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