Line data Source code
1 : /*
2 : * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3 : *
4 : * This source code is subject to the terms of the BSD 2 Clause License and
5 : * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6 : * was not distributed with this source code in the LICENSE file, you can
7 : * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8 : * Media Patent License 1.0 was not distributed with this source code in the
9 : * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 : */
11 :
12 : #ifndef AV1_TXFM_H_
13 : #define AV1_TXFM_H_
14 :
15 : #include <assert.h>
16 : #include <math.h>
17 : #include <stdio.h>
18 :
19 : #include "av1/common/enums.h"
20 : #include "aom/aom_integer.h"
21 : #include "aom_dsp/aom_dsp_common.h"
22 :
23 : static const int cos_bit_min = 10;
24 : static const int cos_bit_max = 16;
25 :
26 : // cospi_arr[i][j] = (int)round(cos(M_PI*j/128) * (1<<(cos_bit_min+i)));
27 : static const int32_t cospi_arr_data[7][64] = {
28 : { 1024, 1024, 1023, 1021, 1019, 1016, 1013, 1009, 1004, 999, 993, 987, 980,
29 : 972, 964, 955, 946, 936, 926, 915, 903, 891, 878, 865, 851, 837,
30 : 822, 807, 792, 775, 759, 742, 724, 706, 688, 669, 650, 630, 610,
31 : 590, 569, 548, 526, 505, 483, 460, 438, 415, 392, 369, 345, 321,
32 : 297, 273, 249, 224, 200, 175, 150, 125, 100, 75, 50, 25 },
33 : { 2048, 2047, 2046, 2042, 2038, 2033, 2026, 2018, 2009, 1998, 1987,
34 : 1974, 1960, 1945, 1928, 1911, 1892, 1872, 1851, 1829, 1806, 1782,
35 : 1757, 1730, 1703, 1674, 1645, 1615, 1583, 1551, 1517, 1483, 1448,
36 : 1412, 1375, 1338, 1299, 1260, 1220, 1179, 1138, 1096, 1053, 1009,
37 : 965, 921, 876, 830, 784, 737, 690, 642, 595, 546, 498,
38 : 449, 400, 350, 301, 251, 201, 151, 100, 50 },
39 : { 4096, 4095, 4091, 4085, 4076, 4065, 4052, 4036, 4017, 3996, 3973,
40 : 3948, 3920, 3889, 3857, 3822, 3784, 3745, 3703, 3659, 3612, 3564,
41 : 3513, 3461, 3406, 3349, 3290, 3229, 3166, 3102, 3035, 2967, 2896,
42 : 2824, 2751, 2675, 2598, 2520, 2440, 2359, 2276, 2191, 2106, 2019,
43 : 1931, 1842, 1751, 1660, 1567, 1474, 1380, 1285, 1189, 1092, 995,
44 : 897, 799, 700, 601, 501, 401, 301, 201, 101 },
45 : { 8192, 8190, 8182, 8170, 8153, 8130, 8103, 8071, 8035, 7993, 7946,
46 : 7895, 7839, 7779, 7713, 7643, 7568, 7489, 7405, 7317, 7225, 7128,
47 : 7027, 6921, 6811, 6698, 6580, 6458, 6333, 6203, 6070, 5933, 5793,
48 : 5649, 5501, 5351, 5197, 5040, 4880, 4717, 4551, 4383, 4212, 4038,
49 : 3862, 3683, 3503, 3320, 3135, 2948, 2760, 2570, 2378, 2185, 1990,
50 : 1795, 1598, 1401, 1202, 1003, 803, 603, 402, 201 },
51 : { 16384, 16379, 16364, 16340, 16305, 16261, 16207, 16143, 16069, 15986, 15893,
52 : 15791, 15679, 15557, 15426, 15286, 15137, 14978, 14811, 14635, 14449, 14256,
53 : 14053, 13842, 13623, 13395, 13160, 12916, 12665, 12406, 12140, 11866, 11585,
54 : 11297, 11003, 10702, 10394, 10080, 9760, 9434, 9102, 8765, 8423, 8076,
55 : 7723, 7366, 7005, 6639, 6270, 5897, 5520, 5139, 4756, 4370, 3981,
56 : 3590, 3196, 2801, 2404, 2006, 1606, 1205, 804, 402 },
57 : { 32768, 32758, 32729, 32679, 32610, 32522, 32413, 32286, 32138, 31972, 31786,
58 : 31581, 31357, 31114, 30853, 30572, 30274, 29957, 29622, 29269, 28899, 28511,
59 : 28106, 27684, 27246, 26791, 26320, 25833, 25330, 24812, 24279, 23732, 23170,
60 : 22595, 22006, 21403, 20788, 20160, 19520, 18868, 18205, 17531, 16846, 16151,
61 : 15447, 14733, 14010, 13279, 12540, 11793, 11039, 10279, 9512, 8740, 7962,
62 : 7180, 6393, 5602, 4808, 4011, 3212, 2411, 1608, 804 },
63 : { 65536, 65516, 65457, 65358, 65220, 65043, 64827, 64571, 64277, 63944, 63572,
64 : 63162, 62714, 62228, 61705, 61145, 60547, 59914, 59244, 58538, 57798, 57022,
65 : 56212, 55368, 54491, 53581, 52639, 51665, 50660, 49624, 48559, 47464, 46341,
66 : 45190, 44011, 42806, 41576, 40320, 39040, 37736, 36410, 35062, 33692, 32303,
67 : 30893, 29466, 28020, 26558, 25080, 23586, 22078, 20557, 19024, 17479, 15924,
68 : 14359, 12785, 11204, 9616, 8022, 6424, 4821, 3216, 1608 }
69 : };
70 :
71 0 : static INLINE const int32_t *cospi_arr(int n) {
72 0 : return cospi_arr_data[n - cos_bit_min];
73 : }
74 :
75 0 : static INLINE int32_t round_shift(int32_t value, int bit) {
76 0 : assert(bit >= 1);
77 0 : return (value + (1 << (bit - 1))) >> bit;
78 : }
79 :
80 0 : static INLINE void round_shift_array(int32_t *arr, int size, int bit) {
81 : int i;
82 0 : if (bit == 0) {
83 0 : return;
84 : } else {
85 0 : if (bit > 0) {
86 0 : for (i = 0; i < size; i++) {
87 0 : arr[i] = round_shift(arr[i], bit);
88 : }
89 : } else {
90 0 : for (i = 0; i < size; i++) {
91 0 : arr[i] = arr[i] * (1 << (-bit));
92 : }
93 : }
94 : }
95 : }
96 :
97 0 : static INLINE int32_t half_btf(int32_t w0, int32_t in0, int32_t w1, int32_t in1,
98 : int bit) {
99 0 : int32_t result_32 = w0 * in0 + w1 * in1;
100 : #if CONFIG_COEFFICIENT_RANGE_CHECKING
101 : int64_t result_64 = (int64_t)w0 * (int64_t)in0 + (int64_t)w1 * (int64_t)in1;
102 : if (result_64 < INT32_MIN || result_64 > INT32_MAX) {
103 : printf("%s %d overflow result_32: %d result_64: %" PRId64
104 : " w0: %d in0: %d w1: %d in1: "
105 : "%d\n",
106 : __FILE__, __LINE__, result_32, result_64, w0, in0, w1, in1);
107 : assert(0 && "half_btf overflow");
108 : }
109 : #endif
110 0 : return round_shift(result_32, bit);
111 : }
112 :
113 : static INLINE int get_max_bit(int x) {
114 : int max_bit = -1;
115 : while (x) {
116 : x = x >> 1;
117 : max_bit++;
118 : }
119 : return max_bit;
120 : }
121 :
122 : // TODO(angiebird): implement SSE
123 0 : static INLINE void clamp_block(int16_t *block, int block_size, int stride,
124 : int low, int high) {
125 : int i, j;
126 0 : for (i = 0; i < block_size; ++i) {
127 0 : for (j = 0; j < block_size; ++j) {
128 0 : block[i * stride + j] = clamp(block[i * stride + j], low, high);
129 : }
130 : }
131 0 : }
132 :
133 : typedef void (*TxfmFunc)(const int32_t *input, int32_t *output,
134 : const int8_t *cos_bit, const int8_t *stage_range);
135 :
136 : typedef enum TXFM_TYPE {
137 : TXFM_TYPE_DCT4,
138 : TXFM_TYPE_DCT8,
139 : TXFM_TYPE_DCT16,
140 : TXFM_TYPE_DCT32,
141 : TXFM_TYPE_DCT64,
142 : TXFM_TYPE_ADST4,
143 : TXFM_TYPE_ADST8,
144 : TXFM_TYPE_ADST16,
145 : TXFM_TYPE_ADST32,
146 : TXFM_TYPE_IDENTITY4,
147 : TXFM_TYPE_IDENTITY8,
148 : TXFM_TYPE_IDENTITY16,
149 : TXFM_TYPE_IDENTITY32,
150 : } TXFM_TYPE;
151 :
152 : typedef struct TXFM_1D_CFG {
153 : const int txfm_size;
154 : const int stage_num;
155 :
156 : const int8_t *shift;
157 : const int8_t *stage_range;
158 : const int8_t *cos_bit;
159 : const TXFM_TYPE txfm_type;
160 : } TXFM_1D_CFG;
161 :
162 : typedef struct TXFM_2D_FLIP_CFG {
163 : int ud_flip; // flip upside down
164 : int lr_flip; // flip left to right
165 : const TXFM_1D_CFG *col_cfg;
166 : const TXFM_1D_CFG *row_cfg;
167 : } TXFM_2D_FLIP_CFG;
168 :
169 0 : static INLINE void set_flip_cfg(int tx_type, TXFM_2D_FLIP_CFG *cfg) {
170 0 : switch (tx_type) {
171 : case DCT_DCT:
172 : case ADST_DCT:
173 : case DCT_ADST:
174 : case ADST_ADST:
175 0 : cfg->ud_flip = 0;
176 0 : cfg->lr_flip = 0;
177 0 : break;
178 : #if CONFIG_EXT_TX
179 : case IDTX:
180 : case V_DCT:
181 : case H_DCT:
182 : case V_ADST:
183 : case H_ADST:
184 0 : cfg->ud_flip = 0;
185 0 : cfg->lr_flip = 0;
186 0 : break;
187 : case FLIPADST_DCT:
188 : case FLIPADST_ADST:
189 : case V_FLIPADST:
190 0 : cfg->ud_flip = 1;
191 0 : cfg->lr_flip = 0;
192 0 : break;
193 : case DCT_FLIPADST:
194 : case ADST_FLIPADST:
195 : case H_FLIPADST:
196 0 : cfg->ud_flip = 0;
197 0 : cfg->lr_flip = 1;
198 0 : break;
199 : case FLIPADST_FLIPADST:
200 0 : cfg->ud_flip = 1;
201 0 : cfg->lr_flip = 1;
202 0 : break;
203 : #endif // CONFIG_EXT_TX
204 : default:
205 0 : cfg->ud_flip = 0;
206 0 : cfg->lr_flip = 0;
207 0 : assert(0);
208 : }
209 0 : }
210 :
211 : #ifdef __cplusplus
212 : extern "C" {
213 : #endif
214 : TXFM_2D_FLIP_CFG av1_get_fwd_txfm_cfg(int tx_type, int tx_size);
215 : TXFM_2D_FLIP_CFG av1_get_fwd_txfm_64x64_cfg(int tx_type);
216 : #ifdef __cplusplus
217 : }
218 : #endif // __cplusplus
219 :
220 : #endif // AV1_TXFM_H_
|
