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
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
|
/*
* Copyright (c) 2014 The WebM 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 <arm_neon.h>
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"
#include "vpx/vpx_integer.h"
#include "vpx_dsp/arm/mem_neon.h"
#include "vpx_dsp/arm/sum_neon.h"
static INLINE unsigned int sad64xh_neon(const uint8_t *src_ptr, int src_stride,
const uint8_t *ref_ptr, int ref_stride,
int h) {
uint16x8_t sum[4] = { vdupq_n_u16(0), vdupq_n_u16(0), vdupq_n_u16(0),
vdupq_n_u16(0) };
uint32x4_t sum_u32;
int i = h;
do {
uint8x16_t s0, s1, s2, s3, r0, r1, r2, r3;
uint8x16_t diff0, diff1, diff2, diff3;
s0 = vld1q_u8(src_ptr);
r0 = vld1q_u8(ref_ptr);
diff0 = vabdq_u8(s0, r0);
sum[0] = vpadalq_u8(sum[0], diff0);
s1 = vld1q_u8(src_ptr + 16);
r1 = vld1q_u8(ref_ptr + 16);
diff1 = vabdq_u8(s1, r1);
sum[1] = vpadalq_u8(sum[1], diff1);
s2 = vld1q_u8(src_ptr + 32);
r2 = vld1q_u8(ref_ptr + 32);
diff2 = vabdq_u8(s2, r2);
sum[2] = vpadalq_u8(sum[2], diff2);
s3 = vld1q_u8(src_ptr + 48);
r3 = vld1q_u8(ref_ptr + 48);
diff3 = vabdq_u8(s3, r3);
sum[3] = vpadalq_u8(sum[3], diff3);
src_ptr += src_stride;
ref_ptr += ref_stride;
} while (--i != 0);
sum_u32 = vpaddlq_u16(sum[0]);
sum_u32 = vpadalq_u16(sum_u32, sum[1]);
sum_u32 = vpadalq_u16(sum_u32, sum[2]);
sum_u32 = vpadalq_u16(sum_u32, sum[3]);
return horizontal_add_uint32x4(sum_u32);
}
static INLINE unsigned int sad32xh_neon(const uint8_t *src_ptr, int src_stride,
const uint8_t *ref_ptr, int ref_stride,
int h) {
uint32x4_t sum = vdupq_n_u32(0);
int i = h;
do {
uint8x16_t s0 = vld1q_u8(src_ptr);
uint8x16_t r0 = vld1q_u8(ref_ptr);
uint8x16_t diff0 = vabdq_u8(s0, r0);
uint16x8_t sum0 = vpaddlq_u8(diff0);
uint8x16_t s1 = vld1q_u8(src_ptr + 16);
uint8x16_t r1 = vld1q_u8(ref_ptr + 16);
uint8x16_t diff1 = vabdq_u8(s1, r1);
uint16x8_t sum1 = vpaddlq_u8(diff1);
sum = vpadalq_u16(sum, sum0);
sum = vpadalq_u16(sum, sum1);
src_ptr += src_stride;
ref_ptr += ref_stride;
} while (--i != 0);
return horizontal_add_uint32x4(sum);
}
static INLINE unsigned int sad16xh_neon(const uint8_t *src_ptr, int src_stride,
const uint8_t *ref_ptr, int ref_stride,
int h) {
uint16x8_t sum = vdupq_n_u16(0);
int i = h;
do {
uint8x16_t s = vld1q_u8(src_ptr);
uint8x16_t r = vld1q_u8(ref_ptr);
uint8x16_t diff = vabdq_u8(s, r);
sum = vpadalq_u8(sum, diff);
src_ptr += src_stride;
ref_ptr += ref_stride;
} while (--i != 0);
return horizontal_add_uint16x8(sum);
}
static INLINE unsigned int sad8xh_neon(const uint8_t *src_ptr, int src_stride,
const uint8_t *ref_ptr, int ref_stride,
int h) {
uint16x8_t sum = vdupq_n_u16(0);
int i = h;
do {
uint8x8_t s = vld1_u8(src_ptr);
uint8x8_t r = vld1_u8(ref_ptr);
sum = vabal_u8(sum, s, r);
src_ptr += src_stride;
ref_ptr += ref_stride;
} while (--i != 0);
return horizontal_add_uint16x8(sum);
}
static INLINE unsigned int sad4xh_neon(const uint8_t *src_ptr, int src_stride,
const uint8_t *ref_ptr, int ref_stride,
int h) {
uint16x8_t sum = vdupq_n_u16(0);
int i = h / 2;
do {
uint8x8_t s = load_unaligned_u8(src_ptr, src_stride);
uint8x8_t r = load_unaligned_u8(ref_ptr, ref_stride);
sum = vabal_u8(sum, s, r);
src_ptr += 2 * src_stride;
ref_ptr += 2 * ref_stride;
} while (--i != 0);
return horizontal_add_uint16x8(sum);
}
#define SAD_WXH_NEON(w, h) \
unsigned int vpx_sad##w##x##h##_neon(const uint8_t *src, int src_stride, \
const uint8_t *ref, int ref_stride) { \
return sad##w##xh_neon(src, src_stride, ref, ref_stride, (h)); \
}
SAD_WXH_NEON(4, 4)
SAD_WXH_NEON(4, 8)
SAD_WXH_NEON(8, 4)
SAD_WXH_NEON(8, 8)
SAD_WXH_NEON(8, 16)
SAD_WXH_NEON(16, 8)
SAD_WXH_NEON(16, 16)
SAD_WXH_NEON(16, 32)
SAD_WXH_NEON(32, 16)
SAD_WXH_NEON(32, 32)
SAD_WXH_NEON(32, 64)
SAD_WXH_NEON(64, 32)
SAD_WXH_NEON(64, 64)
#undef SAD_WXH_NEON
#define SAD_SKIP_WXH_NEON(w, h) \
unsigned int vpx_sad_skip_##w##x##h##_neon( \
const uint8_t *src, int src_stride, const uint8_t *ref, \
int ref_stride) { \
return 2 * \
sad##w##xh_neon(src, 2 * src_stride, ref, 2 * ref_stride, (h) / 2); \
}
SAD_SKIP_WXH_NEON(4, 4)
SAD_SKIP_WXH_NEON(4, 8)
SAD_SKIP_WXH_NEON(8, 4)
SAD_SKIP_WXH_NEON(8, 8)
SAD_SKIP_WXH_NEON(8, 16)
SAD_SKIP_WXH_NEON(16, 8)
SAD_SKIP_WXH_NEON(16, 16)
SAD_SKIP_WXH_NEON(16, 32)
SAD_SKIP_WXH_NEON(32, 16)
SAD_SKIP_WXH_NEON(32, 32)
SAD_SKIP_WXH_NEON(32, 64)
SAD_SKIP_WXH_NEON(64, 32)
SAD_SKIP_WXH_NEON(64, 64)
#undef SAD_SKIP_WXH_NEON
static INLINE unsigned int sad64xh_avg_neon(const uint8_t *src_ptr,
int src_stride,
const uint8_t *ref_ptr,
int ref_stride, int h,
const uint8_t *second_pred) {
uint16x8_t sum[4] = { vdupq_n_u16(0), vdupq_n_u16(0), vdupq_n_u16(0),
vdupq_n_u16(0) };
uint32x4_t sum_u32;
int i = h;
do {
uint8x16_t s0, s1, s2, s3, r0, r1, r2, r3, p0, p1, p2, p3;
uint8x16_t avg0, avg1, avg2, avg3, diff0, diff1, diff2, diff3;
s0 = vld1q_u8(src_ptr);
r0 = vld1q_u8(ref_ptr);
p0 = vld1q_u8(second_pred);
avg0 = vrhaddq_u8(r0, p0);
diff0 = vabdq_u8(s0, avg0);
sum[0] = vpadalq_u8(sum[0], diff0);
s1 = vld1q_u8(src_ptr + 16);
r1 = vld1q_u8(ref_ptr + 16);
p1 = vld1q_u8(second_pred + 16);
avg1 = vrhaddq_u8(r1, p1);
diff1 = vabdq_u8(s1, avg1);
sum[1] = vpadalq_u8(sum[1], diff1);
s2 = vld1q_u8(src_ptr + 32);
r2 = vld1q_u8(ref_ptr + 32);
p2 = vld1q_u8(second_pred + 32);
avg2 = vrhaddq_u8(r2, p2);
diff2 = vabdq_u8(s2, avg2);
sum[2] = vpadalq_u8(sum[2], diff2);
s3 = vld1q_u8(src_ptr + 48);
r3 = vld1q_u8(ref_ptr + 48);
p3 = vld1q_u8(second_pred + 48);
avg3 = vrhaddq_u8(r3, p3);
diff3 = vabdq_u8(s3, avg3);
sum[3] = vpadalq_u8(sum[3], diff3);
src_ptr += src_stride;
ref_ptr += ref_stride;
second_pred += 64;
} while (--i != 0);
sum_u32 = vpaddlq_u16(sum[0]);
sum_u32 = vpadalq_u16(sum_u32, sum[1]);
sum_u32 = vpadalq_u16(sum_u32, sum[2]);
sum_u32 = vpadalq_u16(sum_u32, sum[3]);
return horizontal_add_uint32x4(sum_u32);
}
static INLINE unsigned int sad32xh_avg_neon(const uint8_t *src_ptr,
int src_stride,
const uint8_t *ref_ptr,
int ref_stride, int h,
const uint8_t *second_pred) {
uint32x4_t sum = vdupq_n_u32(0);
int i = h;
do {
uint8x16_t s0 = vld1q_u8(src_ptr);
uint8x16_t r0 = vld1q_u8(ref_ptr);
uint8x16_t p0 = vld1q_u8(second_pred);
uint8x16_t avg0 = vrhaddq_u8(r0, p0);
uint8x16_t diff0 = vabdq_u8(s0, avg0);
uint16x8_t sum0 = vpaddlq_u8(diff0);
uint8x16_t s1 = vld1q_u8(src_ptr + 16);
uint8x16_t r1 = vld1q_u8(ref_ptr + 16);
uint8x16_t p1 = vld1q_u8(second_pred + 16);
uint8x16_t avg1 = vrhaddq_u8(r1, p1);
uint8x16_t diff1 = vabdq_u8(s1, avg1);
uint16x8_t sum1 = vpaddlq_u8(diff1);
sum = vpadalq_u16(sum, sum0);
sum = vpadalq_u16(sum, sum1);
src_ptr += src_stride;
ref_ptr += ref_stride;
second_pred += 32;
} while (--i != 0);
return horizontal_add_uint32x4(sum);
}
static INLINE unsigned int sad16xh_avg_neon(const uint8_t *src_ptr,
int src_stride,
const uint8_t *ref_ptr,
int ref_stride, int h,
const uint8_t *second_pred) {
uint16x8_t sum = vdupq_n_u16(0);
int i = h;
do {
uint8x16_t s = vld1q_u8(src_ptr);
uint8x16_t r = vld1q_u8(ref_ptr);
uint8x16_t p = vld1q_u8(second_pred);
uint8x16_t avg = vrhaddq_u8(r, p);
uint8x16_t diff = vabdq_u8(s, avg);
sum = vpadalq_u8(sum, diff);
src_ptr += src_stride;
ref_ptr += ref_stride;
second_pred += 16;
} while (--i != 0);
return horizontal_add_uint16x8(sum);
}
static INLINE unsigned int sad8xh_avg_neon(const uint8_t *src_ptr,
int src_stride,
const uint8_t *ref_ptr,
int ref_stride, int h,
const uint8_t *second_pred) {
uint16x8_t sum = vdupq_n_u16(0);
int i = h;
do {
uint8x8_t s = vld1_u8(src_ptr);
uint8x8_t r = vld1_u8(ref_ptr);
uint8x8_t p = vld1_u8(second_pred);
uint8x8_t avg = vrhadd_u8(r, p);
sum = vabal_u8(sum, s, avg);
src_ptr += src_stride;
ref_ptr += ref_stride;
second_pred += 8;
} while (--i != 0);
return horizontal_add_uint16x8(sum);
}
static INLINE unsigned int sad4xh_avg_neon(const uint8_t *src_ptr,
int src_stride,
const uint8_t *ref_ptr,
int ref_stride, int h,
const uint8_t *second_pred) {
uint16x8_t sum = vdupq_n_u16(0);
int i = h / 2;
do {
uint8x8_t s = load_unaligned_u8(src_ptr, src_stride);
uint8x8_t r = load_unaligned_u8(ref_ptr, ref_stride);
uint8x8_t p = vld1_u8(second_pred);
uint8x8_t avg = vrhadd_u8(r, p);
sum = vabal_u8(sum, s, avg);
src_ptr += 2 * src_stride;
ref_ptr += 2 * ref_stride;
second_pred += 8;
} while (--i != 0);
return horizontal_add_uint16x8(sum);
}
#define SAD_WXH_AVG_NEON(w, h) \
uint32_t vpx_sad##w##x##h##_avg_neon(const uint8_t *src, int src_stride, \
const uint8_t *ref, int ref_stride, \
const uint8_t *second_pred) { \
return sad##w##xh_avg_neon(src, src_stride, ref, ref_stride, (h), \
second_pred); \
}
SAD_WXH_AVG_NEON(4, 4)
SAD_WXH_AVG_NEON(4, 8)
SAD_WXH_AVG_NEON(8, 4)
SAD_WXH_AVG_NEON(8, 8)
SAD_WXH_AVG_NEON(8, 16)
SAD_WXH_AVG_NEON(16, 8)
SAD_WXH_AVG_NEON(16, 16)
SAD_WXH_AVG_NEON(16, 32)
SAD_WXH_AVG_NEON(32, 16)
SAD_WXH_AVG_NEON(32, 32)
SAD_WXH_AVG_NEON(32, 64)
SAD_WXH_AVG_NEON(64, 32)
SAD_WXH_AVG_NEON(64, 64)
#undef SAD_WXH_AVG_NEON
|
