FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
ivi_dsp.c
Go to the documentation of this file.
1 /*
2  * DSP functions for Indeo Video Interactive codecs (Indeo4 and Indeo5)
3  *
4  * Copyright (c) 2009-2011 Maxim Poliakovski
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  * DSP functions (inverse transforms, motion compensation, wavelet recompostions)
26  * for Indeo Video Interactive codecs.
27  */
28 
29 #include "avcodec.h"
30 #include "ivi_common.h"
31 #include "ivi_dsp.h"
32 
33 void ff_ivi_recompose53(const IVIPlaneDesc *plane, uint8_t *dst,
34  const int dst_pitch)
35 {
36  int x, y, indx;
37  int32_t p0, p1, p2, p3, tmp0, tmp1, tmp2;
38  int32_t b0_1, b0_2, b1_1, b1_2, b1_3, b2_1, b2_2, b2_3, b2_4, b2_5, b2_6;
39  int32_t b3_1, b3_2, b3_3, b3_4, b3_5, b3_6, b3_7, b3_8, b3_9;
40  int32_t pitch, back_pitch;
41  const short *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
42  const int num_bands = 4;
43 
44  /* all bands should have the same pitch */
45  pitch = plane->bands[0].pitch;
46 
47  /* pixels at the position "y-1" will be set to pixels at the "y" for the 1st iteration */
48  back_pitch = 0;
49 
50  /* get pointers to the wavelet bands */
51  b0_ptr = plane->bands[0].buf;
52  b1_ptr = plane->bands[1].buf;
53  b2_ptr = plane->bands[2].buf;
54  b3_ptr = plane->bands[3].buf;
55 
56  for (y = 0; y < plane->height; y += 2) {
57 
58  if (y+2 >= plane->height)
59  pitch= 0;
60  /* load storage variables with values */
61  if (num_bands > 0) {
62  b0_1 = b0_ptr[0];
63  b0_2 = b0_ptr[pitch];
64  }
65 
66  if (num_bands > 1) {
67  b1_1 = b1_ptr[back_pitch];
68  b1_2 = b1_ptr[0];
69  b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch];
70  }
71 
72  if (num_bands > 2) {
73  b2_2 = b2_ptr[0]; // b2[x, y ]
74  b2_3 = b2_2; // b2[x+1,y ] = b2[x,y]
75  b2_5 = b2_ptr[pitch]; // b2[x ,y+1]
76  b2_6 = b2_5; // b2[x+1,y+1] = b2[x,y+1]
77  }
78 
79  if (num_bands > 3) {
80  b3_2 = b3_ptr[back_pitch]; // b3[x ,y-1]
81  b3_3 = b3_2; // b3[x+1,y-1] = b3[x ,y-1]
82  b3_5 = b3_ptr[0]; // b3[x ,y ]
83  b3_6 = b3_5; // b3[x+1,y ] = b3[x ,y ]
84  b3_8 = b3_2 - b3_5*6 + b3_ptr[pitch];
85  b3_9 = b3_8;
86  }
87 
88  for (x = 0, indx = 0; x < plane->width; x+=2, indx++) {
89  if (x+2 >= plane->width) {
90  b0_ptr --;
91  b1_ptr --;
92  b2_ptr --;
93  b3_ptr --;
94  }
95 
96  /* some values calculated in the previous iterations can */
97  /* be reused in the next ones, so do appropriate copying */
98  b2_1 = b2_2; // b2[x-1,y ] = b2[x, y ]
99  b2_2 = b2_3; // b2[x ,y ] = b2[x+1,y ]
100  b2_4 = b2_5; // b2[x-1,y+1] = b2[x ,y+1]
101  b2_5 = b2_6; // b2[x ,y+1] = b2[x+1,y+1]
102  b3_1 = b3_2; // b3[x-1,y-1] = b3[x ,y-1]
103  b3_2 = b3_3; // b3[x ,y-1] = b3[x+1,y-1]
104  b3_4 = b3_5; // b3[x-1,y ] = b3[x ,y ]
105  b3_5 = b3_6; // b3[x ,y ] = b3[x+1,y ]
106  b3_7 = b3_8; // vert_HPF(x-1)
107  b3_8 = b3_9; // vert_HPF(x )
108 
109  p0 = p1 = p2 = p3 = 0;
110 
111  /* process the LL-band by applying LPF both vertically and horizontally */
112  if (num_bands > 0) {
113  tmp0 = b0_1;
114  tmp2 = b0_2;
115  b0_1 = b0_ptr[indx+1];
116  b0_2 = b0_ptr[pitch+indx+1];
117  tmp1 = tmp0 + b0_1;
118 
119  p0 = tmp0 << 4;
120  p1 = tmp1 << 3;
121  p2 = (tmp0 + tmp2) << 3;
122  p3 = (tmp1 + tmp2 + b0_2) << 2;
123  }
124 
125  /* process the HL-band by applying HPF vertically and LPF horizontally */
126  if (num_bands > 1) {
127  tmp0 = b1_2;
128  tmp1 = b1_1;
129  b1_2 = b1_ptr[indx+1];
130  b1_1 = b1_ptr[back_pitch+indx+1];
131 
132  tmp2 = tmp1 - tmp0*6 + b1_3;
133  b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch+indx+1];
134 
135  p0 += (tmp0 + tmp1) << 3;
136  p1 += (tmp0 + tmp1 + b1_1 + b1_2) << 2;
137  p2 += tmp2 << 2;
138  p3 += (tmp2 + b1_3) << 1;
139  }
140 
141  /* process the LH-band by applying LPF vertically and HPF horizontally */
142  if (num_bands > 2) {
143  b2_3 = b2_ptr[indx+1];
144  b2_6 = b2_ptr[pitch+indx+1];
145 
146  tmp0 = b2_1 + b2_2;
147  tmp1 = b2_1 - b2_2*6 + b2_3;
148 
149  p0 += tmp0 << 3;
150  p1 += tmp1 << 2;
151  p2 += (tmp0 + b2_4 + b2_5) << 2;
152  p3 += (tmp1 + b2_4 - b2_5*6 + b2_6) << 1;
153  }
154 
155  /* process the HH-band by applying HPF both vertically and horizontally */
156  if (num_bands > 3) {
157  b3_6 = b3_ptr[indx+1]; // b3[x+1,y ]
158  b3_3 = b3_ptr[back_pitch+indx+1]; // b3[x+1,y-1]
159 
160  tmp0 = b3_1 + b3_4;
161  tmp1 = b3_2 + b3_5;
162  tmp2 = b3_3 + b3_6;
163 
164  b3_9 = b3_3 - b3_6*6 + b3_ptr[pitch+indx+1];
165 
166  p0 += (tmp0 + tmp1) << 2;
167  p1 += (tmp0 - tmp1*6 + tmp2) << 1;
168  p2 += (b3_7 + b3_8) << 1;
169  p3 += b3_7 - b3_8*6 + b3_9;
170  }
171 
172  /* output four pixels */
173  dst[x] = av_clip_uint8((p0 >> 6) + 128);
174  dst[x+1] = av_clip_uint8((p1 >> 6) + 128);
175  dst[dst_pitch+x] = av_clip_uint8((p2 >> 6) + 128);
176  dst[dst_pitch+x+1] = av_clip_uint8((p3 >> 6) + 128);
177  }// for x
178 
179  dst += dst_pitch << 1;
180 
181  back_pitch = -pitch;
182 
183  b0_ptr += pitch + 1;
184  b1_ptr += pitch + 1;
185  b2_ptr += pitch + 1;
186  b3_ptr += pitch + 1;
187  }
188 }
189 
190 void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst,
191  const int dst_pitch)
192 {
193  int x, y, indx, b0, b1, b2, b3, p0, p1, p2, p3;
194  const short *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
195  int32_t pitch;
196 
197  /* all bands should have the same pitch */
198  pitch = plane->bands[0].pitch;
199 
200  /* get pointers to the wavelet bands */
201  b0_ptr = plane->bands[0].buf;
202  b1_ptr = plane->bands[1].buf;
203  b2_ptr = plane->bands[2].buf;
204  b3_ptr = plane->bands[3].buf;
205 
206  for (y = 0; y < plane->height; y += 2) {
207  for (x = 0, indx = 0; x < plane->width; x += 2, indx++) {
208  /* load coefficients */
209  b0 = b0_ptr[indx]; //should be: b0 = (num_bands > 0) ? b0_ptr[indx] : 0;
210  b1 = b1_ptr[indx]; //should be: b1 = (num_bands > 1) ? b1_ptr[indx] : 0;
211  b2 = b2_ptr[indx]; //should be: b2 = (num_bands > 2) ? b2_ptr[indx] : 0;
212  b3 = b3_ptr[indx]; //should be: b3 = (num_bands > 3) ? b3_ptr[indx] : 0;
213 
214  /* haar wavelet recomposition */
215  p0 = (b0 + b1 + b2 + b3 + 2) >> 2;
216  p1 = (b0 + b1 - b2 - b3 + 2) >> 2;
217  p2 = (b0 - b1 + b2 - b3 + 2) >> 2;
218  p3 = (b0 - b1 - b2 + b3 + 2) >> 2;
219 
220  /* bias, convert and output four pixels */
221  dst[x] = av_clip_uint8(p0 + 128);
222  dst[x + 1] = av_clip_uint8(p1 + 128);
223  dst[dst_pitch + x] = av_clip_uint8(p2 + 128);
224  dst[dst_pitch + x + 1] = av_clip_uint8(p3 + 128);
225  }// for x
226 
227  dst += dst_pitch << 1;
228 
229  b0_ptr += pitch;
230  b1_ptr += pitch;
231  b2_ptr += pitch;
232  b3_ptr += pitch;
233  }// for y
234 }
235 
236 /** butterfly operation for the inverse Haar transform */
237 #define IVI_HAAR_BFLY(s1, s2, o1, o2, t) \
238  t = (s1 - s2) >> 1;\
239  o1 = (s1 + s2) >> 1;\
240  o2 = t;\
241 
242 /** inverse 8-point Haar transform */
243 #define INV_HAAR8(s1, s5, s3, s7, s2, s4, s6, s8,\
244  d1, d2, d3, d4, d5, d6, d7, d8,\
245  t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
246  t1 = s1 << 1; t5 = s5 << 1;\
247  IVI_HAAR_BFLY(t1, t5, t1, t5, t0); IVI_HAAR_BFLY(t1, s3, t1, t3, t0);\
248  IVI_HAAR_BFLY(t5, s7, t5, t7, t0); IVI_HAAR_BFLY(t1, s2, t1, t2, t0);\
249  IVI_HAAR_BFLY(t3, s4, t3, t4, t0); IVI_HAAR_BFLY(t5, s6, t5, t6, t0);\
250  IVI_HAAR_BFLY(t7, s8, t7, t8, t0);\
251  d1 = COMPENSATE(t1);\
252  d2 = COMPENSATE(t2);\
253  d3 = COMPENSATE(t3);\
254  d4 = COMPENSATE(t4);\
255  d5 = COMPENSATE(t5);\
256  d6 = COMPENSATE(t6);\
257  d7 = COMPENSATE(t7);\
258  d8 = COMPENSATE(t8); }
259 
260 /** inverse 4-point Haar transform */
261 #define INV_HAAR4(s1, s3, s5, s7) {\
262  HAAR_BFLY(s1, s5); HAAR_BFLY(s1, s3); HAAR_BFLY(s5, s7);\
263  s1 = COMPENSATE(s1);\
264  s3 = COMPENSATE(s3);\
265  s5 = COMPENSATE(s5);\
266  s7 = COMPENSATE(s7); }
267 
268 void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
269  const uint8_t *flags)
270 {
271  int i, shift, sp1, sp2, sp3, sp4;
272  const int32_t *src;
273  int32_t *dst;
274  int tmp[64];
275  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
276 
277  /* apply the InvHaar8 to all columns */
278 #define COMPENSATE(x) (x)
279  src = in;
280  dst = tmp;
281  for (i = 0; i < 8; i++) {
282  if (flags[i]) {
283  /* pre-scaling */
284  shift = !(i & 4);
285  sp1 = src[ 0] << shift;
286  sp2 = src[ 8] << shift;
287  sp3 = src[16] << shift;
288  sp4 = src[24] << shift;
289  INV_HAAR8( sp1, sp2, sp3, sp4,
290  src[32], src[40], src[48], src[56],
291  dst[ 0], dst[ 8], dst[16], dst[24],
292  dst[32], dst[40], dst[48], dst[56],
293  t0, t1, t2, t3, t4, t5, t6, t7, t8);
294  } else
295  dst[ 0] = dst[ 8] = dst[16] = dst[24] =
296  dst[32] = dst[40] = dst[48] = dst[56] = 0;
297 
298  src++;
299  dst++;
300  }
301 #undef COMPENSATE
302 
303  /* apply the InvHaar8 to all rows */
304 #define COMPENSATE(x) (x)
305  src = tmp;
306  for (i = 0; i < 8; i++) {
307  if ( !src[0] && !src[1] && !src[2] && !src[3]
308  && !src[4] && !src[5] && !src[6] && !src[7]) {
309  memset(out, 0, 8 * sizeof(out[0]));
310  } else {
311  INV_HAAR8(src[0], src[1], src[2], src[3],
312  src[4], src[5], src[6], src[7],
313  out[0], out[1], out[2], out[3],
314  out[4], out[5], out[6], out[7],
315  t0, t1, t2, t3, t4, t5, t6, t7, t8);
316  }
317  src += 8;
318  out += pitch;
319  }
320 #undef COMPENSATE
321 }
322 
323 void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, uint32_t pitch,
324  int blk_size)
325 {
326  int x, y;
327  int16_t dc_coeff;
328 
329  dc_coeff = (*in + 0) >> 3;
330 
331  for (y = 0; y < blk_size; out += pitch, y++) {
332  for (x = 0; x < blk_size; x++)
333  out[x] = dc_coeff;
334  }
335 }
336 
337 /** butterfly operation for the inverse slant transform */
338 #define IVI_SLANT_BFLY(s1, s2, o1, o2, t) \
339  t = s1 - s2;\
340  o1 = s1 + s2;\
341  o2 = t;\
342 
343 /** This is a reflection a,b = 1/2, 5/4 for the inverse slant transform */
344 #define IVI_IREFLECT(s1, s2, o1, o2, t) \
345  t = ((s1 + s2*2 + 2) >> 2) + s1;\
346  o2 = ((s1*2 - s2 + 2) >> 2) - s2;\
347  o1 = t;\
348 
349 /** This is a reflection a,b = 1/2, 7/8 for the inverse slant transform */
350 #define IVI_SLANT_PART4(s1, s2, o1, o2, t) \
351  t = s2 + ((s1*4 - s2 + 4) >> 3);\
352  o2 = s1 + ((-s1 - s2*4 + 4) >> 3);\
353  o1 = t;\
354 
355 /** inverse slant8 transform */
356 #define IVI_INV_SLANT8(s1, s4, s8, s5, s2, s6, s3, s7,\
357  d1, d2, d3, d4, d5, d6, d7, d8,\
358  t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
359  IVI_SLANT_PART4(s4, s5, t4, t5, t0);\
360 \
361  IVI_SLANT_BFLY(s1, t5, t1, t5, t0); IVI_SLANT_BFLY(s2, s6, t2, t6, t0);\
362  IVI_SLANT_BFLY(s7, s3, t7, t3, t0); IVI_SLANT_BFLY(t4, s8, t4, t8, t0);\
363 \
364  IVI_SLANT_BFLY(t1, t2, t1, t2, t0); IVI_IREFLECT (t4, t3, t4, t3, t0);\
365  IVI_SLANT_BFLY(t5, t6, t5, t6, t0); IVI_IREFLECT (t8, t7, t8, t7, t0);\
366  IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
367  IVI_SLANT_BFLY(t5, t8, t5, t8, t0); IVI_SLANT_BFLY(t6, t7, t6, t7, t0);\
368  d1 = COMPENSATE(t1);\
369  d2 = COMPENSATE(t2);\
370  d3 = COMPENSATE(t3);\
371  d4 = COMPENSATE(t4);\
372  d5 = COMPENSATE(t5);\
373  d6 = COMPENSATE(t6);\
374  d7 = COMPENSATE(t7);\
375  d8 = COMPENSATE(t8);}
376 
377 /** inverse slant4 transform */
378 #define IVI_INV_SLANT4(s1, s4, s2, s3, d1, d2, d3, d4, t0, t1, t2, t3, t4) {\
379  IVI_SLANT_BFLY(s1, s2, t1, t2, t0); IVI_IREFLECT (s4, s3, t4, t3, t0);\
380 \
381  IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
382  d1 = COMPENSATE(t1);\
383  d2 = COMPENSATE(t2);\
384  d3 = COMPENSATE(t3);\
385  d4 = COMPENSATE(t4);}
386 
387 void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
388 {
389  int i;
390  const int32_t *src;
391  int32_t *dst;
392  int tmp[64];
393  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
394 
395 #define COMPENSATE(x) (x)
396  src = in;
397  dst = tmp;
398  for (i = 0; i < 8; i++) {
399  if (flags[i]) {
400  IVI_INV_SLANT8(src[0], src[8], src[16], src[24], src[32], src[40], src[48], src[56],
401  dst[0], dst[8], dst[16], dst[24], dst[32], dst[40], dst[48], dst[56],
402  t0, t1, t2, t3, t4, t5, t6, t7, t8);
403  } else
404  dst[0] = dst[8] = dst[16] = dst[24] = dst[32] = dst[40] = dst[48] = dst[56] = 0;
405 
406  src++;
407  dst++;
408  }
409 #undef COMPENSATE
410 
411 #define COMPENSATE(x) ((x + 1)>>1)
412  src = tmp;
413  for (i = 0; i < 8; i++) {
414  if (!src[0] && !src[1] && !src[2] && !src[3] && !src[4] && !src[5] && !src[6] && !src[7]) {
415  memset(out, 0, 8*sizeof(out[0]));
416  } else {
417  IVI_INV_SLANT8(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
418  out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
419  t0, t1, t2, t3, t4, t5, t6, t7, t8);
420  }
421  src += 8;
422  out += pitch;
423  }
424 #undef COMPENSATE
425 }
426 
427 void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
428 {
429  int i;
430  const int32_t *src;
431  int32_t *dst;
432  int tmp[16];
433  int t0, t1, t2, t3, t4;
434 
435 #define COMPENSATE(x) (x)
436  src = in;
437  dst = tmp;
438  for (i = 0; i < 4; i++) {
439  if (flags[i]) {
440  IVI_INV_SLANT4(src[0], src[4], src[8], src[12],
441  dst[0], dst[4], dst[8], dst[12],
442  t0, t1, t2, t3, t4);
443  } else
444  dst[0] = dst[4] = dst[8] = dst[12] = 0;
445 
446  src++;
447  dst++;
448  }
449 #undef COMPENSATE
450 
451 #define COMPENSATE(x) ((x + 1)>>1)
452  src = tmp;
453  for (i = 0; i < 4; i++) {
454  if (!src[0] && !src[1] && !src[2] && !src[3]) {
455  out[0] = out[1] = out[2] = out[3] = 0;
456  } else {
457  IVI_INV_SLANT4(src[0], src[1], src[2], src[3],
458  out[0], out[1], out[2], out[3],
459  t0, t1, t2, t3, t4);
460  }
461  src += 4;
462  out += pitch;
463  }
464 #undef COMPENSATE
465 }
466 
467 void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
468 {
469  int x, y;
470  int16_t dc_coeff;
471 
472  dc_coeff = (*in + 1) >> 1;
473 
474  for (y = 0; y < blk_size; out += pitch, y++) {
475  for (x = 0; x < blk_size; x++)
476  out[x] = dc_coeff;
477  }
478 }
479 
480 void ff_ivi_row_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
481 {
482  int i;
483  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
484 
485 #define COMPENSATE(x) ((x + 1)>>1)
486  for (i = 0; i < 8; i++) {
487  if (!in[0] && !in[1] && !in[2] && !in[3] && !in[4] && !in[5] && !in[6] && !in[7]) {
488  memset(out, 0, 8*sizeof(out[0]));
489  } else {
490  IVI_INV_SLANT8( in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7],
491  out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
492  t0, t1, t2, t3, t4, t5, t6, t7, t8);
493  }
494  in += 8;
495  out += pitch;
496  }
497 #undef COMPENSATE
498 }
499 
500 void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
501 {
502  int x, y;
503  int16_t dc_coeff;
504 
505  dc_coeff = (*in + 1) >> 1;
506 
507  for (x = 0; x < blk_size; x++)
508  out[x] = dc_coeff;
509 
510  out += pitch;
511 
512  for (y = 1; y < blk_size; out += pitch, y++) {
513  for (x = 0; x < blk_size; x++)
514  out[x] = 0;
515  }
516 }
517 
518 void ff_ivi_col_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
519 {
520  int i, row2, row4, row8;
521  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
522 
523  row2 = pitch << 1;
524  row4 = pitch << 2;
525  row8 = pitch << 3;
526 
527 #define COMPENSATE(x) ((x + 1)>>1)
528  for (i = 0; i < 8; i++) {
529  if (flags[i]) {
530  IVI_INV_SLANT8(in[0], in[8], in[16], in[24], in[32], in[40], in[48], in[56],
531  out[0], out[pitch], out[row2], out[row2 + pitch], out[row4],
532  out[row4 + pitch], out[row4 + row2], out[row8 - pitch],
533  t0, t1, t2, t3, t4, t5, t6, t7, t8);
534  } else {
535  out[0] = out[pitch] = out[row2] = out[row2 + pitch] = out[row4] =
536  out[row4 + pitch] = out[row4 + row2] = out[row8 - pitch] = 0;
537  }
538 
539  in++;
540  out++;
541  }
542 #undef COMPENSATE
543 }
544 
545 void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
546 {
547  int x, y;
548  int16_t dc_coeff;
549 
550  dc_coeff = (*in + 1) >> 1;
551 
552  for (y = 0; y < blk_size; out += pitch, y++) {
553  out[0] = dc_coeff;
554  for (x = 1; x < blk_size; x++)
555  out[x] = 0;
556  }
557 }
558 
559 void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
560  const uint8_t *flags)
561 {
562  int x, y;
563 
564  for (y = 0; y < 8; out += pitch, in += 8, y++)
565  for (x = 0; x < 8; x++)
566  out[x] = in[x];
567 }
568 
569 void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
570  int blk_size)
571 {
572  int y;
573 
574  out[0] = in[0];
575  memset(out + 1, 0, 7*sizeof(out[0]));
576  out += pitch;
577 
578  for (y = 1; y < 8; out += pitch, y++)
579  memset(out, 0, 8*sizeof(out[0]));
580 }
581 
582 #define IVI_MC_TEMPLATE(size, suffix, OP) \
583 void ff_ivi_mc_ ## size ##x## size ## suffix (int16_t *buf, const int16_t *ref_buf, \
584  uint32_t pitch, int mc_type) \
585 { \
586  int i, j; \
587  const int16_t *wptr; \
588 \
589  switch (mc_type) { \
590  case 0: /* fullpel (no interpolation) */ \
591  for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) { \
592  for (j = 0; j < size; j++) {\
593  OP(buf[j], ref_buf[j]); \
594  } \
595  } \
596  break; \
597  case 1: /* horizontal halfpel interpolation */ \
598  for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) \
599  for (j = 0; j < size; j++) \
600  OP(buf[j], (ref_buf[j] + ref_buf[j+1]) >> 1); \
601  break; \
602  case 2: /* vertical halfpel interpolation */ \
603  wptr = ref_buf + pitch; \
604  for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
605  for (j = 0; j < size; j++) \
606  OP(buf[j], (ref_buf[j] + wptr[j]) >> 1); \
607  break; \
608  case 3: /* vertical and horizontal halfpel interpolation */ \
609  wptr = ref_buf + pitch; \
610  for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
611  for (j = 0; j < size; j++) \
612  OP(buf[j], (ref_buf[j] + ref_buf[j+1] + wptr[j] + wptr[j+1]) >> 2); \
613  break; \
614  } \
615 } \
616 
617 #define OP_PUT(a, b) (a) = (b)
618 #define OP_ADD(a, b) (a) += (b)
619 
620 IVI_MC_TEMPLATE(8, _no_delta, OP_PUT)
621 IVI_MC_TEMPLATE(8, _delta, OP_ADD)
622 IVI_MC_TEMPLATE(4, _no_delta, OP_PUT)
623 IVI_MC_TEMPLATE(4, _delta, OP_ADD)
Generated on Sat May 25 2013 04:01:05 for FFmpeg by   doxygen 1.8.2