add rounding bias earlier in the idct, avoids many adds
[libav.git] / libavcodec / cavsdsp.c
1 /*
2 * Chinese AVS video (AVS1-P2, JiZhun profile) decoder.
3 *
4 * DSP functions
5 *
6 * Copyright (c) 2006 Stefan Gehrer <stefan.gehrer@gmx.de>
7 *
8 * This library 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 of the License, or (at your option) any later version.
12 *
13 * This library 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 this library; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 #include <stdio.h>
24 #include "dsputil.h"
25
26 /*****************************************************************************
27 *
28 * in-loop deblocking filter
29 *
30 ****************************************************************************/
31
32 #define P2 p0_p[-3*stride]
33 #define P1 p0_p[-2*stride]
34 #define P0 p0_p[-1*stride]
35 #define Q0 p0_p[ 0*stride]
36 #define Q1 p0_p[ 1*stride]
37 #define Q2 p0_p[ 2*stride]
38
39 static inline void loop_filter_l2(uint8_t *p0_p,int stride,int alpha, int beta) {
40 int p0 = P0;
41 int q0 = Q0;
42
43 if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
44 int s = p0 + q0 + 2;
45 alpha = (alpha>>2) + 2;
46 if(abs(P2-p0) < beta && abs(p0-q0) < alpha) {
47 P0 = (P1 + p0 + s) >> 2;
48 P1 = (2*P1 + s) >> 2;
49 } else
50 P0 = (2*P1 + s) >> 2;
51 if(abs(Q2-q0) < beta && abs(q0-p0) < alpha) {
52 Q0 = (Q1 + q0 + s) >> 2;
53 Q1 = (2*Q1 + s) >> 2;
54 } else
55 Q0 = (2*Q1 + s) >> 2;
56 }
57 }
58
59 static inline void loop_filter_l1(uint8_t *p0_p, int stride, int alpha, int beta, int tc) {
60 int p0 = P0;
61 int q0 = Q0;
62
63 if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
64 int delta = clip(((q0-p0)*3+P1-Q1+4)>>3,-tc, tc);
65 P0 = clip_uint8(p0+delta);
66 Q0 = clip_uint8(q0-delta);
67 if(abs(P2-p0)<beta) {
68 delta = clip(((P0-P1)*3+P2-Q0+4)>>3, -tc, tc);
69 P1 = clip_uint8(P1+delta);
70 }
71 if(abs(Q2-q0)<beta) {
72 delta = clip(((Q1-Q0)*3+P0-Q2+4)>>3, -tc, tc);
73 Q1 = clip_uint8(Q1-delta);
74 }
75 }
76 }
77
78 static inline void loop_filter_c2(uint8_t *p0_p,int stride,int alpha, int beta) {
79 int p0 = P0;
80 int q0 = Q0;
81
82 if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
83 int s = p0 + q0 + 2;
84 alpha = (alpha>>2) + 2;
85 if(abs(P2-p0) < beta && abs(p0-q0) < alpha) {
86 P0 = (P1 + p0 + s) >> 2;
87 } else
88 P0 = (2*P1 + s) >> 2;
89 if(abs(Q2-q0) < beta && abs(q0-p0) < alpha) {
90 Q0 = (Q1 + q0 + s) >> 2;
91 } else
92 Q0 = (2*Q1 + s) >> 2;
93 }
94 }
95
96 static inline void loop_filter_c1(uint8_t *p0_p,int stride,int alpha, int beta,
97 int tc) {
98 if(abs(P0-Q0)<alpha && abs(P1-P0)<beta && abs(Q1-Q0)<beta) {
99 int delta = clip(((Q0-P0)*3+P1-Q1+4)>>3, -tc, tc);
100 P0 = clip_uint8(P0+delta);
101 Q0 = clip_uint8(Q0-delta);
102 }
103 }
104
105 #undef P0
106 #undef P1
107 #undef P2
108 #undef Q0
109 #undef Q1
110 #undef Q2
111
112 static void cavs_filter_lv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
113 int bs1, int bs2) {
114 int i;
115 if(bs1==2)
116 for(i=0;i<16;i++)
117 loop_filter_l2(d + i*stride,1,alpha,beta);
118 else {
119 if(bs1)
120 for(i=0;i<8;i++)
121 loop_filter_l1(d + i*stride,1,alpha,beta,tc);
122 if (bs2)
123 for(i=8;i<16;i++)
124 loop_filter_l1(d + i*stride,1,alpha,beta,tc);
125 }
126 }
127
128 static void cavs_filter_lh_c(uint8_t *d, int stride, int alpha, int beta, int tc,
129 int bs1, int bs2) {
130 int i;
131 if(bs1==2)
132 for(i=0;i<16;i++)
133 loop_filter_l2(d + i,stride,alpha,beta);
134 else {
135 if(bs1)
136 for(i=0;i<8;i++)
137 loop_filter_l1(d + i,stride,alpha,beta,tc);
138 if (bs2)
139 for(i=8;i<16;i++)
140 loop_filter_l1(d + i,stride,alpha,beta,tc);
141 }
142 }
143
144 static void cavs_filter_cv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
145 int bs1, int bs2) {
146 int i;
147 if(bs1==2)
148 for(i=0;i<8;i++)
149 loop_filter_c2(d + i*stride,1,alpha,beta);
150 else {
151 if(bs1)
152 for(i=0;i<4;i++)
153 loop_filter_c1(d + i*stride,1,alpha,beta,tc);
154 if (bs2)
155 for(i=4;i<8;i++)
156 loop_filter_c1(d + i*stride,1,alpha,beta,tc);
157 }
158 }
159
160 static void cavs_filter_ch_c(uint8_t *d, int stride, int alpha, int beta, int tc,
161 int bs1, int bs2) {
162 int i;
163 if(bs1==2)
164 for(i=0;i<8;i++)
165 loop_filter_c2(d + i,stride,alpha,beta);
166 else {
167 if(bs1)
168 for(i=0;i<4;i++)
169 loop_filter_c1(d + i,stride,alpha,beta,tc);
170 if (bs2)
171 for(i=4;i<8;i++)
172 loop_filter_c1(d + i,stride,alpha,beta,tc);
173 }
174 }
175
176 /*****************************************************************************
177 *
178 * inverse transform
179 *
180 ****************************************************************************/
181
182 static void cavs_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride) {
183 int i;
184 DCTELEM (*src)[8] = (DCTELEM(*)[8])block;
185 uint8_t *cm = cropTbl + MAX_NEG_CROP;
186
187 src[0][0] += 8;
188
189 for( i = 0; i < 8; i++ ) {
190 const int a0 = 3*src[i][1] - (src[i][7]<<1);
191 const int a1 = 3*src[i][3] + (src[i][5]<<1);
192 const int a2 = (src[i][3]<<1) - 3*src[i][5];
193 const int a3 = (src[i][1]<<1) + 3*src[i][7];
194
195 const int b4 = ((a0 + a1 + a3)<<1) + a1;
196 const int b5 = ((a0 - a1 + a2)<<1) + a0;
197 const int b6 = ((a3 - a2 - a1)<<1) + a3;
198 const int b7 = ((a0 - a2 - a3)<<1) - a2;
199
200 const int a7 = (src[i][2]<<2) - 10*src[i][6];
201 const int a6 = (src[i][6]<<2) + 10*src[i][2];
202 const int a5 = ((src[i][0] - src[i][4]) << 3) + 4;
203 const int a4 = ((src[i][0] + src[i][4]) << 3) + 4;
204
205 const int b0 = a4 + a6;
206 const int b1 = a5 + a7;
207 const int b2 = a5 - a7;
208 const int b3 = a4 - a6;
209
210 src[i][0] = (b0 + b4) >> 3;
211 src[i][1] = (b1 + b5) >> 3;
212 src[i][2] = (b2 + b6) >> 3;
213 src[i][3] = (b3 + b7) >> 3;
214 src[i][4] = (b3 - b7) >> 3;
215 src[i][5] = (b2 - b6) >> 3;
216 src[i][6] = (b1 - b5) >> 3;
217 src[i][7] = (b0 - b4) >> 3;
218 }
219 for( i = 0; i < 8; i++ ) {
220 const int a0 = 3*src[1][i] - (src[7][i]<<1);
221 const int a1 = 3*src[3][i] + (src[5][i]<<1);
222 const int a2 = (src[3][i]<<1) - 3*src[5][i];
223 const int a3 = (src[1][i]<<1) + 3*src[7][i];
224
225 const int b4 = ((a0 + a1 + a3)<<1) + a1;
226 const int b5 = ((a0 - a1 + a2)<<1) + a0;
227 const int b6 = ((a3 - a2 - a1)<<1) + a3;
228 const int b7 = ((a0 - a2 - a3)<<1) - a2;
229
230 const int a7 = (src[2][i]<<2) - 10*src[6][i];
231 const int a6 = (src[6][i]<<2) + 10*src[2][i];
232 const int a5 = (src[0][i] - src[4][i]) << 3;
233 const int a4 = (src[0][i] + src[4][i]) << 3;
234
235 const int b0 = a4 + a6;
236 const int b1 = a5 + a7;
237 const int b2 = a5 - a7;
238 const int b3 = a4 - a6;
239
240 dst[i + 0*stride] = cm[ dst[i + 0*stride] + ((b0 + b4) >> 7)];
241 dst[i + 1*stride] = cm[ dst[i + 1*stride] + ((b1 + b5) >> 7)];
242 dst[i + 2*stride] = cm[ dst[i + 2*stride] + ((b2 + b6) >> 7)];
243 dst[i + 3*stride] = cm[ dst[i + 3*stride] + ((b3 + b7) >> 7)];
244 dst[i + 4*stride] = cm[ dst[i + 4*stride] + ((b3 - b7) >> 7)];
245 dst[i + 5*stride] = cm[ dst[i + 5*stride] + ((b2 - b6) >> 7)];
246 dst[i + 6*stride] = cm[ dst[i + 6*stride] + ((b1 - b5) >> 7)];
247 dst[i + 7*stride] = cm[ dst[i + 7*stride] + ((b0 - b4) >> 7)];
248 }
249 }
250
251 /*****************************************************************************
252 *
253 * motion compensation
254 *
255 ****************************************************************************/
256
257 #define CAVS_SUBPIX(OPNAME, OP, NAME, A, B, C, D, E, F) \
258 static void OPNAME ## cavs_filt8_h_ ## NAME(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
259 const int h=8;\
260 uint8_t *cm = cropTbl + MAX_NEG_CROP;\
261 int i;\
262 for(i=0; i<h; i++)\
263 {\
264 OP(dst[0], A*src[-2] + B*src[-1] + C*src[0] + D*src[1] + E*src[2] + F*src[3]);\
265 OP(dst[1], A*src[-1] + B*src[ 0] + C*src[1] + D*src[2] + E*src[3] + F*src[4]);\
266 OP(dst[2], A*src[ 0] + B*src[ 1] + C*src[2] + D*src[3] + E*src[4] + F*src[5]);\
267 OP(dst[3], A*src[ 1] + B*src[ 2] + C*src[3] + D*src[4] + E*src[5] + F*src[6]);\
268 OP(dst[4], A*src[ 2] + B*src[ 3] + C*src[4] + D*src[5] + E*src[6] + F*src[7]);\
269 OP(dst[5], A*src[ 3] + B*src[ 4] + C*src[5] + D*src[6] + E*src[7] + F*src[8]);\
270 OP(dst[6], A*src[ 4] + B*src[ 5] + C*src[6] + D*src[7] + E*src[8] + F*src[9]);\
271 OP(dst[7], A*src[ 5] + B*src[ 6] + C*src[7] + D*src[8] + E*src[9] + F*src[10]);\
272 dst+=dstStride;\
273 src+=srcStride;\
274 }\
275 }\
276 \
277 static void OPNAME ## cavs_filt8_v_ ## NAME(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
278 const int w=8;\
279 uint8_t *cm = cropTbl + MAX_NEG_CROP;\
280 int i;\
281 for(i=0; i<w; i++)\
282 {\
283 const int srcB= src[-2*srcStride];\
284 const int srcA= src[-1*srcStride];\
285 const int src0= src[0 *srcStride];\
286 const int src1= src[1 *srcStride];\
287 const int src2= src[2 *srcStride];\
288 const int src3= src[3 *srcStride];\
289 const int src4= src[4 *srcStride];\
290 const int src5= src[5 *srcStride];\
291 const int src6= src[6 *srcStride];\
292 const int src7= src[7 *srcStride];\
293 const int src8= src[8 *srcStride];\
294 const int src9= src[9 *srcStride];\
295 const int src10= src[10 *srcStride];\
296 OP(dst[0*dstStride], A*srcB + B*srcA + C*src0 + D*src1 + E*src2 + F*src3);\
297 OP(dst[1*dstStride], A*srcA + B*src0 + C*src1 + D*src2 + E*src3 + F*src4);\
298 OP(dst[2*dstStride], A*src0 + B*src1 + C*src2 + D*src3 + E*src4 + F*src5);\
299 OP(dst[3*dstStride], A*src1 + B*src2 + C*src3 + D*src4 + E*src5 + F*src6);\
300 OP(dst[4*dstStride], A*src2 + B*src3 + C*src4 + D*src5 + E*src6 + F*src7);\
301 OP(dst[5*dstStride], A*src3 + B*src4 + C*src5 + D*src6 + E*src7 + F*src8);\
302 OP(dst[6*dstStride], A*src4 + B*src5 + C*src6 + D*src7 + E*src8 + F*src9);\
303 OP(dst[7*dstStride], A*src5 + B*src6 + C*src7 + D*src8 + E*src9 + F*src10);\
304 dst++;\
305 src++;\
306 }\
307 }\
308 \
309 static void OPNAME ## cavs_filt16_v_ ## NAME(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
310 OPNAME ## cavs_filt8_v_ ## NAME(dst , src , dstStride, srcStride);\
311 OPNAME ## cavs_filt8_v_ ## NAME(dst+8, src+8, dstStride, srcStride);\
312 src += 8*srcStride;\
313 dst += 8*dstStride;\
314 OPNAME ## cavs_filt8_v_ ## NAME(dst , src , dstStride, srcStride);\
315 OPNAME ## cavs_filt8_v_ ## NAME(dst+8, src+8, dstStride, srcStride);\
316 }\
317 \
318 static void OPNAME ## cavs_filt16_h_ ## NAME(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
319 OPNAME ## cavs_filt8_h_ ## NAME(dst , src , dstStride, srcStride);\
320 OPNAME ## cavs_filt8_h_ ## NAME(dst+8, src+8, dstStride, srcStride);\
321 src += 8*srcStride;\
322 dst += 8*dstStride;\
323 OPNAME ## cavs_filt8_h_ ## NAME(dst , src , dstStride, srcStride);\
324 OPNAME ## cavs_filt8_h_ ## NAME(dst+8, src+8, dstStride, srcStride);\
325 }\
326
327 #define CAVS_SUBPIX_HV(OPNAME, OP, NAME, AH, BH, CH, DH, EH, FH, AV, BV, CV, DV, EV, FV, FULL) \
328 static void OPNAME ## cavs_filt8_hv_ ## NAME(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int srcStride){\
329 int16_t temp[8*(8+5)];\
330 int16_t *tmp = temp;\
331 const int h=8;\
332 const int w=8;\
333 uint8_t *cm = cropTbl + MAX_NEG_CROP;\
334 int i;\
335 src1 -= 2*srcStride;\
336 for(i=0; i<h+5; i++)\
337 {\
338 tmp[0]= AH*src1[-2] + BH*src1[-1] + CH*src1[0] + DH*src1[1] + EH*src1[2] + FH*src1[3];\
339 tmp[1]= AH*src1[-1] + BH*src1[ 0] + CH*src1[1] + DH*src1[2] + EH*src1[3] + FH*src1[4];\
340 tmp[2]= AH*src1[ 0] + BH*src1[ 1] + CH*src1[2] + DH*src1[3] + EH*src1[4] + FH*src1[5];\
341 tmp[3]= AH*src1[ 1] + BH*src1[ 2] + CH*src1[3] + DH*src1[4] + EH*src1[5] + FH*src1[6];\
342 tmp[4]= AH*src1[ 2] + BH*src1[ 3] + CH*src1[4] + DH*src1[5] + EH*src1[6] + FH*src1[7];\
343 tmp[5]= AH*src1[ 3] + BH*src1[ 4] + CH*src1[5] + DH*src1[6] + EH*src1[7] + FH*src1[8];\
344 tmp[6]= AH*src1[ 4] + BH*src1[ 5] + CH*src1[6] + DH*src1[7] + EH*src1[8] + FH*src1[9];\
345 tmp[7]= AH*src1[ 5] + BH*src1[ 6] + CH*src1[7] + DH*src1[8] + EH*src1[9] + FH*src1[10];\
346 tmp+=8;\
347 src1+=srcStride;\
348 }\
349 if(FULL) {\
350 tmp = temp+8*2; \
351 for(i=0; i<w; i++) \
352 { \
353 const int tmpB= tmp[-2*8]; \
354 const int tmpA= tmp[-1*8]; \
355 const int tmp0= tmp[0 *8]; \
356 const int tmp1= tmp[1 *8]; \
357 const int tmp2= tmp[2 *8]; \
358 const int tmp3= tmp[3 *8]; \
359 const int tmp4= tmp[4 *8]; \
360 const int tmp5= tmp[5 *8]; \
361 const int tmp6= tmp[6 *8]; \
362 const int tmp7= tmp[7 *8]; \
363 const int tmp8= tmp[8 *8]; \
364 const int tmp9= tmp[9 *8]; \
365 const int tmp10=tmp[10*8]; \
366 OP(dst[0*dstStride], AV*tmpB + BV*tmpA + CV*tmp0 + DV*tmp1 + EV*tmp2 + FV*tmp3 + 64*src2[0*srcStride]); \
367 OP(dst[1*dstStride], AV*tmpA + BV*tmp0 + CV*tmp1 + DV*tmp2 + EV*tmp3 + FV*tmp4 + 64*src2[1*srcStride]); \
368 OP(dst[2*dstStride], AV*tmp0 + BV*tmp1 + CV*tmp2 + DV*tmp3 + EV*tmp4 + FV*tmp5 + 64*src2[2*srcStride]); \
369 OP(dst[3*dstStride], AV*tmp1 + BV*tmp2 + CV*tmp3 + DV*tmp4 + EV*tmp5 + FV*tmp6 + 64*src2[3*srcStride]); \
370 OP(dst[4*dstStride], AV*tmp2 + BV*tmp3 + CV*tmp4 + DV*tmp5 + EV*tmp6 + FV*tmp7 + 64*src2[4*srcStride]); \
371 OP(dst[5*dstStride], AV*tmp3 + BV*tmp4 + CV*tmp5 + DV*tmp6 + EV*tmp7 + FV*tmp8 + 64*src2[5*srcStride]); \
372 OP(dst[6*dstStride], AV*tmp4 + BV*tmp5 + CV*tmp6 + DV*tmp7 + EV*tmp8 + FV*tmp9 + 64*src2[6*srcStride]); \
373 OP(dst[7*dstStride], AV*tmp5 + BV*tmp6 + CV*tmp7 + DV*tmp8 + EV*tmp9 + FV*tmp10 + 64*src2[7*srcStride]); \
374 dst++; \
375 tmp++; \
376 src2++; \
377 } \
378 } else {\
379 tmp = temp+8*2; \
380 for(i=0; i<w; i++) \
381 { \
382 const int tmpB= tmp[-2*8]; \
383 const int tmpA= tmp[-1*8]; \
384 const int tmp0= tmp[0 *8]; \
385 const int tmp1= tmp[1 *8]; \
386 const int tmp2= tmp[2 *8]; \
387 const int tmp3= tmp[3 *8]; \
388 const int tmp4= tmp[4 *8]; \
389 const int tmp5= tmp[5 *8]; \
390 const int tmp6= tmp[6 *8]; \
391 const int tmp7= tmp[7 *8]; \
392 const int tmp8= tmp[8 *8]; \
393 const int tmp9= tmp[9 *8]; \
394 const int tmp10=tmp[10*8]; \
395 OP(dst[0*dstStride], AV*tmpB + BV*tmpA + CV*tmp0 + DV*tmp1 + EV*tmp2 + FV*tmp3); \
396 OP(dst[1*dstStride], AV*tmpA + BV*tmp0 + CV*tmp1 + DV*tmp2 + EV*tmp3 + FV*tmp4); \
397 OP(dst[2*dstStride], AV*tmp0 + BV*tmp1 + CV*tmp2 + DV*tmp3 + EV*tmp4 + FV*tmp5); \
398 OP(dst[3*dstStride], AV*tmp1 + BV*tmp2 + CV*tmp3 + DV*tmp4 + EV*tmp5 + FV*tmp6); \
399 OP(dst[4*dstStride], AV*tmp2 + BV*tmp3 + CV*tmp4 + DV*tmp5 + EV*tmp6 + FV*tmp7); \
400 OP(dst[5*dstStride], AV*tmp3 + BV*tmp4 + CV*tmp5 + DV*tmp6 + EV*tmp7 + FV*tmp8); \
401 OP(dst[6*dstStride], AV*tmp4 + BV*tmp5 + CV*tmp6 + DV*tmp7 + EV*tmp8 + FV*tmp9); \
402 OP(dst[7*dstStride], AV*tmp5 + BV*tmp6 + CV*tmp7 + DV*tmp8 + EV*tmp9 + FV*tmp10); \
403 dst++; \
404 tmp++; \
405 } \
406 }\
407 }\
408 \
409 static void OPNAME ## cavs_filt16_hv_ ## NAME(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int srcStride){ \
410 OPNAME ## cavs_filt8_hv_ ## NAME(dst , src1, src2 , dstStride, srcStride); \
411 OPNAME ## cavs_filt8_hv_ ## NAME(dst+8, src1+8, src2+8, dstStride, srcStride); \
412 src1 += 8*srcStride;\
413 src2 += 8*srcStride;\
414 dst += 8*dstStride;\
415 OPNAME ## cavs_filt8_hv_ ## NAME(dst , src1, src2 , dstStride, srcStride); \
416 OPNAME ## cavs_filt8_hv_ ## NAME(dst+8, src1+8, src2+8, dstStride, srcStride); \
417 }\
418
419 #define CAVS_MC(OPNAME, SIZE) \
420 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
421 OPNAME ## cavs_filt ## SIZE ## _h_qpel_l(dst, src, stride, stride);\
422 }\
423 \
424 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\
425 OPNAME ## cavs_filt ## SIZE ## _h_hpel(dst, src, stride, stride);\
426 }\
427 \
428 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
429 OPNAME ## cavs_filt ## SIZE ## _h_qpel_r(dst, src, stride, stride);\
430 }\
431 \
432 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
433 OPNAME ## cavs_filt ## SIZE ## _v_qpel_l(dst, src, stride, stride);\
434 }\
435 \
436 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\
437 OPNAME ## cavs_filt ## SIZE ## _v_hpel(dst, src, stride, stride);\
438 }\
439 \
440 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
441 OPNAME ## cavs_filt ## SIZE ## _v_qpel_r(dst, src, stride, stride);\
442 }\
443 \
444 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
445 OPNAME ## cavs_filt ## SIZE ## _hv_jj(dst, src, NULL, stride, stride); \
446 }\
447 \
448 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
449 OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src, stride, stride); \
450 }\
451 \
452 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
453 OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+stride, stride, stride); \
454 }\
455 \
456 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
457 OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+1, stride, stride); \
458 }\
459 \
460 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\
461 OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+stride+1,stride, stride); \
462 }\
463 \
464 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
465 OPNAME ## cavs_filt ## SIZE ## _hv_ff(dst, src, src+stride+1,stride, stride); \
466 }\
467 \
468 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
469 OPNAME ## cavs_filt ## SIZE ## _hv_ii(dst, src, src+stride+1,stride, stride); \
470 }\
471 \
472 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
473 OPNAME ## cavs_filt ## SIZE ## _hv_kk(dst, src, src+stride+1,stride, stride); \
474 }\
475 \
476 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
477 OPNAME ## cavs_filt ## SIZE ## _hv_qq(dst, src, src+stride+1,stride, stride); \
478 }\
479
480 #define op_put1(a, b) a = cm[((b)+4)>>3]
481 #define op_put2(a, b) a = cm[((b)+64)>>7]
482 #define op_put3(a, b) a = cm[((b)+32)>>6]
483 #define op_put4(a, b) a = cm[((b)+512)>>10]
484 #define op_avg1(a, b) a = ((a)+cm[((b)+4)>>3] +1)>>1
485 #define op_avg2(a, b) a = ((a)+cm[((b)+64)>>7] +1)>>1
486 #define op_avg3(a, b) a = ((a)+cm[((b)+32)>>6] +1)>>1
487 #define op_avg4(a, b) a = ((a)+cm[((b)+512)>>10]+1)>>1
488 CAVS_SUBPIX(put_ , op_put1, hpel, 0, -1, 5, 5, -1, 0)
489 CAVS_SUBPIX(put_ , op_put2, qpel_l, -1, -2, 96, 42, -7, 0)
490 CAVS_SUBPIX(put_ , op_put2, qpel_r, 0, -7, 42, 96, -2, -1)
491 CAVS_SUBPIX_HV(put_, op_put3, jj, 0, -1, 5, 5, -1, 0, 0, -1, 5, 5, -1, 0, 0)
492 CAVS_SUBPIX_HV(put_, op_put4, ff, 0, -1, 5, 5, -1, 0, -1, -2, 96, 42, -7, 0, 0)
493 CAVS_SUBPIX_HV(put_, op_put4, ii, -1, -2, 96, 42, -7, 0, 0, -1, 5, 5, -1, 0, 0)
494 CAVS_SUBPIX_HV(put_, op_put4, kk, 0, -7, 42, 96, -2, -1, 0, -1, 5, 5, -1, 0, 0)
495 CAVS_SUBPIX_HV(put_, op_put4, qq, 0, -1, 5, 5, -1, 0, 0, -7, 42, 96, -2,-1, 0)
496 CAVS_SUBPIX_HV(put_, op_put2, egpr, 0, -1, 5, 5, -1, 0, 0, -1, 5, 5, -1, 0, 1)
497 CAVS_SUBPIX(avg_ , op_avg1, hpel, 0, -1, 5, 5, -1, 0)
498 CAVS_SUBPIX(avg_ , op_avg2, qpel_l, -1, -2, 96, 42, -7, 0)
499 CAVS_SUBPIX(avg_ , op_avg2, qpel_r, 0, -7, 42, 96, -2, -1)
500 CAVS_SUBPIX_HV(avg_, op_avg3, jj, 0, -1, 5, 5, -1, 0, 0, -1, 5, 5, -1, 0, 0)
501 CAVS_SUBPIX_HV(avg_, op_avg4, ff, 0, -1, 5, 5, -1, 0, -1, -2, 96, 42, -7, 0, 0)
502 CAVS_SUBPIX_HV(avg_, op_avg4, ii, -1, -2, 96, 42, -7, 0, 0, -1, 5, 5, -1, 0, 0)
503 CAVS_SUBPIX_HV(avg_, op_avg4, kk, 0, -7, 42, 96, -2, -1, 0, -1, 5, 5, -1, 0, 0)
504 CAVS_SUBPIX_HV(avg_, op_avg4, qq, 0, -1, 5, 5, -1, 0, 0, -7, 42, 96, -2,-1, 0)
505 CAVS_SUBPIX_HV(avg_, op_avg2, egpr, 0, -1, 5, 5, -1, 0, 0, -1, 5, 5, -1, 0, 1)
506 CAVS_MC(put_, 8)
507 CAVS_MC(put_, 16)
508 CAVS_MC(avg_, 8)
509 CAVS_MC(avg_, 16)
510
511 void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
512 void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
513 void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
514 void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
515
516 void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx) {
517 #define dspfunc(PFX, IDX, NUM) \
518 c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_c; \
519 c->PFX ## _pixels_tab[IDX][ 1] = ff_ ## PFX ## NUM ## _mc10_c; \
520 c->PFX ## _pixels_tab[IDX][ 2] = ff_ ## PFX ## NUM ## _mc20_c; \
521 c->PFX ## _pixels_tab[IDX][ 3] = ff_ ## PFX ## NUM ## _mc30_c; \
522 c->PFX ## _pixels_tab[IDX][ 4] = ff_ ## PFX ## NUM ## _mc01_c; \
523 c->PFX ## _pixels_tab[IDX][ 5] = ff_ ## PFX ## NUM ## _mc11_c; \
524 c->PFX ## _pixels_tab[IDX][ 6] = ff_ ## PFX ## NUM ## _mc21_c; \
525 c->PFX ## _pixels_tab[IDX][ 7] = ff_ ## PFX ## NUM ## _mc31_c; \
526 c->PFX ## _pixels_tab[IDX][ 8] = ff_ ## PFX ## NUM ## _mc02_c; \
527 c->PFX ## _pixels_tab[IDX][ 9] = ff_ ## PFX ## NUM ## _mc12_c; \
528 c->PFX ## _pixels_tab[IDX][10] = ff_ ## PFX ## NUM ## _mc22_c; \
529 c->PFX ## _pixels_tab[IDX][11] = ff_ ## PFX ## NUM ## _mc32_c; \
530 c->PFX ## _pixels_tab[IDX][12] = ff_ ## PFX ## NUM ## _mc03_c; \
531 c->PFX ## _pixels_tab[IDX][13] = ff_ ## PFX ## NUM ## _mc13_c; \
532 c->PFX ## _pixels_tab[IDX][14] = ff_ ## PFX ## NUM ## _mc23_c; \
533 c->PFX ## _pixels_tab[IDX][15] = ff_ ## PFX ## NUM ## _mc33_c
534 dspfunc(put_cavs_qpel, 0, 16);
535 dspfunc(put_cavs_qpel, 1, 8);
536 dspfunc(avg_cavs_qpel, 0, 16);
537 dspfunc(avg_cavs_qpel, 1, 8);
538 c->cavs_filter_lv = cavs_filter_lv_c;
539 c->cavs_filter_lh = cavs_filter_lh_c;
540 c->cavs_filter_cv = cavs_filter_cv_c;
541 c->cavs_filter_ch = cavs_filter_ch_c;
542 c->cavs_idct8_add = cavs_idct8_add_c;
543 }