FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
simple_idct_template.c
Go to the documentation of this file.
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 /*
29  based upon some outcommented c code from mpeg2dec (idct_mmx.c
30  written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>)
31  */
32 
33 #include "bit_depth_template.c"
34 
35 #undef W1
36 #undef W2
37 #undef W3
38 #undef W4
39 #undef W5
40 #undef W6
41 #undef W7
42 #undef ROW_SHIFT
43 #undef COL_SHIFT
44 #undef DC_SHIFT
45 #undef MUL
46 #undef MAC
47 
48 #if BIT_DEPTH == 8
49 
50 #define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
51 #define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
52 #define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
53 #define W4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
54 #define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
55 #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
56 #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
57 
58 #define ROW_SHIFT 11
59 #define COL_SHIFT 20
60 #define DC_SHIFT 3
61 
62 #define MUL(a, b) MUL16(a, b)
63 #define MAC(a, b, c) MAC16(a, b, c)
64 
65 #elif BIT_DEPTH == 10
66 
67 #define W1 90901
68 #define W2 85627
69 #define W3 77062
70 #define W4 65535
71 #define W5 51491
72 #define W6 35468
73 #define W7 18081
74 
75 #define ROW_SHIFT 15
76 #define COL_SHIFT 20
77 #define DC_SHIFT 1
78 
79 #define MUL(a, b) ((a) * (b))
80 #define MAC(a, b, c) ((a) += (b) * (c))
81 
82 #else
83 
84 #error "Unsupported bitdepth"
85 
86 #endif
87 
88 static inline void FUNC(idctRowCondDC)(DCTELEM *row, int extra_shift)
89 {
90  int a0, a1, a2, a3, b0, b1, b2, b3;
91 
92 #if HAVE_FAST_64BIT
93 #define ROW0_MASK (0xffffLL << 48 * HAVE_BIGENDIAN)
94  if (((((uint64_t *)row)[0] & ~ROW0_MASK) | ((uint64_t *)row)[1]) == 0) {
95  uint64_t temp;
96  if (DC_SHIFT - extra_shift > 0) {
97  temp = (row[0] << (DC_SHIFT - extra_shift)) & 0xffff;
98  } else {
99  temp = (row[0] >> (extra_shift - DC_SHIFT)) & 0xffff;
100  }
101  temp += temp << 16;
102  temp += temp << 32;
103  ((uint64_t *)row)[0] = temp;
104  ((uint64_t *)row)[1] = temp;
105  return;
106  }
107 #else
108  if (!(((uint32_t*)row)[1] |
109  ((uint32_t*)row)[2] |
110  ((uint32_t*)row)[3] |
111  row[1])) {
112  uint32_t temp;
113  if (DC_SHIFT - extra_shift > 0) {
114  temp = (row[0] << (DC_SHIFT - extra_shift)) & 0xffff;
115  } else {
116  temp = (row[0] >> (extra_shift - DC_SHIFT)) & 0xffff;
117  }
118  temp += temp << 16;
119  ((uint32_t*)row)[0]=((uint32_t*)row)[1] =
120  ((uint32_t*)row)[2]=((uint32_t*)row)[3] = temp;
121  return;
122  }
123 #endif
124 
125  a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
126  a1 = a0;
127  a2 = a0;
128  a3 = a0;
129 
130  a0 += W2 * row[2];
131  a1 += W6 * row[2];
132  a2 -= W6 * row[2];
133  a3 -= W2 * row[2];
134 
135  b0 = MUL(W1, row[1]);
136  MAC(b0, W3, row[3]);
137  b1 = MUL(W3, row[1]);
138  MAC(b1, -W7, row[3]);
139  b2 = MUL(W5, row[1]);
140  MAC(b2, -W1, row[3]);
141  b3 = MUL(W7, row[1]);
142  MAC(b3, -W5, row[3]);
143 
144  if (AV_RN64A(row + 4)) {
145  a0 += W4*row[4] + W6*row[6];
146  a1 += - W4*row[4] - W2*row[6];
147  a2 += - W4*row[4] + W2*row[6];
148  a3 += W4*row[4] - W6*row[6];
149 
150  MAC(b0, W5, row[5]);
151  MAC(b0, W7, row[7]);
152 
153  MAC(b1, -W1, row[5]);
154  MAC(b1, -W5, row[7]);
155 
156  MAC(b2, W7, row[5]);
157  MAC(b2, W3, row[7]);
158 
159  MAC(b3, W3, row[5]);
160  MAC(b3, -W1, row[7]);
161  }
162 
163  row[0] = (a0 + b0) >> (ROW_SHIFT + extra_shift);
164  row[7] = (a0 - b0) >> (ROW_SHIFT + extra_shift);
165  row[1] = (a1 + b1) >> (ROW_SHIFT + extra_shift);
166  row[6] = (a1 - b1) >> (ROW_SHIFT + extra_shift);
167  row[2] = (a2 + b2) >> (ROW_SHIFT + extra_shift);
168  row[5] = (a2 - b2) >> (ROW_SHIFT + extra_shift);
169  row[3] = (a3 + b3) >> (ROW_SHIFT + extra_shift);
170  row[4] = (a3 - b3) >> (ROW_SHIFT + extra_shift);
171 }
172 
173 #define IDCT_COLS do { \
174  a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); \
175  a1 = a0; \
176  a2 = a0; \
177  a3 = a0; \
178  \
179  a0 += W2*col[8*2]; \
180  a1 += W6*col[8*2]; \
181  a2 += -W6*col[8*2]; \
182  a3 += -W2*col[8*2]; \
183  \
184  b0 = MUL(W1, col[8*1]); \
185  b1 = MUL(W3, col[8*1]); \
186  b2 = MUL(W5, col[8*1]); \
187  b3 = MUL(W7, col[8*1]); \
188  \
189  MAC(b0, W3, col[8*3]); \
190  MAC(b1, -W7, col[8*3]); \
191  MAC(b2, -W1, col[8*3]); \
192  MAC(b3, -W5, col[8*3]); \
193  \
194  if (col[8*4]) { \
195  a0 += W4*col[8*4]; \
196  a1 += -W4*col[8*4]; \
197  a2 += -W4*col[8*4]; \
198  a3 += W4*col[8*4]; \
199  } \
200  \
201  if (col[8*5]) { \
202  MAC(b0, W5, col[8*5]); \
203  MAC(b1, -W1, col[8*5]); \
204  MAC(b2, W7, col[8*5]); \
205  MAC(b3, W3, col[8*5]); \
206  } \
207  \
208  if (col[8*6]) { \
209  a0 += W6*col[8*6]; \
210  a1 += -W2*col[8*6]; \
211  a2 += W2*col[8*6]; \
212  a3 += -W6*col[8*6]; \
213  } \
214  \
215  if (col[8*7]) { \
216  MAC(b0, W7, col[8*7]); \
217  MAC(b1, -W5, col[8*7]); \
218  MAC(b2, W3, col[8*7]); \
219  MAC(b3, -W1, col[8*7]); \
220  } \
221  } while (0)
222 
223 static inline void FUNC(idctSparseColPut)(pixel *dest, int line_size,
224  DCTELEM *col)
225 {
226  int a0, a1, a2, a3, b0, b1, b2, b3;
227 
228  IDCT_COLS;
229 
230  dest[0] = av_clip_pixel((a0 + b0) >> COL_SHIFT);
231  dest += line_size;
232  dest[0] = av_clip_pixel((a1 + b1) >> COL_SHIFT);
233  dest += line_size;
234  dest[0] = av_clip_pixel((a2 + b2) >> COL_SHIFT);
235  dest += line_size;
236  dest[0] = av_clip_pixel((a3 + b3) >> COL_SHIFT);
237  dest += line_size;
238  dest[0] = av_clip_pixel((a3 - b3) >> COL_SHIFT);
239  dest += line_size;
240  dest[0] = av_clip_pixel((a2 - b2) >> COL_SHIFT);
241  dest += line_size;
242  dest[0] = av_clip_pixel((a1 - b1) >> COL_SHIFT);
243  dest += line_size;
244  dest[0] = av_clip_pixel((a0 - b0) >> COL_SHIFT);
245 }
246 
247 static inline void FUNC(idctSparseColAdd)(pixel *dest, int line_size,
248  DCTELEM *col)
249 {
250  int a0, a1, a2, a3, b0, b1, b2, b3;
251 
252  IDCT_COLS;
253 
254  dest[0] = av_clip_pixel(dest[0] + ((a0 + b0) >> COL_SHIFT));
255  dest += line_size;
256  dest[0] = av_clip_pixel(dest[0] + ((a1 + b1) >> COL_SHIFT));
257  dest += line_size;
258  dest[0] = av_clip_pixel(dest[0] + ((a2 + b2) >> COL_SHIFT));
259  dest += line_size;
260  dest[0] = av_clip_pixel(dest[0] + ((a3 + b3) >> COL_SHIFT));
261  dest += line_size;
262  dest[0] = av_clip_pixel(dest[0] + ((a3 - b3) >> COL_SHIFT));
263  dest += line_size;
264  dest[0] = av_clip_pixel(dest[0] + ((a2 - b2) >> COL_SHIFT));
265  dest += line_size;
266  dest[0] = av_clip_pixel(dest[0] + ((a1 - b1) >> COL_SHIFT));
267  dest += line_size;
268  dest[0] = av_clip_pixel(dest[0] + ((a0 - b0) >> COL_SHIFT));
269 }
270 
271 static inline void FUNC(idctSparseCol)(DCTELEM *col)
272 {
273  int a0, a1, a2, a3, b0, b1, b2, b3;
274 
275  IDCT_COLS;
276 
277  col[0 ] = ((a0 + b0) >> COL_SHIFT);
278  col[8 ] = ((a1 + b1) >> COL_SHIFT);
279  col[16] = ((a2 + b2) >> COL_SHIFT);
280  col[24] = ((a3 + b3) >> COL_SHIFT);
281  col[32] = ((a3 - b3) >> COL_SHIFT);
282  col[40] = ((a2 - b2) >> COL_SHIFT);
283  col[48] = ((a1 - b1) >> COL_SHIFT);
284  col[56] = ((a0 - b0) >> COL_SHIFT);
285 }
286 
287 void FUNC(ff_simple_idct_put)(uint8_t *dest_, int line_size, DCTELEM *block)
288 {
289  pixel *dest = (pixel *)dest_;
290  int i;
291 
292  line_size /= sizeof(pixel);
293 
294  for (i = 0; i < 8; i++)
295  FUNC(idctRowCondDC)(block + i*8, 0);
296 
297  for (i = 0; i < 8; i++)
298  FUNC(idctSparseColPut)(dest + i, line_size, block + i);
299 }
300 
301 void FUNC(ff_simple_idct_add)(uint8_t *dest_, int line_size, DCTELEM *block)
302 {
303  pixel *dest = (pixel *)dest_;
304  int i;
305 
306  line_size /= sizeof(pixel);
307 
308  for (i = 0; i < 8; i++)
309  FUNC(idctRowCondDC)(block + i*8, 0);
310 
311  for (i = 0; i < 8; i++)
312  FUNC(idctSparseColAdd)(dest + i, line_size, block + i);
313 }
314 
316 {
317  int i;
318 
319  for (i = 0; i < 8; i++)
320  FUNC(idctRowCondDC)(block + i*8, 0);
321 
322  for (i = 0; i < 8; i++)
323  FUNC(idctSparseCol)(block + i);
324 }