Libav 0.7.1
libavcodec/h264idct_template.c
Go to the documentation of this file.
00001 /*
00002  * H.264 IDCT
00003  * Copyright (c) 2004-2011 Michael Niedermayer <michaelni@gmx.at>
00004  *
00005  * This file is part of Libav.
00006  *
00007  * Libav is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * Libav is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with Libav; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00028 #include "high_bit_depth.h"
00029 
00030 #ifndef AVCODEC_H264IDCT_INTERNAL_H
00031 #define AVCODEC_H264IDCT_INTERNAL_H
00032 //FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split
00033 static const uint8_t scan8[16*3]={
00034  4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8,
00035  6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8,
00036  4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8,
00037  6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8,
00038  4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8,
00039  6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8,
00040  4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8,
00041  6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8,
00042  4+11*8, 5+11*8, 4+12*8, 5+12*8,
00043  6+11*8, 7+11*8, 6+12*8, 7+12*8,
00044  4+13*8, 5+13*8, 4+14*8, 5+14*8,
00045  6+13*8, 7+13*8, 6+14*8, 7+14*8
00046 };
00047 #endif
00048 
00049 static av_always_inline void FUNCC(idct_internal)(uint8_t *_dst, DCTELEM *_block, int stride, int block_stride, int shift, int add){
00050     int i;
00051     INIT_CLIP
00052     pixel *dst = (pixel*)_dst;
00053     dctcoef *block = (dctcoef*)_block;
00054     stride /= sizeof(pixel);
00055 
00056     block[0] += 1<<(shift-1);
00057 
00058     for(i=0; i<4; i++){
00059         const int z0=  block[i + block_stride*0]     +  block[i + block_stride*2];
00060         const int z1=  block[i + block_stride*0]     -  block[i + block_stride*2];
00061         const int z2= (block[i + block_stride*1]>>1) -  block[i + block_stride*3];
00062         const int z3=  block[i + block_stride*1]     + (block[i + block_stride*3]>>1);
00063 
00064         block[i + block_stride*0]= z0 + z3;
00065         block[i + block_stride*1]= z1 + z2;
00066         block[i + block_stride*2]= z1 - z2;
00067         block[i + block_stride*3]= z0 - z3;
00068     }
00069 
00070     for(i=0; i<4; i++){
00071         const int z0=  block[0 + block_stride*i]     +  block[2 + block_stride*i];
00072         const int z1=  block[0 + block_stride*i]     -  block[2 + block_stride*i];
00073         const int z2= (block[1 + block_stride*i]>>1) -  block[3 + block_stride*i];
00074         const int z3=  block[1 + block_stride*i]     + (block[3 + block_stride*i]>>1);
00075 
00076         dst[i + 0*stride]= CLIP(add*dst[i + 0*stride] + ((z0 + z3) >> shift));
00077         dst[i + 1*stride]= CLIP(add*dst[i + 1*stride] + ((z1 + z2) >> shift));
00078         dst[i + 2*stride]= CLIP(add*dst[i + 2*stride] + ((z1 - z2) >> shift));
00079         dst[i + 3*stride]= CLIP(add*dst[i + 3*stride] + ((z0 - z3) >> shift));
00080     }
00081 }
00082 
00083 void FUNCC(ff_h264_idct_add)(uint8_t *dst, DCTELEM *block, int stride){
00084     FUNCC(idct_internal)(dst, block, stride, 4, 6, 1);
00085 }
00086 
00087 void FUNCC(ff_h264_lowres_idct_add)(uint8_t *dst, int stride, DCTELEM *block){
00088     FUNCC(idct_internal)(dst, block, stride, 8, 3, 1);
00089 }
00090 
00091 void FUNCC(ff_h264_lowres_idct_put)(uint8_t *dst, int stride, DCTELEM *block){
00092     FUNCC(idct_internal)(dst, block, stride, 8, 3, 0);
00093 }
00094 
00095 void FUNCC(ff_h264_idct8_add)(uint8_t *_dst, DCTELEM *_block, int stride){
00096     int i;
00097     INIT_CLIP
00098     pixel *dst = (pixel*)_dst;
00099     dctcoef *block = (dctcoef*)_block;
00100     stride /= sizeof(pixel);
00101 
00102     block[0] += 32;
00103 
00104     for( i = 0; i < 8; i++ )
00105     {
00106         const int a0 =  block[i+0*8] + block[i+4*8];
00107         const int a2 =  block[i+0*8] - block[i+4*8];
00108         const int a4 = (block[i+2*8]>>1) - block[i+6*8];
00109         const int a6 = (block[i+6*8]>>1) + block[i+2*8];
00110 
00111         const int b0 = a0 + a6;
00112         const int b2 = a2 + a4;
00113         const int b4 = a2 - a4;
00114         const int b6 = a0 - a6;
00115 
00116         const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
00117         const int a3 =  block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
00118         const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
00119         const int a7 =  block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
00120 
00121         const int b1 = (a7>>2) + a1;
00122         const int b3 =  a3 + (a5>>2);
00123         const int b5 = (a3>>2) - a5;
00124         const int b7 =  a7 - (a1>>2);
00125 
00126         block[i+0*8] = b0 + b7;
00127         block[i+7*8] = b0 - b7;
00128         block[i+1*8] = b2 + b5;
00129         block[i+6*8] = b2 - b5;
00130         block[i+2*8] = b4 + b3;
00131         block[i+5*8] = b4 - b3;
00132         block[i+3*8] = b6 + b1;
00133         block[i+4*8] = b6 - b1;
00134     }
00135     for( i = 0; i < 8; i++ )
00136     {
00137         const int a0 =  block[0+i*8] + block[4+i*8];
00138         const int a2 =  block[0+i*8] - block[4+i*8];
00139         const int a4 = (block[2+i*8]>>1) - block[6+i*8];
00140         const int a6 = (block[6+i*8]>>1) + block[2+i*8];
00141 
00142         const int b0 = a0 + a6;
00143         const int b2 = a2 + a4;
00144         const int b4 = a2 - a4;
00145         const int b6 = a0 - a6;
00146 
00147         const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
00148         const int a3 =  block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
00149         const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
00150         const int a7 =  block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
00151 
00152         const int b1 = (a7>>2) + a1;
00153         const int b3 =  a3 + (a5>>2);
00154         const int b5 = (a3>>2) - a5;
00155         const int b7 =  a7 - (a1>>2);
00156 
00157         dst[i + 0*stride] = CLIP( dst[i + 0*stride] + ((b0 + b7) >> 6) );
00158         dst[i + 1*stride] = CLIP( dst[i + 1*stride] + ((b2 + b5) >> 6) );
00159         dst[i + 2*stride] = CLIP( dst[i + 2*stride] + ((b4 + b3) >> 6) );
00160         dst[i + 3*stride] = CLIP( dst[i + 3*stride] + ((b6 + b1) >> 6) );
00161         dst[i + 4*stride] = CLIP( dst[i + 4*stride] + ((b6 - b1) >> 6) );
00162         dst[i + 5*stride] = CLIP( dst[i + 5*stride] + ((b4 - b3) >> 6) );
00163         dst[i + 6*stride] = CLIP( dst[i + 6*stride] + ((b2 - b5) >> 6) );
00164         dst[i + 7*stride] = CLIP( dst[i + 7*stride] + ((b0 - b7) >> 6) );
00165     }
00166 }
00167 
00168 // assumes all AC coefs are 0
00169 void FUNCC(ff_h264_idct_dc_add)(uint8_t *_dst, DCTELEM *block, int stride){
00170     int i, j;
00171     int dc = (((dctcoef*)block)[0] + 32) >> 6;
00172     INIT_CLIP
00173     pixel *dst = (pixel*)_dst;
00174     stride /= sizeof(pixel);
00175     for( j = 0; j < 4; j++ )
00176     {
00177         for( i = 0; i < 4; i++ )
00178             dst[i] = CLIP( dst[i] + dc );
00179         dst += stride;
00180     }
00181 }
00182 
00183 void FUNCC(ff_h264_idct8_dc_add)(uint8_t *_dst, DCTELEM *block, int stride){
00184     int i, j;
00185     int dc = (((dctcoef*)block)[0] + 32) >> 6;
00186     INIT_CLIP
00187     pixel *dst = (pixel*)_dst;
00188     stride /= sizeof(pixel);
00189     for( j = 0; j < 8; j++ )
00190     {
00191         for( i = 0; i < 8; i++ )
00192             dst[i] = CLIP( dst[i] + dc );
00193         dst += stride;
00194     }
00195 }
00196 
00197 void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
00198     int i;
00199     for(i=0; i<16; i++){
00200         int nnz = nnzc[ scan8[i] ];
00201         if(nnz){
00202             if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
00203             else                                  FUNCC(idct_internal      )(dst + block_offset[i], block + i*16*sizeof(pixel), stride, 4, 6, 1);
00204         }
00205     }
00206 }
00207 
00208 void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
00209     int i;
00210     for(i=0; i<16; i++){
00211         if(nnzc[ scan8[i] ])             FUNCC(idct_internal      )(dst + block_offset[i], block + i*16*sizeof(pixel), stride, 4, 6, 1);
00212         else if(((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
00213     }
00214 }
00215 
00216 void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
00217     int i;
00218     for(i=0; i<16; i+=4){
00219         int nnz = nnzc[ scan8[i] ];
00220         if(nnz){
00221             if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct8_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
00222             else                                  FUNCC(ff_h264_idct8_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
00223         }
00224     }
00225 }
00226 
00227 void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
00228     int i, j;
00229     for(j=1; j<3; j++){
00230         for(i=j*16; i<j*16+4; i++){
00231             if(nnzc[ scan8[i] ])
00232                 FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
00233             else if(((dctcoef*)block)[i*16])
00234                 FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
00235         }
00236     }
00237 }
00241 void FUNCC(ff_h264_luma_dc_dequant_idct)(DCTELEM *_output, DCTELEM *_input, int qmul){
00242 #define stride 16
00243     int i;
00244     int temp[16];
00245     static const uint8_t x_offset[4]={0, 2*stride, 8*stride, 10*stride};
00246     dctcoef *input = (dctcoef*)_input;
00247     dctcoef *output = (dctcoef*)_output;
00248 
00249     for(i=0; i<4; i++){
00250         const int z0= input[4*i+0] + input[4*i+1];
00251         const int z1= input[4*i+0] - input[4*i+1];
00252         const int z2= input[4*i+2] - input[4*i+3];
00253         const int z3= input[4*i+2] + input[4*i+3];
00254 
00255         temp[4*i+0]= z0+z3;
00256         temp[4*i+1]= z0-z3;
00257         temp[4*i+2]= z1-z2;
00258         temp[4*i+3]= z1+z2;
00259     }
00260 
00261     for(i=0; i<4; i++){
00262         const int offset= x_offset[i];
00263         const int z0= temp[4*0+i] + temp[4*2+i];
00264         const int z1= temp[4*0+i] - temp[4*2+i];
00265         const int z2= temp[4*1+i] - temp[4*3+i];
00266         const int z3= temp[4*1+i] + temp[4*3+i];
00267 
00268         output[stride* 0+offset]= ((((z0 + z3)*qmul + 128 ) >> 8));
00269         output[stride* 1+offset]= ((((z1 + z2)*qmul + 128 ) >> 8));
00270         output[stride* 4+offset]= ((((z1 - z2)*qmul + 128 ) >> 8));
00271         output[stride* 5+offset]= ((((z0 - z3)*qmul + 128 ) >> 8));
00272     }
00273 #undef stride
00274 }
00275 
00276 void FUNCC(ff_h264_chroma_dc_dequant_idct)(DCTELEM *_block, int qmul){
00277     const int stride= 16*2;
00278     const int xStride= 16;
00279     int a,b,c,d,e;
00280     dctcoef *block = (dctcoef*)_block;
00281 
00282     a= block[stride*0 + xStride*0];
00283     b= block[stride*0 + xStride*1];
00284     c= block[stride*1 + xStride*0];
00285     d= block[stride*1 + xStride*1];
00286 
00287     e= a-b;
00288     a= a+b;
00289     b= c-d;
00290     c= c+d;
00291 
00292     block[stride*0 + xStride*0]= ((a+c)*qmul) >> 7;
00293     block[stride*0 + xStride*1]= ((e+b)*qmul) >> 7;
00294     block[stride*1 + xStride*0]= ((a-c)*qmul) >> 7;
00295     block[stride*1 + xStride*1]= ((e-b)*qmul) >> 7;
00296 }