm6 \n"
"vpand %%ymm2,%%ymm4,%%ymm4 \n"
"vpshufb %%ymm4,%%ymm3,%%ymm4 \n"
"vpaddb %%ymm4,%%ymm6,%%ymm6 \n"
"vpxor 0x20(%0,%1),%%ymm5,%%ymm4 \n"
"add $0x40,%0 \n"
"vpand %%ymm2,%%ymm4,%%ymm5 \n"
"vpsrlw $0x4,%%ymm4,%%ymm4 \n"
"vpshufb %%ymm5,%%ymm3,%%ymm5 \n"
"vpand %%ymm2,%%ymm4,%%ymm4 \n"
"vpshufb %%ymm4,%%ymm3,%%ymm4 \n"
"vpaddb %%ymm5,%%ymm4,%%ymm4 \n"
"vpaddb %%ymm6,%%ymm4,%%ymm4 \n"
"vpsadbw %%ymm1,%%ymm4,%%ymm4 \n"
"vpaddd %%ymm0,%%ymm4,%%ymm0 \n"
"sub $0x40,%2 \n"
"jg 1b \n"
"vpermq $0xb1,%%ymm0,%%ymm1 \n"
"vpaddd %%ymm1,%%ymm0,%%ymm0 \n"
"vpermq $0xaa,%%ymm0,%%ymm1 \n"
"vpaddd %%ymm1,%%ymm0,%%ymm0 \n"
"vmovd %%xmm0, %3 \n"
"vzeroupper \n"
: "+r"(src_a), // %0
"+r"(src_b), // %1
"+r"(count), // %2
"=r"(diff) // %3
: "m"(kNibbleMask), // %4
"m"(kBitCount) // %5
: "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6");
return diff;
}
#endif // HAS_HAMMINGDISTANCE_AVX2
uint32_t SumSquareError_SSE2(const uint8_t* src_a,
const uint8_t* src_b,
int count) {
uint32_t sse;
asm volatile(
"pxor %%xmm0,%%xmm0 \n"
"pxor %%xmm5,%%xmm5 \n"
LABELALIGN
"1: \n"
"movdqu (%0),%%xmm1 \n"
"lea 0x10(%0),%0 \n"
"movdqu (%1),%%xmm2 \n"
"lea 0x10(%1),%1 \n"
"movdqa %%xmm1,%%xmm3 \n"
"psubusb %%xmm2,%%xmm1 \n"
"psubusb %%xmm3,%%xmm2 \n"
"por %%xmm2,%%xmm1 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklbw %%xmm5,%%xmm1 \n"
"punpckhbw %%xmm5,%%xmm2 \n"
"pmaddwd %%xmm1,%%xmm1 \n"
"pmaddwd %%xmm2,%%xmm2 \n"
"paddd %%xmm1,%%xmm0 \n"
"paddd %%xmm2,%%xmm0 \n"
"sub $0x10,%2 \n"
"jg 1b \n"
"pshufd $0xee,%%xmm0,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"pshufd $0x1,%%xmm0,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"movd %%xmm0,%3 \n"
: "+r"(src_a), // %0
"+r"(src_b), // %1
"+r"(count), // %2
"=g"(sse) // %3
::"memory",
"cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm5");
return sse;
}
static const uvec32 kHash16x33 = {0x92d9e201, 0, 0, 0}; // 33 ^ 16
static const uvec32 kHashMul0 = {
0x0c3525e1, // 33 ^ 15
0xa3476dc1, // 33 ^ 14
0x3b4039a1, // 33 ^ 13
0x4f5f0981, // 33 ^ 12
};
static const uvec32 kHashMul1 = {
0x30f35d61, // 33 ^ 11
0x855cb541, // 33 ^ 10
0x040a9121, // 33 ^ 9
0x747c7101, // 33 ^ 8
};
static const uvec32 kHashMul2 = {
0xec41d4e1, // 33 ^ 7
0x4cfa3cc1, // 33 ^ 6
0x025528a1, // 33 ^ 5
0x00121881, // 33 ^ 4
};
static const uvec32 kHashMul3 = {
0x00008c61, // 33 ^ 3
0x00000441, // 33 ^ 2
0x00000021, // 33 ^ 1
0x00000001, // 33 ^ 0
};
uint32_t HashDjb2_SSE41(const uint8_t* src, int count, uint32_t seed) {
uint32_t hash;
asm volatile(
"movd %2,%%xmm0 \n"
"pxor %%xmm7,%%xmm7 \n"
"movdqa %4,%%xmm6 \n"
LABELALIGN
"1: \n"
"movdqu (%0),%%xmm1 \n"
"lea 0x10(%0),%0 \n"
"pmulld %%xmm6,%%xmm0 \n"
"movdqa %5,%%xmm5 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklbw %%xmm7,%%xmm2 \n"
"movdqa %%xmm2,%%xmm3 \n"
"punpcklwd %%xmm7,%%xmm3 \n"
"pmulld %%xmm5,%%xmm3 \n"
"movdqa %6,%%xmm5 \n"
"movdqa %%xmm2,%%xmm4 \n"
"punpckhwd %%xmm7,%%xmm4 \n"
"pmulld %%xmm5,%%xmm4 \n"
"movdqa %7,%%xmm5 \n"
"punpckhbw %%xmm7,%%xmm1 \n"
"movdqa %%xmm1,%%xmm2 \n"
"punpcklwd %%xmm7,%%xmm2 \n"
"pmulld %%xmm5,%%xmm2 \n"
"movdqa %8,%%xmm5 \n"
"punpckhwd %%xmm7,%%xmm1 \n"
"pmulld %%xmm5,%%xmm1 \n"
"paddd %%xmm4,%%xmm3 \n"
"paddd %%xmm2,%%xmm1 \n"
"paddd %%xmm3,%%xmm1 \n"
"pshufd $0xe,%%xmm1,%%xmm2 \n"
"paddd %%xmm2,%%xmm1 \n"
"pshufd $0x1,%%xmm1,%%xmm2 \n"
"paddd %%xmm2,%%xmm1 \n"
"paddd %%xmm1,%%xmm0 \n"
"sub $0x10,%1 \n"
"jg 1b \n"
"movd %%xmm0,%3 \n"
: "+r"(src), // %0
"+r"(count), // %1
"+rm"(seed), // %2
"=g"(hash) // %3
: "m"(kHash16x33), // %4
"m"(kHashMul0), // %5
"m"(kHashMul1), // %6
"m"(kHashMul2), // %7
"m"(kHashMul3) // %8
: "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
"xmm7");
return hash;
}
#endif // defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
|
