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
|
// Copyright 2019 The libgav1 Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "src/dsp/average_blend.h"
#include "src/utils/cpu.h"
#if LIBGAV1_ENABLE_NEON
#include <arm_neon.h>
#include <cassert>
#include <cstddef>
#include <cstdint>
#include "src/dsp/arm/common_neon.h"
#include "src/dsp/constants.h"
#include "src/dsp/dsp.h"
#include "src/utils/common.h"
namespace libgav1 {
namespace dsp {
namespace {
constexpr int kInterPostRoundBit =
kInterRoundBitsVertical - kInterRoundBitsCompoundVertical;
} // namespace
namespace low_bitdepth {
namespace {
inline uint8x8_t AverageBlend8Row(const int16_t* prediction_0,
const int16_t* prediction_1) {
const int16x8_t pred0 = vld1q_s16(prediction_0);
const int16x8_t pred1 = vld1q_s16(prediction_1);
const int16x8_t res = vaddq_s16(pred0, pred1);
return vqrshrun_n_s16(res, kInterPostRoundBit + 1);
}
inline void AverageBlendLargeRow(const int16_t* prediction_0,
const int16_t* prediction_1, const int width,
uint8_t* dest) {
int x = width;
do {
const int16x8_t pred_00 = vld1q_s16(prediction_0);
const int16x8_t pred_01 = vld1q_s16(prediction_1);
prediction_0 += 8;
prediction_1 += 8;
const int16x8_t res0 = vaddq_s16(pred_00, pred_01);
const uint8x8_t res_out0 = vqrshrun_n_s16(res0, kInterPostRoundBit + 1);
const int16x8_t pred_10 = vld1q_s16(prediction_0);
const int16x8_t pred_11 = vld1q_s16(prediction_1);
prediction_0 += 8;
prediction_1 += 8;
const int16x8_t res1 = vaddq_s16(pred_10, pred_11);
const uint8x8_t res_out1 = vqrshrun_n_s16(res1, kInterPostRoundBit + 1);
vst1q_u8(dest, vcombine_u8(res_out0, res_out1));
dest += 16;
x -= 16;
} while (x != 0);
}
void AverageBlend_NEON(const void* prediction_0, const void* prediction_1,
const int width, const int height, void* const dest,
const ptrdiff_t dest_stride) {
auto* dst = static_cast<uint8_t*>(dest);
const auto* pred_0 = static_cast<const int16_t*>(prediction_0);
const auto* pred_1 = static_cast<const int16_t*>(prediction_1);
int y = height;
if (width == 4) {
do {
const uint8x8_t result = AverageBlend8Row(pred_0, pred_1);
pred_0 += 8;
pred_1 += 8;
StoreLo4(dst, result);
dst += dest_stride;
StoreHi4(dst, result);
dst += dest_stride;
y -= 2;
} while (y != 0);
return;
}
if (width == 8) {
do {
vst1_u8(dst, AverageBlend8Row(pred_0, pred_1));
dst += dest_stride;
pred_0 += 8;
pred_1 += 8;
vst1_u8(dst, AverageBlend8Row(pred_0, pred_1));
dst += dest_stride;
pred_0 += 8;
pred_1 += 8;
y -= 2;
} while (y != 0);
return;
}
do {
AverageBlendLargeRow(pred_0, pred_1, width, dst);
dst += dest_stride;
pred_0 += width;
pred_1 += width;
AverageBlendLargeRow(pred_0, pred_1, width, dst);
dst += dest_stride;
pred_0 += width;
pred_1 += width;
y -= 2;
} while (y != 0);
}
void Init8bpp() {
Dsp* const dsp = dsp_internal::GetWritableDspTable(kBitdepth8);
assert(dsp != nullptr);
dsp->average_blend = AverageBlend_NEON;
}
} // namespace
} // namespace low_bitdepth
#if LIBGAV1_MAX_BITDEPTH >= 10
namespace high_bitdepth {
namespace {
inline uint16x8_t AverageBlend8Row(const uint16_t* prediction_0,
const uint16_t* prediction_1,
const int32x4_t compound_offset,
const uint16x8_t v_bitdepth) {
const uint16x8_t pred0 = vld1q_u16(prediction_0);
const uint16x8_t pred1 = vld1q_u16(prediction_1);
const uint32x4_t pred_lo =
vaddl_u16(vget_low_u16(pred0), vget_low_u16(pred1));
const uint32x4_t pred_hi =
vaddl_u16(vget_high_u16(pred0), vget_high_u16(pred1));
const int32x4_t offset_lo =
vsubq_s32(vreinterpretq_s32_u32(pred_lo), compound_offset);
const int32x4_t offset_hi =
vsubq_s32(vreinterpretq_s32_u32(pred_hi), compound_offset);
const uint16x4_t res_lo = vqrshrun_n_s32(offset_lo, kInterPostRoundBit + 1);
const uint16x4_t res_hi = vqrshrun_n_s32(offset_hi, kInterPostRoundBit + 1);
return vminq_u16(vcombine_u16(res_lo, res_hi), v_bitdepth);
}
inline void AverageBlendLargeRow(const uint16_t* prediction_0,
const uint16_t* prediction_1, const int width,
uint16_t* dest,
const int32x4_t compound_offset,
const uint16x8_t v_bitdepth) {
int x = width;
do {
vst1q_u16(dest, AverageBlend8Row(prediction_0, prediction_1,
compound_offset, v_bitdepth));
prediction_0 += 8;
prediction_1 += 8;
dest += 8;
vst1q_u16(dest, AverageBlend8Row(prediction_0, prediction_1,
compound_offset, v_bitdepth));
prediction_0 += 8;
prediction_1 += 8;
dest += 8;
x -= 16;
} while (x != 0);
}
void AverageBlend_NEON(const void* prediction_0, const void* prediction_1,
const int width, const int height, void* const dest,
const ptrdiff_t dest_stride) {
auto* dst = static_cast<uint16_t*>(dest);
const auto* pred_0 = static_cast<const uint16_t*>(prediction_0);
const auto* pred_1 = static_cast<const uint16_t*>(prediction_1);
int y = height;
const ptrdiff_t dst_stride = dest_stride >> 1;
const int32x4_t compound_offset =
vdupq_n_s32(static_cast<int32_t>(kCompoundOffset + kCompoundOffset));
const uint16x8_t v_bitdepth = vdupq_n_u16((1 << kBitdepth10) - 1);
if (width == 4) {
do {
const uint16x8_t result =
AverageBlend8Row(pred_0, pred_1, compound_offset, v_bitdepth);
pred_0 += 8;
pred_1 += 8;
vst1_u16(dst, vget_low_u16(result));
dst += dst_stride;
vst1_u16(dst, vget_high_u16(result));
dst += dst_stride;
y -= 2;
} while (y != 0);
return;
}
if (width == 8) {
do {
vst1q_u16(dst,
AverageBlend8Row(pred_0, pred_1, compound_offset, v_bitdepth));
dst += dst_stride;
pred_0 += 8;
pred_1 += 8;
vst1q_u16(dst,
AverageBlend8Row(pred_0, pred_1, compound_offset, v_bitdepth));
dst += dst_stride;
pred_0 += 8;
pred_1 += 8;
y -= 2;
} while (y != 0);
return;
}
do {
AverageBlendLargeRow(pred_0, pred_1, width, dst, compound_offset,
v_bitdepth);
dst += dst_stride;
pred_0 += width;
pred_1 += width;
AverageBlendLargeRow(pred_0, pred_1, width, dst, compound_offset,
v_bitdepth);
dst += dst_stride;
pred_0 += width;
pred_1 += width;
y -= 2;
} while (y != 0);
}
void Init10bpp() {
Dsp* const dsp = dsp_internal::GetWritableDspTable(kBitdepth10);
assert(dsp != nullptr);
dsp->average_blend = AverageBlend_NEON;
}
} // namespace
} // namespace high_bitdepth
#endif // LIBGAV1_MAX_BITDEPTH >= 10
void AverageBlendInit_NEON() {
low_bitdepth::Init8bpp();
#if LIBGAV1_MAX_BITDEPTH >= 10
high_bitdepth::Init10bpp();
#endif // LIBGAV1_MAX_BITDEPTH >= 10
}
} // namespace dsp
} // namespace libgav1
#else // !LIBGAV1_ENABLE_NEON
namespace libgav1 {
namespace dsp {
void AverageBlendInit_NEON() {}
} // namespace dsp
} // namespace libgav1
#endif // LIBGAV1_ENABLE_NEON
|
