Mercurial > cgi-bin > hgwebdir.cgi > PR > Applications > VSs > VSs__H264__App

annotate libavcodec/h264_idct.c @ 9:ea1ba68cf0ed

find changesets by author, revision, files, or words in the commit message
update to match api changes + add sscc produced source
author Nina Engelhardt <nengel@mailbox.tu-berlin.de>
date Wed, 05 Jun 2013 14:43:26 +0200
parents
children
rev   line source
nengel@2 1 /*
nengel@2 2 * H.264 IDCT
nengel@2 3 * Copyright (c) 2004 Michael Niedermayer <michaelni@gmx.at>
nengel@2 4 *
nengel@2 5 * This file is part of FFmpeg.
nengel@2 6 *
nengel@2 7 * FFmpeg is free software; you can redistribute it and/or
nengel@2 8 * modify it under the terms of the GNU Lesser General Public
nengel@2 9 * License as published by the Free Software Foundation; either
nengel@2 10 * version 2.1 of the License, or (at your option) any later version.
nengel@2 11 *
nengel@2 12 * FFmpeg is distributed in the hope that it will be useful,
nengel@2 13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
nengel@2 14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
nengel@2 15 * Lesser General Public License for more details.
nengel@2 16 *
nengel@2 17 * You should have received a copy of the GNU Lesser General Public
nengel@2 18 * License along with FFmpeg; if not, write to the Free Software
nengel@2 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
nengel@2 20 */
nengel@2 21
nengel@2 22 /**
nengel@2 23 * @file
nengel@2 24 * H.264 IDCT.
nengel@2 25 * @author Michael Niedermayer <michaelni@gmx.at>
nengel@2 26 */
nengel@2 27
nengel@2 28 #include "dsputil.h"
nengel@2 29 #include "h264_data.h"
nengel@2 30
nengel@2 31 static av_always_inline void idct_internal(uint8_t *dst, DCTELEM *block, int stride, int block_stride, int shift, int add){
nengel@2 32 int i;
nengel@2 33 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
nengel@2 34
nengel@2 35 block[0] += 1<<(shift-1);
nengel@2 36
nengel@2 37 for(i=0; i<4; i++){
nengel@2 38 const int z0= block[0 + block_stride*i] + block[2 + block_stride*i];
nengel@2 39 const int z1= block[0 + block_stride*i] - block[2 + block_stride*i];
nengel@2 40 const int z2= (block[1 + block_stride*i]>>1) - block[3 + block_stride*i];
nengel@2 41 const int z3= block[1 + block_stride*i] + (block[3 + block_stride*i]>>1);
nengel@2 42
nengel@2 43 block[0 + block_stride*i]= z0 + z3;
nengel@2 44 block[1 + block_stride*i]= z1 + z2;
nengel@2 45 block[2 + block_stride*i]= z1 - z2;
nengel@2 46 block[3 + block_stride*i]= z0 - z3;
nengel@2 47 }
nengel@2 48
nengel@2 49 for(i=0; i<4; i++){
nengel@2 50 const int z0= block[i + block_stride*0] + block[i + block_stride*2];
nengel@2 51 const int z1= block[i + block_stride*0] - block[i + block_stride*2];
nengel@2 52 const int z2= (block[i + block_stride*1]>>1) - block[i + block_stride*3];
nengel@2 53 const int z3= block[i + block_stride*1] + (block[i + block_stride*3]>>1);
nengel@2 54
nengel@2 55 dst[i + 0*stride]= cm[ add*dst[i + 0*stride] + ((z0 + z3) >> shift) ];
nengel@2 56 dst[i + 1*stride]= cm[ add*dst[i + 1*stride] + ((z1 + z2) >> shift) ];
nengel@2 57 dst[i + 2*stride]= cm[ add*dst[i + 2*stride] + ((z1 - z2) >> shift) ];
nengel@2 58 dst[i + 3*stride]= cm[ add*dst[i + 3*stride] + ((z0 - z3) >> shift) ];
nengel@2 59 }
nengel@2 60 }
nengel@2 61
nengel@2 62 void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride){
nengel@2 63 idct_internal(dst, block, stride, 4, 6, 1);
nengel@2 64 }
nengel@2 65
nengel@2 66 void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block){
nengel@2 67 idct_internal(dst, block, stride, 8, 3, 1);
nengel@2 68 }
nengel@2 69
nengel@2 70 void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block){
nengel@2 71 idct_internal(dst, block, stride, 8, 3, 0);
nengel@2 72 }
nengel@2 73
nengel@2 74 void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride){
nengel@2 75 int i;
nengel@2 76 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
nengel@2 77
nengel@2 78 block[0] += 32;
nengel@2 79
nengel@2 80 for( i = 0; i < 8; i++ )
nengel@2 81 {
nengel@2 82 const int a0 = block[0+i*8] + block[4+i*8];
nengel@2 83 const int a2 = block[0+i*8] - block[4+i*8];
nengel@2 84 const int a4 = (block[2+i*8]>>1) - block[6+i*8];
nengel@2 85 const int a6 = (block[6+i*8]>>1) + block[2+i*8];
nengel@2 86
nengel@2 87 const int b0 = a0 + a6;
nengel@2 88 const int b2 = a2 + a4;
nengel@2 89 const int b4 = a2 - a4;
nengel@2 90 const int b6 = a0 - a6;
nengel@2 91
nengel@2 92 const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
nengel@2 93 const int a3 = block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
nengel@2 94 const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
nengel@2 95 const int a7 = block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
nengel@2 96
nengel@2 97 const int b1 = (a7>>2) + a1;
nengel@2 98 const int b3 = a3 + (a5>>2);
nengel@2 99 const int b5 = (a3>>2) - a5;
nengel@2 100 const int b7 = a7 - (a1>>2);
nengel@2 101
nengel@2 102 block[0+i*8] = b0 + b7;
nengel@2 103 block[7+i*8] = b0 - b7;
nengel@2 104 block[1+i*8] = b2 + b5;
nengel@2 105 block[6+i*8] = b2 - b5;
nengel@2 106 block[2+i*8] = b4 + b3;
nengel@2 107 block[5+i*8] = b4 - b3;
nengel@2 108 block[3+i*8] = b6 + b1;
nengel@2 109 block[4+i*8] = b6 - b1;
nengel@2 110 }
nengel@2 111 for( i = 0; i < 8; i++ )
nengel@2 112 {
nengel@2 113 const int a0 = block[i+0*8] + block[i+4*8];
nengel@2 114 const int a2 = block[i+0*8] - block[i+4*8];
nengel@2 115 const int a4 = (block[i+2*8]>>1) - block[i+6*8];
nengel@2 116 const int a6 = (block[i+6*8]>>1) + block[i+2*8];
nengel@2 117
nengel@2 118 const int b0 = a0 + a6;
nengel@2 119 const int b2 = a2 + a4;
nengel@2 120 const int b4 = a2 - a4;
nengel@2 121 const int b6 = a0 - a6;
nengel@2 122
nengel@2 123 const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
nengel@2 124 const int a3 = block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
nengel@2 125 const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
nengel@2 126 const int a7 = block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
nengel@2 127
nengel@2 128 const int b1 = (a7>>2) + a1;
nengel@2 129 const int b3 = a3 + (a5>>2);
nengel@2 130 const int b5 = (a3>>2) - a5;
nengel@2 131 const int b7 = a7 - (a1>>2);
nengel@2 132
nengel@2 133 dst[i + 0*stride] = cm[ dst[i + 0*stride] + ((b0 + b7) >> 6) ];
nengel@2 134 dst[i + 1*stride] = cm[ dst[i + 1*stride] + ((b2 + b5) >> 6) ];
nengel@2 135 dst[i + 2*stride] = cm[ dst[i + 2*stride] + ((b4 + b3) >> 6) ];
nengel@2 136 dst[i + 3*stride] = cm[ dst[i + 3*stride] + ((b6 + b1) >> 6) ];
nengel@2 137 dst[i + 4*stride] = cm[ dst[i + 4*stride] + ((b6 - b1) >> 6) ];
nengel@2 138 dst[i + 5*stride] = cm[ dst[i + 5*stride] + ((b4 - b3) >> 6) ];
nengel@2 139 dst[i + 6*stride] = cm[ dst[i + 6*stride] + ((b2 - b5) >> 6) ];
nengel@2 140 dst[i + 7*stride] = cm[ dst[i + 7*stride] + ((b0 - b7) >> 6) ];
nengel@2 141 }
nengel@2 142 }
nengel@2 143
nengel@2 144 // assumes all AC coefs are 0
nengel@2 145 void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride){
nengel@2 146 int i, j;
nengel@2 147 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
nengel@2 148 int dc = (block[0] + 32) >> 6;
nengel@2 149 for( j = 0; j < 4; j++ )
nengel@2 150 {
nengel@2 151 for( i = 0; i < 4; i++ )
nengel@2 152 dst[i] = cm[ dst[i] + dc ];
nengel@2 153 dst += stride;
nengel@2 154 }
nengel@2 155 }
nengel@2 156
nengel@2 157 void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride){
nengel@2 158 int i, j;
nengel@2 159 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
nengel@2 160 int dc = (block[0] + 32) >> 6;
nengel@2 161 for( j = 0; j < 8; j++ )
nengel@2 162 {
nengel@2 163 for( i = 0; i < 8; i++ )
nengel@2 164 dst[i] = cm[ dst[i] + dc ];
nengel@2 165 dst += stride;
nengel@2 166 }
nengel@2 167 }
nengel@2 168
nengel@2 169 void ff_h264_idct_add16_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
nengel@2 170 int i;
nengel@2 171 for(i=0; i<16; i++){
nengel@2 172 int nnz = nnzc[ scan8[i] ];
nengel@2 173 if(nnz){
nengel@2 174 if(nnz==1 && block[i*16]) ff_h264_idct_dc_add_c(dst + block_offset[i], block + i*16, stride);
nengel@2 175 else idct_internal (dst + block_offset[i], block + i*16, stride, 4, 6, 1);
nengel@2 176 }
nengel@2 177 }
nengel@2 178 }
nengel@2 179
nengel@2 180 void ff_h264_idct_add16intra_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
nengel@2 181 int i;
nengel@2 182 for(i=0; i<16; i++){
nengel@2 183 if(nnzc[ scan8[i] ]) idct_internal (dst + block_offset[i], block + i*16, stride, 4, 6, 1);
nengel@2 184 else if(block[i*16]) ff_h264_idct_dc_add_c(dst + block_offset[i], block + i*16, stride);
nengel@2 185 }
nengel@2 186 }
nengel@2 187
nengel@2 188 void ff_h264_idct8_add4_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
nengel@2 189 int i;
nengel@2 190 for(i=0; i<16; i+=4){
nengel@2 191 int nnz = nnzc[ scan8[i] ];
nengel@2 192 if(nnz){
nengel@2 193 if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_c(dst + block_offset[i], block + i*16, stride);
nengel@2 194 else ff_h264_idct8_add_c (dst + block_offset[i], block + i*16, stride);
nengel@2 195 }
nengel@2 196 }
nengel@2 197 }
nengel@2 198
nengel@2 199 void ff_h264_idct_add8_c(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
nengel@2 200 int i;
nengel@2 201 for(i=16; i<16+8; i++){
nengel@2 202 if(nnzc[ scan8[i] ])
nengel@2 203 ff_h264_idct_add_c (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
nengel@2 204 else if(block[i*16])
nengel@2 205 ff_h264_idct_dc_add_c(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
nengel@2 206 }
nengel@2 207 }
nengel@2 208
nengel@2 209 /**
nengel@2 210 * IDCT transforms the 16 dc values and dequantizes them.
nengel@2 211 * @param qp quantization parameter
nengel@2 212 */
nengel@2 213 void h264_luma_dc_dequant_idct_c(DCTELEM *block, int qmul){
nengel@2 214 #define stride 16
nengel@2 215 int i;
nengel@2 216 int temp[16]; //FIXME check if this is a good idea
nengel@2 217 static const int x_offset[4]={0, 1*stride, 4* stride, 5*stride};
nengel@2 218 static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride};
nengel@2 219
nengel@2 220 //return;
nengel@2 221 for(i=0; i<4; i++){
nengel@2 222 const int offset= y_offset[i];
nengel@2 223 const int z0= block[offset+stride*0] + block[offset+stride*4];
nengel@2 224 const int z1= block[offset+stride*0] - block[offset+stride*4];
nengel@2 225 const int z2= block[offset+stride*1] - block[offset+stride*5];
nengel@2 226 const int z3= block[offset+stride*1] + block[offset+stride*5];
nengel@2 227
nengel@2 228 temp[4*i+0]= z0+z3;
nengel@2 229 temp[4*i+1]= z1+z2;
nengel@2 230 temp[4*i+2]= z1-z2;
nengel@2 231 temp[4*i+3]= z0-z3;
nengel@2 232 }
nengel@2 233
nengel@2 234 for(i=0; i<4; i++){
nengel@2 235 const int offset= x_offset[i];
nengel@2 236 const int z0= temp[4*0+i] + temp[4*2+i];
nengel@2 237 const int z1= temp[4*0+i] - temp[4*2+i];
nengel@2 238 const int z2= temp[4*1+i] - temp[4*3+i];
nengel@2 239 const int z3= temp[4*1+i] + temp[4*3+i];
nengel@2 240
nengel@2 241 block[stride*0 +offset]= ((((z0 + z3)*qmul + 128 ) >> 8)); //FIXME think about merging this into decode_residual
nengel@2 242 block[stride*2 +offset]= ((((z1 + z2)*qmul + 128 ) >> 8));
nengel@2 243 block[stride*8 +offset]= ((((z1 - z2)*qmul + 128 ) >> 8));
nengel@2 244 block[stride*10+offset]= ((((z0 - z3)*qmul + 128 ) >> 8));
nengel@2 245 }
nengel@2 246 }
nengel@2 247
nengel@2 248 #undef xStride
nengel@2 249 #undef stride
nengel@2 250
nengel@2 251 void chroma_dc_dequant_idct_c(DCTELEM *block, int qmul){
nengel@2 252 const int stride= 16*2;
nengel@2 253 const int xStride= 16;
nengel@2 254 int a,b,c,d,e;
nengel@2 255
nengel@2 256 a= block[stride*0 + xStride*0];
nengel@2 257 b= block[stride*0 + xStride*1];
nengel@2 258 c= block[stride*1 + xStride*0];
nengel@2 259 d= block[stride*1 + xStride*1];
nengel@2 260
nengel@2 261 e= a-b;
nengel@2 262 a= a+b;
nengel@2 263 b= c-d;
nengel@2 264 c= c+d;
nengel@2 265
nengel@2 266 block[stride*0 + xStride*0]= ((a+c)*qmul) >> 7;
nengel@2 267 block[stride*0 + xStride*1]= ((e+b)*qmul) >> 7;
nengel@2 268 block[stride*1 + xStride*0]= ((a-c)*qmul) >> 7;
nengel@2 269 block[stride*1 + xStride*1]= ((e-b)*qmul) >> 7;
nengel@2 270 }