libavcodec/h264idct_template.c

   1 /*
   2  * H.264 IDCT
   3  * Copyright (c) 2004-2011 Michael Niedermayer <michaelni@gmx.at>
   4  *
   5  * This file is part of Libav.
   6  *
   7  * Libav is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * Libav is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with Libav; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 /**
  23  * @file
  24  * H.264 IDCT.
  25  * @author Michael Niedermayer <michaelni@gmx.at>
  26  */
  27
  28 #include "bit_depth_template.c"
  29 #include "libavutil/common.h"
  30 #include "h264.h"
  31
  32 void FUNCC(ff_h264_idct_add)(uint8_t *_dst, int16_t *_block, int stride)
  33 {
  34     int i;
  35     pixel *dst = (pixel*)_dst;
  36     dctcoef *block = (dctcoef*)_block;
  37     stride /= sizeof(pixel);
  38
  39     block[0] += 1 << 5;
  40
  41     for(i=0; i<4; i++){
  42         const int z0=  block[i + 4*0]     +  block[i + 4*2];
  43         const int z1=  block[i + 4*0]     -  block[i + 4*2];
  44         const int z2= (block[i + 4*1]>>1) -  block[i + 4*3];
  45         const int z3=  block[i + 4*1]     + (block[i + 4*3]>>1);
  46
  47         block[i + 4*0]= z0 + z3;
  48         block[i + 4*1]= z1 + z2;
  49         block[i + 4*2]= z1 - z2;
  50         block[i + 4*3]= z0 - z3;
  51     }
  52
  53     for(i=0; i<4; i++){
  54         const int z0=  block[0 + 4*i]     +  block[2 + 4*i];
  55         const int z1=  block[0 + 4*i]     -  block[2 + 4*i];
  56         const int z2= (block[1 + 4*i]>>1) -  block[3 + 4*i];
  57         const int z3=  block[1 + 4*i]     + (block[3 + 4*i]>>1);
  58
  59         dst[i + 0*stride]= av_clip_pixel(dst[i + 0*stride] + ((z0 + z3) >> 6));
  60         dst[i + 1*stride]= av_clip_pixel(dst[i + 1*stride] + ((z1 + z2) >> 6));
  61         dst[i + 2*stride]= av_clip_pixel(dst[i + 2*stride] + ((z1 - z2) >> 6));
  62         dst[i + 3*stride]= av_clip_pixel(dst[i + 3*stride] + ((z0 - z3) >> 6));
  63     }
  64 }
  65
  66 void FUNCC(ff_h264_idct8_add)(uint8_t *_dst, int16_t *_block, int stride){
  67     int i;
  68     pixel *dst = (pixel*)_dst;
  69     dctcoef *block = (dctcoef*)_block;
  70     stride /= sizeof(pixel);
  71
  72     block[0] += 32;
  73
  74     for( i = 0; i < 8; i++ )
  75     {
  76         const int a0 =  block[i+0*8] + block[i+4*8];
  77         const int a2 =  block[i+0*8] - block[i+4*8];
  78         const int a4 = (block[i+2*8]>>1) - block[i+6*8];
  79         const int a6 = (block[i+6*8]>>1) + block[i+2*8];
  80
  81         const int b0 = a0 + a6;
  82         const int b2 = a2 + a4;
  83         const int b4 = a2 - a4;
  84         const int b6 = a0 - a6;
  85
  86         const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
  87         const int a3 =  block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
  88         const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
  89         const int a7 =  block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
  90
  91         const int b1 = (a7>>2) + a1;
  92         const int b3 =  a3 + (a5>>2);
  93         const int b5 = (a3>>2) - a5;
  94         const int b7 =  a7 - (a1>>2);
  95
  96         block[i+0*8] = b0 + b7;
  97         block[i+7*8] = b0 - b7;
  98         block[i+1*8] = b2 + b5;
  99         block[i+6*8] = b2 - b5;
 100         block[i+2*8] = b4 + b3;
 101         block[i+5*8] = b4 - b3;
 102         block[i+3*8] = b6 + b1;
 103         block[i+4*8] = b6 - b1;
 104     }
 105     for( i = 0; i < 8; i++ )
 106     {
 107         const int a0 =  block[0+i*8] + block[4+i*8];
 108         const int a2 =  block[0+i*8] - block[4+i*8];
 109         const int a4 = (block[2+i*8]>>1) - block[6+i*8];
 110         const int a6 = (block[6+i*8]>>1) + block[2+i*8];
 111
 112         const int b0 = a0 + a6;
 113         const int b2 = a2 + a4;
 114         const int b4 = a2 - a4;
 115         const int b6 = a0 - a6;
 116
 117         const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
 118         const int a3 =  block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
 119         const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
 120         const int a7 =  block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
 121
 122         const int b1 = (a7>>2) + a1;
 123         const int b3 =  a3 + (a5>>2);
 124         const int b5 = (a3>>2) - a5;
 125         const int b7 =  a7 - (a1>>2);
 126
 127         dst[i + 0*stride] = av_clip_pixel( dst[i + 0*stride] + ((b0 + b7) >> 6) );
 128         dst[i + 1*stride] = av_clip_pixel( dst[i + 1*stride] + ((b2 + b5) >> 6) );
 129         dst[i + 2*stride] = av_clip_pixel( dst[i + 2*stride] + ((b4 + b3) >> 6) );
 130         dst[i + 3*stride] = av_clip_pixel( dst[i + 3*stride] + ((b6 + b1) >> 6) );
 131         dst[i + 4*stride] = av_clip_pixel( dst[i + 4*stride] + ((b6 - b1) >> 6) );
 132         dst[i + 5*stride] = av_clip_pixel( dst[i + 5*stride] + ((b4 - b3) >> 6) );
 133         dst[i + 6*stride] = av_clip_pixel( dst[i + 6*stride] + ((b2 - b5) >> 6) );
 134         dst[i + 7*stride] = av_clip_pixel( dst[i + 7*stride] + ((b0 - b7) >> 6) );
 135     }
 136 }
 137
 138 // assumes all AC coefs are 0
 139 void FUNCC(ff_h264_idct_dc_add)(uint8_t *_dst, int16_t *block, int stride){
 140     int i, j;
 141     int dc = (((dctcoef*)block)[0] + 32) >> 6;
 142     pixel *dst = (pixel*)_dst;
 143     stride /= sizeof(pixel);
 144     for( j = 0; j < 4; j++ )
 145     {
 146         for( i = 0; i < 4; i++ )
 147             dst[i] = av_clip_pixel( dst[i] + dc );
 148         dst += stride;
 149     }
 150 }
 151
 152 void FUNCC(ff_h264_idct8_dc_add)(uint8_t *_dst, int16_t *block, int stride){
 153     int i, j;
 154     int dc = (((dctcoef*)block)[0] + 32) >> 6;
 155     pixel *dst = (pixel*)_dst;
 156     stride /= sizeof(pixel);
 157     for( j = 0; j < 8; j++ )
 158     {
 159         for( i = 0; i < 8; i++ )
 160             dst[i] = av_clip_pixel( dst[i] + dc );
 161         dst += stride;
 162     }
 163 }
 164
 165 void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
 166     int i;
 167     for(i=0; i<16; i++){
 168         int nnz = nnzc[ scan8[i] ];
 169         if(nnz){
 170             if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
 171             else                                  FUNCC(ff_h264_idct_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
 172         }
 173     }
 174 }
 175
 176 void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
 177     int i;
 178     for(i=0; i<16; i++){
 179         if(nnzc[ scan8[i] ])             FUNCC(ff_h264_idct_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
 180         else if(((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
 181     }
 182 }
 183
 184 void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
 185     int i;
 186     for(i=0; i<16; i+=4){
 187         int nnz = nnzc[ scan8[i] ];
 188         if(nnz){
 189             if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct8_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
 190             else                                  FUNCC(ff_h264_idct8_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
 191         }
 192     }
 193 }
 194
 195 void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
 196     int i, j;
 197     for(j=1; j<3; j++){
 198         for(i=j*16; i<j*16+4; i++){
 199             if(nnzc[ scan8[i] ])
 200                 FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
 201             else if(((dctcoef*)block)[i*16])
 202                 FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
 203         }
 204     }
 205 }
 206
 207 void FUNCC(ff_h264_idct_add8_422)(uint8_t **dest, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
 208     int i, j;
 209
 210     for(j=1; j<3; j++){
 211         for(i=j*16; i<j*16+4; i++){
 212             if(nnzc[ scan8[i] ])
 213                 FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
 214             else if(((dctcoef*)block)[i*16])
 215                 FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
 216         }
 217     }
 218
 219     for(j=1; j<3; j++){
 220         for(i=j*16+4; i<j*16+8; i++){
 221             if(nnzc[ scan8[i+4] ])
 222                 FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
 223             else if(((dctcoef*)block)[i*16])
 224                 FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
 225         }
 226     }
 227 }
 228
 229 /**
 230  * IDCT transforms the 16 dc values and dequantizes them.
 231  * @param qmul quantization parameter
 232  */
 233 void FUNCC(ff_h264_luma_dc_dequant_idct)(int16_t *_output, int16_t *_input, int qmul){
 234 #define stride 16
 235     int i;
 236     int temp[16];
 237     static const uint8_t x_offset[4]={0, 2*stride, 8*stride, 10*stride};
 238     dctcoef *input = (dctcoef*)_input;
 239     dctcoef *output = (dctcoef*)_output;
 240
 241     for(i=0; i<4; i++){
 242         const int z0= input[4*i+0] + input[4*i+1];
 243         const int z1= input[4*i+0] - input[4*i+1];
 244         const int z2= input[4*i+2] - input[4*i+3];
 245         const int z3= input[4*i+2] + input[4*i+3];
 246
 247         temp[4*i+0]= z0+z3;
 248         temp[4*i+1]= z0-z3;
 249         temp[4*i+2]= z1-z2;
 250         temp[4*i+3]= z1+z2;
 251     }
 252
 253     for(i=0; i<4; i++){
 254         const int offset= x_offset[i];
 255         const int z0= temp[4*0+i] + temp[4*2+i];
 256         const int z1= temp[4*0+i] - temp[4*2+i];
 257         const int z2= temp[4*1+i] - temp[4*3+i];
 258         const int z3= temp[4*1+i] + temp[4*3+i];
 259
 260         output[stride* 0+offset]= ((((z0 + z3)*qmul + 128 ) >> 8));
 261         output[stride* 1+offset]= ((((z1 + z2)*qmul + 128 ) >> 8));
 262         output[stride* 4+offset]= ((((z1 - z2)*qmul + 128 ) >> 8));
 263         output[stride* 5+offset]= ((((z0 - z3)*qmul + 128 ) >> 8));
 264     }
 265 #undef stride
 266 }
 267
 268 void FUNCC(ff_h264_chroma422_dc_dequant_idct)(int16_t *_block, int qmul){
 269     const int stride= 16*2;
 270     const int xStride= 16;
 271     int i;
 272     int temp[8];
 273     static const uint8_t x_offset[2]={0, 16};
 274     dctcoef *block = (dctcoef*)_block;
 275
 276     for(i=0; i<4; i++){
 277         temp[2*i+0] = block[stride*i + xStride*0] + block[stride*i + xStride*1];
 278         temp[2*i+1] = block[stride*i + xStride*0] - block[stride*i + xStride*1];
 279     }
 280
 281     for(i=0; i<2; i++){
 282         const int offset= x_offset[i];
 283         const int z0= temp[2*0+i] + temp[2*2+i];
 284         const int z1= temp[2*0+i] - temp[2*2+i];
 285         const int z2= temp[2*1+i] - temp[2*3+i];
 286         const int z3= temp[2*1+i] + temp[2*3+i];
 287
 288         block[stride*0+offset]= ((z0 + z3)*qmul + 128) >> 8;
 289         block[stride*1+offset]= ((z1 + z2)*qmul + 128) >> 8;
 290         block[stride*2+offset]= ((z1 - z2)*qmul + 128) >> 8;
 291         block[stride*3+offset]= ((z0 - z3)*qmul + 128) >> 8;
 292     }
 293 }
 294
 295 void FUNCC(ff_h264_chroma_dc_dequant_idct)(int16_t *_block, int qmul){
 296     const int stride= 16*2;
 297     const int xStride= 16;
 298     int a,b,c,d,e;
 299     dctcoef *block = (dctcoef*)_block;
 300
 301     a= block[stride*0 + xStride*0];
 302     b= block[stride*0 + xStride*1];
 303     c= block[stride*1 + xStride*0];
 304     d= block[stride*1 + xStride*1];
 305
 306     e= a-b;
 307     a= a+b;
 308     b= c-d;
 309     c= c+d;
 310
 311     block[stride*0 + xStride*0]= ((a+c)*qmul) >> 7;
 312     block[stride*0 + xStride*1]= ((e+b)*qmul) >> 7;
 313     block[stride*1 + xStride*0]= ((a-c)*qmul) >> 7;
 314     block[stride*1 + xStride*1]= ((e-b)*qmul) >> 7;
 315 }