Mercurial > libavcodec.hg

comparison cavsdsp.c @ 3412:26dc0057a8f8 libavcodec

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
add rounding bias earlier in the idct, avoids many adds
author michael
date Tue, 04 Jul 2006 22:43:35 +0000
parents adccbf4a1040
children 419409926166
comparison
equal deleted inserted replaced
3411:5d17415fb535 3412:26dc0057a8f8
182 static void cavs_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride) { 182 static void cavs_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride) {
183 int i; 183 int i;
184 DCTELEM (*src)[8] = (DCTELEM(*)[8])block; 184 DCTELEM (*src)[8] = (DCTELEM(*)[8])block;
185 uint8_t *cm = cropTbl + MAX_NEG_CROP; 185 uint8_t *cm = cropTbl + MAX_NEG_CROP;
186 186
187 src[0][0] += 8;
188
187 for( i = 0; i < 8; i++ ) { 189 for( i = 0; i < 8; i++ ) {
188 const int a0 = 3*src[i][1] - (src[i][7]<<1); 190 const int a0 = 3*src[i][1] - (src[i][7]<<1);
189 const int a1 = 3*src[i][3] + (src[i][5]<<1); 191 const int a1 = 3*src[i][3] + (src[i][5]<<1);
190 const int a2 = (src[i][3]<<1) - 3*src[i][5]; 192 const int a2 = (src[i][3]<<1) - 3*src[i][5];
191 const int a3 = (src[i][1]<<1) + 3*src[i][7]; 193 const int a3 = (src[i][1]<<1) + 3*src[i][7];
195 const int b6 = ((a3 - a2 - a1)<<1) + a3; 197 const int b6 = ((a3 - a2 - a1)<<1) + a3;
196 const int b7 = ((a0 - a2 - a3)<<1) - a2; 198 const int b7 = ((a0 - a2 - a3)<<1) - a2;
197 199
198 const int a7 = (src[i][2]<<2) - 10*src[i][6]; 200 const int a7 = (src[i][2]<<2) - 10*src[i][6];
199 const int a6 = (src[i][6]<<2) + 10*src[i][2]; 201 const int a6 = (src[i][6]<<2) + 10*src[i][2];
200 const int a5 = (src[i][0] - src[i][4]) << 3; 202 const int a5 = ((src[i][0] - src[i][4]) << 3) + 4;
201 const int a4 = (src[i][0] + src[i][4]) << 3; 203 const int a4 = ((src[i][0] + src[i][4]) << 3) + 4;
202 204
203 const int b0 = a4 + a6; 205 const int b0 = a4 + a6;
204 const int b1 = a5 + a7; 206 const int b1 = a5 + a7;
205 const int b2 = a5 - a7; 207 const int b2 = a5 - a7;
206 const int b3 = a4 - a6; 208 const int b3 = a4 - a6;
207 209
208 src[i][0] = (b0 + b4 + 4) >> 3; 210 src[i][0] = (b0 + b4) >> 3;
209 src[i][1] = (b1 + b5 + 4) >> 3; 211 src[i][1] = (b1 + b5) >> 3;
210 src[i][2] = (b2 + b6 + 4) >> 3; 212 src[i][2] = (b2 + b6) >> 3;
211 src[i][3] = (b3 + b7 + 4) >> 3; 213 src[i][3] = (b3 + b7) >> 3;
212 src[i][4] = (b3 - b7 + 4) >> 3; 214 src[i][4] = (b3 - b7) >> 3;
213 src[i][5] = (b2 - b6 + 4) >> 3; 215 src[i][5] = (b2 - b6) >> 3;
214 src[i][6] = (b1 - b5 + 4) >> 3; 216 src[i][6] = (b1 - b5) >> 3;
215 src[i][7] = (b0 - b4 + 4) >> 3; 217 src[i][7] = (b0 - b4) >> 3;
216 } 218 }
217 for( i = 0; i < 8; i++ ) { 219 for( i = 0; i < 8; i++ ) {
218 const int a0 = 3*src[1][i] - (src[7][i]<<1); 220 const int a0 = 3*src[1][i] - (src[7][i]<<1);
219 const int a1 = 3*src[3][i] + (src[5][i]<<1); 221 const int a1 = 3*src[3][i] + (src[5][i]<<1);
220 const int a2 = (src[3][i]<<1) - 3*src[5][i]; 222 const int a2 = (src[3][i]<<1) - 3*src[5][i];
233 const int b0 = a4 + a6; 235 const int b0 = a4 + a6;
234 const int b1 = a5 + a7; 236 const int b1 = a5 + a7;
235 const int b2 = a5 - a7; 237 const int b2 = a5 - a7;
236 const int b3 = a4 - a6; 238 const int b3 = a4 - a6;
237 239
238 dst[i + 0*stride] = cm[ dst[i + 0*stride] + ((b0 + b4 + 64) >> 7)]; 240 dst[i + 0*stride] = cm[ dst[i + 0*stride] + ((b0 + b4) >> 7)];
239 dst[i + 1*stride] = cm[ dst[i + 1*stride] + ((b1 + b5 + 64) >> 7)]; 241 dst[i + 1*stride] = cm[ dst[i + 1*stride] + ((b1 + b5) >> 7)];
240 dst[i + 2*stride] = cm[ dst[i + 2*stride] + ((b2 + b6 + 64) >> 7)]; 242 dst[i + 2*stride] = cm[ dst[i + 2*stride] + ((b2 + b6) >> 7)];
241 dst[i + 3*stride] = cm[ dst[i + 3*stride] + ((b3 + b7 + 64) >> 7)]; 243 dst[i + 3*stride] = cm[ dst[i + 3*stride] + ((b3 + b7) >> 7)];
242 dst[i + 4*stride] = cm[ dst[i + 4*stride] + ((b3 - b7 + 64) >> 7)]; 244 dst[i + 4*stride] = cm[ dst[i + 4*stride] + ((b3 - b7) >> 7)];
243 dst[i + 5*stride] = cm[ dst[i + 5*stride] + ((b2 - b6 + 64) >> 7)]; 245 dst[i + 5*stride] = cm[ dst[i + 5*stride] + ((b2 - b6) >> 7)];
244 dst[i + 6*stride] = cm[ dst[i + 6*stride] + ((b1 - b5 + 64) >> 7)]; 246 dst[i + 6*stride] = cm[ dst[i + 6*stride] + ((b1 - b5) >> 7)];
245 dst[i + 7*stride] = cm[ dst[i + 7*stride] + ((b0 - b4 + 64) >> 7)]; 247 dst[i + 7*stride] = cm[ dst[i + 7*stride] + ((b0 - b4) >> 7)];
246 } 248 }
247 } 249 }
248 250
249 /***************************************************************************** 251 /*****************************************************************************
250 * 252 *