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