1 /*
2 * Simple IDCT
3 *
4 * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 /**
24 * @file
25 * simpleidct in C.
26 */
27
28 #include "libavutil/intreadwrite.h"
29 #include "avcodec.h"
30 #include "mathops.h"
31 #include "simple_idct.h"
32
33 #define IN_IDCT_DEPTH 16
34
35 #define BIT_DEPTH 8
36 #include "simple_idct_template.c"
37 #undef BIT_DEPTH
38
39 #define BIT_DEPTH 10
40 #include "simple_idct_template.c"
41
42 #define EXTRA_SHIFT 2
43 #include "simple_idct_template.c"
44
45 #undef EXTRA_SHIFT
46 #undef BIT_DEPTH
47
48 #define BIT_DEPTH 12
49 #include "simple_idct_template.c"
50 #undef BIT_DEPTH
51 #undef IN_IDCT_DEPTH
52
53 #define IN_IDCT_DEPTH 32
54 #define BIT_DEPTH 10
55 #include "simple_idct_template.c"
56 #undef BIT_DEPTH
57 #undef IN_IDCT_DEPTH
58
59 /* 2x4x8 idct */
60
61 #define CN_SHIFT 12
62 #define C_FIX(x) ((int)((x) * (1 << CN_SHIFT) + 0.5))
63 #define C1 C_FIX(0.6532814824)
64 #define C2 C_FIX(0.2705980501)
65
66 /* row idct is multiple by 16 * sqrt(2.0), col idct4 is normalized,
67 and the butterfly must be multiplied by 0.5 * sqrt(2.0) */
68 #define C_SHIFT (4+1+12)
69
idct4col_put(uint8_t * dest,ptrdiff_t line_size,const int16_t * col)70 static inline void idct4col_put(uint8_t *dest, ptrdiff_t line_size, const int16_t *col)
71 {
72 int c0, c1, c2, c3, a0, a1, a2, a3;
73
74 a0 = col[8*0];
75 a1 = col[8*2];
76 a2 = col[8*4];
77 a3 = col[8*6];
78 c0 = ((a0 + a2) * (1 << CN_SHIFT - 1)) + (1 << (C_SHIFT - 1));
79 c2 = ((a0 - a2) * (1 << CN_SHIFT - 1)) + (1 << (C_SHIFT - 1));
80 c1 = a1 * C1 + a3 * C2;
81 c3 = a1 * C2 - a3 * C1;
82 dest[0] = av_clip_uint8((c0 + c1) >> C_SHIFT);
83 dest += line_size;
84 dest[0] = av_clip_uint8((c2 + c3) >> C_SHIFT);
85 dest += line_size;
86 dest[0] = av_clip_uint8((c2 - c3) >> C_SHIFT);
87 dest += line_size;
88 dest[0] = av_clip_uint8((c0 - c1) >> C_SHIFT);
89 }
90
91 #define BF(k) \
92 {\
93 int a0, a1;\
94 a0 = ptr[k];\
95 a1 = ptr[8 + k];\
96 ptr[k] = a0 + a1;\
97 ptr[8 + k] = a0 - a1;\
98 }
99
100 /* only used by DV codec. The input must be interlaced. 128 is added
101 to the pixels before clamping to avoid systematic error
102 (1024*sqrt(2)) offset would be needed otherwise. */
103 /* XXX: I think a 1.0/sqrt(2) normalization should be needed to
104 compensate the extra butterfly stage - I don't have the full DV
105 specification */
ff_simple_idct248_put(uint8_t * dest,ptrdiff_t line_size,int16_t * block)106 void ff_simple_idct248_put(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
107 {
108 int i;
109 int16_t *ptr;
110
111 /* butterfly */
112 ptr = block;
113 for(i=0;i<4;i++) {
114 BF(0);
115 BF(1);
116 BF(2);
117 BF(3);
118 BF(4);
119 BF(5);
120 BF(6);
121 BF(7);
122 ptr += 2 * 8;
123 }
124
125 /* IDCT8 on each line */
126 for(i=0; i<8; i++) {
127 idctRowCondDC_int16_8bit(block + i*8, 0);
128 }
129
130 /* IDCT4 and store */
131 for(i=0;i<8;i++) {
132 idct4col_put(dest + i, 2 * line_size, block + i);
133 idct4col_put(dest + line_size + i, 2 * line_size, block + 8 + i);
134 }
135 }
136
137 /* 8x4 & 4x8 WMV2 IDCT */
138 #undef CN_SHIFT
139 #undef C_SHIFT
140 #undef C_FIX
141 #undef C1
142 #undef C2
143 #define CN_SHIFT 12
144 #define C_FIX(x) ((int)((x) * M_SQRT2 * (1 << CN_SHIFT) + 0.5))
145 #define C1 C_FIX(0.6532814824)
146 #define C2 C_FIX(0.2705980501)
147 #define C3 C_FIX(0.5)
148 #define C_SHIFT (4+1+12)
idct4col_add(uint8_t * dest,ptrdiff_t line_size,const int16_t * col)149 static inline void idct4col_add(uint8_t *dest, ptrdiff_t line_size, const int16_t *col)
150 {
151 int c0, c1, c2, c3, a0, a1, a2, a3;
152
153 a0 = col[8*0];
154 a1 = col[8*1];
155 a2 = col[8*2];
156 a3 = col[8*3];
157 c0 = (a0 + a2)*C3 + (1 << (C_SHIFT - 1));
158 c2 = (a0 - a2)*C3 + (1 << (C_SHIFT - 1));
159 c1 = a1 * C1 + a3 * C2;
160 c3 = a1 * C2 - a3 * C1;
161 dest[0] = av_clip_uint8(dest[0] + ((c0 + c1) >> C_SHIFT));
162 dest += line_size;
163 dest[0] = av_clip_uint8(dest[0] + ((c2 + c3) >> C_SHIFT));
164 dest += line_size;
165 dest[0] = av_clip_uint8(dest[0] + ((c2 - c3) >> C_SHIFT));
166 dest += line_size;
167 dest[0] = av_clip_uint8(dest[0] + ((c0 - c1) >> C_SHIFT));
168 }
169
170 #define RN_SHIFT 15
171 #define R_FIX(x) ((int)((x) * M_SQRT2 * (1 << RN_SHIFT) + 0.5))
172 #define R1 R_FIX(0.6532814824)
173 #define R2 R_FIX(0.2705980501)
174 #define R3 R_FIX(0.5)
175 #define R_SHIFT 11
idct4row(int16_t * row)176 static inline void idct4row(int16_t *row)
177 {
178 unsigned c0, c1, c2, c3;
179 int a0, a1, a2, a3;
180
181 a0 = row[0];
182 a1 = row[1];
183 a2 = row[2];
184 a3 = row[3];
185 c0 = (a0 + a2)*R3 + (1 << (R_SHIFT - 1));
186 c2 = (a0 - a2)*R3 + (1 << (R_SHIFT - 1));
187 c1 = a1 * R1 + a3 * R2;
188 c3 = a1 * R2 - a3 * R1;
189 row[0]= (c0 + c1) >> R_SHIFT;
190 row[1]= (c2 + c3) >> R_SHIFT;
191 row[2]= (c2 - c3) >> R_SHIFT;
192 row[3]= (c0 - c1) >> R_SHIFT;
193 }
194
ff_simple_idct84_add(uint8_t * dest,ptrdiff_t line_size,int16_t * block)195 void ff_simple_idct84_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
196 {
197 int i;
198
199 /* IDCT8 on each line */
200 for(i=0; i<4; i++) {
201 idctRowCondDC_int16_8bit(block + i*8, 0);
202 }
203
204 /* IDCT4 and store */
205 for(i=0;i<8;i++) {
206 idct4col_add(dest + i, line_size, block + i);
207 }
208 }
209
ff_simple_idct48_add(uint8_t * dest,ptrdiff_t line_size,int16_t * block)210 void ff_simple_idct48_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
211 {
212 int i;
213
214 /* IDCT4 on each line */
215 for(i=0; i<8; i++) {
216 idct4row(block + i*8);
217 }
218
219 /* IDCT8 and store */
220 for(i=0; i<4; i++){
221 idctSparseColAdd_int16_8bit(dest + i, line_size, block + i);
222 }
223 }
224
ff_simple_idct44_add(uint8_t * dest,ptrdiff_t line_size,int16_t * block)225 void ff_simple_idct44_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
226 {
227 int i;
228
229 /* IDCT4 on each line */
230 for(i=0; i<4; i++) {
231 idct4row(block + i*8);
232 }
233
234 /* IDCT4 and store */
235 for(i=0; i<4; i++){
236 idct4col_add(dest + i, line_size, block + i);
237 }
238 }
239
ff_prores_idct_10(int16_t * block,const int16_t * qmat)240 void ff_prores_idct_10(int16_t *block, const int16_t *qmat)
241 {
242 int i;
243
244 for (i = 0; i < 64; i++)
245 block[i] *= qmat[i];
246
247 for (i = 0; i < 8; i++)
248 idctRowCondDC_extrashift_10(block + i*8, 2);
249
250 for (i = 0; i < 8; i++) {
251 block[i] += 8192;
252 idctSparseCol_extrashift_10(block + i);
253 }
254 }
255
ff_prores_idct_12(int16_t * block,const int16_t * qmat)256 void ff_prores_idct_12(int16_t *block, const int16_t *qmat)
257 {
258 int i;
259
260 for (i = 0; i < 64; i++)
261 block[i] *= qmat[i];
262
263 for (i = 0; i < 8; i++)
264 idctRowCondDC_int16_12bit(block + i*8, 0);
265
266 for (i = 0; i < 8; i++) {
267 block[i] += 8192;
268 idctSparseCol_int16_12bit(block + i);
269 }
270 }
271