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 "dsputil.h"
31 #include "dwt.h"
32 #include "ivi_common.h"
33 #include "ivi_dsp.h"
34 
35 void ff_ivi_recompose53(const IVIPlaneDesc *plane, uint8_t *dst,
36  const int dst_pitch)
37 {
38  int x, y, indx;
39  int32_t p0, p1, p2, p3, tmp0, tmp1, tmp2;
40  int32_t b0_1, b0_2, b1_1, b1_2, b1_3, b2_1, b2_2, b2_3, b2_4, b2_5, b2_6;
41  int32_t b3_1, b3_2, b3_3, b3_4, b3_5, b3_6, b3_7, b3_8, b3_9;
42  int32_t pitch, back_pitch;
43  const IDWTELEM *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
44  const int num_bands = 4;
45 
46  /* all bands should have the same pitch */
47  pitch = plane->bands[0].pitch;
48 
49  /* pixels at the position "y-1" will be set to pixels at the "y" for the 1st iteration */
50  back_pitch = 0;
51 
52  /* get pointers to the wavelet bands */
53  b0_ptr = plane->bands[0].buf;
54  b1_ptr = plane->bands[1].buf;
55  b2_ptr = plane->bands[2].buf;
56  b3_ptr = plane->bands[3].buf;
57 
58  for (y = 0; y < plane->height; y += 2) {
59 
60  if (y+2 >= plane->height)
61  pitch= 0;
62  /* load storage variables with values */
63  if (num_bands > 0) {
64  b0_1 = b0_ptr[0];
65  b0_2 = b0_ptr[pitch];
66  }
67 
68  if (num_bands > 1) {
69  b1_1 = b1_ptr[back_pitch];
70  b1_2 = b1_ptr[0];
71  b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch];
72  }
73 
74  if (num_bands > 2) {
75  b2_2 = b2_ptr[0]; // b2[x, y ]
76  b2_3 = b2_2; // b2[x+1,y ] = b2[x,y]
77  b2_5 = b2_ptr[pitch]; // b2[x ,y+1]
78  b2_6 = b2_5; // b2[x+1,y+1] = b2[x,y+1]
79  }
80 
81  if (num_bands > 3) {
82  b3_2 = b3_ptr[back_pitch]; // b3[x ,y-1]
83  b3_3 = b3_2; // b3[x+1,y-1] = b3[x ,y-1]
84  b3_5 = b3_ptr[0]; // b3[x ,y ]
85  b3_6 = b3_5; // b3[x+1,y ] = b3[x ,y ]
86  b3_8 = b3_2 - b3_5*6 + b3_ptr[pitch];
87  b3_9 = b3_8;
88  }
89 
90  for (x = 0, indx = 0; x < plane->width; x+=2, indx++) {
91  if (x+2 >= plane->width) {
92  b0_ptr --;
93  b1_ptr --;
94  b2_ptr --;
95  b3_ptr --;
96  }
97 
98  /* some values calculated in the previous iterations can */
99  /* be reused in the next ones, so do appropriate copying */
100  b2_1 = b2_2; // b2[x-1,y ] = b2[x, y ]
101  b2_2 = b2_3; // b2[x ,y ] = b2[x+1,y ]
102  b2_4 = b2_5; // b2[x-1,y+1] = b2[x ,y+1]
103  b2_5 = b2_6; // b2[x ,y+1] = b2[x+1,y+1]
104  b3_1 = b3_2; // b3[x-1,y-1] = b3[x ,y-1]
105  b3_2 = b3_3; // b3[x ,y-1] = b3[x+1,y-1]
106  b3_4 = b3_5; // b3[x-1,y ] = b3[x ,y ]
107  b3_5 = b3_6; // b3[x ,y ] = b3[x+1,y ]
108  b3_7 = b3_8; // vert_HPF(x-1)
109  b3_8 = b3_9; // vert_HPF(x )
110 
111  p0 = p1 = p2 = p3 = 0;
112 
113  /* process the LL-band by applying LPF both vertically and horizontally */
114  if (num_bands > 0) {
115  tmp0 = b0_1;
116  tmp2 = b0_2;
117  b0_1 = b0_ptr[indx+1];
118  b0_2 = b0_ptr[pitch+indx+1];
119  tmp1 = tmp0 + b0_1;
120 
121  p0 = tmp0 << 4;
122  p1 = tmp1 << 3;
123  p2 = (tmp0 + tmp2) << 3;
124  p3 = (tmp1 + tmp2 + b0_2) << 2;
125  }
126 
127  /* process the HL-band by applying HPF vertically and LPF horizontally */
128  if (num_bands > 1) {
129  tmp0 = b1_2;
130  tmp1 = b1_1;
131  b1_2 = b1_ptr[indx+1];
132  b1_1 = b1_ptr[back_pitch+indx+1];
133 
134  tmp2 = tmp1 - tmp0*6 + b1_3;
135  b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch+indx+1];
136 
137  p0 += (tmp0 + tmp1) << 3;
138  p1 += (tmp0 + tmp1 + b1_1 + b1_2) << 2;
139  p2 += tmp2 << 2;
140  p3 += (tmp2 + b1_3) << 1;
141  }
142 
143  /* process the LH-band by applying LPF vertically and HPF horizontally */
144  if (num_bands > 2) {
145  b2_3 = b2_ptr[indx+1];
146  b2_6 = b2_ptr[pitch+indx+1];
147 
148  tmp0 = b2_1 + b2_2;
149  tmp1 = b2_1 - b2_2*6 + b2_3;
150 
151  p0 += tmp0 << 3;
152  p1 += tmp1 << 2;
153  p2 += (tmp0 + b2_4 + b2_5) << 2;
154  p3 += (tmp1 + b2_4 - b2_5*6 + b2_6) << 1;
155  }
156 
157  /* process the HH-band by applying HPF both vertically and horizontally */
158  if (num_bands > 3) {
159  b3_6 = b3_ptr[indx+1]; // b3[x+1,y ]
160  b3_3 = b3_ptr[back_pitch+indx+1]; // b3[x+1,y-1]
161 
162  tmp0 = b3_1 + b3_4;
163  tmp1 = b3_2 + b3_5;
164  tmp2 = b3_3 + b3_6;
165 
166  b3_9 = b3_3 - b3_6*6 + b3_ptr[pitch+indx+1];
167 
168  p0 += (tmp0 + tmp1) << 2;
169  p1 += (tmp0 - tmp1*6 + tmp2) << 1;
170  p2 += (b3_7 + b3_8) << 1;
171  p3 += b3_7 - b3_8*6 + b3_9;
172  }
173 
174  /* output four pixels */
175  dst[x] = av_clip_uint8((p0 >> 6) + 128);
176  dst[x+1] = av_clip_uint8((p1 >> 6) + 128);
177  dst[dst_pitch+x] = av_clip_uint8((p2 >> 6) + 128);
178  dst[dst_pitch+x+1] = av_clip_uint8((p3 >> 6) + 128);
179  }// for x
180 
181  dst += dst_pitch << 1;
182 
183  back_pitch = -pitch;
184 
185  b0_ptr += pitch + 1;
186  b1_ptr += pitch + 1;
187  b2_ptr += pitch + 1;
188  b3_ptr += pitch + 1;
189  }
190 }
191 
192 void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst,
193  const int dst_pitch)
194 {
195  int x, y, indx, b0, b1, b2, b3, p0, p1, p2, p3;
196  const IDWTELEM *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
197  int32_t pitch;
198 
199  /* all bands should have the same pitch */
200  pitch = plane->bands[0].pitch;
201 
202  /* get pointers to the wavelet bands */
203  b0_ptr = plane->bands[0].buf;
204  b1_ptr = plane->bands[1].buf;
205  b2_ptr = plane->bands[2].buf;
206  b3_ptr = plane->bands[3].buf;
207 
208  for (y = 0; y < plane->height; y += 2) {
209  for (x = 0, indx = 0; x < plane->width; x += 2, indx++) {
210  /* load coefficients */
211  b0 = b0_ptr[indx]; //should be: b0 = (num_bands > 0) ? b0_ptr[indx] : 0;
212  b1 = b1_ptr[indx]; //should be: b1 = (num_bands > 1) ? b1_ptr[indx] : 0;
213  b2 = b2_ptr[indx]; //should be: b2 = (num_bands > 2) ? b2_ptr[indx] : 0;
214  b3 = b3_ptr[indx]; //should be: b3 = (num_bands > 3) ? b3_ptr[indx] : 0;
215 
216  /* haar wavelet recomposition */
217  p0 = (b0 + b1 + b2 + b3 + 2) >> 2;
218  p1 = (b0 + b1 - b2 - b3 + 2) >> 2;
219  p2 = (b0 - b1 + b2 - b3 + 2) >> 2;
220  p3 = (b0 - b1 - b2 + b3 + 2) >> 2;
221 
222  /* bias, convert and output four pixels */
223  dst[x] = av_clip_uint8(p0 + 128);
224  dst[x + 1] = av_clip_uint8(p1 + 128);
225  dst[dst_pitch + x] = av_clip_uint8(p2 + 128);
226  dst[dst_pitch + x + 1] = av_clip_uint8(p3 + 128);
227  }// for x
228 
229  dst += dst_pitch << 1;
230 
231  b0_ptr += pitch;
232  b1_ptr += pitch;
233  b2_ptr += pitch;
234  b3_ptr += pitch;
235  }// for y
236 }
237 
238 /** butterfly operation for the inverse Haar transform */
239 #define IVI_HAAR_BFLY(s1, s2, o1, o2, t) \
240  t = (s1 - s2) >> 1;\
241  o1 = (s1 + s2) >> 1;\
242  o2 = t;\
243 
244 /** inverse 8-point Haar transform */
245 #define INV_HAAR8(s1, s5, s3, s7, s2, s4, s6, s8,\
246  d1, d2, d3, d4, d5, d6, d7, d8,\
247  t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
248  t1 = s1 << 1; t5 = s5 << 1;\
249  IVI_HAAR_BFLY(t1, t5, t1, t5, t0); IVI_HAAR_BFLY(t1, s3, t1, t3, t0);\
250  IVI_HAAR_BFLY(t5, s7, t5, t7, t0); IVI_HAAR_BFLY(t1, s2, t1, t2, t0);\
251  IVI_HAAR_BFLY(t3, s4, t3, t4, t0); IVI_HAAR_BFLY(t5, s6, t5, t6, t0);\
252  IVI_HAAR_BFLY(t7, s8, t7, t8, t0);\
253  d1 = COMPENSATE(t1);\
254  d2 = COMPENSATE(t2);\
255  d3 = COMPENSATE(t3);\
256  d4 = COMPENSATE(t4);\
257  d5 = COMPENSATE(t5);\
258  d6 = COMPENSATE(t6);\
259  d7 = COMPENSATE(t7);\
260  d8 = COMPENSATE(t8); }
261 
262 /** inverse 4-point Haar transform */
263 #define INV_HAAR4(s1, s3, s5, s7) {\
264  HAAR_BFLY(s1, s5); HAAR_BFLY(s1, s3); HAAR_BFLY(s5, s7);\
265  s1 = COMPENSATE(s1);\
266  s3 = COMPENSATE(s3);\
267  s5 = COMPENSATE(s5);\
268  s7 = COMPENSATE(s7); }
269 
270 void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
271  const uint8_t *flags)
272 {
273  int i, shift, sp1, sp2, sp3, sp4;
274  const int32_t *src;
275  int32_t *dst;
276  int tmp[64];
277  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
278 
279  /* apply the InvHaar8 to all columns */
280 #define COMPENSATE(x) (x)
281  src = in;
282  dst = tmp;
283  for (i = 0; i < 8; i++) {
284  if (flags[i]) {
285  /* pre-scaling */
286  shift = !(i & 4);
287  sp1 = src[ 0] << shift;
288  sp2 = src[ 8] << shift;
289  sp3 = src[16] << shift;
290  sp4 = src[24] << shift;
291  INV_HAAR8( sp1, sp2, sp3, sp4,
292  src[32], src[40], src[48], src[56],
293  dst[ 0], dst[ 8], dst[16], dst[24],
294  dst[32], dst[40], dst[48], dst[56],
295  t0, t1, t2, t3, t4, t5, t6, t7, t8);
296  } else
297  dst[ 0] = dst[ 8] = dst[16] = dst[24] =
298  dst[32] = dst[40] = dst[48] = dst[56] = 0;
299 
300  src++;
301  dst++;
302  }
303 #undef COMPENSATE
304 
305  /* apply the InvHaar8 to all rows */
306 #define COMPENSATE(x) (x)
307  src = tmp;
308  for (i = 0; i < 8; i++) {
309  if ( !src[0] && !src[1] && !src[2] && !src[3]
310  && !src[4] && !src[5] && !src[6] && !src[7]) {
311  memset(out, 0, 8 * sizeof(out[0]));
312  } else {
313  INV_HAAR8(src[0], src[1], src[2], src[3],
314  src[4], src[5], src[6], src[7],
315  out[0], out[1], out[2], out[3],
316  out[4], out[5], out[6], out[7],
317  t0, t1, t2, t3, t4, t5, t6, t7, t8);
318  }
319  src += 8;
320  out += pitch;
321  }
322 #undef COMPENSATE
323 }
324 
325 void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, uint32_t pitch,
326  int blk_size)
327 {
328  int x, y;
329  int16_t dc_coeff;
330 
331  dc_coeff = (*in + 0) >> 3;
332 
333  for (y = 0; y < blk_size; out += pitch, y++) {
334  for (x = 0; x < blk_size; x++)
335  out[x] = dc_coeff;
336  }
337 }
338 
339 /** butterfly operation for the inverse slant transform */
340 #define IVI_SLANT_BFLY(s1, s2, o1, o2, t) \
341  t = s1 - s2;\
342  o1 = s1 + s2;\
343  o2 = t;\
344 
345 /** This is a reflection a,b = 1/2, 5/4 for the inverse slant transform */
346 #define IVI_IREFLECT(s1, s2, o1, o2, t) \
347  t = ((s1 + s2*2 + 2) >> 2) + s1;\
348  o2 = ((s1*2 - s2 + 2) >> 2) - s2;\
349  o1 = t;\
350 
351 /** This is a reflection a,b = 1/2, 7/8 for the inverse slant transform */
352 #define IVI_SLANT_PART4(s1, s2, o1, o2, t) \
353  t = s2 + ((s1*4 - s2 + 4) >> 3);\
354  o2 = s1 + ((-s1 - s2*4 + 4) >> 3);\
355  o1 = t;\
356 
357 /** inverse slant8 transform */
358 #define IVI_INV_SLANT8(s1, s4, s8, s5, s2, s6, s3, s7,\
359  d1, d2, d3, d4, d5, d6, d7, d8,\
360  t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
361  IVI_SLANT_PART4(s4, s5, t4, t5, t0);\
362 \
363  IVI_SLANT_BFLY(s1, t5, t1, t5, t0); IVI_SLANT_BFLY(s2, s6, t2, t6, t0);\
364  IVI_SLANT_BFLY(s7, s3, t7, t3, t0); IVI_SLANT_BFLY(t4, s8, t4, t8, t0);\
365 \
366  IVI_SLANT_BFLY(t1, t2, t1, t2, t0); IVI_IREFLECT (t4, t3, t4, t3, t0);\
367  IVI_SLANT_BFLY(t5, t6, t5, t6, t0); IVI_IREFLECT (t8, t7, t8, t7, t0);\
368  IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
369  IVI_SLANT_BFLY(t5, t8, t5, t8, t0); IVI_SLANT_BFLY(t6, t7, t6, t7, t0);\
370  d1 = COMPENSATE(t1);\
371  d2 = COMPENSATE(t2);\
372  d3 = COMPENSATE(t3);\
373  d4 = COMPENSATE(t4);\
374  d5 = COMPENSATE(t5);\
375  d6 = COMPENSATE(t6);\
376  d7 = COMPENSATE(t7);\
377  d8 = COMPENSATE(t8);}
378 
379 /** inverse slant4 transform */
380 #define IVI_INV_SLANT4(s1, s4, s2, s3, d1, d2, d3, d4, t0, t1, t2, t3, t4) {\
381  IVI_SLANT_BFLY(s1, s2, t1, t2, t0); IVI_IREFLECT (s4, s3, t4, t3, t0);\
382 \
383  IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
384  d1 = COMPENSATE(t1);\
385  d2 = COMPENSATE(t2);\
386  d3 = COMPENSATE(t3);\
387  d4 = COMPENSATE(t4);}
388 
389 void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
390 {
391  int i;
392  const int32_t *src;
393  int32_t *dst;
394  int tmp[64];
395  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
396 
397 #define COMPENSATE(x) (x)
398  src = in;
399  dst = tmp;
400  for (i = 0; i < 8; i++) {
401  if (flags[i]) {
402  IVI_INV_SLANT8(src[0], src[8], src[16], src[24], src[32], src[40], src[48], src[56],
403  dst[0], dst[8], dst[16], dst[24], dst[32], dst[40], dst[48], dst[56],
404  t0, t1, t2, t3, t4, t5, t6, t7, t8);
405  } else
406  dst[0] = dst[8] = dst[16] = dst[24] = dst[32] = dst[40] = dst[48] = dst[56] = 0;
407 
408  src++;
409  dst++;
410  }
411 #undef COMPENSATE
412 
413 #define COMPENSATE(x) ((x + 1)>>1)
414  src = tmp;
415  for (i = 0; i < 8; i++) {
416  if (!src[0] && !src[1] && !src[2] && !src[3] && !src[4] && !src[5] && !src[6] && !src[7]) {
417  memset(out, 0, 8*sizeof(out[0]));
418  } else {
419  IVI_INV_SLANT8(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
420  out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
421  t0, t1, t2, t3, t4, t5, t6, t7, t8);
422  }
423  src += 8;
424  out += pitch;
425  }
426 #undef COMPENSATE
427 }
428 
429 void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
430 {
431  int i;
432  const int32_t *src;
433  int32_t *dst;
434  int tmp[16];
435  int t0, t1, t2, t3, t4;
436 
437 #define COMPENSATE(x) (x)
438  src = in;
439  dst = tmp;
440  for (i = 0; i < 4; i++) {
441  if (flags[i]) {
442  IVI_INV_SLANT4(src[0], src[4], src[8], src[12],
443  dst[0], dst[4], dst[8], dst[12],
444  t0, t1, t2, t3, t4);
445  } else
446  dst[0] = dst[4] = dst[8] = dst[12] = 0;
447 
448  src++;
449  dst++;
450  }
451 #undef COMPENSATE
452 
453 #define COMPENSATE(x) ((x + 1)>>1)
454  src = tmp;
455  for (i = 0; i < 4; i++) {
456  if (!src[0] && !src[1] && !src[2] && !src[3]) {
457  out[0] = out[1] = out[2] = out[3] = 0;
458  } else {
459  IVI_INV_SLANT4(src[0], src[1], src[2], src[3],
460  out[0], out[1], out[2], out[3],
461  t0, t1, t2, t3, t4);
462  }
463  src += 4;
464  out += pitch;
465  }
466 #undef COMPENSATE
467 }
468 
469 void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
470 {
471  int x, y;
472  int16_t dc_coeff;
473 
474  dc_coeff = (*in + 1) >> 1;
475 
476  for (y = 0; y < blk_size; out += pitch, y++) {
477  for (x = 0; x < blk_size; x++)
478  out[x] = dc_coeff;
479  }
480 }
481 
482 void ff_ivi_row_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
483 {
484  int i;
485  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
486 
487 #define COMPENSATE(x) ((x + 1)>>1)
488  for (i = 0; i < 8; i++) {
489  if (!in[0] && !in[1] && !in[2] && !in[3] && !in[4] && !in[5] && !in[6] && !in[7]) {
490  memset(out, 0, 8*sizeof(out[0]));
491  } else {
492  IVI_INV_SLANT8( in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7],
493  out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
494  t0, t1, t2, t3, t4, t5, t6, t7, t8);
495  }
496  in += 8;
497  out += pitch;
498  }
499 #undef COMPENSATE
500 }
501 
502 void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
503 {
504  int x, y;
505  int16_t dc_coeff;
506 
507  dc_coeff = (*in + 1) >> 1;
508 
509  for (x = 0; x < blk_size; x++)
510  out[x] = dc_coeff;
511 
512  out += pitch;
513 
514  for (y = 1; y < blk_size; out += pitch, y++) {
515  for (x = 0; x < blk_size; x++)
516  out[x] = 0;
517  }
518 }
519 
520 void ff_ivi_col_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
521 {
522  int i, row2, row4, row8;
523  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
524 
525  row2 = pitch << 1;
526  row4 = pitch << 2;
527  row8 = pitch << 3;
528 
529 #define COMPENSATE(x) ((x + 1)>>1)
530  for (i = 0; i < 8; i++) {
531  if (flags[i]) {
532  IVI_INV_SLANT8(in[0], in[8], in[16], in[24], in[32], in[40], in[48], in[56],
533  out[0], out[pitch], out[row2], out[row2 + pitch], out[row4],
534  out[row4 + pitch], out[row4 + row2], out[row8 - pitch],
535  t0, t1, t2, t3, t4, t5, t6, t7, t8);
536  } else {
537  out[0] = out[pitch] = out[row2] = out[row2 + pitch] = out[row4] =
538  out[row4 + pitch] = out[row4 + row2] = out[row8 - pitch] = 0;
539  }
540 
541  in++;
542  out++;
543  }
544 #undef COMPENSATE
545 }
546 
547 void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
548 {
549  int x, y;
550  int16_t dc_coeff;
551 
552  dc_coeff = (*in + 1) >> 1;
553 
554  for (y = 0; y < blk_size; out += pitch, y++) {
555  out[0] = dc_coeff;
556  for (x = 1; x < blk_size; x++)
557  out[x] = 0;
558  }
559 }
560 
561 void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
562  const uint8_t *flags)
563 {
564  int x, y;
565 
566  for (y = 0; y < 8; out += pitch, in += 8, y++)
567  for (x = 0; x < 8; x++)
568  out[x] = in[x];
569 }
570 
571 void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
572  int blk_size)
573 {
574  int y;
575 
576  out[0] = in[0];
577  memset(out + 1, 0, 7*sizeof(out[0]));
578  out += pitch;
579 
580  for (y = 1; y < 8; out += pitch, y++)
581  memset(out, 0, 8*sizeof(out[0]));
582 }
583 
584 #define IVI_MC_TEMPLATE(size, suffix, OP) \
585 void ff_ivi_mc_ ## size ##x## size ## suffix (int16_t *buf, const int16_t *ref_buf, \
586  uint32_t pitch, int mc_type) \
587 { \
588  int i, j; \
589  const int16_t *wptr; \
590 \
591  switch (mc_type) { \
592  case 0: /* fullpel (no interpolation) */ \
593  for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) { \
594  for (j = 0; j < size; j++) {\
595  OP(buf[j], ref_buf[j]); \
596  } \
597  } \
598  break; \
599  case 1: /* horizontal halfpel interpolation */ \
600  for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) \
601  for (j = 0; j < size; j++) \
602  OP(buf[j], (ref_buf[j] + ref_buf[j+1]) >> 1); \
603  break; \
604  case 2: /* vertical halfpel interpolation */ \
605  wptr = ref_buf + pitch; \
606  for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
607  for (j = 0; j < size; j++) \
608  OP(buf[j], (ref_buf[j] + wptr[j]) >> 1); \
609  break; \
610  case 3: /* vertical and horizontal halfpel interpolation */ \
611  wptr = ref_buf + pitch; \
612  for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
613  for (j = 0; j < size; j++) \
614  OP(buf[j], (ref_buf[j] + ref_buf[j+1] + wptr[j] + wptr[j+1]) >> 2); \
615  break; \
616  } \
617 } \
618 
619 #define OP_PUT(a, b) (a) = (b)
620 #define OP_ADD(a, b) (a) += (b)
621 
622 IVI_MC_TEMPLATE(8, _no_delta, OP_PUT)
623 IVI_MC_TEMPLATE(8, _delta, OP_ADD)
624 IVI_MC_TEMPLATE(4, _no_delta, OP_PUT)
625 IVI_MC_TEMPLATE(4, _delta, OP_ADD)
Generated on Sat May 25 2013 03:58:35 for FFmpeg by   doxygen 1.8.2