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
|
/*
* Copyright (c) 2016, Alliance for Open Media. All rights reserved
*
* This source code is subject to the terms of the BSD 2 Clause License and
* the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
* was not distributed with this source code in the LICENSE file, you can
* obtain it at www.aomedia.org/license/software. If the Alliance for Open
* Media Patent License 1.0 was not distributed with this source code in the
* PATENTS file, you can obtain it at www.aomedia.org/license/patent.
*/
#include <immintrin.h>
#include "./aom_dsp_rtcd.h"
typedef void (*get_var_avx2)(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse, int *sum);
void aom_get32x32var_avx2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride, unsigned int *sse,
int *sum);
static void variance_avx2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride, int w, int h,
unsigned int *sse, int *sum, get_var_avx2 var_fn,
int block_size) {
int i, j;
*sse = 0;
*sum = 0;
for (i = 0; i < h; i += 16) {
for (j = 0; j < w; j += block_size) {
unsigned int sse0;
int sum0;
var_fn(&src[src_stride * i + j], src_stride, &ref[ref_stride * i + j],
ref_stride, &sse0, &sum0);
*sse += sse0;
*sum += sum0;
}
}
}
unsigned int aom_variance16x16_avx2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
unsigned int variance;
variance_avx2(src, src_stride, ref, ref_stride, 16, 16, sse, &sum,
aom_get16x16var_avx2, 16);
variance = *sse - (((uint32_t)((int64_t)sum * sum)) >> 8);
_mm256_zeroupper();
return variance;
}
unsigned int aom_mse16x16_avx2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
aom_get16x16var_avx2(src, src_stride, ref, ref_stride, sse, &sum);
_mm256_zeroupper();
return *sse;
}
unsigned int aom_variance32x16_avx2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
unsigned int variance;
variance_avx2(src, src_stride, ref, ref_stride, 32, 16, sse, &sum,
aom_get32x32var_avx2, 32);
variance = *sse - (uint32_t)(((int64_t)sum * sum) >> 9);
_mm256_zeroupper();
return variance;
}
unsigned int aom_variance32x32_avx2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
unsigned int variance;
variance_avx2(src, src_stride, ref, ref_stride, 32, 32, sse, &sum,
aom_get32x32var_avx2, 32);
variance = *sse - (uint32_t)(((int64_t)sum * sum) >> 10);
_mm256_zeroupper();
return variance;
}
unsigned int aom_variance64x64_avx2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
unsigned int variance;
variance_avx2(src, src_stride, ref, ref_stride, 64, 64, sse, &sum,
aom_get32x32var_avx2, 32);
variance = *sse - (uint32_t)(((int64_t)sum * sum) >> 12);
_mm256_zeroupper();
return variance;
}
unsigned int aom_variance64x32_avx2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
unsigned int variance;
variance_avx2(src, src_stride, ref, ref_stride, 64, 32, sse, &sum,
aom_get32x32var_avx2, 32);
variance = *sse - (uint32_t)(((int64_t)sum * sum) >> 11);
_mm256_zeroupper();
return variance;
}
unsigned int aom_sub_pixel_variance32xh_avx2(const uint8_t *src, int src_stride,
int x_offset, int y_offset,
const uint8_t *dst, int dst_stride,
int height, unsigned int *sse);
unsigned int aom_sub_pixel_avg_variance32xh_avx2(
const uint8_t *src, int src_stride, int x_offset, int y_offset,
const uint8_t *dst, int dst_stride, const uint8_t *sec, int sec_stride,
int height, unsigned int *sseptr);
unsigned int aom_sub_pixel_variance64x64_avx2(const uint8_t *src,
int src_stride, int x_offset,
int y_offset, const uint8_t *dst,
int dst_stride,
unsigned int *sse) {
unsigned int sse1;
const int se1 = aom_sub_pixel_variance32xh_avx2(
src, src_stride, x_offset, y_offset, dst, dst_stride, 64, &sse1);
unsigned int sse2;
const int se2 =
aom_sub_pixel_variance32xh_avx2(src + 32, src_stride, x_offset, y_offset,
dst + 32, dst_stride, 64, &sse2);
const int se = se1 + se2;
unsigned int variance;
*sse = sse1 + sse2;
variance = *sse - (uint32_t)(((int64_t)se * se) >> 12);
_mm256_zeroupper();
return variance;
}
unsigned int aom_sub_pixel_variance32x32_avx2(const uint8_t *src,
int src_stride, int x_offset,
int y_offset, const uint8_t *dst,
int dst_stride,
unsigned int *sse) {
const int se = aom_sub_pixel_variance32xh_avx2(
src, src_stride, x_offset, y_offset, dst, dst_stride, 32, sse);
const unsigned int variance = *sse - (uint32_t)(((int64_t)se * se) >> 10);
_mm256_zeroupper();
return variance;
}
unsigned int aom_sub_pixel_avg_variance64x64_avx2(
const uint8_t *src, int src_stride, int x_offset, int y_offset,
const uint8_t *dst, int dst_stride, unsigned int *sse, const uint8_t *sec) {
unsigned int sse1;
const int se1 = aom_sub_pixel_avg_variance32xh_avx2(
src, src_stride, x_offset, y_offset, dst, dst_stride, sec, 64, 64, &sse1);
unsigned int sse2;
const int se2 = aom_sub_pixel_avg_variance32xh_avx2(
src + 32, src_stride, x_offset, y_offset, dst + 32, dst_stride, sec + 32,
64, 64, &sse2);
const int se = se1 + se2;
unsigned int variance;
*sse = sse1 + sse2;
variance = *sse - (uint32_t)(((int64_t)se * se) >> 12);
_mm256_zeroupper();
return variance;
}
unsigned int aom_sub_pixel_avg_variance32x32_avx2(
const uint8_t *src, int src_stride, int x_offset, int y_offset,
const uint8_t *dst, int dst_stride, unsigned int *sse, const uint8_t *sec) {
// Process 32 elements in parallel.
const int se = aom_sub_pixel_avg_variance32xh_avx2(
src, src_stride, x_offset, y_offset, dst, dst_stride, sec, 32, 32, sse);
const unsigned int variance = *sse - (uint32_t)(((int64_t)se * se) >> 10);
_mm256_zeroupper();
return variance;
}
|
