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 file is part of FFmpeg.
   9  *
  10  * FFmpeg is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU Lesser General Public
  12  * License as published by the Free Software Foundation; either
  13  * version 2.1 of the License, or (at your option) any later version.
  14  *
  15  * FFmpeg is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * Lesser General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public
  21  * License along with FFmpeg; if not, write to the Free Software
  22  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  23  */
  24
  25 #include <stdio.h>
  26 #include "dsputil.h"
  27
  28 /*****************************************************************************
  29  *
  30  * in-loop deblocking filter
  31  *
  32  ****************************************************************************/
  33
  34 #define P2 p0_p[-3*stride]
  35 #define P1 p0_p[-2*stride]
  36 #define P0 p0_p[-1*stride]
  37 #define Q0 p0_p[ 0*stride]
  38 #define Q1 p0_p[ 1*stride]
  39 #define Q2 p0_p[ 2*stride]
  40
  41 static inline void loop_filter_l2(uint8_t *p0_p,int stride,int alpha, int beta) {
  42     int p0 = P0;
  43     int q0 = Q0;
  44
  45     if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
  46         int s = p0 + q0 + 2;
  47         alpha = (alpha>>2) + 2;
  48         if(abs(P2-p0) < beta && abs(p0-q0) < alpha) {
  49             P0 = (P1 + p0 + s) >> 2;
  50             P1 = (2*P1 + s) >> 2;
  51         } else
  52             P0 = (2*P1 + s) >> 2;
  53         if(abs(Q2-q0) < beta && abs(q0-p0) < alpha) {
  54             Q0 = (Q1 + q0 + s) >> 2;
  55             Q1 = (2*Q1 + s) >> 2;
  56         } else
  57             Q0 = (2*Q1 + s) >> 2;
  58     }
  59 }
  60
  61 static inline void loop_filter_l1(uint8_t *p0_p, int stride, int alpha, int beta, int tc) {
  62     int p0 = P0;
  63     int q0 = Q0;
  64
  65     if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
  66         int delta = av_clip(((q0-p0)*3+P1-Q1+4)>>3,-tc, tc);
  67         P0 = av_clip_uint8(p0+delta);
  68         Q0 = av_clip_uint8(q0-delta);
  69         if(abs(P2-p0)<beta) {
  70             delta = av_clip(((P0-P1)*3+P2-Q0+4)>>3, -tc, tc);
  71             P1 = av_clip_uint8(P1+delta);
  72         }
  73         if(abs(Q2-q0)<beta) {
  74             delta = av_clip(((Q1-Q0)*3+P0-Q2+4)>>3, -tc, tc);
  75             Q1 = av_clip_uint8(Q1-delta);
  76         }
  77     }
  78 }
  79
  80 static inline void loop_filter_c2(uint8_t *p0_p,int stride,int alpha, int beta) {
  81     int p0 = P0;
  82     int q0 = Q0;
  83
  84     if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
  85         int s = p0 + q0 + 2;
  86         alpha = (alpha>>2) + 2;
  87         if(abs(P2-p0) < beta && abs(p0-q0) < alpha) {
  88             P0 = (P1 + p0 + s) >> 2;
  89         } else
  90             P0 = (2*P1 + s) >> 2;
  91         if(abs(Q2-q0) < beta && abs(q0-p0) < alpha) {
  92             Q0 = (Q1 + q0 + s) >> 2;
  93         } else
  94             Q0 = (2*Q1 + s) >> 2;
  95     }
  96 }
  97
  98 static inline void loop_filter_c1(uint8_t *p0_p,int stride,int alpha, int beta,
  99                                   int tc) {
 100     if(abs(P0-Q0)<alpha && abs(P1-P0)<beta && abs(Q1-Q0)<beta) {
 101         int delta = av_clip(((Q0-P0)*3+P1-Q1+4)>>3, -tc, tc);
 102         P0 = av_clip_uint8(P0+delta);
 103         Q0 = av_clip_uint8(Q0-delta);
 104     }
 105 }
 106
 107 #undef P0
 108 #undef P1
 109 #undef P2
 110 #undef Q0
 111 #undef Q1
 112 #undef Q2
 113
 114 static void cavs_filter_lv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
 115                            int bs1, int bs2) {
 116     int i;
 117     if(bs1==2)
 118         for(i=0;i<16;i++)
 119             loop_filter_l2(d + i*stride,1,alpha,beta);
 120     else {
 121         if(bs1)
 122             for(i=0;i<8;i++)
 123                 loop_filter_l1(d + i*stride,1,alpha,beta,tc);
 124         if (bs2)
 125             for(i=8;i<16;i++)
 126                 loop_filter_l1(d + i*stride,1,alpha,beta,tc);
 127     }
 128 }
 129
 130 static void cavs_filter_lh_c(uint8_t *d, int stride, int alpha, int beta, int tc,
 131                            int bs1, int bs2) {
 132     int i;
 133     if(bs1==2)
 134         for(i=0;i<16;i++)
 135             loop_filter_l2(d + i,stride,alpha,beta);
 136     else {
 137         if(bs1)
 138             for(i=0;i<8;i++)
 139                 loop_filter_l1(d + i,stride,alpha,beta,tc);
 140         if (bs2)
 141             for(i=8;i<16;i++)
 142                 loop_filter_l1(d + i,stride,alpha,beta,tc);
 143     }
 144 }
 145
 146 static void cavs_filter_cv_c(uint8_t *d, int stride, int alpha, int beta, int tc,
 147                            int bs1, int bs2) {
 148     int i;
 149     if(bs1==2)
 150         for(i=0;i<8;i++)
 151             loop_filter_c2(d + i*stride,1,alpha,beta);
 152     else {
 153         if(bs1)
 154             for(i=0;i<4;i++)
 155                 loop_filter_c1(d + i*stride,1,alpha,beta,tc);
 156         if (bs2)
 157             for(i=4;i<8;i++)
 158                 loop_filter_c1(d + i*stride,1,alpha,beta,tc);
 159     }
 160 }
 161
 162 static void cavs_filter_ch_c(uint8_t *d, int stride, int alpha, int beta, int tc,
 163                            int bs1, int bs2) {
 164     int i;
 165     if(bs1==2)
 166         for(i=0;i<8;i++)
 167             loop_filter_c2(d + i,stride,alpha,beta);
 168     else {
 169         if(bs1)
 170             for(i=0;i<4;i++)
 171                 loop_filter_c1(d + i,stride,alpha,beta,tc);
 172         if (bs2)
 173             for(i=4;i<8;i++)
 174                 loop_filter_c1(d + i,stride,alpha,beta,tc);
 175     }
 176 }
 177
 178 /*****************************************************************************
 179  *
 180  * inverse transform
 181  *
 182  ****************************************************************************/
 183
 184 static void cavs_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride) {
 185     int i;
 186     DCTELEM (*src)[8] = (DCTELEM(*)[8])block;
 187     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 188
 189     src[0][0] += 8;
 190
 191     for( i = 0; i < 8; i++ ) {
 192         const int a0 =  3*src[i][1] - (src[i][7]<<1);
 193         const int a1 =  3*src[i][3] + (src[i][5]<<1);
 194         const int a2 =  (src[i][3]<<1) - 3*src[i][5];
 195         const int a3 =  (src[i][1]<<1) + 3*src[i][7];
 196
 197         const int b4 = ((a0 + a1 + a3)<<1) + a1;
 198         const int b5 = ((a0 - a1 + a2)<<1) + a0;
 199         const int b6 = ((a3 - a2 - a1)<<1) + a3;
 200         const int b7 = ((a0 - a2 - a3)<<1) - a2;
 201
 202         const int a7 = (src[i][2]<<2) - 10*src[i][6];
 203         const int a6 = (src[i][6]<<2) + 10*src[i][2];
 204         const int a5 = ((src[i][0] - src[i][4]) << 3) + 4;
 205         const int a4 = ((src[i][0] + src[i][4]) << 3) + 4;
 206
 207         const int b0 = a4 + a6;
 208         const int b1 = a5 + a7;
 209         const int b2 = a5 - a7;
 210         const int b3 = a4 - a6;
 211
 212         src[i][0] = (b0 + b4) >> 3;
 213         src[i][1] = (b1 + b5) >> 3;
 214         src[i][2] = (b2 + b6) >> 3;
 215         src[i][3] = (b3 + b7) >> 3;
 216         src[i][4] = (b3 - b7) >> 3;
 217         src[i][5] = (b2 - b6) >> 3;
 218         src[i][6] = (b1 - b5) >> 3;
 219         src[i][7] = (b0 - b4) >> 3;
 220     }
 221     for( i = 0; i < 8; i++ ) {
 222         const int a0 =  3*src[1][i] - (src[7][i]<<1);
 223         const int a1 =  3*src[3][i] + (src[5][i]<<1);
 224         const int a2 =  (src[3][i]<<1) - 3*src[5][i];
 225         const int a3 =  (src[1][i]<<1) + 3*src[7][i];
 226
 227         const int b4 = ((a0 + a1 + a3)<<1) + a1;
 228         const int b5 = ((a0 - a1 + a2)<<1) + a0;
 229         const int b6 = ((a3 - a2 - a1)<<1) + a3;
 230         const int b7 = ((a0 - a2 - a3)<<1) - a2;
 231
 232         const int a7 = (src[2][i]<<2) - 10*src[6][i];
 233         const int a6 = (src[6][i]<<2) + 10*src[2][i];
 234         const int a5 = (src[0][i] - src[4][i]) << 3;
 235         const int a4 = (src[0][i] + src[4][i]) << 3;
 236
 237         const int b0 = a4 + a6;
 238         const int b1 = a5 + a7;
 239         const int b2 = a5 - a7;
 240         const int b3 = a4 - a6;
 241
 242         dst[i + 0*stride] = cm[ dst[i + 0*stride] + ((b0 + b4) >> 7)];
 243         dst[i + 1*stride] = cm[ dst[i + 1*stride] + ((b1 + b5) >> 7)];
 244         dst[i + 2*stride] = cm[ dst[i + 2*stride] + ((b2 + b6) >> 7)];
 245         dst[i + 3*stride] = cm[ dst[i + 3*stride] + ((b3 + b7) >> 7)];
 246         dst[i + 4*stride] = cm[ dst[i + 4*stride] + ((b3 - b7) >> 7)];
 247         dst[i + 5*stride] = cm[ dst[i + 5*stride] + ((b2 - b6) >> 7)];
 248         dst[i + 6*stride] = cm[ dst[i + 6*stride] + ((b1 - b5) >> 7)];
 249         dst[i + 7*stride] = cm[ dst[i + 7*stride] + ((b0 - b4) >> 7)];
 250     }
 251     memset(block,0,64*sizeof(DCTELEM));
 252 }
 253
 254 /*****************************************************************************
 255  *
 256  * motion compensation
 257  *
 258  ****************************************************************************/
 259
 260 #define CAVS_SUBPIX(OPNAME, OP, NAME, A, B, C, D, E, F) \
 261 static void OPNAME ## cavs_filt8_h_ ## NAME(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
 262     const int h=8;\
 263     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
 264     int i;\
 265     for(i=0; i<h; i++)\
 266     {\
 267         OP(dst[0], A*src[-2] + B*src[-1] + C*src[0] + D*src[1] + E*src[2] + F*src[3]);\
 268         OP(dst[1], A*src[-1] + B*src[ 0] + C*src[1] + D*src[2] + E*src[3] + F*src[4]);\
 269         OP(dst[2], A*src[ 0] + B*src[ 1] + C*src[2] + D*src[3] + E*src[4] + F*src[5]);\
 270         OP(dst[3], A*src[ 1] + B*src[ 2] + C*src[3] + D*src[4] + E*src[5] + F*src[6]);\
 271         OP(dst[4], A*src[ 2] + B*src[ 3] + C*src[4] + D*src[5] + E*src[6] + F*src[7]);\
 272         OP(dst[5], A*src[ 3] + B*src[ 4] + C*src[5] + D*src[6] + E*src[7] + F*src[8]);\
 273         OP(dst[6], A*src[ 4] + B*src[ 5] + C*src[6] + D*src[7] + E*src[8] + F*src[9]);\
 274         OP(dst[7], A*src[ 5] + B*src[ 6] + C*src[7] + D*src[8] + E*src[9] + F*src[10]);\
 275         dst+=dstStride;\
 276         src+=srcStride;\
 277     }\
 278 }\
 279 \
 280 static void OPNAME ## cavs_filt8_v_  ## NAME(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
 281     const int w=8;\
 282     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
 283     int i;\
 284     for(i=0; i<w; i++)\
 285     {\
 286         const int srcB= src[-2*srcStride];\
 287         const int srcA= src[-1*srcStride];\
 288         const int src0= src[0 *srcStride];\
 289         const int src1= src[1 *srcStride];\
 290         const int src2= src[2 *srcStride];\
 291         const int src3= src[3 *srcStride];\
 292         const int src4= src[4 *srcStride];\
 293         const int src5= src[5 *srcStride];\
 294         const int src6= src[6 *srcStride];\
 295         const int src7= src[7 *srcStride];\
 296         const int src8= src[8 *srcStride];\
 297         const int src9= src[9 *srcStride];\
 298         const int src10= src[10 *srcStride];\
 299         OP(dst[0*dstStride], A*srcB + B*srcA + C*src0 + D*src1 + E*src2 + F*src3);\
 300         OP(dst[1*dstStride], A*srcA + B*src0 + C*src1 + D*src2 + E*src3 + F*src4);\
 301         OP(dst[2*dstStride], A*src0 + B*src1 + C*src2 + D*src3 + E*src4 + F*src5);\
 302         OP(dst[3*dstStride], A*src1 + B*src2 + C*src3 + D*src4 + E*src5 + F*src6);\
 303         OP(dst[4*dstStride], A*src2 + B*src3 + C*src4 + D*src5 + E*src6 + F*src7);\
 304         OP(dst[5*dstStride], A*src3 + B*src4 + C*src5 + D*src6 + E*src7 + F*src8);\
 305         OP(dst[6*dstStride], A*src4 + B*src5 + C*src6 + D*src7 + E*src8 + F*src9);\
 306         OP(dst[7*dstStride], A*src5 + B*src6 + C*src7 + D*src8 + E*src9 + F*src10);\
 307         dst++;\
 308         src++;\
 309     }\
 310 }\
 311 \
 312 static void OPNAME ## cavs_filt16_v_ ## NAME(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
 313     OPNAME ## cavs_filt8_v_ ## NAME(dst  , src  , dstStride, srcStride);\
 314     OPNAME ## cavs_filt8_v_ ## NAME(dst+8, src+8, dstStride, srcStride);\
 315     src += 8*srcStride;\
 316     dst += 8*dstStride;\
 317     OPNAME ## cavs_filt8_v_ ## NAME(dst  , src  , dstStride, srcStride);\
 318     OPNAME ## cavs_filt8_v_ ## NAME(dst+8, src+8, dstStride, srcStride);\
 319 }\
 320 \
 321 static void OPNAME ## cavs_filt16_h_ ## NAME(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
 322     OPNAME ## cavs_filt8_h_ ## NAME(dst  , src  , dstStride, srcStride);\
 323     OPNAME ## cavs_filt8_h_ ## NAME(dst+8, src+8, dstStride, srcStride);\
 324     src += 8*srcStride;\
 325     dst += 8*dstStride;\
 326     OPNAME ## cavs_filt8_h_ ## NAME(dst  , src  , dstStride, srcStride);\
 327     OPNAME ## cavs_filt8_h_ ## NAME(dst+8, src+8, dstStride, srcStride);\
 328 }\
 329
 330 #define CAVS_SUBPIX_HV(OPNAME, OP, NAME, AH, BH, CH, DH, EH, FH, AV, BV, CV, DV, EV, FV, FULL) \
 331 static void OPNAME ## cavs_filt8_hv_ ## NAME(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int srcStride){\
 332     int16_t temp[8*(8+5)];\
 333     int16_t *tmp = temp;\
 334     const int h=8;\
 335     const int w=8;\
 336     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
 337     int i;\
 338     src1 -= 2*srcStride;\
 339     for(i=0; i<h+5; i++)\
 340     {\
 341         tmp[0]= AH*src1[-2] + BH*src1[-1] + CH*src1[0] + DH*src1[1] + EH*src1[2] + FH*src1[3];\
 342         tmp[1]= AH*src1[-1] + BH*src1[ 0] + CH*src1[1] + DH*src1[2] + EH*src1[3] + FH*src1[4];\
 343         tmp[2]= AH*src1[ 0] + BH*src1[ 1] + CH*src1[2] + DH*src1[3] + EH*src1[4] + FH*src1[5];\
 344         tmp[3]= AH*src1[ 1] + BH*src1[ 2] + CH*src1[3] + DH*src1[4] + EH*src1[5] + FH*src1[6];\
 345         tmp[4]= AH*src1[ 2] + BH*src1[ 3] + CH*src1[4] + DH*src1[5] + EH*src1[6] + FH*src1[7];\
 346         tmp[5]= AH*src1[ 3] + BH*src1[ 4] + CH*src1[5] + DH*src1[6] + EH*src1[7] + FH*src1[8];\
 347         tmp[6]= AH*src1[ 4] + BH*src1[ 5] + CH*src1[6] + DH*src1[7] + EH*src1[8] + FH*src1[9];\
 348         tmp[7]= AH*src1[ 5] + BH*src1[ 6] + CH*src1[7] + DH*src1[8] + EH*src1[9] + FH*src1[10];\
 349         tmp+=8;\
 350         src1+=srcStride;\
 351     }\
 352     if(FULL) {\
 353       tmp = temp+8*2;                           \
 354       for(i=0; i<w; i++)                        \
 355         {                                       \
 356           const int tmpB= tmp[-2*8];    \
 357           const int tmpA= tmp[-1*8];    \
 358           const int tmp0= tmp[0 *8];    \
 359           const int tmp1= tmp[1 *8];    \
 360           const int tmp2= tmp[2 *8];    \
 361           const int tmp3= tmp[3 *8];    \
 362           const int tmp4= tmp[4 *8];    \
 363           const int tmp5= tmp[5 *8];    \
 364           const int tmp6= tmp[6 *8];    \
 365           const int tmp7= tmp[7 *8];    \
 366           const int tmp8= tmp[8 *8];    \
 367           const int tmp9= tmp[9 *8];    \
 368           const int tmp10=tmp[10*8];                            \
 369           OP(dst[0*dstStride], AV*tmpB + BV*tmpA + CV*tmp0 + DV*tmp1 + EV*tmp2 + FV*tmp3 + 64*src2[0*srcStride]); \
 370           OP(dst[1*dstStride], AV*tmpA + BV*tmp0 + CV*tmp1 + DV*tmp2 + EV*tmp3 + FV*tmp4 + 64*src2[1*srcStride]); \
 371           OP(dst[2*dstStride], AV*tmp0 + BV*tmp1 + CV*tmp2 + DV*tmp3 + EV*tmp4 + FV*tmp5 + 64*src2[2*srcStride]); \
 372           OP(dst[3*dstStride], AV*tmp1 + BV*tmp2 + CV*tmp3 + DV*tmp4 + EV*tmp5 + FV*tmp6 + 64*src2[3*srcStride]); \
 373           OP(dst[4*dstStride], AV*tmp2 + BV*tmp3 + CV*tmp4 + DV*tmp5 + EV*tmp6 + FV*tmp7 + 64*src2[4*srcStride]); \
 374           OP(dst[5*dstStride], AV*tmp3 + BV*tmp4 + CV*tmp5 + DV*tmp6 + EV*tmp7 + FV*tmp8 + 64*src2[5*srcStride]); \
 375           OP(dst[6*dstStride], AV*tmp4 + BV*tmp5 + CV*tmp6 + DV*tmp7 + EV*tmp8 + FV*tmp9 + 64*src2[6*srcStride]); \
 376           OP(dst[7*dstStride], AV*tmp5 + BV*tmp6 + CV*tmp7 + DV*tmp8 + EV*tmp9 + FV*tmp10 + 64*src2[7*srcStride]); \
 377           dst++;                                                        \
 378           tmp++;                                                        \
 379           src2++;                                                       \
 380         }                                                               \
 381     } else {\
 382       tmp = temp+8*2;                           \
 383       for(i=0; i<w; i++)                        \
 384         {                                       \
 385           const int tmpB= tmp[-2*8];    \
 386           const int tmpA= tmp[-1*8];    \
 387           const int tmp0= tmp[0 *8];    \
 388           const int tmp1= tmp[1 *8];    \
 389           const int tmp2= tmp[2 *8];    \
 390           const int tmp3= tmp[3 *8];    \
 391           const int tmp4= tmp[4 *8];    \
 392           const int tmp5= tmp[5 *8];    \
 393           const int tmp6= tmp[6 *8];    \
 394           const int tmp7= tmp[7 *8];    \
 395           const int tmp8= tmp[8 *8];    \
 396           const int tmp9= tmp[9 *8];    \
 397           const int tmp10=tmp[10*8];                            \
 398           OP(dst[0*dstStride], AV*tmpB + BV*tmpA + CV*tmp0 + DV*tmp1 + EV*tmp2 + FV*tmp3); \
 399           OP(dst[1*dstStride], AV*tmpA + BV*tmp0 + CV*tmp1 + DV*tmp2 + EV*tmp3 + FV*tmp4); \
 400           OP(dst[2*dstStride], AV*tmp0 + BV*tmp1 + CV*tmp2 + DV*tmp3 + EV*tmp4 + FV*tmp5); \
 401           OP(dst[3*dstStride], AV*tmp1 + BV*tmp2 + CV*tmp3 + DV*tmp4 + EV*tmp5 + FV*tmp6); \
 402           OP(dst[4*dstStride], AV*tmp2 + BV*tmp3 + CV*tmp4 + DV*tmp5 + EV*tmp6 + FV*tmp7); \
 403           OP(dst[5*dstStride], AV*tmp3 + BV*tmp4 + CV*tmp5 + DV*tmp6 + EV*tmp7 + FV*tmp8); \
 404           OP(dst[6*dstStride], AV*tmp4 + BV*tmp5 + CV*tmp6 + DV*tmp7 + EV*tmp8 + FV*tmp9); \
 405           OP(dst[7*dstStride], AV*tmp5 + BV*tmp6 + CV*tmp7 + DV*tmp8 + EV*tmp9 + FV*tmp10); \
 406           dst++;                                                        \
 407           tmp++;                                                        \
 408         }                                                               \
 409     }\
 410 }\
 411 \
 412 static void OPNAME ## cavs_filt16_hv_ ## NAME(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int srcStride){ \
 413     OPNAME ## cavs_filt8_hv_ ## NAME(dst  , src1,   src2  , dstStride, srcStride); \
 414     OPNAME ## cavs_filt8_hv_ ## NAME(dst+8, src1+8, src2+8, dstStride, srcStride); \
 415     src1 += 8*srcStride;\
 416     src2 += 8*srcStride;\
 417     dst += 8*dstStride;\
 418     OPNAME ## cavs_filt8_hv_ ## NAME(dst  , src1,   src2  , dstStride, srcStride); \
 419     OPNAME ## cavs_filt8_hv_ ## NAME(dst+8, src1+8, src2+8, dstStride, srcStride); \
 420 }\
 421
 422 #define CAVS_MC(OPNAME, SIZE) \
 423 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
 424     OPNAME ## cavs_filt ## SIZE ## _h_qpel_l(dst, src, stride, stride);\
 425 }\
 426 \
 427 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\
 428     OPNAME ## cavs_filt ## SIZE ## _h_hpel(dst, src, stride, stride);\
 429 }\
 430 \
 431 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
 432     OPNAME ## cavs_filt ## SIZE ## _h_qpel_r(dst, src, stride, stride);\
 433 }\
 434 \
 435 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
 436     OPNAME ## cavs_filt ## SIZE ## _v_qpel_l(dst, src, stride, stride);\
 437 }\
 438 \
 439 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\
 440     OPNAME ## cavs_filt ## SIZE ## _v_hpel(dst, src, stride, stride);\
 441 }\
 442 \
 443 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
 444     OPNAME ## cavs_filt ## SIZE ## _v_qpel_r(dst, src, stride, stride);\
 445 }\
 446 \
 447 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
 448   OPNAME ## cavs_filt ## SIZE ## _hv_jj(dst, src, NULL, stride, stride); \
 449 }\
 450 \
 451 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
 452   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src, stride, stride); \
 453 }\
 454 \
 455 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
 456   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+stride, stride, stride); \
 457 }\
 458 \
 459 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
 460   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+1, stride, stride); \
 461 }\
 462 \
 463 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\
 464   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+stride+1,stride, stride); \
 465 }\
 466 \
 467 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
 468   OPNAME ## cavs_filt ## SIZE ## _hv_ff(dst, src, src+stride+1,stride, stride); \
 469 }\
 470 \
 471 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
 472   OPNAME ## cavs_filt ## SIZE ## _hv_ii(dst, src, src+stride+1,stride, stride); \
 473 }\
 474 \
 475 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
 476   OPNAME ## cavs_filt ## SIZE ## _hv_kk(dst, src, src+stride+1,stride, stride); \
 477 }\
 478 \
 479 static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
 480   OPNAME ## cavs_filt ## SIZE ## _hv_qq(dst, src, src+stride+1,stride, stride); \
 481 }\
 482
 483 #define op_put1(a, b)  a = cm[((b)+4)>>3]
 484 #define op_put2(a, b)  a = cm[((b)+64)>>7]
 485 #define op_put3(a, b)  a = cm[((b)+32)>>6]
 486 #define op_put4(a, b)  a = cm[((b)+512)>>10]
 487 #define op_avg1(a, b)  a = ((a)+cm[((b)+4)>>3]   +1)>>1
 488 #define op_avg2(a, b)  a = ((a)+cm[((b)+64)>>7]  +1)>>1
 489 #define op_avg3(a, b)  a = ((a)+cm[((b)+32)>>6]  +1)>>1
 490 #define op_avg4(a, b)  a = ((a)+cm[((b)+512)>>10]+1)>>1
 491 CAVS_SUBPIX(put_   , op_put1, hpel,    0, -1,  5,  5, -1,  0)
 492 CAVS_SUBPIX(put_   , op_put2, qpel_l, -1, -2, 96, 42, -7,  0)
 493 CAVS_SUBPIX(put_   , op_put2, qpel_r,  0, -7, 42, 96, -2, -1)
 494 CAVS_SUBPIX_HV(put_, op_put3, jj,      0, -1,  5,  5, -1,  0,  0, -1,  5,  5, -1, 0, 0)
 495 CAVS_SUBPIX_HV(put_, op_put4, ff,      0, -1,  5,  5, -1,  0, -1, -2, 96, 42, -7, 0, 0)
 496 CAVS_SUBPIX_HV(put_, op_put4, ii,     -1, -2, 96, 42, -7,  0,  0, -1,  5,  5, -1, 0, 0)
 497 CAVS_SUBPIX_HV(put_, op_put4, kk,      0, -7, 42, 96, -2, -1,  0, -1,  5,  5, -1, 0, 0)
 498 CAVS_SUBPIX_HV(put_, op_put4, qq,      0, -1,  5,  5, -1,  0,  0, -7, 42, 96, -2,-1, 0)
 499 CAVS_SUBPIX_HV(put_, op_put2, egpr,    0, -1,  5,  5, -1,  0,  0, -1,  5,  5, -1, 0, 1)
 500 CAVS_SUBPIX(avg_   , op_avg1, hpel,    0, -1,  5,  5, -1,  0)
 501 CAVS_SUBPIX(avg_   , op_avg2, qpel_l, -1, -2, 96, 42, -7,  0)
 502 CAVS_SUBPIX(avg_   , op_avg2, qpel_r,  0, -7, 42, 96, -2, -1)
 503 CAVS_SUBPIX_HV(avg_, op_avg3, jj,      0, -1,  5,  5, -1,  0,  0, -1,  5,  5, -1, 0, 0)
 504 CAVS_SUBPIX_HV(avg_, op_avg4, ff,      0, -1,  5,  5, -1,  0, -1, -2, 96, 42, -7, 0, 0)
 505 CAVS_SUBPIX_HV(avg_, op_avg4, ii,     -1, -2, 96, 42, -7,  0,  0, -1,  5,  5, -1, 0, 0)
 506 CAVS_SUBPIX_HV(avg_, op_avg4, kk,      0, -7, 42, 96, -2, -1,  0, -1,  5,  5, -1, 0, 0)
 507 CAVS_SUBPIX_HV(avg_, op_avg4, qq,      0, -1,  5,  5, -1,  0,  0, -7, 42, 96, -2,-1, 0)
 508 CAVS_SUBPIX_HV(avg_, op_avg2, egpr,    0, -1,  5,  5, -1,  0,  0, -1,  5,  5, -1, 0, 1)
 509 CAVS_MC(put_, 8)
 510 CAVS_MC(put_, 16)
 511 CAVS_MC(avg_, 8)
 512 CAVS_MC(avg_, 16)
 513
 514 void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
 515 void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
 516 void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
 517 void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
 518
 519 void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx) {
 520 #define dspfunc(PFX, IDX, NUM) \
 521     c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_c; \
 522     c->PFX ## _pixels_tab[IDX][ 1] = ff_ ## PFX ## NUM ## _mc10_c; \
 523     c->PFX ## _pixels_tab[IDX][ 2] = ff_ ## PFX ## NUM ## _mc20_c; \
 524     c->PFX ## _pixels_tab[IDX][ 3] = ff_ ## PFX ## NUM ## _mc30_c; \
 525     c->PFX ## _pixels_tab[IDX][ 4] = ff_ ## PFX ## NUM ## _mc01_c; \
 526     c->PFX ## _pixels_tab[IDX][ 5] = ff_ ## PFX ## NUM ## _mc11_c; \
 527     c->PFX ## _pixels_tab[IDX][ 6] = ff_ ## PFX ## NUM ## _mc21_c; \
 528     c->PFX ## _pixels_tab[IDX][ 7] = ff_ ## PFX ## NUM ## _mc31_c; \
 529     c->PFX ## _pixels_tab[IDX][ 8] = ff_ ## PFX ## NUM ## _mc02_c; \
 530     c->PFX ## _pixels_tab[IDX][ 9] = ff_ ## PFX ## NUM ## _mc12_c; \
 531     c->PFX ## _pixels_tab[IDX][10] = ff_ ## PFX ## NUM ## _mc22_c; \
 532     c->PFX ## _pixels_tab[IDX][11] = ff_ ## PFX ## NUM ## _mc32_c; \
 533     c->PFX ## _pixels_tab[IDX][12] = ff_ ## PFX ## NUM ## _mc03_c; \
 534     c->PFX ## _pixels_tab[IDX][13] = ff_ ## PFX ## NUM ## _mc13_c; \
 535     c->PFX ## _pixels_tab[IDX][14] = ff_ ## PFX ## NUM ## _mc23_c; \
 536     c->PFX ## _pixels_tab[IDX][15] = ff_ ## PFX ## NUM ## _mc33_c
 537     dspfunc(put_cavs_qpel, 0, 16);
 538     dspfunc(put_cavs_qpel, 1, 8);
 539     dspfunc(avg_cavs_qpel, 0, 16);
 540     dspfunc(avg_cavs_qpel, 1, 8);
 541     c->cavs_filter_lv = cavs_filter_lv_c;
 542     c->cavs_filter_lh = cavs_filter_lh_c;
 543     c->cavs_filter_cv = cavs_filter_cv_c;
 544     c->cavs_filter_ch = cavs_filter_ch_c;
 545     c->cavs_idct8_add = cavs_idct8_add_c;
 546 }