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
|
/*
* Copyright (c) 2018, 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.
*/
#ifndef AOM_AV1_ENCODER_X86_AV1_TXFM1D_SSE4_H_
#define AOM_AV1_ENCODER_X86_AV1_TXFM1D_SSE4_H_
#include <smmintrin.h>
#include "av1/common/av1_txfm.h"
#include "av1/common/x86/av1_txfm_sse4.h"
#ifdef __cplusplus
extern "C" {
#endif
void av1_fdct4_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_fdct8_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_fdct16_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_fdct32_new_sse4_1(const __m128i *input, __m128i *output,
int8_t cos_bit);
void av1_fdct64_new_sse4_1(const __m128i *input, __m128i *output,
int8_t cos_bit, const int instride,
const int outstride);
void av1_fadst4_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_fadst8_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_fadst16_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_idct4_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_idct8_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_idct16_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_idct32_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_idct64_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_iadst4_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_iadst8_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
void av1_iadst16_new_sse4_1(const __m128i *input, __m128i *output,
const int8_t cos_bit, const int8_t *stage_range);
static INLINE void transpose_32_4x4(int stride, const __m128i *input,
__m128i *output) {
__m128i temp0 = _mm_unpacklo_epi32(input[0 * stride], input[2 * stride]);
__m128i temp1 = _mm_unpackhi_epi32(input[0 * stride], input[2 * stride]);
__m128i temp2 = _mm_unpacklo_epi32(input[1 * stride], input[3 * stride]);
__m128i temp3 = _mm_unpackhi_epi32(input[1 * stride], input[3 * stride]);
output[0 * stride] = _mm_unpacklo_epi32(temp0, temp2);
output[1 * stride] = _mm_unpackhi_epi32(temp0, temp2);
output[2 * stride] = _mm_unpacklo_epi32(temp1, temp3);
output[3 * stride] = _mm_unpackhi_epi32(temp1, temp3);
}
// the entire input block can be represent by a grid of 4x4 blocks
// each 4x4 blocks can be represent by 4 vertical __m128i
// we first transpose each 4x4 block internally
// then transpose the grid
static INLINE void transpose_32(int txfm_size, const __m128i *input,
__m128i *output) {
const int num_per_128 = 4;
const int row_size = txfm_size;
const int col_size = txfm_size / num_per_128;
int r, c;
// transpose each 4x4 block internally
for (r = 0; r < row_size; r += 4) {
for (c = 0; c < col_size; c++) {
transpose_32_4x4(col_size, &input[r * col_size + c],
&output[c * 4 * col_size + r / 4]);
}
}
}
// out0 = in0*w0 + in1*w1
// out1 = -in1*w0 + in0*w1
#define btf_32_sse4_1_type0(w0, w1, in0, in1, out0, out1, bit) \
do { \
const __m128i ww0 = _mm_set1_epi32(w0); \
const __m128i ww1 = _mm_set1_epi32(w1); \
const __m128i in0_w0 = _mm_mullo_epi32(in0, ww0); \
const __m128i in1_w1 = _mm_mullo_epi32(in1, ww1); \
out0 = _mm_add_epi32(in0_w0, in1_w1); \
out0 = av1_round_shift_32_sse4_1(out0, bit); \
const __m128i in0_w1 = _mm_mullo_epi32(in0, ww1); \
const __m128i in1_w0 = _mm_mullo_epi32(in1, ww0); \
out1 = _mm_sub_epi32(in0_w1, in1_w0); \
out1 = av1_round_shift_32_sse4_1(out1, bit); \
} while (0)
// out0 = in0*w0 + in1*w1
// out1 = in1*w0 - in0*w1
#define btf_32_sse4_1_type1(w0, w1, in0, in1, out0, out1, bit) \
do { \
btf_32_sse4_1_type0(w1, w0, in1, in0, out0, out1, bit); \
} while (0)
// out0 = in0*w0 + in1*w1
// out1 = -in1*w0 + in0*w1
#define btf_32_type0_sse4_1_new(ww0, ww1, in0, in1, out0, out1, r, bit) \
do { \
const __m128i in0_w0 = _mm_mullo_epi32(in0, ww0); \
const __m128i in1_w1 = _mm_mullo_epi32(in1, ww1); \
out0 = _mm_add_epi32(in0_w0, in1_w1); \
out0 = _mm_add_epi32(out0, r); \
out0 = _mm_srai_epi32(out0, bit); \
const __m128i in0_w1 = _mm_mullo_epi32(in0, ww1); \
const __m128i in1_w0 = _mm_mullo_epi32(in1, ww0); \
out1 = _mm_sub_epi32(in0_w1, in1_w0); \
out1 = _mm_add_epi32(out1, r); \
out1 = _mm_srai_epi32(out1, bit); \
} while (0)
// out0 = in0*w0 + in1*w1
// out1 = in1*w0 - in0*w1
#define btf_32_type1_sse4_1_new(ww0, ww1, in0, in1, out0, out1, r, bit) \
do { \
btf_32_type0_sse4_1_new(ww1, ww0, in1, in0, out0, out1, r, bit); \
} while (0)
#ifdef __cplusplus
}
#endif
#endif // AOM_AV1_ENCODER_X86_AV1_TXFM1D_SSE4_H_
|
