--- decoder.c 2002/04/28 22:12:45 1.12 +++ decoder.c 2003/02/11 21:56:31 1.46 @@ -3,17 +3,13 @@ * XVID MPEG-4 VIDEO CODEC * - Decoder main module - * - * This program is an implementation of a part of one or more MPEG-4 - * Video tools as specified in ISO/IEC 14496-2 standard. Those intending - * to use this software module in hardware or software products are - * advised that its use may infringe existing patents or copyrights, and - * any such use would be at such party's own risk. The original - * developer of this software module and his/her company, and subsequent - * editors and their companies, will have no liability for use of this - * software or modifications or derivatives thereof. + * Copyright(C) 2002 MinChen + * 2002 Peter Ross * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by + * This file is part of XviD, a free MPEG-4 video encoder/decoder + * + * XviD is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * @@ -26,29 +22,46 @@ * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * - *************************************************************************/ - -/************************************************************************** + * Under section 8 of the GNU General Public License, the copyright + * holders of XVID explicitly forbid distribution in the following + * countries: + * + * - Japan + * - United States of America * - * History: + * Linking XviD statically or dynamically with other modules is making a + * combined work based on XviD. Thus, the terms and conditions of the + * GNU General Public License cover the whole combination. * - * 22.04.2002 add some B-frame decode support; chenm001 - * 29.03.2002 interlacing fix - compensated block wasn't being used when - * reconstructing blocks, thus artifacts - * interlacing speedup - used transfers to re-interlace - * interlaced decoding should be as fast as progressive now - * 26.03.2002 interlacing support - moved transfers outside decode loop - * 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block - * 22.12.2001 lock based interpolation - * 01.12.2001 inital version; (c)2001 peter ross + * As a special exception, the copyright holders of XviD give you + * permission to link XviD with independent modules that communicate with + * XviD solely through the VFW1.1 and DShow interfaces, regardless of the + * license terms of these independent modules, and to copy and distribute + * the resulting combined work under terms of your choice, provided that + * every copy of the combined work is accompanied by a complete copy of + * the source code of XviD (the version of XviD used to produce the + * combined work), being distributed under the terms of the GNU General + * Public License plus this exception. An independent module is a module + * which is not derived from or based on XviD. * - * $Id: decoder.c,v 1.12 2002/04/28 22:12:45 edgomez Exp $ + * Note that people who make modified versions of XviD are not obligated + * to grant this special exception for their modified versions; it is + * their choice whether to do so. The GNU General Public License gives + * permission to release a modified version without this exception; this + * exception also makes it possible to release a modified version which + * carries forward this exception. + * + * $Id: decoder.c,v 1.46 2003/02/11 21:56:31 edgomez Exp $ * *************************************************************************/ #include #include +#ifdef BFRAMES_DEC_DEBUG + #define BFRAMES_DEC +#endif + #include "xvid.h" #include "portab.h" @@ -72,13 +85,13 @@ #include "image/colorspace.h" #include "utils/mem_align.h" -int decoder_create(XVID_DEC_PARAM * param) +int +decoder_create(XVID_DEC_PARAM * param) { - DECODER * dec; + DECODER *dec; dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); - if (dec == NULL) - { + if (dec == NULL) { return XVID_ERR_MEMORY; } param->handle = dec; @@ -91,47 +104,84 @@ dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; - - if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) - { + dec->low_delay = 0; + + if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { xvid_free(dec); return XVID_ERR_MEMORY; } - if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) - { + if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { + image_destroy(&dec->cur, dec->edged_width, dec->edged_height); + xvid_free(dec); + return XVID_ERR_MEMORY; + } + /* add by chenm001 */ + /* for support B-frame to reference last 2 frame */ + if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { + image_destroy(&dec->cur, dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); + xvid_free(dec); + return XVID_ERR_MEMORY; + } + if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { image_destroy(&dec->cur, dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); xvid_free(dec); return XVID_ERR_MEMORY; } - // add by chenm001 - // for support B-frame to reference last 2 frame - if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) - { + + dec->mbs = + xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, + CACHE_LINE); + if (dec->mbs == NULL) { image_destroy(&dec->cur, dec->edged_width, dec->edged_height); image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); xvid_free(dec); return XVID_ERR_MEMORY; } - dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); - if (dec->mbs == NULL) - { + memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); + + /* add by chenm001 */ + /* for skip MB flag */ + dec->last_mbs = + xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, + CACHE_LINE); + if (dec->last_mbs == NULL) { + xvid_free(dec->mbs); image_destroy(&dec->cur, dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); xvid_free(dec); return XVID_ERR_MEMORY; } + memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); + init_timer(); + /* add by chenm001 */ + /* for support B-frame to save reference frame's time */ + dec->frames = -1; + dec->time = dec->time_base = dec->last_time_base = 0; + return XVID_ERR_OK; } -int decoder_destroy(DECODER * dec) +int +decoder_destroy(DECODER * dec) { + xvid_free(dec->last_mbs); xvid_free(dec->mbs); image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); + image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); image_destroy(&dec->cur, dec->edged_width, dec->edged_height); xvid_free(dec); @@ -141,27 +191,30 @@ -static const int32_t dquant_table[4] = -{ +static const int32_t dquant_table[4] = { -1, -2, 1, 2 }; -// decode an intra macroblock -void decoder_mbintra(DECODER * dec, - MACROBLOCK * pMB, - const uint32_t x_pos, - const uint32_t y_pos, - const uint32_t acpred_flag, - const uint32_t cbp, - Bitstream * bs, - const uint32_t quant, - const uint32_t intra_dc_threshold) + +/* decode an intra macroblock */ + +void +decoder_mbintra(DECODER * dec, + MACROBLOCK * pMB, + const uint32_t x_pos, + const uint32_t y_pos, + const uint32_t acpred_flag, + const uint32_t cbp, + Bitstream * bs, + const uint32_t quant, + const uint32_t intra_dc_threshold, + const unsigned int bound) { DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); - DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); + DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); uint32_t stride = dec->edged_width; uint32_t stride2 = stride / 2; @@ -174,83 +227,77 @@ pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); - memset(block, 0, 6*64*sizeof(int16_t)); // clear + memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ - for (i = 0; i < 6; i++) - { + for (i = 0; i < 6; i++) { uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); int16_t predictors[8]; int start_coeff; start_timer(); - predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i*64], iQuant, iDcScaler, predictors); - if (!acpred_flag) - { + predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], + iQuant, iDcScaler, predictors, bound); + if (!acpred_flag) { pMB->acpred_directions[i] = 0; } stop_prediction_timer(); - if (quant < intra_dc_threshold) - { + if (quant < intra_dc_threshold) { int dc_size; int dc_dif; - dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs); - dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; + dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs); + dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; - if (dc_size > 8) - { - BitstreamSkip(bs, 1); // marker + if (dc_size > 8) { + BitstreamSkip(bs, 1); /* marker */ } - - block[i*64 + 0] = dc_dif; + + block[i * 64 + 0] = dc_dif; start_coeff = 1; - } - else - { + + DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); + } else { start_coeff = 0; } start_timer(); - if (cbp & (1 << (5-i))) // coded + if (cbp & (1 << (5 - i))) /* coded */ { - get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); + get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], + start_coeff); } stop_coding_timer(); start_timer(); - add_acdc(pMB, i, &block[i*64], iDcScaler, predictors); + add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); stop_prediction_timer(); start_timer(); - if (dec->quant_type == 0) - { - dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); - } - else - { - dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); + if (dec->quant_type == 0) { + dequant_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); + } else { + dequant4_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); } stop_iquant_timer(); start_timer(); - idct(&data[i*64]); + idct(&data[i * 64]); stop_idct_timer(); } - if (dec->interlacing && pMB->field_dct) - { + if (dec->interlacing && pMB->field_dct) { next_block = stride; stride *= 2; } start_timer(); - transfer_16to8copy(pY_Cur, &data[0*64], stride); - transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); - transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); - transfer_16to8copy(pY_Cur + 8 + next_block, &data[3*64], stride); - transfer_16to8copy(pU_Cur, &data[4*64], stride2); - transfer_16to8copy(pV_Cur, &data[5*64], stride2); + transfer_16to8copy(pY_Cur, &data[0 * 64], stride); + transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); + transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); + transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); + transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); + transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); stop_transfer_timer(); } @@ -261,23 +308,24 @@ #define SIGN(X) (((X)>0)?1:-1) #define ABS(X) (((X)>0)?(X):-(X)) static const uint32_t roundtab[16] = -{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; + { 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; -// decode an inter macroblock +/* decode an inter macroblock */ -void decoder_mbinter(DECODER * dec, - const MACROBLOCK * pMB, - const uint32_t x_pos, - const uint32_t y_pos, - const uint32_t acpred_flag, - const uint32_t cbp, - Bitstream * bs, - const uint32_t quant, - const uint32_t rounding) +void +decoder_mbinter(DECODER * dec, + const MACROBLOCK * pMB, + const uint32_t x_pos, + const uint32_t y_pos, + const uint32_t acpred_flag, + const uint32_t cbp, + Bitstream * bs, + const uint32_t quant, + const uint32_t rounding) { - DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); + DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); uint32_t stride = dec->edged_width; @@ -292,143 +340,164 @@ pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); - if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) - { + if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { uv_dx = pMB->mvs[0].x; uv_dy = pMB->mvs[0].y; uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; - } - else - { + } else { int sum; sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; - uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); + + uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; - uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); + + uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); } start_timer(); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); - interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); - interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); + interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, + pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); + interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, + pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); + interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, + pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); + interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, + pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); + interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, + uv_dx, uv_dy, stride2, rounding); + interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, + uv_dx, uv_dy, stride2, rounding); stop_comp_timer(); - for (i = 0; i < 6; i++) - { - if (cbp & (1 << (5-i))) // coded + for (i = 0; i < 6; i++) { + if (cbp & (1 << (5 - i))) /* coded */ { - memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear + memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ start_timer(); - get_inter_block(bs, &block[i*64]); + get_inter_block(bs, &block[i * 64]); stop_coding_timer(); start_timer(); - if (dec->quant_type == 0) - { - dequant_inter(&data[i*64], &block[i*64], iQuant); - } - else - { - dequant4_inter(&data[i*64], &block[i*64], iQuant); + if (dec->quant_type == 0) { + dequant_inter(&data[i * 64], &block[i * 64], iQuant); + } else { + dequant4_inter(&data[i * 64], &block[i * 64], iQuant); } stop_iquant_timer(); start_timer(); - idct(&data[i*64]); + idct(&data[i * 64]); stop_idct_timer(); } } - if (dec->interlacing && pMB->field_dct) - { + if (dec->interlacing && pMB->field_dct) { next_block = stride; stride *= 2; } start_timer(); if (cbp & 32) - transfer_16to8add(pY_Cur, &data[0*64], stride); + transfer_16to8add(pY_Cur, &data[0 * 64], stride); if (cbp & 16) - transfer_16to8add(pY_Cur + 8, &data[1*64], stride); + transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); if (cbp & 8) - transfer_16to8add(pY_Cur + next_block, &data[2*64], stride); + transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); if (cbp & 4) - transfer_16to8add(pY_Cur + 8 + next_block, &data[3*64], stride); + transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); if (cbp & 2) - transfer_16to8add(pU_Cur, &data[4*64], stride2); + transfer_16to8add(pU_Cur, &data[4 * 64], stride2); if (cbp & 1) - transfer_16to8add(pV_Cur, &data[5*64], stride2); + transfer_16to8add(pV_Cur, &data[5 * 64], stride2); stop_transfer_timer(); } -void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) +void +decoder_iframe(DECODER * dec, + Bitstream * bs, + int quant, + int intra_dc_threshold) { - + uint32_t bound; uint32_t x, y; - for (y = 0; y < dec->mb_height; y++) - { - for (x = 0; x < dec->mb_width; x++) - { - MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; + bound = 0; + for (y = 0; y < dec->mb_height; y++) { + for (x = 0; x < dec->mb_width; x++) { + MACROBLOCK *mb; uint32_t mcbpc; uint32_t cbpc; uint32_t acpred_flag; uint32_t cbpy; uint32_t cbp; + while (BitstreamShowBits(bs, 9) == 1) + BitstreamSkip(bs, 9); + + if (check_resync_marker(bs, 0)) + { + bound = read_video_packet_header(bs, 0, &quant); + x = bound % dec->mb_width; + y = bound / dec->mb_width; + } + mb = &dec->mbs[y * dec->mb_width + x]; + + DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); + mcbpc = get_mcbpc_intra(bs); mb->mode = mcbpc & 7; cbpc = (mcbpc >> 4); acpred_flag = BitstreamGetBit(bs); - if (mb->mode == MODE_STUFFING) - { - DEBUG("-- STUFFING ?"); - continue; - } - cbpy = get_cbpy(bs, 1); cbp = (cbpy << 2) | cbpc; - if (mb->mode == MODE_INTRA_Q) - { - quant += dquant_table[BitstreamGetBits(bs,2)]; - if (quant > 31) - { + if (mb->mode == MODE_INTRA_Q) { + quant += dquant_table[BitstreamGetBits(bs, 2)]; + if (quant > 31) { quant = 31; - } - else if (quant < 1) - { + } else if (quant < 1) { quant = 1; } } mb->quant = quant; + mb->mvs[0].x = mb->mvs[0].y = + mb->mvs[1].x = mb->mvs[1].y = + mb->mvs[2].x = mb->mvs[2].y = + mb->mvs[3].x = mb->mvs[3].y =0; - if (dec->interlacing) - { + if (dec->interlacing) { mb->field_dct = BitstreamGetBit(bs); - DEBUG1("deci: field_dct: ", mb->field_dct); + DPRINTF(DPRINTF_DEBUG, "deci: field_dct: %d", mb->field_dct); } - decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); + decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, + intra_dc_threshold, bound); } + if(dec->out_frm) + output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width); + } } -void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) +void +get_motion_vector(DECODER * dec, + Bitstream * bs, + int x, + int y, + int k, + VECTOR * mv, + int fcode, + const int bound) { int scale_fac = 1 << (fcode - 1); @@ -436,39 +505,28 @@ int low = ((-32) * scale_fac); int range = (64 * scale_fac); - VECTOR pmv[4]; - uint32_t psad[4]; - + VECTOR pmv; int mv_x, mv_y; - int pmv_x, pmv_y; - - get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad); - - pmv_x = pmv[0].x; - pmv_y = pmv[0].y; + pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); mv_x = get_mv(bs, fcode); mv_y = get_mv(bs, fcode); - - mv_x += pmv_x; - mv_y += pmv_y; - if (mv_x < low) - { + DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y); + + mv_x += pmv.x; + mv_y += pmv.y; + + if (mv_x < low) { mv_x += range; - } - else if (mv_x > high) - { + } else if (mv_x > high) { mv_x -= range; } - if (mv_y < low) - { + if (mv_y < low) { mv_y += range; - } - else if (mv_y > high) - { + } else if (mv_y > high) { mv_y -= range; } @@ -478,22 +536,47 @@ } -void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) +void +decoder_pframe(DECODER * dec, + Bitstream * bs, + int rounding, + int quant, + int fcode, + int intra_dc_threshold) { uint32_t x, y; + uint32_t bound; + int cp_mb, st_mb; start_timer(); - image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); + image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, + dec->width, dec->height); stop_edges_timer(); - for (y = 0; y < dec->mb_height; y++) - { - for (x = 0; x < dec->mb_width; x++) - { - MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; + bound = 0; + + for (y = 0; y < dec->mb_height; y++) { + cp_mb = st_mb = 0; + for (x = 0; x < dec->mb_width; x++) { + MACROBLOCK *mb; + + /* skip stuffing */ + while (BitstreamShowBits(bs, 10) == 1) + BitstreamSkip(bs, 10); + + if (check_resync_marker(bs, fcode - 1)) + { + bound = read_video_packet_header(bs, fcode - 1, &quant); + x = bound % dec->mb_width; + y = bound / dec->mb_width; + } + mb = &dec->mbs[y * dec->mb_width + x]; + + DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); - if (!BitstreamGetBit(bs)) // not_coded + /*if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) not_coded */ + if (!(BitstreamGetBit(bs))) /* not_coded */ { uint32_t mcbpc; uint32_t cbpc; @@ -502,188 +585,233 @@ uint32_t cbp; uint32_t intra; + cp_mb++; mcbpc = get_mcbpc_inter(bs); mb->mode = mcbpc & 7; cbpc = (mcbpc >> 4); + + DPRINTF(DPRINTF_MB, "mode %i", mb->mode); + DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); acpred_flag = 0; intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); - - if (intra) - { - acpred_flag = BitstreamGetBit(bs); - } - if (mb->mode == MODE_STUFFING) - { - DEBUG("-- STUFFING ?"); - continue; + if (intra) { + acpred_flag = BitstreamGetBit(bs); } cbpy = get_cbpy(bs, intra); + DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); + cbp = (cbpy << 2) | cbpc; - if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) - { - quant += dquant_table[BitstreamGetBits(bs,2)]; - if (quant > 31) - { + if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { + int dquant = dquant_table[BitstreamGetBits(bs, 2)]; + DPRINTF(DPRINTF_MB, "dquant %i", dquant); + quant += dquant; + if (quant > 31) { quant = 31; - } - else if (mb->quant < 1) - { + } else if (quant < 1) { quant = 1; } + DPRINTF(DPRINTF_MB, "quant %i", quant); } mb->quant = quant; - if (dec->interlacing) - { - mb->field_dct = BitstreamGetBit(bs); - DEBUG1("decp: field_dct: ", mb->field_dct); + if (dec->interlacing) { + if (cbp || intra) { + mb->field_dct = BitstreamGetBit(bs); + DPRINTF(DPRINTF_DEBUG, "decp: field_dct: %d", mb->field_dct); + } - if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) - { + if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { mb->field_pred = BitstreamGetBit(bs); - DEBUG1("decp: field_pred: ", mb->field_pred); + DPRINTF(DPRINTF_DEBUG, "decp: field_pred: %d", mb->field_pred); - if (mb->field_pred) - { + if (mb->field_pred) { mb->field_for_top = BitstreamGetBit(bs); - DEBUG1("decp: field_for_top: ", mb->field_for_top); + DPRINTF(DPRINTF_DEBUG, "decp: field_for_top: %d", mb->field_for_top); mb->field_for_bot = BitstreamGetBit(bs); - DEBUG1("decp: field_for_bot: ", mb->field_for_bot); + DPRINTF(DPRINTF_DEBUG, "decp: field_for_bot: %d", mb->field_for_bot); } } } - if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) - { - if (dec->interlacing && mb->field_pred) - { - get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); - get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode); + if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { + if (dec->interlacing && mb->field_pred) { + get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], + fcode, bound); + get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], + fcode, bound); + } else { + get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], + fcode, bound); + mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = + mb->mvs[0].x; + mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = + mb->mvs[0].y; } - else - { - get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); - mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; - mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; - } - } - else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) - { - get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); - get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); - get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); - get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); - } - else // MODE_INTRA, MODE_INTRA_Q + } else if (mb->mode == MODE_INTER4V ) { + + get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); + get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); + get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); + get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); + } else /* MODE_INTRA, MODE_INTRA_Q */ { - mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; - mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; - decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); + mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = + 0; + mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = + 0; + decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, + intra_dc_threshold, bound); continue; } - decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); - } - else // not coded + decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, + rounding); + } else /* not coded */ { + DPRINTF(DPRINTF_DEBUG, "P-frame MB at (X,Y)=(%d,%d)", x, y); - mb->mode = MODE_NOT_CODED; + mb->mode = MODE_NOT_CODED; mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; - - // copy macroblock directly from ref to cur - start_timer(); + /* copy macroblock directly from ref to cur */ - transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), - dec->refn[0].y + (16*y)*dec->edged_width + (16*x), - dec->edged_width); - - transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), - dec->refn[0].y + (16*y)*dec->edged_width + (16*x+8), - dec->edged_width); - - transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), - dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x), - dec->edged_width); - - transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), - dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x+8), - dec->edged_width); - - transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), - dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), - dec->edged_width/2); - - transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), - dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), - dec->edged_width/2); + start_timer(); + transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + + (16 * x), + dec->refn[0].y + (16 * y) * dec->edged_width + + (16 * x), dec->edged_width); + + transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + + (16 * x + 8), + dec->refn[0].y + (16 * y) * dec->edged_width + + (16 * x + 8), dec->edged_width); + + transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + + (16 * x), + dec->refn[0].y + (16 * y + + 8) * dec->edged_width + + (16 * x), dec->edged_width); + + transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + + (16 * x + 8), + dec->refn[0].y + (16 * y + + 8) * dec->edged_width + + (16 * x + 8), dec->edged_width); + + transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + + (8 * x), + dec->refn[0].u + + (8 * y) * dec->edged_width / 2 + (8 * x), + dec->edged_width / 2); + + transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + + (8 * x), + dec->refn[0].v + + (8 * y) * dec->edged_width / 2 + (8 * x), + dec->edged_width / 2); stop_transfer_timer(); + if(dec->out_frm && cp_mb > 0) { + output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); + cp_mb = 0; + } + st_mb = x+1; } } + if(dec->out_frm && cp_mb > 0) + output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); } } -int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) +/* swap two MACROBLOCK array */ +void +mb_swap(MACROBLOCK ** mb1, + MACROBLOCK ** mb2) +{ + MACROBLOCK *temp = *mb1; + + *mb1 = *mb2; + *mb2 = temp; +} + +int +decoder_decode(DECODER * dec, + XVID_DEC_FRAME * frame) { Bitstream bs; uint32_t rounding; uint32_t quant; - uint32_t fcode; + uint32_t fcode_forward; + uint32_t fcode_backward; uint32_t intra_dc_threshold; uint32_t vop_type; start_global_timer(); - - BitstreamInit(&bs, frame->bitstream, frame->length); - // add by chenm001 - // for support B-frame to reference last 2 frame - vop_type=BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold); + dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; - if (vop_type==I_VOP || vop_type==P_VOP){ - image_swap(&dec->refn[0], &dec->refn[1]); - image_swap(&dec->cur, &dec->refn[0]); - } + BitstreamInit(&bs, frame->bitstream, frame->length); - switch (vop_type) - { - case P_VOP : - decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold); + /* add by chenm001 */ + /* for support B-frame to reference last 2 frame */ + dec->frames++; + vop_type = + BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, + &fcode_backward, &intra_dc_threshold); + + dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ + + switch (vop_type) { + case P_VOP: + decoder_pframe(dec, &bs, rounding, quant, fcode_forward, + intra_dc_threshold); break; - case I_VOP : - //DEBUG1("",intra_dc_threshold); + case I_VOP: decoder_iframe(dec, &bs, quant, intra_dc_threshold); break; - - case B_VOP : // ignore + case B_VOP: + image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); break; - - case N_VOP : // vop not coded + case N_VOP: + /* when low_delay==0, N_VOP's should interpolate between the past and future frames */ + image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); break; - default : + default: return XVID_ERR_FAIL; } frame->length = BitstreamPos(&bs) / 8; - start_timer(); image_output(&dec->cur, dec->width, dec->height, dec->edged_width, - frame->image, frame->stride, frame->colorspace); - stop_conv_timer(); - + frame->image, frame->stride, frame->colorspace); + + if (vop_type == I_VOP || vop_type == P_VOP) { + image_swap(&dec->refn[0], &dec->refn[1]); + image_swap(&dec->cur, &dec->refn[0]); + + /* swap MACROBLOCK */ + /* the Divx will not set the low_delay flage some times */ + /* so follow code will wrong to not swap at that time */ + /* this will broken bitstream! so I'm change it, */ + /* But that is not the best way! can anyone tell me how */ + /* to do another way? */ + /* 18-07-2002 MinChen */ + /*if (!dec->low_delay && vop_type == P_VOP) */ + if (vop_type == P_VOP) + mb_swap(&dec->mbs, &dec->last_mbs); + } + emms(); stop_global_timer(); return XVID_ERR_OK; - }