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 : #include <assert.h>
13 :
14 : #include "./av1_rtcd.h"
15 : #include "av1/common/enums.h"
16 : #include "av1/common/av1_fwd_txfm1d.h"
17 : #include "av1/common/av1_fwd_txfm1d_cfg.h"
18 : #include "av1/common/av1_txfm.h"
19 :
20 0 : static INLINE TxfmFunc fwd_txfm_type_to_func(TXFM_TYPE txfm_type) {
21 0 : switch (txfm_type) {
22 0 : case TXFM_TYPE_DCT4: return av1_fdct4_new;
23 0 : case TXFM_TYPE_DCT8: return av1_fdct8_new;
24 0 : case TXFM_TYPE_DCT16: return av1_fdct16_new;
25 0 : case TXFM_TYPE_DCT32: return av1_fdct32_new;
26 0 : case TXFM_TYPE_ADST4: return av1_fadst4_new;
27 0 : case TXFM_TYPE_ADST8: return av1_fadst8_new;
28 0 : case TXFM_TYPE_ADST16: return av1_fadst16_new;
29 0 : case TXFM_TYPE_ADST32: return av1_fadst32_new;
30 : #if CONFIG_EXT_TX
31 0 : case TXFM_TYPE_IDENTITY4: return av1_fidentity4_c;
32 0 : case TXFM_TYPE_IDENTITY8: return av1_fidentity8_c;
33 0 : case TXFM_TYPE_IDENTITY16: return av1_fidentity16_c;
34 0 : case TXFM_TYPE_IDENTITY32: return av1_fidentity32_c;
35 : #endif // CONFIG_EXT_TX
36 0 : default: assert(0); return NULL;
37 : }
38 : }
39 :
40 0 : static INLINE void fwd_txfm2d_c(const int16_t *input, int32_t *output,
41 : const int stride, const TXFM_2D_FLIP_CFG *cfg,
42 : int32_t *buf) {
43 : int c, r;
44 : // TODO(sarahparker) must correct for rectangular transforms in follow up
45 0 : const int txfm_size = cfg->row_cfg->txfm_size;
46 0 : const int8_t *shift = cfg->row_cfg->shift;
47 0 : const int8_t *stage_range_col = cfg->col_cfg->stage_range;
48 0 : const int8_t *stage_range_row = cfg->row_cfg->stage_range;
49 0 : const int8_t *cos_bit_col = cfg->col_cfg->cos_bit;
50 0 : const int8_t *cos_bit_row = cfg->row_cfg->cos_bit;
51 0 : const TxfmFunc txfm_func_col = fwd_txfm_type_to_func(cfg->col_cfg->txfm_type);
52 0 : const TxfmFunc txfm_func_row = fwd_txfm_type_to_func(cfg->row_cfg->txfm_type);
53 :
54 : // use output buffer as temp buffer
55 0 : int32_t *temp_in = output;
56 0 : int32_t *temp_out = output + txfm_size;
57 :
58 : // Columns
59 0 : for (c = 0; c < txfm_size; ++c) {
60 0 : if (cfg->ud_flip == 0) {
61 0 : for (r = 0; r < txfm_size; ++r) temp_in[r] = input[r * stride + c];
62 : } else {
63 0 : for (r = 0; r < txfm_size; ++r)
64 : // flip upside down
65 0 : temp_in[r] = input[(txfm_size - r - 1) * stride + c];
66 : }
67 0 : round_shift_array(temp_in, txfm_size, -shift[0]);
68 0 : txfm_func_col(temp_in, temp_out, cos_bit_col, stage_range_col);
69 0 : round_shift_array(temp_out, txfm_size, -shift[1]);
70 0 : if (cfg->lr_flip == 0) {
71 0 : for (r = 0; r < txfm_size; ++r) buf[r * txfm_size + c] = temp_out[r];
72 : } else {
73 0 : for (r = 0; r < txfm_size; ++r)
74 : // flip from left to right
75 0 : buf[r * txfm_size + (txfm_size - c - 1)] = temp_out[r];
76 : }
77 : }
78 :
79 : // Rows
80 0 : for (r = 0; r < txfm_size; ++r) {
81 0 : txfm_func_row(buf + r * txfm_size, output + r * txfm_size, cos_bit_row,
82 : stage_range_row);
83 0 : round_shift_array(output + r * txfm_size, txfm_size, -shift[2]);
84 : }
85 0 : }
86 :
87 0 : void av1_fwd_txfm2d_4x4_c(const int16_t *input, int32_t *output, int stride,
88 : int tx_type, int bd) {
89 : int32_t txfm_buf[4 * 4];
90 0 : TXFM_2D_FLIP_CFG cfg = av1_get_fwd_txfm_cfg(tx_type, TX_4X4);
91 : (void)bd;
92 0 : fwd_txfm2d_c(input, output, stride, &cfg, txfm_buf);
93 0 : }
94 :
95 0 : void av1_fwd_txfm2d_8x8_c(const int16_t *input, int32_t *output, int stride,
96 : int tx_type, int bd) {
97 : int32_t txfm_buf[8 * 8];
98 0 : TXFM_2D_FLIP_CFG cfg = av1_get_fwd_txfm_cfg(tx_type, TX_8X8);
99 : (void)bd;
100 0 : fwd_txfm2d_c(input, output, stride, &cfg, txfm_buf);
101 0 : }
102 :
103 0 : void av1_fwd_txfm2d_16x16_c(const int16_t *input, int32_t *output, int stride,
104 : int tx_type, int bd) {
105 : int32_t txfm_buf[16 * 16];
106 0 : TXFM_2D_FLIP_CFG cfg = av1_get_fwd_txfm_cfg(tx_type, TX_16X16);
107 : (void)bd;
108 0 : fwd_txfm2d_c(input, output, stride, &cfg, txfm_buf);
109 0 : }
110 :
111 0 : void av1_fwd_txfm2d_32x32_c(const int16_t *input, int32_t *output, int stride,
112 : int tx_type, int bd) {
113 : int32_t txfm_buf[32 * 32];
114 0 : TXFM_2D_FLIP_CFG cfg = av1_get_fwd_txfm_cfg(tx_type, TX_32X32);
115 : (void)bd;
116 0 : fwd_txfm2d_c(input, output, stride, &cfg, txfm_buf);
117 0 : }
118 :
119 0 : void av1_fwd_txfm2d_64x64_c(const int16_t *input, int32_t *output, int stride,
120 : int tx_type, int bd) {
121 : int32_t txfm_buf[64 * 64];
122 0 : TXFM_2D_FLIP_CFG cfg = av1_get_fwd_txfm_64x64_cfg(tx_type);
123 : (void)bd;
124 0 : fwd_txfm2d_c(input, output, stride, &cfg, txfm_buf);
125 0 : }
126 :
127 : static const TXFM_1D_CFG *fwd_txfm_col_cfg_ls[TX_TYPES_1D][TX_SIZES] = {
128 : // DCT
129 : {
130 : #if CONFIG_CHROMA_2X2
131 : NULL,
132 : #endif
133 : &fwd_txfm_1d_col_cfg_dct_4, &fwd_txfm_1d_col_cfg_dct_8,
134 : &fwd_txfm_1d_col_cfg_dct_16, &fwd_txfm_1d_col_cfg_dct_32 },
135 : // ADST
136 : {
137 : #if CONFIG_CHROMA_2X2
138 : NULL,
139 : #endif
140 : &fwd_txfm_1d_col_cfg_adst_4, &fwd_txfm_1d_col_cfg_adst_8,
141 : &fwd_txfm_1d_col_cfg_adst_16, &fwd_txfm_1d_col_cfg_adst_32 },
142 : #if CONFIG_EXT_TX
143 : // FLIPADST
144 : {
145 : #if CONFIG_CHROMA_2X2
146 : NULL,
147 : #endif
148 : &fwd_txfm_1d_col_cfg_adst_4, &fwd_txfm_1d_col_cfg_adst_8,
149 : &fwd_txfm_1d_col_cfg_adst_16, &fwd_txfm_1d_col_cfg_adst_32 },
150 : // IDENTITY
151 : {
152 : #if CONFIG_CHROMA_2X2
153 : NULL,
154 : #endif
155 : &fwd_txfm_1d_cfg_identity_4, &fwd_txfm_1d_cfg_identity_8,
156 : &fwd_txfm_1d_cfg_identity_16, &fwd_txfm_1d_cfg_identity_32 },
157 : #endif // CONFIG_EXT_TX
158 : };
159 :
160 : static const TXFM_1D_CFG *fwd_txfm_row_cfg_ls[TX_TYPES_1D][TX_SIZES] = {
161 : // DCT
162 : {
163 : #if CONFIG_CHROMA_2X2
164 : NULL,
165 : #endif
166 : &fwd_txfm_1d_row_cfg_dct_4, &fwd_txfm_1d_row_cfg_dct_8,
167 : &fwd_txfm_1d_row_cfg_dct_16, &fwd_txfm_1d_row_cfg_dct_32 },
168 : // ADST
169 : {
170 : #if CONFIG_CHROMA_2X2
171 : NULL,
172 : #endif
173 : &fwd_txfm_1d_row_cfg_adst_4, &fwd_txfm_1d_row_cfg_adst_8,
174 : &fwd_txfm_1d_row_cfg_adst_16, &fwd_txfm_1d_row_cfg_adst_32 },
175 : #if CONFIG_EXT_TX
176 : // FLIPADST
177 : {
178 : #if CONFIG_CHROMA_2X2
179 : NULL,
180 : #endif
181 : &fwd_txfm_1d_row_cfg_adst_4, &fwd_txfm_1d_row_cfg_adst_8,
182 : &fwd_txfm_1d_row_cfg_adst_16, &fwd_txfm_1d_row_cfg_adst_32 },
183 : // IDENTITY
184 : {
185 : #if CONFIG_CHROMA_2X2
186 : NULL,
187 : #endif
188 : &fwd_txfm_1d_cfg_identity_4, &fwd_txfm_1d_cfg_identity_8,
189 : &fwd_txfm_1d_cfg_identity_16, &fwd_txfm_1d_cfg_identity_32 },
190 : #endif // CONFIG_EXT_TX
191 : };
192 :
193 0 : TXFM_2D_FLIP_CFG av1_get_fwd_txfm_cfg(int tx_type, int tx_size) {
194 : TXFM_2D_FLIP_CFG cfg;
195 0 : set_flip_cfg(tx_type, &cfg);
196 0 : int tx_type_col = vtx_tab[tx_type];
197 0 : int tx_type_row = htx_tab[tx_type];
198 0 : cfg.col_cfg = fwd_txfm_col_cfg_ls[tx_type_col][tx_size];
199 0 : cfg.row_cfg = fwd_txfm_row_cfg_ls[tx_type_row][tx_size];
200 0 : return cfg;
201 : }
202 :
203 0 : TXFM_2D_FLIP_CFG av1_get_fwd_txfm_64x64_cfg(int tx_type) {
204 : TXFM_2D_FLIP_CFG cfg;
205 0 : switch (tx_type) {
206 : case DCT_DCT:
207 0 : cfg.col_cfg = &fwd_txfm_1d_col_cfg_dct_64;
208 0 : cfg.row_cfg = &fwd_txfm_1d_row_cfg_dct_64;
209 0 : cfg.ud_flip = 0;
210 0 : cfg.lr_flip = 0;
211 0 : break;
212 : default:
213 0 : cfg.ud_flip = 0;
214 0 : cfg.lr_flip = 0;
215 0 : assert(0);
216 : }
217 0 : return cfg;
218 : }
|
