Parent Directory
| 
Revision Log
Revision 851 - (view) (download)
| 1 : | edgomez | 851 | /****************************************************************************** |
| 2 : | * * | ||
| 3 : | * This file is part of XviD, a free MPEG-4 video encoder/decoder * | ||
| 4 : | * * | ||
| 5 : | * XviD is an implementation of a part of one or more MPEG-4 Video tools * | ||
| 6 : | * as specified in ISO/IEC 14496-2 standard. Those intending to use this * | ||
| 7 : | * software module in hardware or software products are advised that its * | ||
| 8 : | * use may infringe existing patents or copyrights, and any such use * | ||
| 9 : | * would be at such party's own risk. The original developer of this * | ||
| 10 : | * software module and his/her company, and subsequent editors and their * | ||
| 11 : | * companies, will have no liability for use of this software or * | ||
| 12 : | * modifications or derivatives thereof. * | ||
| 13 : | * * | ||
| 14 : | * XviD is free software; you can redistribute it and/or modify it * | ||
| 15 : | * under the terms of the GNU General Public License as published by * | ||
| 16 : | * the Free Software Foundation; either version 2 of the License, or * | ||
| 17 : | * (at your option) any later version. * | ||
| 18 : | * * | ||
| 19 : | * XviD is distributed in the hope that it will be useful, but * | ||
| 20 : | * WITHOUT ANY WARRANTY; without even the implied warranty of * | ||
| 21 : | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * | ||
| 22 : | * GNU General Public License for more details. * | ||
| 23 : | * * | ||
| 24 : | * You should have received a copy of the GNU General Public License * | ||
| 25 : | * along with this program; if not, write to the Free Software * | ||
| 26 : | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * | ||
| 27 : | * * | ||
| 28 : | ******************************************************************************/ | ||
| 29 : | Isibaar | 3 | |
| 30 : | edgomez | 851 | /****************************************************************************** |
| 31 : | * * | ||
| 32 : | * mbtransquant.c * | ||
| 33 : | * * | ||
| 34 : | * Copyright (C) 2001 - Peter Ross <pross@cs.rmit.edu.au> * | ||
| 35 : | * Copyright (C) 2001 - Michael Militzer <isibaar@xvid.org> * | ||
| 36 : | * * | ||
| 37 : | * For more information visit the XviD homepage: http://www.xvid.org * | ||
| 38 : | * * | ||
| 39 : | ******************************************************************************/ | ||
| 40 : | |||
| 41 : | /****************************************************************************** | ||
| 42 : | * * | ||
| 43 : | * Revision history: * | ||
| 44 : | * * | ||
| 45 : | * 29.03.2002 interlacing speedup - used transfer strides instead of * | ||
| 46 : | * manual field-to-frame conversion * | ||
| 47 : | * 26.03.2002 interlacing support - moved transfers outside loops * | ||
| 48 : | * 22.12.2001 get_dc_scaler() moved to common.h * | ||
| 49 : | * 19.11.2001 introduced coefficient thresholding (Isibaar) * | ||
| 50 : | * 17.11.2001 initial version * | ||
| 51 : | * * | ||
| 52 : | ******************************************************************************/ | ||
| 53 : | |||
| 54 : | edgomez | 78 | #include <string.h> |
| 55 : | |||
| 56 : | Isibaar | 3 | #include "../portab.h" |
| 57 : | #include "mbfunctions.h" | ||
| 58 : | |||
| 59 : | #include "../global.h" | ||
| 60 : | #include "mem_transfer.h" | ||
| 61 : | #include "timer.h" | ||
| 62 : | #include "../dct/fdct.h" | ||
| 63 : | #include "../dct/idct.h" | ||
| 64 : | #include "../quant/quant_mpeg4.h" | ||
| 65 : | #include "../quant/quant_h263.h" | ||
| 66 : | #include "../encoder.h" | ||
| 67 : | |||
| 68 : | edgomez | 851 | #include "../image/reduced.h" |
| 69 : | Isibaar | 3 | |
| 70 : | edgomez | 851 | MBFIELDTEST_PTR MBFieldTest; |
| 71 : | Isibaar | 3 | |
| 72 : | edgomez | 851 | #define TOOSMALL_LIMIT 1 /* skip blocks having a coefficient sum below this value */ |
| 73 : | Isibaar | 3 | |
| 74 : | edgomez | 195 | void |
| 75 : | MBTransQuantIntra(const MBParam * pParam, | ||
| 76 : | FRAMEINFO * frame, | ||
| 77 : | MACROBLOCK * pMB, | ||
| 78 : | const uint32_t x_pos, | ||
| 79 : | const uint32_t y_pos, | ||
| 80 : | int16_t data[6 * 64], | ||
| 81 : | int16_t qcoeff[6 * 64]) | ||
| 82 : | Isibaar | 3 | { |
| 83 : | edgomez | 78 | |
| 84 : | h | 82 | uint32_t stride = pParam->edged_width; |
| 85 : | uint32_t stride2 = stride / 2; | ||
| 86 : | edgomez | 851 | uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
| 87 : | Isibaar | 3 | uint32_t i; |
| 88 : | suxen_drol | 136 | uint32_t iQuant = frame->quant; |
| 89 : | Isibaar | 3 | uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
| 90 : | edgomez | 195 | IMAGE *pCurrent = &frame->image; |
| 91 : | Isibaar | 3 | |
| 92 : | edgomez | 851 | start_timer(); |
| 93 : | if ((frame->global_flags & XVID_REDUCED)) | ||
| 94 : | { | ||
| 95 : | pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); | ||
| 96 : | pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); | ||
| 97 : | pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); | ||
| 98 : | h | 69 | |
| 99 : | edgomez | 851 | filter_18x18_to_8x8(&data[0 * 64], pY_Cur, stride); |
| 100 : | filter_18x18_to_8x8(&data[1 * 64], pY_Cur + 16, stride); | ||
| 101 : | filter_18x18_to_8x8(&data[2 * 64], pY_Cur + next_block, stride); | ||
| 102 : | filter_18x18_to_8x8(&data[3 * 64], pY_Cur + next_block + 16, stride); | ||
| 103 : | filter_18x18_to_8x8(&data[4 * 64], pU_Cur, stride2); | ||
| 104 : | filter_18x18_to_8x8(&data[5 * 64], pV_Cur, stride2); | ||
| 105 : | }else{ | ||
| 106 : | pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); | ||
| 107 : | pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 108 : | pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 109 : | |||
| 110 : | transfer_8to16copy(&data[0 * 64], pY_Cur, stride); | ||
| 111 : | transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); | ||
| 112 : | transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); | ||
| 113 : | transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); | ||
| 114 : | transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); | ||
| 115 : | transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); | ||
| 116 : | } | ||
| 117 : | h | 69 | stop_transfer_timer(); |
| 118 : | |||
| 119 : | edgomez | 851 | /* XXX: rrv+interlacing is buggy */ |
| 120 : | h | 69 | start_timer(); |
| 121 : | pMB->field_dct = 0; | ||
| 122 : | h | 390 | if ((frame->global_flags & XVID_INTERLACING) && |
| 123 : | (x_pos>0) && (x_pos<pParam->mb_width-1) && | ||
| 124 : | (y_pos>0) && (y_pos<pParam->mb_height-1)) { | ||
| 125 : | h | 69 | pMB->field_dct = MBDecideFieldDCT(data); |
| 126 : | } | ||
| 127 : | stop_interlacing_timer(); | ||
| 128 : | |||
| 129 : | edgomez | 195 | for (i = 0; i < 6; i++) { |
| 130 : | Isibaar | 3 | uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
| 131 : | |||
| 132 : | start_timer(); | ||
| 133 : | edgomez | 195 | fdct(&data[i * 64]); |
| 134 : | Isibaar | 3 | stop_dct_timer(); |
| 135 : | |||
| 136 : | edgomez | 195 | if (pParam->m_quant_type == H263_QUANT) { |
| 137 : | Isibaar | 3 | start_timer(); |
| 138 : | edgomez | 195 | quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
| 139 : | Isibaar | 3 | stop_quant_timer(); |
| 140 : | edgomez | 195 | } else { |
| 141 : | Isibaar | 3 | start_timer(); |
| 142 : | edgomez | 195 | quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
| 143 : | Isibaar | 3 | stop_quant_timer(); |
| 144 : | edgomez | 851 | } |
| 145 : | Isibaar | 3 | |
| 146 : | edgomez | 851 | /* speedup: dont decode when encoding only ivops */ |
| 147 : | if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) | ||
| 148 : | { | ||
| 149 : | if (pParam->m_quant_type == H263_QUANT) { | ||
| 150 : | start_timer(); | ||
| 151 : | dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); | ||
| 152 : | stop_iquant_timer(); | ||
| 153 : | } else { | ||
| 154 : | start_timer(); | ||
| 155 : | dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); | ||
| 156 : | stop_iquant_timer(); | ||
| 157 : | } | ||
| 158 : | |||
| 159 : | Isibaar | 3 | start_timer(); |
| 160 : | edgomez | 851 | idct(&data[i * 64]); |
| 161 : | stop_idct_timer(); | ||
| 162 : | Isibaar | 3 | } |
| 163 : | edgomez | 851 | } |
| 164 : | Isibaar | 3 | |
| 165 : | edgomez | 851 | /* speedup: dont decode when encoding only ivops */ |
| 166 : | if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) | ||
| 167 : | { | ||
| 168 : | |||
| 169 : | if (pMB->field_dct) { | ||
| 170 : | next_block = stride; | ||
| 171 : | stride *= 2; | ||
| 172 : | } | ||
| 173 : | |||
| 174 : | Isibaar | 3 | start_timer(); |
| 175 : | edgomez | 851 | if ((frame->global_flags & XVID_REDUCED)) |
| 176 : | { | ||
| 177 : | copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); | ||
| 178 : | copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); | ||
| 179 : | copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); | ||
| 180 : | copy_upsampled_8x8_16to8(pY_Cur + next_block + 16, &data[3 * 64], stride); | ||
| 181 : | copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); | ||
| 182 : | copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); | ||
| 183 : | Isibaar | 3 | |
| 184 : | edgomez | 851 | }else{ |
| 185 : | transfer_16to8copy(pY_Cur, &data[0 * 64], stride); | ||
| 186 : | transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); | ||
| 187 : | transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); | ||
| 188 : | transfer_16to8copy(pY_Cur + next_block + 8, &data[3 * 64], stride); | ||
| 189 : | transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); | ||
| 190 : | transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); | ||
| 191 : | } | ||
| 192 : | stop_transfer_timer(); | ||
| 193 : | h | 69 | } |
| 194 : | |||
| 195 : | Isibaar | 3 | } |
| 196 : | |||
| 197 : | |||
| 198 : | edgomez | 195 | uint8_t |
| 199 : | MBTransQuantInter(const MBParam * pParam, | ||
| 200 : | FRAMEINFO * frame, | ||
| 201 : | MACROBLOCK * pMB, | ||
| 202 : | const uint32_t x_pos, | ||
| 203 : | const uint32_t y_pos, | ||
| 204 : | int16_t data[6 * 64], | ||
| 205 : | int16_t qcoeff[6 * 64]) | ||
| 206 : | Isibaar | 3 | { |
| 207 : | edgomez | 78 | |
| 208 : | h | 82 | uint32_t stride = pParam->edged_width; |
| 209 : | uint32_t stride2 = stride / 2; | ||
| 210 : | edgomez | 851 | uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
| 211 : | edgomez | 78 | uint32_t i; |
| 212 : | suxen_drol | 136 | uint32_t iQuant = frame->quant; |
| 213 : | Isibaar | 3 | uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
| 214 : | edgomez | 78 | uint8_t cbp = 0; |
| 215 : | Isibaar | 3 | uint32_t sum; |
| 216 : | edgomez | 195 | IMAGE *pCurrent = &frame->image; |
| 217 : | |||
| 218 : | edgomez | 851 | if ((frame->global_flags & XVID_REDUCED)) |
| 219 : | { | ||
| 220 : | pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); | ||
| 221 : | pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); | ||
| 222 : | pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); | ||
| 223 : | }else{ | ||
| 224 : | pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); | ||
| 225 : | pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 226 : | pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 227 : | } | ||
| 228 : | Isibaar | 3 | |
| 229 : | h | 69 | start_timer(); |
| 230 : | pMB->field_dct = 0; | ||
| 231 : | h | 390 | if ((frame->global_flags & XVID_INTERLACING) && |
| 232 : | (x_pos>0) && (x_pos<pParam->mb_width-1) && | ||
| 233 : | (y_pos>0) && (y_pos<pParam->mb_height-1)) { | ||
| 234 : | h | 69 | pMB->field_dct = MBDecideFieldDCT(data); |
| 235 : | } | ||
| 236 : | stop_interlacing_timer(); | ||
| 237 : | |||
| 238 : | edgomez | 195 | for (i = 0; i < 6; i++) { |
| 239 : | edgomez | 851 | uint32_t increase_limit = (iQuant == 1) ? 1 : 0; |
| 240 : | |||
| 241 : | Isibaar | 3 | /* |
| 242 : | edgomez | 78 | * no need to transfer 8->16-bit |
| 243 : | * (this is performed already in motion compensation) | ||
| 244 : | */ | ||
| 245 : | Isibaar | 3 | start_timer(); |
| 246 : | edgomez | 195 | fdct(&data[i * 64]); |
| 247 : | Isibaar | 3 | stop_dct_timer(); |
| 248 : | |||
| 249 : | edgomez | 195 | if (pParam->m_quant_type == 0) { |
| 250 : | Isibaar | 3 | start_timer(); |
| 251 : | edgomez | 195 | sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
| 252 : | Isibaar | 3 | stop_quant_timer(); |
| 253 : | edgomez | 195 | } else { |
| 254 : | Isibaar | 3 | start_timer(); |
| 255 : | edgomez | 195 | sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
| 256 : | Isibaar | 3 | stop_quant_timer(); |
| 257 : | } | ||
| 258 : | |||
| 259 : | edgomez | 851 | if ((sum >= TOOSMALL_LIMIT + increase_limit) || (qcoeff[i*64] != 0) || |
| 260 : | Isibaar | 375 | (qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) { |
| 261 : | Isibaar | 3 | |
| 262 : | edgomez | 195 | if (pParam->m_quant_type == H263_QUANT) { |
| 263 : | Isibaar | 3 | start_timer(); |
| 264 : | edgomez | 195 | dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
| 265 : | Isibaar | 3 | stop_iquant_timer(); |
| 266 : | edgomez | 195 | } else { |
| 267 : | Isibaar | 3 | start_timer(); |
| 268 : | edgomez | 195 | dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
| 269 : | Isibaar | 3 | stop_iquant_timer(); |
| 270 : | } | ||
| 271 : | |||
| 272 : | cbp |= 1 << (5 - i); | ||
| 273 : | |||
| 274 : | start_timer(); | ||
| 275 : | edgomez | 195 | idct(&data[i * 64]); |
| 276 : | Isibaar | 3 | stop_idct_timer(); |
| 277 : | } | ||
| 278 : | } | ||
| 279 : | h | 69 | |
| 280 : | edgomez | 195 | if (pMB->field_dct) { |
| 281 : | h | 82 | next_block = stride; |
| 282 : | stride *= 2; | ||
| 283 : | h | 69 | } |
| 284 : | |||
| 285 : | start_timer(); | ||
| 286 : | edgomez | 851 | if ((frame->global_flags & XVID_REDUCED)) |
| 287 : | { | ||
| 288 : | if (cbp & 32) | ||
| 289 : | add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); | ||
| 290 : | if (cbp & 16) | ||
| 291 : | add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); | ||
| 292 : | if (cbp & 8) | ||
| 293 : | add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); | ||
| 294 : | if (cbp & 4) | ||
| 295 : | add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); | ||
| 296 : | if (cbp & 2) | ||
| 297 : | add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); | ||
| 298 : | if (cbp & 1) | ||
| 299 : | add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); | ||
| 300 : | }else{ | ||
| 301 : | if (cbp & 32) | ||
| 302 : | transfer_16to8add(pY_Cur, &data[0 * 64], stride); | ||
| 303 : | if (cbp & 16) | ||
| 304 : | transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); | ||
| 305 : | if (cbp & 8) | ||
| 306 : | transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); | ||
| 307 : | if (cbp & 4) | ||
| 308 : | transfer_16to8add(pY_Cur + next_block + 8, &data[3 * 64], stride); | ||
| 309 : | if (cbp & 2) | ||
| 310 : | transfer_16to8add(pU_Cur, &data[4 * 64], stride2); | ||
| 311 : | if (cbp & 1) | ||
| 312 : | transfer_16to8add(pV_Cur, &data[5 * 64], stride2); | ||
| 313 : | } | ||
| 314 : | h | 69 | stop_transfer_timer(); |
| 315 : | |||
| 316 : | edgomez | 78 | return cbp; |
| 317 : | |||
| 318 : | Isibaar | 3 | } |
| 319 : | h | 69 | |
| 320 : | chl | 368 | void |
| 321 : | MBTransQuantIntra2(const MBParam * pParam, | ||
| 322 : | FRAMEINFO * frame, | ||
| 323 : | MACROBLOCK * pMB, | ||
| 324 : | const uint32_t x_pos, | ||
| 325 : | const uint32_t y_pos, | ||
| 326 : | int16_t data[6 * 64], | ||
| 327 : | int16_t qcoeff[6 * 64]) | ||
| 328 : | { | ||
| 329 : | MBTrans(pParam,frame,pMB,x_pos,y_pos,data); | ||
| 330 : | MBfDCT(pParam,frame,pMB,data); | ||
| 331 : | MBQuantIntra(pParam,frame,pMB,data,qcoeff); | ||
| 332 : | MBDeQuantIntra(pParam,frame->quant,data,qcoeff); | ||
| 333 : | MBiDCT(data,0x3F); | ||
| 334 : | MBTransAdd(pParam,frame,pMB,x_pos,y_pos,data,0x3F); | ||
| 335 : | } | ||
| 336 : | h | 69 | |
| 337 : | chl | 368 | |
| 338 : | uint8_t | ||
| 339 : | MBTransQuantInter2(const MBParam * pParam, | ||
| 340 : | FRAMEINFO * frame, | ||
| 341 : | MACROBLOCK * pMB, | ||
| 342 : | const uint32_t x_pos, | ||
| 343 : | const uint32_t y_pos, | ||
| 344 : | int16_t data[6 * 64], | ||
| 345 : | int16_t qcoeff[6 * 64]) | ||
| 346 : | { | ||
| 347 : | uint8_t cbp; | ||
| 348 : | |||
| 349 : | /* there is no MBTrans for Inter block, that's done in motion compensation already */ | ||
| 350 : | |||
| 351 : | MBfDCT(pParam,frame,pMB,data); | ||
| 352 : | cbp = MBQuantInter(pParam,frame->quant,data,qcoeff); | ||
| 353 : | MBDeQuantInter(pParam,frame->quant,data,qcoeff,cbp); | ||
| 354 : | MBiDCT(data,cbp); | ||
| 355 : | MBTransAdd(pParam,frame,pMB,x_pos,y_pos,data,cbp); | ||
| 356 : | |||
| 357 : | return cbp; | ||
| 358 : | } | ||
| 359 : | |||
| 360 : | uint8_t | ||
| 361 : | MBTransQuantInterBVOP(const MBParam * pParam, | ||
| 362 : | FRAMEINFO * frame, | ||
| 363 : | MACROBLOCK * pMB, | ||
| 364 : | int16_t data[6 * 64], | ||
| 365 : | int16_t qcoeff[6 * 64]) | ||
| 366 : | { | ||
| 367 : | uint8_t cbp; | ||
| 368 : | |||
| 369 : | /* there is no MBTrans for Inter block, that's done in motion compensation already */ | ||
| 370 : | |||
| 371 : | MBfDCT(pParam,frame,pMB,data); | ||
| 372 : | cbp = MBQuantInter(pParam,frame->quant,data,qcoeff); | ||
| 373 : | |||
| 374 : | /* we don't have to DeQuant, iDCT and Transfer back data for B-frames */ | ||
| 375 : | |||
| 376 : | return cbp; | ||
| 377 : | } | ||
| 378 : | |||
| 379 : | |||
| 380 : | void | ||
| 381 : | MBfDCT(const MBParam * pParam, | ||
| 382 : | FRAMEINFO * frame, | ||
| 383 : | MACROBLOCK * pMB, | ||
| 384 : | int16_t data[6 * 64]) | ||
| 385 : | { | ||
| 386 : | int i; | ||
| 387 : | |||
| 388 : | start_timer(); | ||
| 389 : | pMB->field_dct = 0; | ||
| 390 : | if ((frame->global_flags & XVID_INTERLACING)) { | ||
| 391 : | pMB->field_dct = MBDecideFieldDCT(data); | ||
| 392 : | } | ||
| 393 : | stop_interlacing_timer(); | ||
| 394 : | |||
| 395 : | for (i = 0; i < 6; i++) { | ||
| 396 : | start_timer(); | ||
| 397 : | fdct(&data[i * 64]); | ||
| 398 : | stop_dct_timer(); | ||
| 399 : | } | ||
| 400 : | } | ||
| 401 : | |||
| 402 : | void | ||
| 403 : | MBQuantDeQuantIntra(const MBParam * pParam, | ||
| 404 : | FRAMEINFO * frame, | ||
| 405 : | MACROBLOCK * pMB, | ||
| 406 : | int16_t qcoeff[6 * 64], | ||
| 407 : | int16_t data[6*64]) | ||
| 408 : | { | ||
| 409 : | int i; | ||
| 410 : | int iQuant = frame->quant; | ||
| 411 : | |||
| 412 : | start_timer(); | ||
| 413 : | pMB->field_dct = 0; | ||
| 414 : | if ((frame->global_flags & XVID_INTERLACING)) { | ||
| 415 : | pMB->field_dct = MBDecideFieldDCT(data); | ||
| 416 : | } | ||
| 417 : | stop_interlacing_timer(); | ||
| 418 : | |||
| 419 : | for (i = 0; i < 6; i++) { | ||
| 420 : | uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); | ||
| 421 : | |||
| 422 : | if (pParam->m_quant_type == H263_QUANT) { | ||
| 423 : | start_timer(); | ||
| 424 : | quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); | ||
| 425 : | stop_quant_timer(); | ||
| 426 : | |||
| 427 : | start_timer(); | ||
| 428 : | dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); | ||
| 429 : | stop_iquant_timer(); | ||
| 430 : | } else { | ||
| 431 : | start_timer(); | ||
| 432 : | quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); | ||
| 433 : | stop_quant_timer(); | ||
| 434 : | |||
| 435 : | start_timer(); | ||
| 436 : | dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); | ||
| 437 : | stop_iquant_timer(); | ||
| 438 : | } | ||
| 439 : | } | ||
| 440 : | } | ||
| 441 : | |||
| 442 : | void | ||
| 443 : | MBQuantIntra(const MBParam * pParam, | ||
| 444 : | FRAMEINFO * frame, | ||
| 445 : | MACROBLOCK *pMB, | ||
| 446 : | edgomez | 851 | int16_t qcoeff[6 * 64], |
| 447 : | int16_t data[6*64]) | ||
| 448 : | chl | 368 | { |
| 449 : | int i; | ||
| 450 : | int iQuant = frame->quant; | ||
| 451 : | |||
| 452 : | start_timer(); | ||
| 453 : | pMB->field_dct = 0; | ||
| 454 : | if ((frame->global_flags & XVID_INTERLACING)) { | ||
| 455 : | pMB->field_dct = MBDecideFieldDCT(data); | ||
| 456 : | } | ||
| 457 : | stop_interlacing_timer(); | ||
| 458 : | |||
| 459 : | for (i = 0; i < 6; i++) { | ||
| 460 : | uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); | ||
| 461 : | |||
| 462 : | if (pParam->m_quant_type == H263_QUANT) { | ||
| 463 : | start_timer(); | ||
| 464 : | quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); | ||
| 465 : | stop_quant_timer(); | ||
| 466 : | } else { | ||
| 467 : | start_timer(); | ||
| 468 : | quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); | ||
| 469 : | stop_quant_timer(); | ||
| 470 : | } | ||
| 471 : | } | ||
| 472 : | } | ||
| 473 : | |||
| 474 : | void | ||
| 475 : | MBDeQuantIntra(const MBParam * pParam, | ||
| 476 : | const int iQuant, | ||
| 477 : | int16_t qcoeff[6 * 64], | ||
| 478 : | int16_t data[6*64]) | ||
| 479 : | { | ||
| 480 : | int i; | ||
| 481 : | |||
| 482 : | for (i = 0; i < 6; i++) { | ||
| 483 : | uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); | ||
| 484 : | |||
| 485 : | if (pParam->m_quant_type == H263_QUANT) { | ||
| 486 : | start_timer(); | ||
| 487 : | dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); | ||
| 488 : | stop_iquant_timer(); | ||
| 489 : | } else { | ||
| 490 : | start_timer(); | ||
| 491 : | dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); | ||
| 492 : | stop_iquant_timer(); | ||
| 493 : | } | ||
| 494 : | } | ||
| 495 : | } | ||
| 496 : | |||
| 497 : | uint8_t | ||
| 498 : | MBQuantInter(const MBParam * pParam, | ||
| 499 : | const int iQuant, | ||
| 500 : | int16_t data[6 * 64], | ||
| 501 : | int16_t qcoeff[6 * 64]) | ||
| 502 : | { | ||
| 503 : | |||
| 504 : | int i; | ||
| 505 : | uint8_t cbp = 0; | ||
| 506 : | int sum; | ||
| 507 : | |||
| 508 : | for (i = 0; i < 6; i++) { | ||
| 509 : | |||
| 510 : | if (pParam->m_quant_type == 0) { | ||
| 511 : | start_timer(); | ||
| 512 : | sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant); | ||
| 513 : | stop_quant_timer(); | ||
| 514 : | } else { | ||
| 515 : | start_timer(); | ||
| 516 : | sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant); | ||
| 517 : | stop_quant_timer(); | ||
| 518 : | } | ||
| 519 : | |||
| 520 : | edgomez | 851 | if (sum >= TOOSMALL_LIMIT) { // skip block ? |
| 521 : | chl | 368 | cbp |= 1 << (5 - i); |
| 522 : | } | ||
| 523 : | } | ||
| 524 : | return cbp; | ||
| 525 : | } | ||
| 526 : | |||
| 527 : | void | ||
| 528 : | MBDeQuantInter( const MBParam * pParam, | ||
| 529 : | const int iQuant, | ||
| 530 : | int16_t data[6 * 64], | ||
| 531 : | int16_t qcoeff[6 * 64], | ||
| 532 : | const uint8_t cbp) | ||
| 533 : | { | ||
| 534 : | int i; | ||
| 535 : | |||
| 536 : | for (i = 0; i < 6; i++) { | ||
| 537 : | if (cbp & (1 << (5 - i))) | ||
| 538 : | { | ||
| 539 : | if (pParam->m_quant_type == H263_QUANT) { | ||
| 540 : | start_timer(); | ||
| 541 : | dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); | ||
| 542 : | stop_iquant_timer(); | ||
| 543 : | } else { | ||
| 544 : | start_timer(); | ||
| 545 : | dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); | ||
| 546 : | stop_iquant_timer(); | ||
| 547 : | } | ||
| 548 : | } | ||
| 549 : | } | ||
| 550 : | } | ||
| 551 : | |||
| 552 : | void | ||
| 553 : | MBiDCT( int16_t data[6 * 64], | ||
| 554 : | const uint8_t cbp) | ||
| 555 : | { | ||
| 556 : | int i; | ||
| 557 : | |||
| 558 : | for (i = 0; i < 6; i++) { | ||
| 559 : | if (cbp & (1 << (5 - i))) | ||
| 560 : | { | ||
| 561 : | start_timer(); | ||
| 562 : | idct(&data[i * 64]); | ||
| 563 : | stop_idct_timer(); | ||
| 564 : | |||
| 565 : | } | ||
| 566 : | } | ||
| 567 : | } | ||
| 568 : | |||
| 569 : | |||
| 570 : | void | ||
| 571 : | MBTrans(const MBParam * pParam, | ||
| 572 : | FRAMEINFO * frame, | ||
| 573 : | MACROBLOCK * pMB, | ||
| 574 : | const uint32_t x_pos, | ||
| 575 : | const uint32_t y_pos, | ||
| 576 : | int16_t data[6 * 64]) | ||
| 577 : | { | ||
| 578 : | uint32_t stride = pParam->edged_width; | ||
| 579 : | uint32_t stride2 = stride / 2; | ||
| 580 : | uint32_t next_block = stride * 8; | ||
| 581 : | uint8_t *pY_Cur, *pU_Cur, *pV_Cur; | ||
| 582 : | IMAGE *pCurrent = &frame->image; | ||
| 583 : | |||
| 584 : | pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); | ||
| 585 : | pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 586 : | pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 587 : | |||
| 588 : | start_timer(); | ||
| 589 : | transfer_8to16copy(&data[0 * 64], pY_Cur, stride); | ||
| 590 : | transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); | ||
| 591 : | transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); | ||
| 592 : | transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); | ||
| 593 : | transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); | ||
| 594 : | transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); | ||
| 595 : | stop_transfer_timer(); | ||
| 596 : | } | ||
| 597 : | |||
| 598 : | void | ||
| 599 : | MBTransAdd(const MBParam * pParam, | ||
| 600 : | FRAMEINFO * frame, | ||
| 601 : | MACROBLOCK * pMB, | ||
| 602 : | const uint32_t x_pos, | ||
| 603 : | const uint32_t y_pos, | ||
| 604 : | int16_t data[6 * 64], | ||
| 605 : | const uint8_t cbp) | ||
| 606 : | { | ||
| 607 : | uint8_t *pY_Cur, *pU_Cur, *pV_Cur; | ||
| 608 : | uint32_t stride = pParam->edged_width; | ||
| 609 : | uint32_t stride2 = stride / 2; | ||
| 610 : | uint32_t next_block = stride * 8; | ||
| 611 : | IMAGE *pCurrent = &frame->image; | ||
| 612 : | |||
| 613 : | pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); | ||
| 614 : | pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 615 : | pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 616 : | |||
| 617 : | if (pMB->field_dct) { | ||
| 618 : | next_block = stride; | ||
| 619 : | stride *= 2; | ||
| 620 : | } | ||
| 621 : | |||
| 622 : | start_timer(); | ||
| 623 : | if (cbp & 32) | ||
| 624 : | transfer_16to8add(pY_Cur, &data[0 * 64], stride); | ||
| 625 : | if (cbp & 16) | ||
| 626 : | transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); | ||
| 627 : | if (cbp & 8) | ||
| 628 : | transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); | ||
| 629 : | if (cbp & 4) | ||
| 630 : | transfer_16to8add(pY_Cur + next_block + 8, &data[3 * 64], stride); | ||
| 631 : | if (cbp & 2) | ||
| 632 : | transfer_16to8add(pU_Cur, &data[4 * 64], stride2); | ||
| 633 : | if (cbp & 1) | ||
| 634 : | transfer_16to8add(pV_Cur, &data[5 * 64], stride2); | ||
| 635 : | stop_transfer_timer(); | ||
| 636 : | } | ||
| 637 : | |||
| 638 : | |||
| 639 : | |||
| 640 : | edgomez | 851 | /* permute block and return field dct choice */ |
| 641 : | h | 69 | |
| 642 : | |||
| 643 : | edgomez | 195 | uint32_t |
| 644 : | MBDecideFieldDCT(int16_t data[6 * 64]) | ||
| 645 : | h | 69 | { |
| 646 : | edgomez | 851 | uint32_t field = MBFieldTest(data); |
| 647 : | h | 69 | |
| 648 : | edgomez | 851 | if (field) { |
| 649 : | MBFrameToField(data); | ||
| 650 : | } | ||
| 651 : | |||
| 652 : | return field; | ||
| 653 : | } | ||
| 654 : | |||
| 655 : | |||
| 656 : | /* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ | ||
| 657 : | |||
| 658 : | uint32_t | ||
| 659 : | MBFieldTest_c(int16_t data[6 * 64]) | ||
| 660 : | { | ||
| 661 : | edgomez | 195 | const uint8_t blocks[] = |
| 662 : | { 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 }; | ||
| 663 : | const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 }; | ||
| 664 : | edgomez | 78 | |
| 665 : | h | 69 | int frame = 0, field = 0; |
| 666 : | int i, j; | ||
| 667 : | |||
| 668 : | edgomez | 195 | for (i = 0; i < 7; ++i) { |
| 669 : | for (j = 0; j < 8; ++j) { | ||
| 670 : | frame += | ||
| 671 : | ABS(data[0 * 64 + (i + 1) * 8 + j] - data[0 * 64 + i * 8 + j]); | ||
| 672 : | frame += | ||
| 673 : | ABS(data[1 * 64 + (i + 1) * 8 + j] - data[1 * 64 + i * 8 + j]); | ||
| 674 : | frame += | ||
| 675 : | ABS(data[2 * 64 + (i + 1) * 8 + j] - data[2 * 64 + i * 8 + j]); | ||
| 676 : | frame += | ||
| 677 : | ABS(data[3 * 64 + (i + 1) * 8 + j] - data[3 * 64 + i * 8 + j]); | ||
| 678 : | h | 69 | |
| 679 : | edgomez | 195 | field += |
| 680 : | ABS(data[blocks[i + 1] + lines[i + 1] + j] - | ||
| 681 : | data[blocks[i] + lines[i] + j]); | ||
| 682 : | field += | ||
| 683 : | ABS(data[blocks[i + 1] + lines[i + 1] + 8 + j] - | ||
| 684 : | data[blocks[i] + lines[i] + 8 + j]); | ||
| 685 : | field += | ||
| 686 : | ABS(data[blocks[i + 1] + 64 + lines[i + 1] + j] - | ||
| 687 : | data[blocks[i] + 64 + lines[i] + j]); | ||
| 688 : | field += | ||
| 689 : | ABS(data[blocks[i + 1] + 64 + lines[i + 1] + 8 + j] - | ||
| 690 : | data[blocks[i] + 64 + lines[i] + 8 + j]); | ||
| 691 : | h | 69 | } |
| 692 : | } | ||
| 693 : | |||
| 694 : | edgomez | 851 | return (frame >= (field + 350)); |
| 695 : | h | 69 | } |
| 696 : | |||
| 697 : | |||
| 698 : | /* deinterlace Y blocks vertically */ | ||
| 699 : | |||
| 700 : | #define MOVLINE(X,Y) memcpy(X, Y, sizeof(tmp)) | ||
| 701 : | edgomez | 78 | #define LINE(X,Y) &data[X*64 + Y*8] |
| 702 : | h | 69 | |
| 703 : | edgomez | 195 | void |
| 704 : | MBFrameToField(int16_t data[6 * 64]) | ||
| 705 : | h | 69 | { |
| 706 : | int16_t tmp[8]; | ||
| 707 : | |||
| 708 : | /* left blocks */ | ||
| 709 : | |||
| 710 : | edgomez | 851 | // 1=2, 2=4, 4=8, 8=1 |
| 711 : | edgomez | 195 | MOVLINE(tmp, LINE(0, 1)); |
| 712 : | MOVLINE(LINE(0, 1), LINE(0, 2)); | ||
| 713 : | MOVLINE(LINE(0, 2), LINE(0, 4)); | ||
| 714 : | MOVLINE(LINE(0, 4), LINE(2, 0)); | ||
| 715 : | MOVLINE(LINE(2, 0), tmp); | ||
| 716 : | h | 69 | |
| 717 : | edgomez | 851 | // 3=6, 6=12, 12=9, 9=3 |
| 718 : | edgomez | 195 | MOVLINE(tmp, LINE(0, 3)); |
| 719 : | MOVLINE(LINE(0, 3), LINE(0, 6)); | ||
| 720 : | MOVLINE(LINE(0, 6), LINE(2, 4)); | ||
| 721 : | MOVLINE(LINE(2, 4), LINE(2, 1)); | ||
| 722 : | MOVLINE(LINE(2, 1), tmp); | ||
| 723 : | h | 69 | |
| 724 : | edgomez | 851 | // 5=10, 10=5 |
| 725 : | edgomez | 195 | MOVLINE(tmp, LINE(0, 5)); |
| 726 : | MOVLINE(LINE(0, 5), LINE(2, 2)); | ||
| 727 : | MOVLINE(LINE(2, 2), tmp); | ||
| 728 : | h | 69 | |
| 729 : | edgomez | 851 | // 7=14, 14=13, 13=11, 11=7 |
| 730 : | edgomez | 195 | MOVLINE(tmp, LINE(0, 7)); |
| 731 : | MOVLINE(LINE(0, 7), LINE(2, 6)); | ||
| 732 : | MOVLINE(LINE(2, 6), LINE(2, 5)); | ||
| 733 : | MOVLINE(LINE(2, 5), LINE(2, 3)); | ||
| 734 : | MOVLINE(LINE(2, 3), tmp); | ||
| 735 : | h | 69 | |
| 736 : | /* right blocks */ | ||
| 737 : | |||
| 738 : | edgomez | 851 | // 1=2, 2=4, 4=8, 8=1 |
| 739 : | edgomez | 195 | MOVLINE(tmp, LINE(1, 1)); |
| 740 : | MOVLINE(LINE(1, 1), LINE(1, 2)); | ||
| 741 : | MOVLINE(LINE(1, 2), LINE(1, 4)); | ||
| 742 : | MOVLINE(LINE(1, 4), LINE(3, 0)); | ||
| 743 : | MOVLINE(LINE(3, 0), tmp); | ||
| 744 : | h | 69 | |
| 745 : | edgomez | 851 | // 3=6, 6=12, 12=9, 9=3 |
| 746 : | edgomez | 195 | MOVLINE(tmp, LINE(1, 3)); |
| 747 : | MOVLINE(LINE(1, 3), LINE(1, 6)); | ||
| 748 : | MOVLINE(LINE(1, 6), LINE(3, 4)); | ||
| 749 : | MOVLINE(LINE(3, 4), LINE(3, 1)); | ||
| 750 : | MOVLINE(LINE(3, 1), tmp); | ||
| 751 : | h | 69 | |
| 752 : | edgomez | 851 | // 5=10, 10=5 |
| 753 : | edgomez | 195 | MOVLINE(tmp, LINE(1, 5)); |
| 754 : | MOVLINE(LINE(1, 5), LINE(3, 2)); | ||
| 755 : | MOVLINE(LINE(3, 2), tmp); | ||
| 756 : | h | 69 | |
| 757 : | edgomez | 851 | // 7=14, 14=13, 13=11, 11=7 |
| 758 : | edgomez | 195 | MOVLINE(tmp, LINE(1, 7)); |
| 759 : | MOVLINE(LINE(1, 7), LINE(3, 6)); | ||
| 760 : | MOVLINE(LINE(3, 6), LINE(3, 5)); | ||
| 761 : | MOVLINE(LINE(3, 5), LINE(3, 3)); | ||
| 762 : | MOVLINE(LINE(3, 3), tmp); | ||
| 763 : | h | 69 | } |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |