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 Libav.
7  *
8  * Libav 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  * Libav 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 Libav; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
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, const int num_bands)
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 
45  /* all bands should have the same pitch */
46  pitch = plane->bands[0].pitch;
47 
48  /* pixels at the position "y-1" will be set to pixels at the "y" for the 1st iteration */
49  back_pitch = 0;
50 
51  /* get pointers to the wavelet bands */
52  b0_ptr = plane->bands[0].buf;
53  b1_ptr = plane->bands[1].buf;
54  b2_ptr = plane->bands[2].buf;
55  b3_ptr = plane->bands[3].buf;
56 
57  for (y = 0; y < plane->height; y += 2) {
58  /* load storage variables with values */
59  if (num_bands > 0) {
60  b0_1 = b0_ptr[0];
61  b0_2 = b0_ptr[pitch];
62  }
63 
64  if (num_bands > 1) {
65  b1_1 = b1_ptr[back_pitch];
66  b1_2 = b1_ptr[0];
67  b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch];
68  }
69 
70  if (num_bands > 2) {
71  b2_2 = b2_ptr[0]; // b2[x, y ]
72  b2_3 = b2_2; // b2[x+1,y ] = b2[x,y]
73  b2_5 = b2_ptr[pitch]; // b2[x ,y+1]
74  b2_6 = b2_5; // b2[x+1,y+1] = b2[x,y+1]
75  }
76 
77  if (num_bands > 3) {
78  b3_2 = b3_ptr[back_pitch]; // b3[x ,y-1]
79  b3_3 = b3_2; // b3[x+1,y-1] = b3[x ,y-1]
80  b3_5 = b3_ptr[0]; // b3[x ,y ]
81  b3_6 = b3_5; // b3[x+1,y ] = b3[x ,y ]
82  b3_8 = b3_2 - b3_5*6 + b3_ptr[pitch];
83  b3_9 = b3_8;
84  }
85 
86  for (x = 0, indx = 0; x < plane->width; x+=2, indx++) {
87  /* some values calculated in the previous iterations can */
88  /* be reused in the next ones, so do appropriate copying */
89  b2_1 = b2_2; // b2[x-1,y ] = b2[x, y ]
90  b2_2 = b2_3; // b2[x ,y ] = b2[x+1,y ]
91  b2_4 = b2_5; // b2[x-1,y+1] = b2[x ,y+1]
92  b2_5 = b2_6; // b2[x ,y+1] = b2[x+1,y+1]
93  b3_1 = b3_2; // b3[x-1,y-1] = b3[x ,y-1]
94  b3_2 = b3_3; // b3[x ,y-1] = b3[x+1,y-1]
95  b3_4 = b3_5; // b3[x-1,y ] = b3[x ,y ]
96  b3_5 = b3_6; // b3[x ,y ] = b3[x+1,y ]
97  b3_7 = b3_8; // vert_HPF(x-1)
98  b3_8 = b3_9; // vert_HPF(x )
99 
100  p0 = p1 = p2 = p3 = 0;
101 
102  /* process the LL-band by applying LPF both vertically and horizontally */
103  if (num_bands > 0) {
104  tmp0 = b0_1;
105  tmp2 = b0_2;
106  b0_1 = b0_ptr[indx+1];
107  b0_2 = b0_ptr[pitch+indx+1];
108  tmp1 = tmp0 + b0_1;
109 
110  p0 = tmp0 << 4;
111  p1 = tmp1 << 3;
112  p2 = (tmp0 + tmp2) << 3;
113  p3 = (tmp1 + tmp2 + b0_2) << 2;
114  }
115 
116  /* process the HL-band by applying HPF vertically and LPF horizontally */
117  if (num_bands > 1) {
118  tmp0 = b1_2;
119  tmp1 = b1_1;
120  b1_2 = b1_ptr[indx+1];
121  b1_1 = b1_ptr[back_pitch+indx+1];
122 
123  tmp2 = tmp1 - tmp0*6 + b1_3;
124  b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch+indx+1];
125 
126  p0 += (tmp0 + tmp1) << 3;
127  p1 += (tmp0 + tmp1 + b1_1 + b1_2) << 2;
128  p2 += tmp2 << 2;
129  p3 += (tmp2 + b1_3) << 1;
130  }
131 
132  /* process the LH-band by applying LPF vertically and HPF horizontally */
133  if (num_bands > 2) {
134  b2_3 = b2_ptr[indx+1];
135  b2_6 = b2_ptr[pitch+indx+1];
136 
137  tmp0 = b2_1 + b2_2;
138  tmp1 = b2_1 - b2_2*6 + b2_3;
139 
140  p0 += tmp0 << 3;
141  p1 += tmp1 << 2;
142  p2 += (tmp0 + b2_4 + b2_5) << 2;
143  p3 += (tmp1 + b2_4 - b2_5*6 + b2_6) << 1;
144  }
145 
146  /* process the HH-band by applying HPF both vertically and horizontally */
147  if (num_bands > 3) {
148  b3_6 = b3_ptr[indx+1]; // b3[x+1,y ]
149  b3_3 = b3_ptr[back_pitch+indx+1]; // b3[x+1,y-1]
150 
151  tmp0 = b3_1 + b3_4;
152  tmp1 = b3_2 + b3_5;
153  tmp2 = b3_3 + b3_6;
154 
155  b3_9 = b3_3 - b3_6*6 + b3_ptr[pitch+indx+1];
156 
157  p0 += (tmp0 + tmp1) << 2;
158  p1 += (tmp0 - tmp1*6 + tmp2) << 1;
159  p2 += (b3_7 + b3_8) << 1;
160  p3 += b3_7 - b3_8*6 + b3_9;
161  }
162 
163  /* output four pixels */
164  dst[x] = av_clip_uint8((p0 >> 6) + 128);
165  dst[x+1] = av_clip_uint8((p1 >> 6) + 128);
166  dst[dst_pitch+x] = av_clip_uint8((p2 >> 6) + 128);
167  dst[dst_pitch+x+1] = av_clip_uint8((p3 >> 6) + 128);
168  }// for x
169 
170  dst += dst_pitch << 1;
171 
172  back_pitch = -pitch;
173 
174  b0_ptr += pitch;
175  b1_ptr += pitch;
176  b2_ptr += pitch;
177  b3_ptr += pitch;
178  }
179 }
180 
181 void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst,
182  const int dst_pitch, const int num_bands)
183 {
184  int x, y, indx, b0, b1, b2, b3, p0, p1, p2, p3;
185  const IDWTELEM *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
186  int32_t pitch;
187 
188  /* all bands should have the same pitch */
189  pitch = plane->bands[0].pitch;
190 
191  /* get pointers to the wavelet bands */
192  b0_ptr = plane->bands[0].buf;
193  b1_ptr = plane->bands[1].buf;
194  b2_ptr = plane->bands[2].buf;
195  b3_ptr = plane->bands[3].buf;
196 
197  for (y = 0; y < plane->height; y += 2) {
198  for (x = 0, indx = 0; x < plane->width; x += 2, indx++) {
199  /* load coefficients */
200  b0 = b0_ptr[indx]; //should be: b0 = (num_bands > 0) ? b0_ptr[indx] : 0;
201  b1 = b1_ptr[indx]; //should be: b1 = (num_bands > 1) ? b1_ptr[indx] : 0;
202  b2 = b2_ptr[indx]; //should be: b2 = (num_bands > 2) ? b2_ptr[indx] : 0;
203  b3 = b3_ptr[indx]; //should be: b3 = (num_bands > 3) ? b3_ptr[indx] : 0;
204 
205  /* haar wavelet recomposition */
206  p0 = (b0 + b1 + b2 + b3 + 2) >> 2;
207  p1 = (b0 + b1 - b2 - b3 + 2) >> 2;
208  p2 = (b0 - b1 + b2 - b3 + 2) >> 2;
209  p3 = (b0 - b1 - b2 + b3 + 2) >> 2;
210 
211  /* bias, convert and output four pixels */
212  dst[x] = av_clip_uint8(p0 + 128);
213  dst[x + 1] = av_clip_uint8(p1 + 128);
214  dst[dst_pitch + x] = av_clip_uint8(p2 + 128);
215  dst[dst_pitch + x + 1] = av_clip_uint8(p3 + 128);
216  }// for x
217 
218  dst += dst_pitch << 1;
219 
220  b0_ptr += pitch;
221  b1_ptr += pitch;
222  b2_ptr += pitch;
223  b3_ptr += pitch;
224  }// for y
225 }
226 
228 #define IVI_HAAR_BFLY(s1, s2, o1, o2, t) \
229  t = (s1 - s2) >> 1;\
230  o1 = (s1 + s2) >> 1;\
231  o2 = t;\
232 
233 
234 #define INV_HAAR8(s1, s5, s3, s7, s2, s4, s6, s8,\
235  d1, d2, d3, d4, d5, d6, d7, d8,\
236  t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
237  t1 = s1 << 1; t5 = s5 << 1;\
238  IVI_HAAR_BFLY(t1, t5, t1, t5, t0); IVI_HAAR_BFLY(t1, s3, t1, t3, t0);\
239  IVI_HAAR_BFLY(t5, s7, t5, t7, t0); IVI_HAAR_BFLY(t1, s2, t1, t2, t0);\
240  IVI_HAAR_BFLY(t3, s4, t3, t4, t0); IVI_HAAR_BFLY(t5, s6, t5, t6, t0);\
241  IVI_HAAR_BFLY(t7, s8, t7, t8, t0);\
242  d1 = COMPENSATE(t1);\
243  d2 = COMPENSATE(t2);\
244  d3 = COMPENSATE(t3);\
245  d4 = COMPENSATE(t4);\
246  d5 = COMPENSATE(t5);\
247  d6 = COMPENSATE(t6);\
248  d7 = COMPENSATE(t7);\
249  d8 = COMPENSATE(t8); }
250 
252 #define INV_HAAR4(s1, s3, s5, s7) {\
253  HAAR_BFLY(s1, s5); HAAR_BFLY(s1, s3); HAAR_BFLY(s5, s7);\
254  s1 = COMPENSATE(s1);\
255  s3 = COMPENSATE(s3);\
256  s5 = COMPENSATE(s5);\
257  s7 = COMPENSATE(s7); }
258 
259 void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
260  const uint8_t *flags)
261 {
262  int i, shift, sp1, sp2, sp3, sp4;
263  const int32_t *src;
264  int32_t *dst;
265  int tmp[64];
266  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
267 
268  /* apply the InvHaar8 to all columns */
269 #define COMPENSATE(x) (x)
270  src = in;
271  dst = tmp;
272  for (i = 0; i < 8; i++) {
273  if (flags[i]) {
274  /* pre-scaling */
275  shift = !(i & 4);
276  sp1 = src[ 0] << shift;
277  sp2 = src[ 8] << shift;
278  sp3 = src[16] << shift;
279  sp4 = src[24] << shift;
280  INV_HAAR8( sp1, sp2, sp3, sp4,
281  src[32], src[40], src[48], src[56],
282  dst[ 0], dst[ 8], dst[16], dst[24],
283  dst[32], dst[40], dst[48], dst[56],
284  t0, t1, t2, t3, t4, t5, t6, t7, t8);
285  } else
286  dst[ 0] = dst[ 8] = dst[16] = dst[24] =
287  dst[32] = dst[40] = dst[48] = dst[56] = 0;
288 
289  src++;
290  dst++;
291  }
292 #undef COMPENSATE
293 
294  /* apply the InvHaar8 to all rows */
295 #define COMPENSATE(x) (x)
296  src = tmp;
297  for (i = 0; i < 8; i++) {
298  if ( !src[0] && !src[1] && !src[2] && !src[3]
299  && !src[4] && !src[5] && !src[6] && !src[7]) {
300  memset(out, 0, 8 * sizeof(out[0]));
301  } else {
302  INV_HAAR8(src[0], src[1], src[2], src[3],
303  src[4], src[5], src[6], src[7],
304  out[0], out[1], out[2], out[3],
305  out[4], out[5], out[6], out[7],
306  t0, t1, t2, t3, t4, t5, t6, t7, t8);
307  }
308  src += 8;
309  out += pitch;
310  }
311 #undef COMPENSATE
312 }
313 
314 void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, uint32_t pitch,
315  int blk_size)
316 {
317  int x, y;
318  int16_t dc_coeff;
319 
320  dc_coeff = (*in + 0) >> 3;
321 
322  for (y = 0; y < blk_size; out += pitch, y++) {
323  for (x = 0; x < blk_size; x++)
324  out[x] = dc_coeff;
325  }
326 }
327 
329 #define IVI_SLANT_BFLY(s1, s2, o1, o2, t) \
330  t = s1 - s2;\
331  o1 = s1 + s2;\
332  o2 = t;\
333 
334 
335 #define IVI_IREFLECT(s1, s2, o1, o2, t) \
336  t = ((s1 + s2*2 + 2) >> 2) + s1;\
337  o2 = ((s1*2 - s2 + 2) >> 2) - s2;\
338  o1 = t;\
339 
340 
341 #define IVI_SLANT_PART4(s1, s2, o1, o2, t) \
342  t = s2 + ((s1*4 - s2 + 4) >> 3);\
343  o2 = s1 + ((-s1 - s2*4 + 4) >> 3);\
344  o1 = t;\
345 
346 
347 #define IVI_INV_SLANT8(s1, s4, s8, s5, s2, s6, s3, s7,\
348  d1, d2, d3, d4, d5, d6, d7, d8,\
349  t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
350  IVI_SLANT_PART4(s4, s5, t4, t5, t0);\
351 \
352  IVI_SLANT_BFLY(s1, t5, t1, t5, t0); IVI_SLANT_BFLY(s2, s6, t2, t6, t0);\
353  IVI_SLANT_BFLY(s7, s3, t7, t3, t0); IVI_SLANT_BFLY(t4, s8, t4, t8, t0);\
354 \
355  IVI_SLANT_BFLY(t1, t2, t1, t2, t0); IVI_IREFLECT (t4, t3, t4, t3, t0);\
356  IVI_SLANT_BFLY(t5, t6, t5, t6, t0); IVI_IREFLECT (t8, t7, t8, t7, t0);\
357  IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
358  IVI_SLANT_BFLY(t5, t8, t5, t8, t0); IVI_SLANT_BFLY(t6, t7, t6, t7, t0);\
359  d1 = COMPENSATE(t1);\
360  d2 = COMPENSATE(t2);\
361  d3 = COMPENSATE(t3);\
362  d4 = COMPENSATE(t4);\
363  d5 = COMPENSATE(t5);\
364  d6 = COMPENSATE(t6);\
365  d7 = COMPENSATE(t7);\
366  d8 = COMPENSATE(t8);}
367 
369 #define IVI_INV_SLANT4(s1, s4, s2, s3, d1, d2, d3, d4, t0, t1, t2, t3, t4) {\
370  IVI_SLANT_BFLY(s1, s2, t1, t2, t0); IVI_IREFLECT (s4, s3, t4, t3, t0);\
371 \
372  IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
373  d1 = COMPENSATE(t1);\
374  d2 = COMPENSATE(t2);\
375  d3 = COMPENSATE(t3);\
376  d4 = COMPENSATE(t4);}
377 
378 void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
379 {
380  int i;
381  const int32_t *src;
382  int32_t *dst;
383  int tmp[64];
384  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
385 
386 #define COMPENSATE(x) (x)
387  src = in;
388  dst = tmp;
389  for (i = 0; i < 8; i++) {
390  if (flags[i]) {
391  IVI_INV_SLANT8(src[0], src[8], src[16], src[24], src[32], src[40], src[48], src[56],
392  dst[0], dst[8], dst[16], dst[24], dst[32], dst[40], dst[48], dst[56],
393  t0, t1, t2, t3, t4, t5, t6, t7, t8);
394  } else
395  dst[0] = dst[8] = dst[16] = dst[24] = dst[32] = dst[40] = dst[48] = dst[56] = 0;
396 
397  src++;
398  dst++;
399  }
400 #undef COMPENSATE
401 
402 #define COMPENSATE(x) ((x + 1)>>1)
403  src = tmp;
404  for (i = 0; i < 8; i++) {
405  if (!src[0] && !src[1] && !src[2] && !src[3] && !src[4] && !src[5] && !src[6] && !src[7]) {
406  memset(out, 0, 8*sizeof(out[0]));
407  } else {
408  IVI_INV_SLANT8(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
409  out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
410  t0, t1, t2, t3, t4, t5, t6, t7, t8);
411  }
412  src += 8;
413  out += pitch;
414  }
415 #undef COMPENSATE
416 }
417 
418 void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
419 {
420  int i;
421  const int32_t *src;
422  int32_t *dst;
423  int tmp[16];
424  int t0, t1, t2, t3, t4;
425 
426 #define COMPENSATE(x) (x)
427  src = in;
428  dst = tmp;
429  for (i = 0; i < 4; i++) {
430  if (flags[i]) {
431  IVI_INV_SLANT4(src[0], src[4], src[8], src[12],
432  dst[0], dst[4], dst[8], dst[12],
433  t0, t1, t2, t3, t4);
434  } else
435  dst[0] = dst[4] = dst[8] = dst[12] = 0;
436 
437  src++;
438  dst++;
439  }
440 #undef COMPENSATE
441 
442 #define COMPENSATE(x) ((x + 1)>>1)
443  src = tmp;
444  for (i = 0; i < 4; i++) {
445  if (!src[0] && !src[1] && !src[2] && !src[3]) {
446  out[0] = out[1] = out[2] = out[3] = 0;
447  } else {
448  IVI_INV_SLANT4(src[0], src[1], src[2], src[3],
449  out[0], out[1], out[2], out[3],
450  t0, t1, t2, t3, t4);
451  }
452  src += 4;
453  out += pitch;
454  }
455 #undef COMPENSATE
456 }
457 
458 void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
459 {
460  int x, y;
461  int16_t dc_coeff;
462 
463  dc_coeff = (*in + 1) >> 1;
464 
465  for (y = 0; y < blk_size; out += pitch, y++) {
466  for (x = 0; x < blk_size; x++)
467  out[x] = dc_coeff;
468  }
469 }
470 
471 void ff_ivi_row_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
472 {
473  int i;
474  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
475 
476 #define COMPENSATE(x) ((x + 1)>>1)
477  for (i = 0; i < 8; i++) {
478  if (!in[0] && !in[1] && !in[2] && !in[3] && !in[4] && !in[5] && !in[6] && !in[7]) {
479  memset(out, 0, 8*sizeof(out[0]));
480  } else {
481  IVI_INV_SLANT8( in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7],
482  out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
483  t0, t1, t2, t3, t4, t5, t6, t7, t8);
484  }
485  in += 8;
486  out += pitch;
487  }
488 #undef COMPENSATE
489 }
490 
491 void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
492 {
493  int x, y;
494  int16_t dc_coeff;
495 
496  dc_coeff = (*in + 1) >> 1;
497 
498  for (x = 0; x < blk_size; x++)
499  out[x] = dc_coeff;
500 
501  out += pitch;
502 
503  for (y = 1; y < blk_size; out += pitch, y++) {
504  for (x = 0; x < blk_size; x++)
505  out[x] = 0;
506  }
507 }
508 
509 void ff_ivi_col_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
510 {
511  int i, row2, row4, row8;
512  int t0, t1, t2, t3, t4, t5, t6, t7, t8;
513 
514  row2 = pitch << 1;
515  row4 = pitch << 2;
516  row8 = pitch << 3;
517 
518 #define COMPENSATE(x) ((x + 1)>>1)
519  for (i = 0; i < 8; i++) {
520  if (flags[i]) {
521  IVI_INV_SLANT8(in[0], in[8], in[16], in[24], in[32], in[40], in[48], in[56],
522  out[0], out[pitch], out[row2], out[row2 + pitch], out[row4],
523  out[row4 + pitch], out[row4 + row2], out[row8 - pitch],
524  t0, t1, t2, t3, t4, t5, t6, t7, t8);
525  } else {
526  out[0] = out[pitch] = out[row2] = out[row2 + pitch] = out[row4] =
527  out[row4 + pitch] = out[row4 + row2] = out[row8 - pitch] = 0;
528  }
529 
530  in++;
531  out++;
532  }
533 #undef COMPENSATE
534 }
535 
536 void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
537 {
538  int x, y;
539  int16_t dc_coeff;
540 
541  dc_coeff = (*in + 1) >> 1;
542 
543  for (y = 0; y < blk_size; out += pitch, y++) {
544  out[0] = dc_coeff;
545  for (x = 1; x < blk_size; x++)
546  out[x] = 0;
547  }
548 }
549 
550 void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
551  const uint8_t *flags)
552 {
553  int x, y;
554 
555  for (y = 0; y < 8; out += pitch, in += 8, y++)
556  for (x = 0; x < 8; x++)
557  out[x] = in[x];
558 }
559 
560 void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
561  int blk_size)
562 {
563  int y;
564 
565  out[0] = in[0];
566  memset(out + 1, 0, 7*sizeof(out[0]));
567  out += pitch;
568 
569  for (y = 1; y < 8; out += pitch, y++)
570  memset(out, 0, 8*sizeof(out[0]));
571 }
572 
573 #define IVI_MC_TEMPLATE(size, suffix, OP) \
574 void ff_ivi_mc_ ## size ##x## size ## suffix (int16_t *buf, const int16_t *ref_buf, \
575  uint32_t pitch, int mc_type) \
576 { \
577  int i, j; \
578  const int16_t *wptr; \
579 \
580  switch (mc_type) { \
581  case 0: /* fullpel (no interpolation) */ \
582  for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) { \
583  for (j = 0; j < size; j++) {\
584  OP(buf[j], ref_buf[j]); \
585  } \
586  } \
587  break; \
588  case 1: /* horizontal halfpel interpolation */ \
589  for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) \
590  for (j = 0; j < size; j++) \
591  OP(buf[j], (ref_buf[j] + ref_buf[j+1]) >> 1); \
592  break; \
593  case 2: /* vertical halfpel interpolation */ \
594  wptr = ref_buf + pitch; \
595  for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
596  for (j = 0; j < size; j++) \
597  OP(buf[j], (ref_buf[j] + wptr[j]) >> 1); \
598  break; \
599  case 3: /* vertical and horizontal halfpel interpolation */ \
600  wptr = ref_buf + pitch; \
601  for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
602  for (j = 0; j < size; j++) \
603  OP(buf[j], (ref_buf[j] + ref_buf[j+1] + wptr[j] + wptr[j+1]) >> 2); \
604  break; \
605  } \
606 } \
607 
608 #define OP_PUT(a, b) (a) = (b)
609 #define OP_ADD(a, b) (a) += (b)
610 
611 IVI_MC_TEMPLATE(8, _no_delta, OP_PUT)
612 IVI_MC_TEMPLATE(8, _delta, OP_ADD)
613 IVI_MC_TEMPLATE(4, _no_delta, OP_PUT)
614 IVI_MC_TEMPLATE(4, _delta, OP_ADD)