rv40dsp.c
Go to the documentation of this file.
1 /*
2  * RV40 decoder motion compensation functions
3  * Copyright (c) 2008 Konstantin Shishkov
4  *
5  * This file is part of Libav.
6  *
7  * Libav is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * Libav is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with Libav; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
27 #include "avcodec.h"
28 #include "dsputil.h"
29 #include "rv34dsp.h"
30 #include "libavutil/common.h"
31 
32 #define RV40_LOWPASS(OPNAME, OP) \
33 static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
34  const int h, const int C1, const int C2, const int SHIFT){\
35  uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
36  int i;\
37  for(i = 0; i < h; i++)\
38  {\
39  OP(dst[0], (src[-2] + src[ 3] - 5*(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
40  OP(dst[1], (src[-1] + src[ 4] - 5*(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
41  OP(dst[2], (src[ 0] + src[ 5] - 5*(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
42  OP(dst[3], (src[ 1] + src[ 6] - 5*(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
43  OP(dst[4], (src[ 2] + src[ 7] - 5*(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
44  OP(dst[5], (src[ 3] + src[ 8] - 5*(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
45  OP(dst[6], (src[ 4] + src[ 9] - 5*(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
46  OP(dst[7], (src[ 5] + src[10] - 5*(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
47  dst += dstStride;\
48  src += srcStride;\
49  }\
50 }\
51 \
52 static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
53  const int w, const int C1, const int C2, const int SHIFT){\
54  uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
55  int i;\
56  for(i = 0; i < w; i++)\
57  {\
58  const int srcB = src[-2*srcStride];\
59  const int srcA = src[-1*srcStride];\
60  const int src0 = src[0 *srcStride];\
61  const int src1 = src[1 *srcStride];\
62  const int src2 = src[2 *srcStride];\
63  const int src3 = src[3 *srcStride];\
64  const int src4 = src[4 *srcStride];\
65  const int src5 = src[5 *srcStride];\
66  const int src6 = src[6 *srcStride];\
67  const int src7 = src[7 *srcStride];\
68  const int src8 = src[8 *srcStride];\
69  const int src9 = src[9 *srcStride];\
70  const int src10 = src[10*srcStride];\
71  OP(dst[0*dstStride], (srcB + src3 - 5*(srcA+src2) + src0*C1 + src1*C2 + (1<<(SHIFT-1))) >> SHIFT);\
72  OP(dst[1*dstStride], (srcA + src4 - 5*(src0+src3) + src1*C1 + src2*C2 + (1<<(SHIFT-1))) >> SHIFT);\
73  OP(dst[2*dstStride], (src0 + src5 - 5*(src1+src4) + src2*C1 + src3*C2 + (1<<(SHIFT-1))) >> SHIFT);\
74  OP(dst[3*dstStride], (src1 + src6 - 5*(src2+src5) + src3*C1 + src4*C2 + (1<<(SHIFT-1))) >> SHIFT);\
75  OP(dst[4*dstStride], (src2 + src7 - 5*(src3+src6) + src4*C1 + src5*C2 + (1<<(SHIFT-1))) >> SHIFT);\
76  OP(dst[5*dstStride], (src3 + src8 - 5*(src4+src7) + src5*C1 + src6*C2 + (1<<(SHIFT-1))) >> SHIFT);\
77  OP(dst[6*dstStride], (src4 + src9 - 5*(src5+src8) + src6*C1 + src7*C2 + (1<<(SHIFT-1))) >> SHIFT);\
78  OP(dst[7*dstStride], (src5 + src10 - 5*(src6+src9) + src7*C1 + src8*C2 + (1<<(SHIFT-1))) >> SHIFT);\
79  dst++;\
80  src++;\
81  }\
82 }\
83 \
84 static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
85  const int w, const int C1, const int C2, const int SHIFT){\
86  OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\
87  OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
88  src += 8*srcStride;\
89  dst += 8*dstStride;\
90  OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, w-8, C1, C2, SHIFT);\
91  OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\
92 }\
93 \
94 static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
95  const int h, const int C1, const int C2, const int SHIFT){\
96  OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\
97  OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
98  src += 8*srcStride;\
99  dst += 8*dstStride;\
100  OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, h-8, C1, C2, SHIFT);\
101  OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, h-8, C1, C2, SHIFT);\
102 }\
103 \
104 
105 #define RV40_MC(OPNAME, SIZE) \
106 static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
107  OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
108 }\
109 \
110 static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
111  OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
112 }\
113 \
114 static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
115  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
116 }\
117 \
118 static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
119  uint8_t full[SIZE*(SIZE+5)];\
120  uint8_t * const full_mid = full + SIZE*2;\
121  put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
122  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
123 }\
124 \
125 static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
126  uint8_t full[SIZE*(SIZE+5)];\
127  uint8_t * const full_mid = full + SIZE*2;\
128  put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
129  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
130 }\
131 \
132 static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
133  uint8_t full[SIZE*(SIZE+5)];\
134  uint8_t * const full_mid = full + SIZE*2;\
135  put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
136  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
137 }\
138 \
139 static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
140  uint8_t full[SIZE*(SIZE+5)];\
141  uint8_t * const full_mid = full + SIZE*2;\
142  put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
143  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
144 }\
145 \
146 static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
147  uint8_t full[SIZE*(SIZE+5)];\
148  uint8_t * const full_mid = full + SIZE*2;\
149  put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
150  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
151 }\
152 \
153 static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
154  uint8_t full[SIZE*(SIZE+5)];\
155  uint8_t * const full_mid = full + SIZE*2;\
156  put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
157  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
158 }\
159 \
160 static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
161  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
162 }\
163 \
164 static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
165  uint8_t full[SIZE*(SIZE+5)];\
166  uint8_t * const full_mid = full + SIZE*2;\
167  put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
168  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
169 }\
170 \
171 static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
172  uint8_t full[SIZE*(SIZE+5)];\
173  uint8_t * const full_mid = full + SIZE*2;\
174  put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
175  OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
176 }\
177 \
178 
179 #define op_avg(a, b) a = (((a)+cm[b]+1)>>1)
180 #define op_put(a, b) a = cm[b]
181 
182 RV40_LOWPASS(put_ , op_put)
183 RV40_LOWPASS(avg_ , op_avg)
184 
185 #undef op_avg
186 #undef op_put
187 
188 RV40_MC(put_, 8)
189 RV40_MC(put_, 16)
190 RV40_MC(avg_, 8)
191 RV40_MC(avg_, 16)
192 
193 static const int rv40_bias[4][4] = {
194  { 0, 16, 32, 16 },
195  { 32, 28, 32, 28 },
196  { 0, 32, 16, 32 },
197  { 32, 28, 32, 28 }
198 };
199 
200 #define RV40_CHROMA_MC(OPNAME, OP)\
201 static void OPNAME ## rv40_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
202  const int A = (8-x) * (8-y);\
203  const int B = ( x) * (8-y);\
204  const int C = (8-x) * ( y);\
205  const int D = ( x) * ( y);\
206  int i;\
207  int bias = rv40_bias[y>>1][x>>1];\
208  \
209  assert(x<8 && y<8 && x>=0 && y>=0);\
210 \
211  if(D){\
212  for(i = 0; i < h; i++){\
213  OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
214  OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
215  OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
216  OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
217  dst += stride;\
218  src += stride;\
219  }\
220  }else{\
221  const int E = B + C;\
222  const int step = C ? stride : 1;\
223  for(i = 0; i < h; i++){\
224  OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
225  OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
226  OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
227  OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
228  dst += stride;\
229  src += stride;\
230  }\
231  }\
232 }\
233 \
234 static void OPNAME ## rv40_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
235  const int A = (8-x) * (8-y);\
236  const int B = ( x) * (8-y);\
237  const int C = (8-x) * ( y);\
238  const int D = ( x) * ( y);\
239  int i;\
240  int bias = rv40_bias[y>>1][x>>1];\
241  \
242  assert(x<8 && y<8 && x>=0 && y>=0);\
243 \
244  if(D){\
245  for(i = 0; i < h; i++){\
246  OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
247  OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
248  OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
249  OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
250  OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + bias));\
251  OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + bias));\
252  OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + bias));\
253  OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + bias));\
254  dst += stride;\
255  src += stride;\
256  }\
257  }else{\
258  const int E = B + C;\
259  const int step = C ? stride : 1;\
260  for(i = 0; i < h; i++){\
261  OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
262  OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
263  OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
264  OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
265  OP(dst[4], (A*src[4] + E*src[step+4] + bias));\
266  OP(dst[5], (A*src[5] + E*src[step+5] + bias));\
267  OP(dst[6], (A*src[6] + E*src[step+6] + bias));\
268  OP(dst[7], (A*src[7] + E*src[step+7] + bias));\
269  dst += stride;\
270  src += stride;\
271  }\
272  }\
273 }
274 
275 #define op_avg(a, b) a = (((a)+((b)>>6)+1)>>1)
276 #define op_put(a, b) a = ((b)>>6)
277 
278 RV40_CHROMA_MC(put_, op_put)
279 RV40_CHROMA_MC(avg_, op_avg)
280 
281 #define RV40_WEIGHT_FUNC(size) \
282 static void rv40_weight_func_rnd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
283 {\
284  int i, j;\
285 \
286  for (j = 0; j < size; j++) {\
287  for (i = 0; i < size; i++)\
288  dst[i] = (((w2 * src1[i]) >> 9) + ((w1 * src2[i]) >> 9) + 0x10) >> 5;\
289  src1 += stride;\
290  src2 += stride;\
291  dst += stride;\
292  }\
293 }\
294 static void rv40_weight_func_nornd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
295 {\
296  int i, j;\
297 \
298  for (j = 0; j < size; j++) {\
299  for (i = 0; i < size; i++)\
300  dst[i] = (w2 * src1[i] + w1 * src2[i] + 0x10) >> 5;\
301  src1 += stride;\
302  src2 += stride;\
303  dst += stride;\
304  }\
305 }
306 
309 
313 static const uint8_t rv40_dither_l[16] = {
314  0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,
315  0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40
316 };
317 
321 static const uint8_t rv40_dither_r[16] = {
322  0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,
323  0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40
324 };
325 
326 #define CLIP_SYMM(a, b) av_clip(a, -(b), b)
327 
331  const int step,
332  const ptrdiff_t stride,
333  const int filter_p1,
334  const int filter_q1,
335  const int alpha,
336  const int beta,
337  const int lim_p0q0,
338  const int lim_q1,
339  const int lim_p1)
340 {
342  int i, t, u, diff;
343 
344  for (i = 0; i < 4; i++, src += stride) {
345  int diff_p1p0 = src[-2*step] - src[-1*step];
346  int diff_q1q0 = src[ 1*step] - src[ 0*step];
347  int diff_p1p2 = src[-2*step] - src[-3*step];
348  int diff_q1q2 = src[ 1*step] - src[ 2*step];
349 
350  t = src[0*step] - src[-1*step];
351  if (!t)
352  continue;
353 
354  u = (alpha * FFABS(t)) >> 7;
355  if (u > 3 - (filter_p1 && filter_q1))
356  continue;
357 
358  t <<= 2;
359  if (filter_p1 && filter_q1)
360  t += src[-2*step] - src[1*step];
361 
362  diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);
363  src[-1*step] = cm[src[-1*step] + diff];
364  src[ 0*step] = cm[src[ 0*step] - diff];
365 
366  if (filter_p1 && FFABS(diff_p1p2) <= beta) {
367  t = (diff_p1p0 + diff_p1p2 - diff) >> 1;
368  src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];
369  }
370 
371  if (filter_q1 && FFABS(diff_q1q2) <= beta) {
372  t = (diff_q1q0 + diff_q1q2 + diff) >> 1;
373  src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];
374  }
375  }
376 }
377 
378 static void rv40_h_weak_loop_filter(uint8_t *src, const ptrdiff_t stride,
379  const int filter_p1, const int filter_q1,
380  const int alpha, const int beta,
381  const int lim_p0q0, const int lim_q1,
382  const int lim_p1)
383 {
384  rv40_weak_loop_filter(src, stride, 1, filter_p1, filter_q1,
385  alpha, beta, lim_p0q0, lim_q1, lim_p1);
386 }
387 
388 static void rv40_v_weak_loop_filter(uint8_t *src, const ptrdiff_t stride,
389  const int filter_p1, const int filter_q1,
390  const int alpha, const int beta,
391  const int lim_p0q0, const int lim_q1,
392  const int lim_p1)
393 {
394  rv40_weak_loop_filter(src, 1, stride, filter_p1, filter_q1,
395  alpha, beta, lim_p0q0, lim_q1, lim_p1);
396 }
397 
399  const int step,
400  const ptrdiff_t stride,
401  const int alpha,
402  const int lims,
403  const int dmode,
404  const int chroma)
405 {
406  int i;
407 
408  for(i = 0; i < 4; i++, src += stride){
409  int sflag, p0, q0, p1, q1;
410  int t = src[0*step] - src[-1*step];
411 
412  if (!t)
413  continue;
414 
415  sflag = (alpha * FFABS(t)) >> 7;
416  if (sflag > 1)
417  continue;
418 
419  p0 = (25*src[-3*step] + 26*src[-2*step] + 26*src[-1*step] +
420  26*src[ 0*step] + 25*src[ 1*step] +
421  rv40_dither_l[dmode + i]) >> 7;
422 
423  q0 = (25*src[-2*step] + 26*src[-1*step] + 26*src[ 0*step] +
424  26*src[ 1*step] + 25*src[ 2*step] +
425  rv40_dither_r[dmode + i]) >> 7;
426 
427  if (sflag) {
428  p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);
429  q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);
430  }
431 
432  p1 = (25*src[-4*step] + 26*src[-3*step] + 26*src[-2*step] + 26*p0 +
433  25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;
434  q1 = (25*src[-1*step] + 26*q0 + 26*src[ 1*step] + 26*src[ 2*step] +
435  25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;
436 
437  if (sflag) {
438  p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);
439  q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);
440  }
441 
442  src[-2*step] = p1;
443  src[-1*step] = p0;
444  src[ 0*step] = q0;
445  src[ 1*step] = q1;
446 
447  if(!chroma){
448  src[-3*step] = (25*src[-1*step] + 26*src[-2*step] +
449  51*src[-3*step] + 26*src[-4*step] + 64) >> 7;
450  src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] +
451  51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;
452  }
453  }
454 }
455 
456 static void rv40_h_strong_loop_filter(uint8_t *src, const ptrdiff_t stride,
457  const int alpha, const int lims,
458  const int dmode, const int chroma)
459 {
460  rv40_strong_loop_filter(src, stride, 1, alpha, lims, dmode, chroma);
461 }
462 
463 static void rv40_v_strong_loop_filter(uint8_t *src, const ptrdiff_t stride,
464  const int alpha, const int lims,
465  const int dmode, const int chroma)
466 {
467  rv40_strong_loop_filter(src, 1, stride, alpha, lims, dmode, chroma);
468 }
469 
471  int step, ptrdiff_t stride,
472  int beta, int beta2,
473  int edge,
474  int *p1, int *q1)
475 {
476  int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;
477  int strong0 = 0, strong1 = 0;
478  uint8_t *ptr;
479  int i;
480 
481  for (i = 0, ptr = src; i < 4; i++, ptr += stride) {
482  sum_p1p0 += ptr[-2*step] - ptr[-1*step];
483  sum_q1q0 += ptr[ 1*step] - ptr[ 0*step];
484  }
485 
486  *p1 = FFABS(sum_p1p0) < (beta << 2);
487  *q1 = FFABS(sum_q1q0) < (beta << 2);
488 
489  if(!*p1 && !*q1)
490  return 0;
491 
492  if (!edge)
493  return 0;
494 
495  for (i = 0, ptr = src; i < 4; i++, ptr += stride) {
496  sum_p1p2 += ptr[-2*step] - ptr[-3*step];
497  sum_q1q2 += ptr[ 1*step] - ptr[ 2*step];
498  }
499 
500  strong0 = *p1 && (FFABS(sum_p1p2) < beta2);
501  strong1 = *q1 && (FFABS(sum_q1q2) < beta2);
502 
503  return strong0 && strong1;
504 }
505 
506 static int rv40_h_loop_filter_strength(uint8_t *src, ptrdiff_t stride,
507  int beta, int beta2, int edge,
508  int *p1, int *q1)
509 {
510  return rv40_loop_filter_strength(src, stride, 1, beta, beta2, edge, p1, q1);
511 }
512 
513 static int rv40_v_loop_filter_strength(uint8_t *src, ptrdiff_t stride,
514  int beta, int beta2, int edge,
515  int *p1, int *q1)
516 {
517  return rv40_loop_filter_strength(src, 1, stride, beta, beta2, edge, p1, q1);
518 }
519 
521 
522  ff_rv34dsp_init(c, dsp);
523 
524  c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0];
525  c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c;
526  c->put_pixels_tab[0][ 2] = dsp->put_h264_qpel_pixels_tab[0][2];
527  c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c;
528  c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c;
529  c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c;
530  c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c;
531  c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c;
532  c->put_pixels_tab[0][ 8] = dsp->put_h264_qpel_pixels_tab[0][8];
533  c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c;
534  c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c;
535  c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c;
536  c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c;
537  c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c;
538  c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c;
540  c->avg_pixels_tab[0][ 0] = dsp->avg_h264_qpel_pixels_tab[0][0];
541  c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c;
542  c->avg_pixels_tab[0][ 2] = dsp->avg_h264_qpel_pixels_tab[0][2];
543  c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c;
544  c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c;
545  c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c;
546  c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c;
547  c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c;
548  c->avg_pixels_tab[0][ 8] = dsp->avg_h264_qpel_pixels_tab[0][8];
549  c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c;
550  c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c;
551  c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c;
552  c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c;
553  c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c;
554  c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c;
556  c->put_pixels_tab[1][ 0] = dsp->put_h264_qpel_pixels_tab[1][0];
557  c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c;
558  c->put_pixels_tab[1][ 2] = dsp->put_h264_qpel_pixels_tab[1][2];
559  c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c;
560  c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c;
561  c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c;
562  c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c;
563  c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c;
564  c->put_pixels_tab[1][ 8] = dsp->put_h264_qpel_pixels_tab[1][8];
565  c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c;
566  c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c;
567  c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c;
568  c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c;
569  c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c;
570  c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c;
572  c->avg_pixels_tab[1][ 0] = dsp->avg_h264_qpel_pixels_tab[1][0];
573  c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c;
574  c->avg_pixels_tab[1][ 2] = dsp->avg_h264_qpel_pixels_tab[1][2];
575  c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c;
576  c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c;
577  c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c;
578  c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c;
579  c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c;
580  c->avg_pixels_tab[1][ 8] = dsp->avg_h264_qpel_pixels_tab[1][8];
581  c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c;
582  c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c;
583  c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c;
584  c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c;
585  c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c;
586  c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c;
588 
589  c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c;
590  c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c;
591  c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c;
592  c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c;
593 
594  c->rv40_weight_pixels_tab[0][0] = rv40_weight_func_rnd_16;
595  c->rv40_weight_pixels_tab[0][1] = rv40_weight_func_rnd_8;
596  c->rv40_weight_pixels_tab[1][0] = rv40_weight_func_nornd_16;
597  c->rv40_weight_pixels_tab[1][1] = rv40_weight_func_nornd_8;
598 
605 
606  if (ARCH_X86)
607  ff_rv40dsp_init_x86(c, dsp);
608  if (ARCH_ARM)
609  ff_rv40dsp_init_arm(c, dsp);
610 }
qpel_mc_func put_pixels_tab[4][16]
Definition: rv34dsp.h:57
#define ARCH_ARM
Definition: config.h:14
static void rv40_h_weak_loop_filter(uint8_t *src, const ptrdiff_t stride, const int filter_p1, const int filter_q1, const int alpha, const int beta, const int lim_p0q0, const int lim_q1, const int lim_p1)
Definition: rv40dsp.c:378
#define ff_cropTbl
static void rv40_h_strong_loop_filter(uint8_t *src, const ptrdiff_t stride, const int alpha, const int lims, const int dmode, const int chroma)
Definition: rv40dsp.c:456
rv40_loop_filter_strength_func rv40_loop_filter_strength[2]
Definition: rv34dsp.h:73
#define MAX_NEG_CROP
Definition: dsputil.h:83
#define ARCH_X86
Definition: config.h:33
int stride
Definition: mace.c:144
static av_always_inline void rv40_strong_loop_filter(uint8_t *src, const int step, const ptrdiff_t stride, const int alpha, const int lims, const int dmode, const int chroma)
Definition: rv40dsp.c:398
uint8_t
#define RV40_LOWPASS(OPNAME, OP)
Definition: rv40dsp.c:32
av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext *dsp)
Definition: rv40dsp.c:520
rv40_weak_loop_filter_func rv40_weak_loop_filter[2]
Definition: rv34dsp.h:71
rv40_weight_func rv40_weight_pixels_tab[2][2]
Biweight functions, first dimension is transform size (16/8), second is whether the weight is prescal...
Definition: rv34dsp.h:66
static av_always_inline int rv40_loop_filter_strength(uint8_t *src, int step, ptrdiff_t stride, int beta, int beta2, int edge, int *p1, int *q1)
Definition: rv40dsp.c:470
#define cm
Definition: dvbsubdec.c:34
static float t
static void rv40_v_strong_loop_filter(uint8_t *src, const ptrdiff_t stride, const int alpha, const int lims, const int dmode, const int chroma)
Definition: rv40dsp.c:463
qpel_mc_func avg_pixels_tab[4][16]
Definition: rv34dsp.h:58
RV30/40 decoder motion compensation functions.
void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride)
#define RV40_MC(OPNAME, SIZE)
Definition: rv40dsp.c:105
static av_always_inline void rv40_weak_loop_filter(uint8_t *src, const int step, const ptrdiff_t stride, const int filter_p1, const int filter_q1, const int alpha, const int beta, const int lim_p0q0, const int lim_q1, const int lim_p1)
weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1
Definition: rv40dsp.c:330
void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride)
#define op_put(a, b)
Definition: rv40dsp.c:276
static const uint8_t rv40_dither_r[16]
dither values for deblocking filter - right/bottom values
Definition: rv40dsp.c:321
static void rv40_v_weak_loop_filter(uint8_t *src, const ptrdiff_t stride, const int filter_p1, const int filter_q1, const int alpha, const int beta, const int lim_p0q0, const int lim_q1, const int lim_p1)
Definition: rv40dsp.c:388
av_cold void ff_rv34dsp_init(RV34DSPContext *c, DSPContext *dsp)
Definition: rv34dsp.c:131
#define op_avg(a, b)
Definition: rv40dsp.c:275
external API header
void ff_rv40dsp_init_arm(RV34DSPContext *c, DSPContext *dsp)
#define CLIP_SYMM(a, b)
Definition: rv40dsp.c:326
static const int rv40_bias[4][4]
Definition: rv40dsp.c:193
static int step
Definition: avplay.c:252
void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride)
#define RV40_WEIGHT_FUNC(size)
Definition: rv40dsp.c:281
qpel_mc_func put_h264_qpel_pixels_tab[4][16]
Definition: dsputil.h:324
static const uint8_t rv40_dither_l[16]
dither values for deblocking filter - left/top values
Definition: rv40dsp.c:313
qpel_mc_func avg_h264_qpel_pixels_tab[4][16]
Definition: dsputil.h:325
static int rv40_h_loop_filter_strength(uint8_t *src, ptrdiff_t stride, int beta, int beta2, int edge, int *p1, int *q1)
Definition: rv40dsp.c:506
common internal and external API header
rv40_strong_loop_filter_func rv40_strong_loop_filter[2]
Definition: rv34dsp.h:72
static int rv40_v_loop_filter_strength(uint8_t *src, ptrdiff_t stride, int beta, int beta2, int edge, int *p1, int *q1)
Definition: rv40dsp.c:513
DSP utils.
void ff_rv40dsp_init_x86(RV34DSPContext *c, DSPContext *dsp)
Definition: rv40dsp_init.c:190
h264_chroma_mc_func avg_chroma_pixels_tab[3]
Definition: rv34dsp.h:60
h264_chroma_mc_func put_chroma_pixels_tab[3]
Definition: rv34dsp.h:59
#define RV40_CHROMA_MC(OPNAME, OP)
Definition: rv40dsp.c:200
void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride)
DSPContext.
Definition: dsputil.h:194