Mercurial > libavcodec.hg
comparison vc1dsp.c @ 5997:90de28dfd8d6 libavcodec
Switch VC-1 decoder to output decoded residual immediately.
author | kostya |
---|---|
date | Sat, 08 Dec 2007 10:41:18 +0000 |
parents | 90d90aecc83c |
children | 953c8efd5298 |
comparison
equal
deleted
inserted
replaced
5996:9319837c5d02 | 5997:90de28dfd8d6 |
---|---|
150 } | 150 } |
151 } | 151 } |
152 | 152 |
153 /** Do inverse transform on 8x4 part of block | 153 /** Do inverse transform on 8x4 part of block |
154 */ | 154 */ |
155 static void vc1_inv_trans_8x4_c(DCTELEM block[64], int n) | 155 static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, DCTELEM *block) |
156 { | 156 { |
157 int i; | 157 int i; |
158 register int t1,t2,t3,t4,t5,t6,t7,t8; | 158 register int t1,t2,t3,t4,t5,t6,t7,t8; |
159 DCTELEM *src, *dst; | 159 DCTELEM *src, *dst; |
160 int off; | 160 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
161 | 161 |
162 off = n * 32; | 162 src = block; |
163 src = block + off; | 163 dst = block; |
164 dst = block + off; | |
165 for(i = 0; i < 4; i++){ | 164 for(i = 0; i < 4; i++){ |
166 t1 = 12 * (src[0] + src[4]); | 165 t1 = 12 * (src[0] + src[4]); |
167 t2 = 12 * (src[0] - src[4]); | 166 t2 = 12 * (src[0] - src[4]); |
168 t3 = 16 * src[2] + 6 * src[6]; | 167 t3 = 16 * src[2] + 6 * src[6]; |
169 t4 = 6 * src[2] - 16 * src[6]; | 168 t4 = 6 * src[2] - 16 * src[6]; |
189 | 188 |
190 src += 8; | 189 src += 8; |
191 dst += 8; | 190 dst += 8; |
192 } | 191 } |
193 | 192 |
194 src = block + off; | 193 src = block; |
195 dst = block + off; | |
196 for(i = 0; i < 8; i++){ | 194 for(i = 0; i < 8; i++){ |
197 t1 = 17 * (src[ 0] + src[16]); | 195 t1 = 17 * (src[ 0] + src[16]); |
198 t2 = 17 * (src[ 0] - src[16]); | 196 t2 = 17 * (src[ 0] - src[16]); |
199 t3 = 22 * src[ 8]; | 197 t3 = 22 * src[ 8]; |
200 t4 = 22 * src[24]; | 198 t4 = 22 * src[24]; |
201 t5 = 10 * src[ 8]; | 199 t5 = 10 * src[ 8]; |
202 t6 = 10 * src[24]; | 200 t6 = 10 * src[24]; |
203 | 201 |
204 dst[ 0] = (t1 + t3 + t6 + 64) >> 7; | 202 dest[0*linesize] = cm[dest[0*linesize] + ((t1 + t3 + t6 + 64) >> 7)]; |
205 dst[ 8] = (t2 - t4 + t5 + 64) >> 7; | 203 dest[1*linesize] = cm[dest[1*linesize] + ((t2 - t4 + t5 + 64) >> 7)]; |
206 dst[16] = (t2 + t4 - t5 + 64) >> 7; | 204 dest[2*linesize] = cm[dest[2*linesize] + ((t2 + t4 - t5 + 64) >> 7)]; |
207 dst[24] = (t1 - t3 - t6 + 64) >> 7; | 205 dest[3*linesize] = cm[dest[3*linesize] + ((t1 - t3 - t6 + 64) >> 7)]; |
208 | 206 |
209 src ++; | 207 src ++; |
210 dst ++; | 208 dest++; |
211 } | 209 } |
212 } | 210 } |
213 | 211 |
214 /** Do inverse transform on 4x8 parts of block | 212 /** Do inverse transform on 4x8 parts of block |
215 */ | 213 */ |
216 static void vc1_inv_trans_4x8_c(DCTELEM block[64], int n) | 214 static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block) |
217 { | 215 { |
218 int i; | 216 int i; |
219 register int t1,t2,t3,t4,t5,t6,t7,t8; | 217 register int t1,t2,t3,t4,t5,t6,t7,t8; |
220 DCTELEM *src, *dst; | 218 DCTELEM *src, *dst; |
221 int off; | 219 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
222 | 220 |
223 off = n * 4; | 221 src = block; |
224 src = block + off; | 222 dst = block; |
225 dst = block + off; | |
226 for(i = 0; i < 8; i++){ | 223 for(i = 0; i < 8; i++){ |
227 t1 = 17 * (src[0] + src[2]); | 224 t1 = 17 * (src[0] + src[2]); |
228 t2 = 17 * (src[0] - src[2]); | 225 t2 = 17 * (src[0] - src[2]); |
229 t3 = 22 * src[1]; | 226 t3 = 22 * src[1]; |
230 t4 = 22 * src[3]; | 227 t4 = 22 * src[3]; |
238 | 235 |
239 src += 8; | 236 src += 8; |
240 dst += 8; | 237 dst += 8; |
241 } | 238 } |
242 | 239 |
243 src = block + off; | 240 src = block; |
244 dst = block + off; | |
245 for(i = 0; i < 4; i++){ | 241 for(i = 0; i < 4; i++){ |
246 t1 = 12 * (src[ 0] + src[32]); | 242 t1 = 12 * (src[ 0] + src[32]); |
247 t2 = 12 * (src[ 0] - src[32]); | 243 t2 = 12 * (src[ 0] - src[32]); |
248 t3 = 16 * src[16] + 6 * src[48]; | 244 t3 = 16 * src[16] + 6 * src[48]; |
249 t4 = 6 * src[16] - 16 * src[48]; | 245 t4 = 6 * src[16] - 16 * src[48]; |
256 t1 = 16 * src[ 8] + 15 * src[24] + 9 * src[40] + 4 * src[56]; | 252 t1 = 16 * src[ 8] + 15 * src[24] + 9 * src[40] + 4 * src[56]; |
257 t2 = 15 * src[ 8] - 4 * src[24] - 16 * src[40] - 9 * src[56]; | 253 t2 = 15 * src[ 8] - 4 * src[24] - 16 * src[40] - 9 * src[56]; |
258 t3 = 9 * src[ 8] - 16 * src[24] + 4 * src[40] + 15 * src[56]; | 254 t3 = 9 * src[ 8] - 16 * src[24] + 4 * src[40] + 15 * src[56]; |
259 t4 = 4 * src[ 8] - 9 * src[24] + 15 * src[40] - 16 * src[56]; | 255 t4 = 4 * src[ 8] - 9 * src[24] + 15 * src[40] - 16 * src[56]; |
260 | 256 |
261 dst[ 0] = (t5 + t1 + 64) >> 7; | 257 dest[0*linesize] = cm[dest[0*linesize] + ((t5 + t1 + 64) >> 7)]; |
262 dst[ 8] = (t6 + t2 + 64) >> 7; | 258 dest[1*linesize] = cm[dest[1*linesize] + ((t6 + t2 + 64) >> 7)]; |
263 dst[16] = (t7 + t3 + 64) >> 7; | 259 dest[2*linesize] = cm[dest[2*linesize] + ((t7 + t3 + 64) >> 7)]; |
264 dst[24] = (t8 + t4 + 64) >> 7; | 260 dest[3*linesize] = cm[dest[3*linesize] + ((t8 + t4 + 64) >> 7)]; |
265 dst[32] = (t8 - t4 + 64 + 1) >> 7; | 261 dest[4*linesize] = cm[dest[4*linesize] + ((t8 - t4 + 64 + 1) >> 7)]; |
266 dst[40] = (t7 - t3 + 64 + 1) >> 7; | 262 dest[5*linesize] = cm[dest[5*linesize] + ((t7 - t3 + 64 + 1) >> 7)]; |
267 dst[48] = (t6 - t2 + 64 + 1) >> 7; | 263 dest[6*linesize] = cm[dest[6*linesize] + ((t6 - t2 + 64 + 1) >> 7)]; |
268 dst[56] = (t5 - t1 + 64 + 1) >> 7; | 264 dest[7*linesize] = cm[dest[7*linesize] + ((t5 - t1 + 64 + 1) >> 7)]; |
269 | 265 |
270 src++; | 266 src ++; |
271 dst++; | 267 dest++; |
272 } | 268 } |
273 } | 269 } |
274 | 270 |
275 /** Do inverse transform on 4x4 part of block | 271 /** Do inverse transform on 4x4 part of block |
276 */ | 272 */ |
277 static void vc1_inv_trans_4x4_c(DCTELEM block[64], int n) | 273 static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, DCTELEM *block) |
278 { | 274 { |
279 int i; | 275 int i; |
280 register int t1,t2,t3,t4,t5,t6; | 276 register int t1,t2,t3,t4,t5,t6; |
281 DCTELEM *src, *dst; | 277 DCTELEM *src, *dst; |
282 int off; | 278 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; |
283 | 279 |
284 off = (n&1) * 4 + (n&2) * 16; | 280 src = block; |
285 src = block + off; | 281 dst = block; |
286 dst = block + off; | |
287 for(i = 0; i < 4; i++){ | 282 for(i = 0; i < 4; i++){ |
288 t1 = 17 * (src[0] + src[2]); | 283 t1 = 17 * (src[0] + src[2]); |
289 t2 = 17 * (src[0] - src[2]); | 284 t2 = 17 * (src[0] - src[2]); |
290 t3 = 22 * src[1]; | 285 t3 = 22 * src[1]; |
291 t4 = 22 * src[3]; | 286 t4 = 22 * src[3]; |
299 | 294 |
300 src += 8; | 295 src += 8; |
301 dst += 8; | 296 dst += 8; |
302 } | 297 } |
303 | 298 |
304 src = block + off; | 299 src = block; |
305 dst = block + off; | |
306 for(i = 0; i < 4; i++){ | 300 for(i = 0; i < 4; i++){ |
307 t1 = 17 * (src[ 0] + src[16]); | 301 t1 = 17 * (src[ 0] + src[16]); |
308 t2 = 17 * (src[ 0] - src[16]); | 302 t2 = 17 * (src[ 0] - src[16]); |
309 t3 = 22 * src[ 8]; | 303 t3 = 22 * src[ 8]; |
310 t4 = 22 * src[24]; | 304 t4 = 22 * src[24]; |
311 t5 = 10 * src[ 8]; | 305 t5 = 10 * src[ 8]; |
312 t6 = 10 * src[24]; | 306 t6 = 10 * src[24]; |
313 | 307 |
314 dst[ 0] = (t1 + t3 + t6 + 64) >> 7; | 308 dest[0*linesize] = cm[dest[0*linesize] + ((t1 + t3 + t6 + 64) >> 7)]; |
315 dst[ 8] = (t2 - t4 + t5 + 64) >> 7; | 309 dest[1*linesize] = cm[dest[1*linesize] + ((t2 - t4 + t5 + 64) >> 7)]; |
316 dst[16] = (t2 + t4 - t5 + 64) >> 7; | 310 dest[2*linesize] = cm[dest[2*linesize] + ((t2 + t4 - t5 + 64) >> 7)]; |
317 dst[24] = (t1 - t3 - t6 + 64) >> 7; | 311 dest[3*linesize] = cm[dest[3*linesize] + ((t1 - t3 - t6 + 64) >> 7)]; |
318 | 312 |
319 src ++; | 313 src ++; |
320 dst ++; | 314 dest++; |
321 } | 315 } |
322 } | 316 } |
323 | 317 |
324 /* motion compensation functions */ | 318 /* motion compensation functions */ |
325 /** Filter in case of 2 filters */ | 319 /** Filter in case of 2 filters */ |