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
|
/*
* 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 <assert.h>
#include <emmintrin.h> // SSE2
#include <tmmintrin.h>
#include "config/aom_config.h"
#include "config/aom_dsp_rtcd.h"
#include "config/av1_rtcd.h"
#include "aom_dsp/x86/synonyms.h"
unsigned int aom_sad4xh_sse2(const uint8_t *a, int a_stride, const uint8_t *b,
int b_stride, int width, int height) {
int i;
assert(width == 4);
(void)width;
__m128i sad = _mm_setzero_si128();
for (i = 0; i < height; i += 4) {
__m128i x0 = xx_loadl_32(a + 0 * a_stride);
__m128i x1 = xx_loadl_32(a + 1 * a_stride);
__m128i x2 = xx_loadl_32(a + 2 * a_stride);
__m128i x3 = xx_loadl_32(a + 3 * a_stride);
__m128i x_lo = _mm_unpacklo_epi32(x0, x1);
__m128i x_hi = _mm_unpacklo_epi32(x2, x3);
__m128i x = _mm_unpacklo_epi64(x_lo, x_hi);
x0 = xx_loadl_32(b + 0 * b_stride);
x1 = xx_loadl_32(b + 1 * b_stride);
x2 = xx_loadl_32(b + 2 * b_stride);
x3 = xx_loadl_32(b + 3 * b_stride);
x_lo = _mm_unpacklo_epi32(x0, x1);
x_hi = _mm_unpacklo_epi32(x2, x3);
__m128i y = _mm_unpacklo_epi64(x_lo, x_hi);
__m128i sad4x4 = _mm_sad_epu8(x, y);
sad = _mm_add_epi32(sad, sad4x4);
a += 4 * a_stride;
b += 4 * b_stride;
}
// At this point, we have two 32-bit partial SADs at bit[0:31] and [64:95].
const unsigned int res =
_mm_cvtsi128_si32(sad) + _mm_cvtsi128_si32(_mm_srli_si128(sad, 8));
return res;
}
unsigned int aom_sad8xh_sse2(const uint8_t *a, int a_stride, const uint8_t *b,
int b_stride, int width, int height) {
int i;
assert(width == 8);
(void)width;
__m128i sad = _mm_setzero_si128();
for (i = 0; i < height; i += 2) {
__m128i x0 = xx_loadl_64(a + 0 * a_stride);
__m128i x1 = xx_loadl_64(a + 1 * a_stride);
__m128i x = _mm_unpacklo_epi64(x0, x1);
x0 = xx_loadl_64(b + 0 * b_stride);
x1 = xx_loadl_64(b + 1 * b_stride);
__m128i y = _mm_unpacklo_epi64(x0, x1);
__m128i sad8x2 = _mm_sad_epu8(x, y);
sad = _mm_add_epi32(sad, sad8x2);
a += 2 * a_stride;
b += 2 * b_stride;
}
const unsigned int res =
_mm_cvtsi128_si32(sad) + _mm_cvtsi128_si32(_mm_srli_si128(sad, 8));
return res;
}
unsigned int aom_sad16xh_sse2(const uint8_t *a, int a_stride, const uint8_t *b,
int b_stride, int width, int height) {
int i;
assert(width == 16);
(void)width;
__m128i sad = _mm_setzero_si128();
for (i = 0; i < height; ++i) {
__m128i x = xx_loadu_128(a);
__m128i y = xx_loadu_128(b);
__m128i sad16x1 = _mm_sad_epu8(x, y);
sad = _mm_add_epi32(sad, sad16x1);
a += a_stride;
b += b_stride;
}
const unsigned int res =
_mm_cvtsi128_si32(sad) + _mm_cvtsi128_si32(_mm_srli_si128(sad, 8));
return res;
}
unsigned int aom_sad32xh_sse2(const uint8_t *a, int a_stride, const uint8_t *b,
int b_stride, int width, int height) {
int i, j;
assert(width == 32);
(void)width;
__m128i sad = _mm_setzero_si128();
for (i = 0; i < height; ++i) {
for (j = 0; j < 2; ++j) {
__m128i x = xx_loadu_128(a + j * 16);
__m128i y = xx_loadu_128(b + j * 16);
__m128i sad32_half = _mm_sad_epu8(x, y);
sad = _mm_add_epi32(sad, sad32_half);
}
a += a_stride;
b += b_stride;
}
const unsigned int res =
_mm_cvtsi128_si32(sad) + _mm_cvtsi128_si32(_mm_srli_si128(sad, 8));
return res;
}
unsigned int aom_sad64xh_sse2(const uint8_t *a, int a_stride, const uint8_t *b,
int b_stride, int width, int height) {
int i, j;
assert(width == 64);
(void)width;
__m128i sad = _mm_setzero_si128();
for (i = 0; i < height; ++i) {
for (j = 0; j < 4; ++j) {
__m128i x = xx_loadu_128(a + j * 16);
__m128i y = xx_loadu_128(b + j * 16);
__m128i sad64_quarter = _mm_sad_epu8(x, y);
sad = _mm_add_epi32(sad, sad64_quarter);
}
a += a_stride;
b += b_stride;
}
const unsigned int res =
_mm_cvtsi128_si32(sad) + _mm_cvtsi128_si32(_mm_srli_si128(sad, 8));
return res;
}
unsigned int aom_sad128xh_sse2(const uint8_t *a, int a_stride, const uint8_t *b,
int b_stride, int width, int height) {
int i, j;
assert(width == 128);
(void)width;
__m128i sad = _mm_setzero_si128();
for (i = 0; i < height; ++i) {
for (j = 0; j < 8; ++j) {
__m128i x = xx_loadu_128(a + j * 16);
__m128i y = xx_loadu_128(b + j * 16);
__m128i sad64_quarter = _mm_sad_epu8(x, y);
sad = _mm_add_epi32(sad, sad64_quarter);
}
a += a_stride;
b += b_stride;
}
const unsigned int res =
_mm_cvtsi128_si32(sad) + _mm_cvtsi128_si32(_mm_srli_si128(sad, 8));
return res;
}
#define jnt_sadMxN_sse2(m, n) \
unsigned int aom_jnt_sad##m##x##n##_avg_ssse3( \
const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, \
const uint8_t *second_pred, const JNT_COMP_PARAMS *jcp_param) { \
uint8_t comp_pred[m * n]; \
aom_jnt_comp_avg_pred(comp_pred, second_pred, m, n, ref, ref_stride, \
jcp_param); \
return aom_sad##m##xh_sse2(src, src_stride, comp_pred, m, m, n); \
}
#define jnt_sadMxN_avx2(m, n) \
unsigned int aom_jnt_sad##m##x##n##_avg_avx2( \
const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, \
const uint8_t *second_pred, const JNT_COMP_PARAMS *jcp_param) { \
uint8_t comp_pred[m * n]; \
aom_jnt_comp_avg_pred(comp_pred, second_pred, m, n, ref, ref_stride, \
jcp_param); \
return aom_sad##m##xh_avx2(src, src_stride, comp_pred, m, m, n); \
}
/* clang-format off */
jnt_sadMxN_sse2(128, 128)
jnt_sadMxN_sse2(128, 64)
jnt_sadMxN_sse2(64, 128)
jnt_sadMxN_sse2(64, 64)
jnt_sadMxN_sse2(64, 32)
jnt_sadMxN_sse2(32, 64)
jnt_sadMxN_sse2(32, 32)
jnt_sadMxN_sse2(32, 16)
jnt_sadMxN_sse2(16, 32)
jnt_sadMxN_sse2(16, 16)
jnt_sadMxN_sse2(16, 8)
jnt_sadMxN_sse2(8, 16)
jnt_sadMxN_sse2(8, 8)
jnt_sadMxN_sse2(8, 4)
jnt_sadMxN_sse2(4, 8)
jnt_sadMxN_sse2(4, 4)
jnt_sadMxN_sse2(4, 16)
jnt_sadMxN_sse2(16, 4)
jnt_sadMxN_sse2(8, 32)
jnt_sadMxN_sse2(32, 8)
jnt_sadMxN_sse2(16, 64)
jnt_sadMxN_sse2(64, 16)
/* clang-format on */
|
