FFmpeg
 All Data Structures Namespaces 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 "simple_idct.h"
34 
35 #include "bit_depth_template.c"
36 
37 #undef W1
38 #undef W2
39 #undef W3
40 #undef W4
41 #undef W5
42 #undef W6
43 #undef W7
44 #undef ROW_SHIFT
45 #undef COL_SHIFT
46 #undef DC_SHIFT
47 #undef MUL
48 #undef MAC
49 
50 #if BIT_DEPTH == 8
51 
52 #define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
53 #define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
54 #define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
55 #define W4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
56 #define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
57 #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
58 #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
59 
60 #define ROW_SHIFT 11
61 #define COL_SHIFT 20
62 #define DC_SHIFT 3
63 
64 #define MUL(a, b) MUL16(a, b)
65 #define MAC(a, b, c) MAC16(a, b, c)
66 
67 #elif BIT_DEPTH == 10 || BIT_DEPTH == 12
68 
69 # if BIT_DEPTH == 10
70 #define W1 22725 // 90901
71 #define W2 21407 // 85627
72 #define W3 19265 // 77062
73 #define W4 16384 // 65535
74 #define W5 12873 // 51491
75 #define W6 8867 // 35468
76 #define W7 4520 // 18081
77 
78 # ifdef EXTRA_SHIFT
79 #define ROW_SHIFT 13
80 #define COL_SHIFT 18
81 #define DC_SHIFT 1
82 # else
83 #define ROW_SHIFT 12
84 #define COL_SHIFT 19
85 #define DC_SHIFT 2
86 # endif
87 
88 # else
89 #define W1 45451
90 #define W2 42813
91 #define W3 38531
92 #define W4 32767
93 #define W5 25746
94 #define W6 17734
95 #define W7 9041
96 
97 #define ROW_SHIFT 16
98 #define COL_SHIFT 17
99 #define DC_SHIFT -1
100 # endif
101 
102 #define MUL(a, b) ((a) * (b))
103 #define MAC(a, b, c) ((a) += (b) * (c))
104 
105 #else
106 
107 #error "Unsupported bitdepth"
108 
109 #endif
110 
111 #ifdef EXTRA_SHIFT
112 static inline void FUNC(idctRowCondDC_extrashift)(int16_t *row, int extra_shift)
113 #else
114 static inline void FUNC(idctRowCondDC)(int16_t *row, int extra_shift)
115 #endif
116 {
117  int a0, a1, a2, a3, b0, b1, b2, b3;
118 
119 #if HAVE_FAST_64BIT
120 #define ROW0_MASK (0xffffLL << 48 * HAVE_BIGENDIAN)
121  if (((AV_RN64A(row) & ~ROW0_MASK) | AV_RN64A(row+4)) == 0) {
122  uint64_t temp;
123  if (DC_SHIFT - extra_shift >= 0) {
124  temp = (row[0] * (1 << (DC_SHIFT - extra_shift))) & 0xffff;
125  } else {
126  temp = ((row[0] + (1<<(extra_shift - DC_SHIFT-1))) >> (extra_shift - DC_SHIFT)) & 0xffff;
127  }
128  temp += temp * (1 << 16);
129  temp += temp * ((uint64_t) 1 << 32);
130  AV_WN64A(row, temp);
131  AV_WN64A(row+4, temp);
132  return;
133  }
134 #else
135  if (!(AV_RN32A(row+2) |
136  AV_RN32A(row+4) |
137  AV_RN32A(row+6) |
138  row[1])) {
139  uint32_t temp;
140  if (DC_SHIFT - extra_shift >= 0) {
141  temp = (row[0] * (1 << (DC_SHIFT - extra_shift))) & 0xffff;
142  } else {
143  temp = ((row[0] + (1<<(extra_shift - DC_SHIFT-1))) >> (extra_shift - DC_SHIFT)) & 0xffff;
144  }
145  temp += temp * (1 << 16);
146  AV_WN32A(row, temp);
147  AV_WN32A(row+2, temp);
148  AV_WN32A(row+4, temp);
149  AV_WN32A(row+6, temp);
150  return;
151  }
152 #endif
153 
154  a0 = (W4 * row[0]) + (1 << (ROW_SHIFT + extra_shift - 1));
155  a1 = a0;
156  a2 = a0;
157  a3 = a0;
158 
159  a0 += W2 * row[2];
160  a1 += W6 * row[2];
161  a2 -= W6 * row[2];
162  a3 -= W2 * row[2];
163 
164  b0 = MUL(W1, row[1]);
165  MAC(b0, W3, row[3]);
166  b1 = MUL(W3, row[1]);
167  MAC(b1, -W7, row[3]);
168  b2 = MUL(W5, row[1]);
169  MAC(b2, -W1, row[3]);
170  b3 = MUL(W7, row[1]);
171  MAC(b3, -W5, row[3]);
172 
173  if (AV_RN64A(row + 4)) {
174  a0 += W4*row[4] + W6*row[6];
175  a1 += - W4*row[4] - W2*row[6];
176  a2 += - W4*row[4] + W2*row[6];
177  a3 += W4*row[4] - W6*row[6];
178 
179  MAC(b0, W5, row[5]);
180  MAC(b0, W7, row[7]);
181 
182  MAC(b1, -W1, row[5]);
183  MAC(b1, -W5, row[7]);
184 
185  MAC(b2, W7, row[5]);
186  MAC(b2, W3, row[7]);
187 
188  MAC(b3, W3, row[5]);
189  MAC(b3, -W1, row[7]);
190  }
191 
192  row[0] = (a0 + b0) >> (ROW_SHIFT + extra_shift);
193  row[7] = (a0 - b0) >> (ROW_SHIFT + extra_shift);
194  row[1] = (a1 + b1) >> (ROW_SHIFT + extra_shift);
195  row[6] = (a1 - b1) >> (ROW_SHIFT + extra_shift);
196  row[2] = (a2 + b2) >> (ROW_SHIFT + extra_shift);
197  row[5] = (a2 - b2) >> (ROW_SHIFT + extra_shift);
198  row[3] = (a3 + b3) >> (ROW_SHIFT + extra_shift);
199  row[4] = (a3 - b3) >> (ROW_SHIFT + extra_shift);
200 }
201 
202 #define IDCT_COLS do { \
203  a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); \
204  a1 = a0; \
205  a2 = a0; \
206  a3 = a0; \
207  \
208  a0 += W2*col[8*2]; \
209  a1 += W6*col[8*2]; \
210  a2 += -W6*col[8*2]; \
211  a3 += -W2*col[8*2]; \
212  \
213  b0 = MUL(W1, col[8*1]); \
214  b1 = MUL(W3, col[8*1]); \
215  b2 = MUL(W5, col[8*1]); \
216  b3 = MUL(W7, col[8*1]); \
217  \
218  MAC(b0, W3, col[8*3]); \
219  MAC(b1, -W7, col[8*3]); \
220  MAC(b2, -W1, col[8*3]); \
221  MAC(b3, -W5, col[8*3]); \
222  \
223  if (col[8*4]) { \
224  a0 += W4*col[8*4]; \
225  a1 += -W4*col[8*4]; \
226  a2 += -W4*col[8*4]; \
227  a3 += W4*col[8*4]; \
228  } \
229  \
230  if (col[8*5]) { \
231  MAC(b0, W5, col[8*5]); \
232  MAC(b1, -W1, col[8*5]); \
233  MAC(b2, W7, col[8*5]); \
234  MAC(b3, W3, col[8*5]); \
235  } \
236  \
237  if (col[8*6]) { \
238  a0 += W6*col[8*6]; \
239  a1 += -W2*col[8*6]; \
240  a2 += W2*col[8*6]; \
241  a3 += -W6*col[8*6]; \
242  } \
243  \
244  if (col[8*7]) { \
245  MAC(b0, W7, col[8*7]); \
246  MAC(b1, -W5, col[8*7]); \
247  MAC(b2, W3, col[8*7]); \
248  MAC(b3, -W1, col[8*7]); \
249  } \
250  } while (0)
251 
252 #ifdef EXTRA_SHIFT
253 static inline void FUNC(idctSparseCol_extrashift)(int16_t *col)
254 #else
255 static inline void FUNC(idctSparseColPut)(pixel *dest, int line_size,
256  int16_t *col)
257 {
258  int a0, a1, a2, a3, b0, b1, b2, b3;
259 
260  IDCT_COLS;
261 
262  dest[0] = av_clip_pixel((a0 + b0) >> COL_SHIFT);
263  dest += line_size;
264  dest[0] = av_clip_pixel((a1 + b1) >> COL_SHIFT);
265  dest += line_size;
266  dest[0] = av_clip_pixel((a2 + b2) >> COL_SHIFT);
267  dest += line_size;
268  dest[0] = av_clip_pixel((a3 + b3) >> COL_SHIFT);
269  dest += line_size;
270  dest[0] = av_clip_pixel((a3 - b3) >> COL_SHIFT);
271  dest += line_size;
272  dest[0] = av_clip_pixel((a2 - b2) >> COL_SHIFT);
273  dest += line_size;
274  dest[0] = av_clip_pixel((a1 - b1) >> COL_SHIFT);
275  dest += line_size;
276  dest[0] = av_clip_pixel((a0 - b0) >> COL_SHIFT);
277 }
278 
279 static inline void FUNC(idctSparseColAdd)(pixel *dest, int line_size,
280  int16_t *col)
281 {
282  int a0, a1, a2, a3, b0, b1, b2, b3;
283 
284  IDCT_COLS;
285 
286  dest[0] = av_clip_pixel(dest[0] + ((a0 + b0) >> COL_SHIFT));
287  dest += line_size;
288  dest[0] = av_clip_pixel(dest[0] + ((a1 + b1) >> COL_SHIFT));
289  dest += line_size;
290  dest[0] = av_clip_pixel(dest[0] + ((a2 + b2) >> COL_SHIFT));
291  dest += line_size;
292  dest[0] = av_clip_pixel(dest[0] + ((a3 + b3) >> COL_SHIFT));
293  dest += line_size;
294  dest[0] = av_clip_pixel(dest[0] + ((a3 - b3) >> COL_SHIFT));
295  dest += line_size;
296  dest[0] = av_clip_pixel(dest[0] + ((a2 - b2) >> COL_SHIFT));
297  dest += line_size;
298  dest[0] = av_clip_pixel(dest[0] + ((a1 - b1) >> COL_SHIFT));
299  dest += line_size;
300  dest[0] = av_clip_pixel(dest[0] + ((a0 - b0) >> COL_SHIFT));
301 }
302 
303 static inline void FUNC(idctSparseCol)(int16_t *col)
304 #endif
305 {
306  int a0, a1, a2, a3, b0, b1, b2, b3;
307 
308  IDCT_COLS;
309 
310  col[0 ] = ((a0 + b0) >> COL_SHIFT);
311  col[8 ] = ((a1 + b1) >> COL_SHIFT);
312  col[16] = ((a2 + b2) >> COL_SHIFT);
313  col[24] = ((a3 + b3) >> COL_SHIFT);
314  col[32] = ((a3 - b3) >> COL_SHIFT);
315  col[40] = ((a2 - b2) >> COL_SHIFT);
316  col[48] = ((a1 - b1) >> COL_SHIFT);
317  col[56] = ((a0 - b0) >> COL_SHIFT);
318 }
319 
320 #ifndef EXTRA_SHIFT
321 void FUNC(ff_simple_idct_put)(uint8_t *dest_, int line_size, int16_t *block)
322 {
323  pixel *dest = (pixel *)dest_;
324  int i;
325 
326  line_size /= sizeof(pixel);
327 
328  for (i = 0; i < 8; i++)
329  FUNC(idctRowCondDC)(block + i*8, 0);
330 
331  for (i = 0; i < 8; i++)
332  FUNC(idctSparseColPut)(dest + i, line_size, block + i);
333 }
334 
335 void FUNC(ff_simple_idct_add)(uint8_t *dest_, int line_size, int16_t *block)
336 {
337  pixel *dest = (pixel *)dest_;
338  int i;
339 
340  line_size /= sizeof(pixel);
341 
342  for (i = 0; i < 8; i++)
343  FUNC(idctRowCondDC)(block + i*8, 0);
344 
345  for (i = 0; i < 8; i++)
346  FUNC(idctSparseColAdd)(dest + i, line_size, block + i);
347 }
348 
349 void FUNC(ff_simple_idct)(int16_t *block)
350 {
351  int i;
352 
353  for (i = 0; i < 8; i++)
354  FUNC(idctRowCondDC)(block + i*8, 0);
355 
356  for (i = 0; i < 8; i++)
357  FUNC(idctSparseCol)(block + i);
358 }
359 #endif
#define W6
else temp
Definition: vf_mcdeint.c:259
#define a0
Definition: regdef.h:46
#define a1
Definition: regdef.h:47
#define AV_WN32A(p, v)
Definition: intreadwrite.h:538
#define W2
#define W5
#define av_clip_pixel(a)
#define AV_RN32A(p)
Definition: intreadwrite.h:526
#define a3
Definition: regdef.h:49
#define W1
#define ROW_SHIFT
uint8_t
#define W7
#define FUNC(a)
#define W3
#define a2
Definition: regdef.h:48
#define W4
#define AV_WN64A(p, v)
Definition: intreadwrite.h:542
void FUNC() ff_simple_idct(int16_t *block)
static void FUNC() idctSparseColPut(pixel *dest, int line_size, int16_t *col)
static void FUNC() idctRowCondDC(int16_t *row, int extra_shift)
static void FUNC() idctSparseColAdd(pixel *dest, int line_size, int16_t *col)
#define MUL(a, b)
#define COL_SHIFT
#define IDCT_COLS
uint8_t pixel
Definition: tiny_ssim.c:42
static void FUNC() idctSparseCol(int16_t *col)
void FUNC() ff_simple_idct_put(uint8_t *dest_, int line_size, int16_t *block)
simple idct header.
void FUNC() ff_simple_idct_add(uint8_t *dest_, int line_size, int16_t *block)
#define AV_RN64A(p)
Definition: intreadwrite.h:530
static int16_t block[64]
Definition: dct-test.c:112