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 "idctdsp.h"
  26 #include "mathops.h"
  27 #include "cavsdsp.h"
  28 #include "libavutil/common.h"
  29
  30 /*****************************************************************************
  31  *
  32  * in-loop deblocking filter
  33  *
  34  ****************************************************************************/
  35
  36 #define P2 p0_p[-3*stride]
  37 #define P1 p0_p[-2*stride]
  38 #define P0 p0_p[-1*stride]
  39 #define Q0 p0_p[ 0*stride]
  40 #define Q1 p0_p[ 1*stride]
  41 #define Q2 p0_p[ 2*stride]
  42
  43 static inline void loop_filter_l2(uint8_t *p0_p, ptrdiff_t stride, int alpha, int beta)
  44 {
  45     int p0 = P0;
  46     int q0 = Q0;
  47
  48     if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
  49         int s = p0 + q0 + 2;
  50         alpha = (alpha>>2) + 2;
  51         if(abs(P2-p0) < beta && abs(p0-q0) < alpha) {
  52             P0 = (P1 + p0 + s) >> 2;
  53             P1 = (2*P1 + s) >> 2;
  54         } else
  55             P0 = (2*P1 + s) >> 2;
  56         if(abs(Q2-q0) < beta && abs(q0-p0) < alpha) {
  57             Q0 = (Q1 + q0 + s) >> 2;
  58             Q1 = (2*Q1 + s) >> 2;
  59         } else
  60             Q0 = (2*Q1 + s) >> 2;
  61     }
  62 }
  63
  64 static inline void loop_filter_l1(uint8_t *p0_p, ptrdiff_t stride, int alpha, int beta, int tc)
  65 {
  66     int p0 = P0;
  67     int q0 = Q0;
  68
  69     if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
  70         int delta = av_clip(((q0-p0)*3+P1-Q1+4)>>3,-tc, tc);
  71         P0 = av_clip_uint8(p0+delta);
  72         Q0 = av_clip_uint8(q0-delta);
  73         if(abs(P2-p0)<beta) {
  74             delta = av_clip(((P0-P1)*3+P2-Q0+4)>>3, -tc, tc);
  75             P1 = av_clip_uint8(P1+delta);
  76         }
  77         if(abs(Q2-q0)<beta) {
  78             delta = av_clip(((Q1-Q0)*3+P0-Q2+4)>>3, -tc, tc);
  79             Q1 = av_clip_uint8(Q1-delta);
  80         }
  81     }
  82 }
  83
  84 static inline void loop_filter_c2(uint8_t *p0_p, ptrdiff_t stride, int alpha, int beta)
  85 {
  86     int p0 = P0;
  87     int q0 = Q0;
  88
  89     if(abs(p0-q0)<alpha && abs(P1-p0)<beta && abs(Q1-q0)<beta) {
  90         int s = p0 + q0 + 2;
  91         alpha = (alpha>>2) + 2;
  92         if(abs(P2-p0) < beta && abs(p0-q0) < alpha) {
  93             P0 = (P1 + p0 + s) >> 2;
  94         } else
  95             P0 = (2*P1 + s) >> 2;
  96         if(abs(Q2-q0) < beta && abs(q0-p0) < alpha) {
  97             Q0 = (Q1 + q0 + s) >> 2;
  98         } else
  99             Q0 = (2*Q1 + s) >> 2;
 100     }
 101 }
 102
 103 static inline void loop_filter_c1(uint8_t *p0_p, ptrdiff_t stride, int alpha, int beta,
 104                                   int tc)
 105 {
 106     if(abs(P0-Q0)<alpha && abs(P1-P0)<beta && abs(Q1-Q0)<beta) {
 107         int delta = av_clip(((Q0-P0)*3+P1-Q1+4)>>3, -tc, tc);
 108         P0 = av_clip_uint8(P0+delta);
 109         Q0 = av_clip_uint8(Q0-delta);
 110     }
 111 }
 112
 113 #undef P0
 114 #undef P1
 115 #undef P2
 116 #undef Q0
 117 #undef Q1
 118 #undef Q2
 119
 120 static void cavs_filter_lv_c(uint8_t *d, ptrdiff_t stride, int alpha, int beta, int tc,
 121                              int bs1, int bs2)
 122 {
 123     int i;
 124     if(bs1==2)
 125         for(i=0;i<16;i++)
 126             loop_filter_l2(d + i*stride,1,alpha,beta);
 127     else {
 128         if(bs1)
 129             for(i=0;i<8;i++)
 130                 loop_filter_l1(d + i*stride,1,alpha,beta,tc);
 131         if (bs2)
 132             for(i=8;i<16;i++)
 133                 loop_filter_l1(d + i*stride,1,alpha,beta,tc);
 134     }
 135 }
 136
 137 static void cavs_filter_lh_c(uint8_t *d, ptrdiff_t stride, int alpha, int beta, int tc,
 138                              int bs1, int bs2)
 139 {
 140     int i;
 141     if(bs1==2)
 142         for(i=0;i<16;i++)
 143             loop_filter_l2(d + i,stride,alpha,beta);
 144     else {
 145         if(bs1)
 146             for(i=0;i<8;i++)
 147                 loop_filter_l1(d + i,stride,alpha,beta,tc);
 148         if (bs2)
 149             for(i=8;i<16;i++)
 150                 loop_filter_l1(d + i,stride,alpha,beta,tc);
 151     }
 152 }
 153
 154 static void cavs_filter_cv_c(uint8_t *d, ptrdiff_t stride, int alpha, int beta, int tc,
 155                              int bs1, int bs2)
 156 {
 157     int i;
 158     if(bs1==2)
 159         for(i=0;i<8;i++)
 160             loop_filter_c2(d + i*stride,1,alpha,beta);
 161     else {
 162         if(bs1)
 163             for(i=0;i<4;i++)
 164                 loop_filter_c1(d + i*stride,1,alpha,beta,tc);
 165         if (bs2)
 166             for(i=4;i<8;i++)
 167                 loop_filter_c1(d + i*stride,1,alpha,beta,tc);
 168     }
 169 }
 170
 171 static void cavs_filter_ch_c(uint8_t *d, ptrdiff_t stride, int alpha, int beta, int tc,
 172                              int bs1, int bs2)
 173 {
 174     int i;
 175     if(bs1==2)
 176         for(i=0;i<8;i++)
 177             loop_filter_c2(d + i,stride,alpha,beta);
 178     else {
 179         if(bs1)
 180             for(i=0;i<4;i++)
 181                 loop_filter_c1(d + i,stride,alpha,beta,tc);
 182         if (bs2)
 183             for(i=4;i<8;i++)
 184                 loop_filter_c1(d + i,stride,alpha,beta,tc);
 185     }
 186 }
 187
 188 /*****************************************************************************
 189  *
 190  * inverse transform
 191  *
 192  ****************************************************************************/
 193
 194 static void cavs_idct8_add_c(uint8_t *dst, int16_t *block, ptrdiff_t stride)
 195 {
 196     int i;
 197     int16_t (*src)[8] = (int16_t(*)[8])block;
 198
 199     src[0][0] += 8;
 200
 201     for( i = 0; i < 8; i++ ) {
 202         const int a0 = 3 * src[i][1] - 2 * src[i][7];
 203         const int a1 = 3 * src[i][3] + 2 * src[i][5];
 204         const int a2 = 2 * src[i][3] - 3 * src[i][5];
 205         const int a3 = 2 * src[i][1] + 3 * src[i][7];
 206
 207         const int b4 = 2 * (a0 + a1 + a3) + a1;
 208         const int b5 = 2 * (a0 - a1 + a2) + a0;
 209         const int b6 = 2 * (a3 - a2 - a1) + a3;
 210         const int b7 = 2 * (a0 - a2 - a3) - a2;
 211
 212         const int a7 = 4 * src[i][2] - 10 * src[i][6];
 213         const int a6 = 4 * src[i][6] + 10 * src[i][2];
 214         const int a5 = 8 * (src[i][0] - src[i][4]) + 4;
 215         const int a4 = 8 * (src[i][0] + src[i][4]) + 4;
 216
 217         const int b0 = a4 + a6;
 218         const int b1 = a5 + a7;
 219         const int b2 = a5 - a7;
 220         const int b3 = a4 - a6;
 221
 222         src[i][0] = (b0 + b4) >> 3;
 223         src[i][1] = (b1 + b5) >> 3;
 224         src[i][2] = (b2 + b6) >> 3;
 225         src[i][3] = (b3 + b7) >> 3;
 226         src[i][4] = (b3 - b7) >> 3;
 227         src[i][5] = (b2 - b6) >> 3;
 228         src[i][6] = (b1 - b5) >> 3;
 229         src[i][7] = (b0 - b4) >> 3;
 230     }
 231     for( i = 0; i < 8; i++ ) {
 232         const int a0 = 3 * src[1][i] - 2 * src[7][i];
 233         const int a1 = 3 * src[3][i] + 2 * src[5][i];
 234         const int a2 = 2 * src[3][i] - 3 * src[5][i];
 235         const int a3 = 2 * src[1][i] + 3 * src[7][i];
 236
 237         const int b4 = 2 * (a0 + a1 + a3) + a1;
 238         const int b5 = 2 * (a0 - a1 + a2) + a0;
 239         const int b6 = 2 * (a3 - a2 - a1) + a3;
 240         const int b7 = 2 * (a0 - a2 - a3) - a2;
 241
 242         const int a7 = 4 * src[2][i] - 10 * src[6][i];
 243         const int a6 = 4 * src[6][i] + 10 * src[2][i];
 244         const int a5 = 8 * (src[0][i] - src[4][i]);
 245         const int a4 = 8 * (src[0][i] + src[4][i]);
 246
 247         const int b0 = a4 + a6;
 248         const int b1 = a5 + a7;
 249         const int b2 = a5 - a7;
 250         const int b3 = a4 - a6;
 251
 252         dst[i + 0*stride] = av_clip_uint8( dst[i + 0*stride] + ((b0 + b4) >> 7));
 253         dst[i + 1*stride] = av_clip_uint8( dst[i + 1*stride] + ((b1 + b5) >> 7));
 254         dst[i + 2*stride] = av_clip_uint8( dst[i + 2*stride] + ((b2 + b6) >> 7));
 255         dst[i + 3*stride] = av_clip_uint8( dst[i + 3*stride] + ((b3 + b7) >> 7));
 256         dst[i + 4*stride] = av_clip_uint8( dst[i + 4*stride] + ((b3 - b7) >> 7));
 257         dst[i + 5*stride] = av_clip_uint8( dst[i + 5*stride] + ((b2 - b6) >> 7));
 258         dst[i + 6*stride] = av_clip_uint8( dst[i + 6*stride] + ((b1 - b5) >> 7));
 259         dst[i + 7*stride] = av_clip_uint8( dst[i + 7*stride] + ((b0 - b4) >> 7));
 260     }
 261 }
 262
 263 /*****************************************************************************
 264  *
 265  * motion compensation
 266  *
 267  ****************************************************************************/
 268
 269 #define CAVS_SUBPIX(OPNAME, OP, NAME, A, B, C, D, E, F) \
 270 static void OPNAME ## cavs_filt8_h_ ## NAME(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
 271 {                                                                       \
 272     const int h=8;\
 273     const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;\
 274     int i;\
 275     for(i=0; i<h; i++)\
 276     {\
 277         OP(dst[0], A*src[-2] + B*src[-1] + C*src[0] + D*src[1] + E*src[2] + F*src[3]);\
 278         OP(dst[1], A*src[-1] + B*src[ 0] + C*src[1] + D*src[2] + E*src[3] + F*src[4]);\
 279         OP(dst[2], A*src[ 0] + B*src[ 1] + C*src[2] + D*src[3] + E*src[4] + F*src[5]);\
 280         OP(dst[3], A*src[ 1] + B*src[ 2] + C*src[3] + D*src[4] + E*src[5] + F*src[6]);\
 281         OP(dst[4], A*src[ 2] + B*src[ 3] + C*src[4] + D*src[5] + E*src[6] + F*src[7]);\
 282         OP(dst[5], A*src[ 3] + B*src[ 4] + C*src[5] + D*src[6] + E*src[7] + F*src[8]);\
 283         OP(dst[6], A*src[ 4] + B*src[ 5] + C*src[6] + D*src[7] + E*src[8] + F*src[9]);\
 284         OP(dst[7], A*src[ 5] + B*src[ 6] + C*src[7] + D*src[8] + E*src[9] + F*src[10]);\
 285         dst+=dstStride;\
 286         src+=srcStride;\
 287     }\
 288 }\
 289 \
 290 static void OPNAME ## cavs_filt8_v_  ## NAME(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
 291 {                                                                       \
 292     const int w=8;\
 293     const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;\
 294     int i;\
 295     for(i=0; i<w; i++)\
 296     {\
 297         const int srcB= src[-2*srcStride];\
 298         const int srcA= src[-1*srcStride];\
 299         const int src0= src[0 *srcStride];\
 300         const int src1= src[1 *srcStride];\
 301         const int src2= src[2 *srcStride];\
 302         const int src3= src[3 *srcStride];\
 303         const int src4= src[4 *srcStride];\
 304         const int src5= src[5 *srcStride];\
 305         const int src6= src[6 *srcStride];\
 306         const int src7= src[7 *srcStride];\
 307         const int src8= src[8 *srcStride];\
 308         const int src9= src[9 *srcStride];\
 309         const int src10= src[10 *srcStride];\
 310         OP(dst[0*dstStride], A*srcB + B*srcA + C*src0 + D*src1 + E*src2 + F*src3);\
 311         OP(dst[1*dstStride], A*srcA + B*src0 + C*src1 + D*src2 + E*src3 + F*src4);\
 312         OP(dst[2*dstStride], A*src0 + B*src1 + C*src2 + D*src3 + E*src4 + F*src5);\
 313         OP(dst[3*dstStride], A*src1 + B*src2 + C*src3 + D*src4 + E*src5 + F*src6);\
 314         OP(dst[4*dstStride], A*src2 + B*src3 + C*src4 + D*src5 + E*src6 + F*src7);\
 315         OP(dst[5*dstStride], A*src3 + B*src4 + C*src5 + D*src6 + E*src7 + F*src8);\
 316         OP(dst[6*dstStride], A*src4 + B*src5 + C*src6 + D*src7 + E*src8 + F*src9);\
 317         OP(dst[7*dstStride], A*src5 + B*src6 + C*src7 + D*src8 + E*src9 + F*src10);\
 318         dst++;\
 319         src++;\
 320     }\
 321 }\
 322 \
 323 static void OPNAME ## cavs_filt16_v_ ## NAME(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
 324 {                                                                       \
 325     OPNAME ## cavs_filt8_v_ ## NAME(dst  , src  , dstStride, srcStride);\
 326     OPNAME ## cavs_filt8_v_ ## NAME(dst+8, src+8, dstStride, srcStride);\
 327     src += 8*srcStride;\
 328     dst += 8*dstStride;\
 329     OPNAME ## cavs_filt8_v_ ## NAME(dst  , src  , dstStride, srcStride);\
 330     OPNAME ## cavs_filt8_v_ ## NAME(dst+8, src+8, dstStride, srcStride);\
 331 }\
 332 \
 333 static void OPNAME ## cavs_filt16_h_ ## NAME(uint8_t *dst, const uint8_t *src, ptrdiff_t dstStride, ptrdiff_t srcStride)\
 334 {                                                                       \
 335     OPNAME ## cavs_filt8_h_ ## NAME(dst  , src  , dstStride, srcStride);\
 336     OPNAME ## cavs_filt8_h_ ## NAME(dst+8, src+8, dstStride, srcStride);\
 337     src += 8*srcStride;\
 338     dst += 8*dstStride;\
 339     OPNAME ## cavs_filt8_h_ ## NAME(dst  , src  , dstStride, srcStride);\
 340     OPNAME ## cavs_filt8_h_ ## NAME(dst+8, src+8, dstStride, srcStride);\
 341 }\
 342
 343 #define CAVS_SUBPIX_HV(OPNAME, OP, NAME, AH, BH, CH, DH, EH, FH, AV, BV, CV, DV, EV, FV, FULL) \
 344 static void OPNAME ## cavs_filt8_hv_ ## NAME(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, ptrdiff_t dstStride, ptrdiff_t srcStride)\
 345 {                                                                       \
 346     int16_t temp[8*(8+5)];\
 347     int16_t *tmp = temp;\
 348     const int h=8;\
 349     const int w=8;\
 350     const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;\
 351     int i;\
 352     src1 -= 2*srcStride;\
 353     for(i=0; i<h+5; i++)\
 354     {\
 355         tmp[0]= AH*src1[-2] + BH*src1[-1] + CH*src1[0] + DH*src1[1] + EH*src1[2] + FH*src1[3];\
 356         tmp[1]= AH*src1[-1] + BH*src1[ 0] + CH*src1[1] + DH*src1[2] + EH*src1[3] + FH*src1[4];\
 357         tmp[2]= AH*src1[ 0] + BH*src1[ 1] + CH*src1[2] + DH*src1[3] + EH*src1[4] + FH*src1[5];\
 358         tmp[3]= AH*src1[ 1] + BH*src1[ 2] + CH*src1[3] + DH*src1[4] + EH*src1[5] + FH*src1[6];\
 359         tmp[4]= AH*src1[ 2] + BH*src1[ 3] + CH*src1[4] + DH*src1[5] + EH*src1[6] + FH*src1[7];\
 360         tmp[5]= AH*src1[ 3] + BH*src1[ 4] + CH*src1[5] + DH*src1[6] + EH*src1[7] + FH*src1[8];\
 361         tmp[6]= AH*src1[ 4] + BH*src1[ 5] + CH*src1[6] + DH*src1[7] + EH*src1[8] + FH*src1[9];\
 362         tmp[7]= AH*src1[ 5] + BH*src1[ 6] + CH*src1[7] + DH*src1[8] + EH*src1[9] + FH*src1[10];\
 363         tmp+=8;\
 364         src1+=srcStride;\
 365     }\
 366     if(FULL) {\
 367       tmp = temp+8*2;                           \
 368       for(i=0; i<w; i++)                        \
 369         {                                       \
 370           const int tmpB= tmp[-2*8];    \
 371           const int tmpA= tmp[-1*8];    \
 372           const int tmp0= tmp[0 *8];    \
 373           const int tmp1= tmp[1 *8];    \
 374           const int tmp2= tmp[2 *8];    \
 375           const int tmp3= tmp[3 *8];    \
 376           const int tmp4= tmp[4 *8];    \
 377           const int tmp5= tmp[5 *8];    \
 378           const int tmp6= tmp[6 *8];    \
 379           const int tmp7= tmp[7 *8];    \
 380           const int tmp8= tmp[8 *8];    \
 381           const int tmp9= tmp[9 *8];    \
 382           const int tmp10=tmp[10*8];                            \
 383           OP(dst[0*dstStride], AV*tmpB + BV*tmpA + CV*tmp0 + DV*tmp1 + EV*tmp2 + FV*tmp3 + 64*src2[0*srcStride]); \
 384           OP(dst[1*dstStride], AV*tmpA + BV*tmp0 + CV*tmp1 + DV*tmp2 + EV*tmp3 + FV*tmp4 + 64*src2[1*srcStride]); \
 385           OP(dst[2*dstStride], AV*tmp0 + BV*tmp1 + CV*tmp2 + DV*tmp3 + EV*tmp4 + FV*tmp5 + 64*src2[2*srcStride]); \
 386           OP(dst[3*dstStride], AV*tmp1 + BV*tmp2 + CV*tmp3 + DV*tmp4 + EV*tmp5 + FV*tmp6 + 64*src2[3*srcStride]); \
 387           OP(dst[4*dstStride], AV*tmp2 + BV*tmp3 + CV*tmp4 + DV*tmp5 + EV*tmp6 + FV*tmp7 + 64*src2[4*srcStride]); \
 388           OP(dst[5*dstStride], AV*tmp3 + BV*tmp4 + CV*tmp5 + DV*tmp6 + EV*tmp7 + FV*tmp8 + 64*src2[5*srcStride]); \
 389           OP(dst[6*dstStride], AV*tmp4 + BV*tmp5 + CV*tmp6 + DV*tmp7 + EV*tmp8 + FV*tmp9 + 64*src2[6*srcStride]); \
 390           OP(dst[7*dstStride], AV*tmp5 + BV*tmp6 + CV*tmp7 + DV*tmp8 + EV*tmp9 + FV*tmp10 + 64*src2[7*srcStride]); \
 391           dst++;                                                        \
 392           tmp++;                                                        \
 393           src2++;                                                       \
 394         }                                                               \
 395     } else {\
 396       tmp = temp+8*2;                           \
 397       for(i=0; i<w; i++)                        \
 398         {                                       \
 399           const int tmpB= tmp[-2*8];    \
 400           const int tmpA= tmp[-1*8];    \
 401           const int tmp0= tmp[0 *8];    \
 402           const int tmp1= tmp[1 *8];    \
 403           const int tmp2= tmp[2 *8];    \
 404           const int tmp3= tmp[3 *8];    \
 405           const int tmp4= tmp[4 *8];    \
 406           const int tmp5= tmp[5 *8];    \
 407           const int tmp6= tmp[6 *8];    \
 408           const int tmp7= tmp[7 *8];    \
 409           const int tmp8= tmp[8 *8];    \
 410           const int tmp9= tmp[9 *8];    \
 411           const int tmp10=tmp[10*8];                            \
 412           OP(dst[0*dstStride], AV*tmpB + BV*tmpA + CV*tmp0 + DV*tmp1 + EV*tmp2 + FV*tmp3); \
 413           OP(dst[1*dstStride], AV*tmpA + BV*tmp0 + CV*tmp1 + DV*tmp2 + EV*tmp3 + FV*tmp4); \
 414           OP(dst[2*dstStride], AV*tmp0 + BV*tmp1 + CV*tmp2 + DV*tmp3 + EV*tmp4 + FV*tmp5); \
 415           OP(dst[3*dstStride], AV*tmp1 + BV*tmp2 + CV*tmp3 + DV*tmp4 + EV*tmp5 + FV*tmp6); \
 416           OP(dst[4*dstStride], AV*tmp2 + BV*tmp3 + CV*tmp4 + DV*tmp5 + EV*tmp6 + FV*tmp7); \
 417           OP(dst[5*dstStride], AV*tmp3 + BV*tmp4 + CV*tmp5 + DV*tmp6 + EV*tmp7 + FV*tmp8); \
 418           OP(dst[6*dstStride], AV*tmp4 + BV*tmp5 + CV*tmp6 + DV*tmp7 + EV*tmp8 + FV*tmp9); \
 419           OP(dst[7*dstStride], AV*tmp5 + BV*tmp6 + CV*tmp7 + DV*tmp8 + EV*tmp9 + FV*tmp10); \
 420           dst++;                                                        \
 421           tmp++;                                                        \
 422         }                                                               \
 423     }\
 424 }\
 425 \
 426 static void OPNAME ## cavs_filt16_hv_ ## NAME(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, ptrdiff_t dstStride, ptrdiff_t srcStride)\
 427 {                                                                       \
 428     OPNAME ## cavs_filt8_hv_ ## NAME(dst  , src1,   FULL ? src2     : NULL, dstStride, srcStride); \
 429     OPNAME ## cavs_filt8_hv_ ## NAME(dst+8, src1+8, FULL ? src2 + 8 : NULL, dstStride, srcStride); \
 430     src1 += 8*srcStride;\
 431     if (FULL) \
 432         src2 += 8*srcStride;\
 433     dst += 8*dstStride;\
 434     OPNAME ## cavs_filt8_hv_ ## NAME(dst  , src1,   FULL ? src2     : NULL, dstStride, srcStride); \
 435     OPNAME ## cavs_filt8_hv_ ## NAME(dst+8, src1+8, FULL ? src2 + 8 : NULL, dstStride, srcStride); \
 436 }\
 437
 438 #define CAVS_MC(OPNAME, SIZE) \
 439 static void OPNAME ## cavs_qpel ## SIZE ## _mc10_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 440 {\
 441     OPNAME ## cavs_filt ## SIZE ## _h_qpel_l(dst, src, stride, stride);\
 442 }\
 443 \
 444 static void OPNAME ## cavs_qpel ## SIZE ## _mc20_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 445 {\
 446     OPNAME ## cavs_filt ## SIZE ## _h_hpel(dst, src, stride, stride);\
 447 }\
 448 \
 449 static void OPNAME ## cavs_qpel ## SIZE ## _mc30_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 450 {\
 451     OPNAME ## cavs_filt ## SIZE ## _h_qpel_r(dst, src, stride, stride);\
 452 }\
 453 \
 454 static void OPNAME ## cavs_qpel ## SIZE ## _mc01_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 455 {\
 456     OPNAME ## cavs_filt ## SIZE ## _v_qpel_l(dst, src, stride, stride);\
 457 }\
 458 \
 459 static void OPNAME ## cavs_qpel ## SIZE ## _mc02_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 460 {\
 461     OPNAME ## cavs_filt ## SIZE ## _v_hpel(dst, src, stride, stride);\
 462 }\
 463 \
 464 static void OPNAME ## cavs_qpel ## SIZE ## _mc03_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 465 {\
 466     OPNAME ## cavs_filt ## SIZE ## _v_qpel_r(dst, src, stride, stride);\
 467 }\
 468 \
 469 static void OPNAME ## cavs_qpel ## SIZE ## _mc22_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 470 {\
 471   OPNAME ## cavs_filt ## SIZE ## _hv_jj(dst, src, NULL, stride, stride); \
 472 }\
 473 \
 474 static void OPNAME ## cavs_qpel ## SIZE ## _mc11_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 475 {\
 476   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src, stride, stride); \
 477 }\
 478 \
 479 static void OPNAME ## cavs_qpel ## SIZE ## _mc13_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 480 {\
 481   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+stride, stride, stride); \
 482 }\
 483 \
 484 static void OPNAME ## cavs_qpel ## SIZE ## _mc31_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 485 {\
 486   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+1, stride, stride); \
 487 }\
 488 \
 489 static void OPNAME ## cavs_qpel ## SIZE ## _mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 490 {\
 491   OPNAME ## cavs_filt ## SIZE ## _hv_egpr(dst, src, src+stride+1,stride, stride); \
 492 }\
 493 \
 494 static void OPNAME ## cavs_qpel ## SIZE ## _mc21_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 495 {\
 496   OPNAME ## cavs_filt ## SIZE ## _hv_ff(dst, src, NULL, stride, stride); \
 497 }\
 498 \
 499 static void OPNAME ## cavs_qpel ## SIZE ## _mc12_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 500 {\
 501   OPNAME ## cavs_filt ## SIZE ## _hv_ii(dst, src, NULL, stride, stride); \
 502 }\
 503 \
 504 static void OPNAME ## cavs_qpel ## SIZE ## _mc32_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 505 {\
 506   OPNAME ## cavs_filt ## SIZE ## _hv_kk(dst, src, NULL, stride, stride); \
 507 }\
 508 \
 509 static void OPNAME ## cavs_qpel ## SIZE ## _mc23_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
 510 {\
 511   OPNAME ## cavs_filt ## SIZE ## _hv_qq(dst, src, NULL, stride, stride); \
 512 }\
 513
 514 #define op_put1(a, b)  a = cm[((b)+4)>>3]
 515 #define op_put2(a, b)  a = cm[((b)+64)>>7]
 516 #define op_put3(a, b)  a = cm[((b)+32)>>6]
 517 #define op_put4(a, b)  a = cm[((b)+512)>>10]
 518 #define op_avg1(a, b)  a = ((a)+cm[((b)+4)>>3]   +1)>>1
 519 #define op_avg2(a, b)  a = ((a)+cm[((b)+64)>>7]  +1)>>1
 520 #define op_avg3(a, b)  a = ((a)+cm[((b)+32)>>6]  +1)>>1
 521 #define op_avg4(a, b)  a = ((a)+cm[((b)+512)>>10]+1)>>1
 522 CAVS_SUBPIX(put_   , op_put1, hpel,    0, -1,  5,  5, -1,  0)
 523 CAVS_SUBPIX(put_   , op_put2, qpel_l, -1, -2, 96, 42, -7,  0)
 524 CAVS_SUBPIX(put_   , op_put2, qpel_r,  0, -7, 42, 96, -2, -1)
 525 CAVS_SUBPIX_HV(put_, op_put3, jj,      0, -1,  5,  5, -1,  0,  0, -1,  5,  5, -1, 0, 0)
 526 CAVS_SUBPIX_HV(put_, op_put4, ff,      0, -1,  5,  5, -1,  0, -1, -2, 96, 42, -7, 0, 0)
 527 CAVS_SUBPIX_HV(put_, op_put4, ii,     -1, -2, 96, 42, -7,  0,  0, -1,  5,  5, -1, 0, 0)
 528 CAVS_SUBPIX_HV(put_, op_put4, kk,      0, -7, 42, 96, -2, -1,  0, -1,  5,  5, -1, 0, 0)
 529 CAVS_SUBPIX_HV(put_, op_put4, qq,      0, -1,  5,  5, -1,  0,  0, -7, 42, 96, -2,-1, 0)
 530 CAVS_SUBPIX_HV(put_, op_put2, egpr,    0, -1,  5,  5, -1,  0,  0, -1,  5,  5, -1, 0, 1)
 531 CAVS_SUBPIX(avg_   , op_avg1, hpel,    0, -1,  5,  5, -1,  0)
 532 CAVS_SUBPIX(avg_   , op_avg2, qpel_l, -1, -2, 96, 42, -7,  0)
 533 CAVS_SUBPIX(avg_   , op_avg2, qpel_r,  0, -7, 42, 96, -2, -1)
 534 CAVS_SUBPIX_HV(avg_, op_avg3, jj,      0, -1,  5,  5, -1,  0,  0, -1,  5,  5, -1, 0, 0)
 535 CAVS_SUBPIX_HV(avg_, op_avg4, ff,      0, -1,  5,  5, -1,  0, -1, -2, 96, 42, -7, 0, 0)
 536 CAVS_SUBPIX_HV(avg_, op_avg4, ii,     -1, -2, 96, 42, -7,  0,  0, -1,  5,  5, -1, 0, 0)
 537 CAVS_SUBPIX_HV(avg_, op_avg4, kk,      0, -7, 42, 96, -2, -1,  0, -1,  5,  5, -1, 0, 0)
 538 CAVS_SUBPIX_HV(avg_, op_avg4, qq,      0, -1,  5,  5, -1,  0,  0, -7, 42, 96, -2,-1, 0)
 539 CAVS_SUBPIX_HV(avg_, op_avg2, egpr,    0, -1,  5,  5, -1,  0,  0, -1,  5,  5, -1, 0, 1)
 540 CAVS_MC(put_, 8)
 541 CAVS_MC(put_, 16)
 542 CAVS_MC(avg_, 8)
 543 CAVS_MC(avg_, 16)
 544
 545 #define put_cavs_qpel8_mc00_c  ff_put_pixels8x8_c
 546 #define avg_cavs_qpel8_mc00_c  ff_avg_pixels8x8_c
 547 #define put_cavs_qpel16_mc00_c ff_put_pixels16x16_c
 548 #define avg_cavs_qpel16_mc00_c ff_avg_pixels16x16_c
 549
 550 av_cold void ff_cavsdsp_init(CAVSDSPContext* c)
 551 {
 552 #define dspfunc(PFX, IDX, NUM) \
 553     c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_c; \
 554     c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_c; \
 555     c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_c; \
 556     c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_c; \
 557     c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_c; \
 558     c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_c; \
 559     c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_c; \
 560     c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_c; \
 561     c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_c; \
 562     c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_c; \
 563     c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_c; \
 564     c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_c; \
 565     c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_c; \
 566     c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_c; \
 567     c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_c; \
 568     c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_c
 569     dspfunc(put_cavs_qpel, 0, 16);
 570     dspfunc(put_cavs_qpel, 1, 8);
 571     dspfunc(avg_cavs_qpel, 0, 16);
 572     dspfunc(avg_cavs_qpel, 1, 8);
 573     c->cavs_filter_lv = cavs_filter_lv_c;
 574     c->cavs_filter_lh = cavs_filter_lh_c;
 575     c->cavs_filter_cv = cavs_filter_cv_c;
 576     c->cavs_filter_ch = cavs_filter_ch_c;
 577     c->cavs_idct8_add = cavs_idct8_add_c;
 578     c->idct_perm = FF_IDCT_PERM_NONE;
 579
 580 #if ARCH_X86
 581     ff_cavsdsp_init_x86(c);
 582 #endif
 583 }