/************************************************************************** * * 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. * * 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 * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * *************************************************************************/ /************************************************************************** * * History: * * 15.07.2002 fix a bug in B-frame decode at DIRECT mode * MinChen * 10.07.2002 added BFRAMES_DEC_DEBUG support * Fix a little bug for low_delay flage * MinChen * 28.06.2002 added basic resync support to iframe/pframe_decode() * 22.06.2002 added primative N_VOP support * #define BFRAMES_DEC now enables Minchen's bframe decoder * 08.05.2002 add low_delay support for B_VOP decode * MinChen * 05.05.2002 fix some B-frame decode problem * 02.05.2002 add B-frame decode support(have some problem); * MinChen * 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 * * $Id: decoder.c,v 1.32 2002-07-19 11:15:21 albeu Exp $ * *************************************************************************/ #include #include #ifdef BFRAMES_DEC_DEBUG #define BFRAMES_DEC #endif #include "xvid.h" #include "portab.h" #include "decoder.h" #include "bitstream/bitstream.h" #include "bitstream/mbcoding.h" #include "quant/quant_h263.h" #include "quant/quant_mpeg4.h" #include "dct/idct.h" #include "dct/fdct.h" #include "utils/mem_transfer.h" #include "image/interpolate8x8.h" #include "bitstream/mbcoding.h" #include "prediction/mbprediction.h" #include "utils/timer.h" #include "utils/emms.h" #include "image/image.h" #include "image/colorspace.h" #include "utils/mem_align.h" int decoder_create(XVID_DEC_PARAM * param) { DECODER *dec; dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); if (dec == NULL) { return XVID_ERR_MEMORY; } param->handle = dec; dec->width = param->width; dec->height = param->height; dec->mb_width = (dec->width + 15) / 16; dec->mb_height = (dec->height + 15) / 16; dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; 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)) { 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; } 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; } 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) { 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); write_timer(); return XVID_ERR_OK; } 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, const unsigned int bound) { 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; uint32_t stride2 = stride / 2; uint32_t next_block = stride * 8; uint32_t i; uint32_t iQuant = pMB->quant; uint8_t *pY_Cur, *pU_Cur, *pV_Cur; pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); 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 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, bound); if (!acpred_flag) { pMB->acpred_directions[i] = 0; } stop_prediction_timer(); 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; if (dc_size > 8) { BitstreamSkip(bs, 1); // marker } block[i * 64 + 0] = dc_dif; start_coeff = 1; DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); } else { start_coeff = 0; } start_timer(); if (cbp & (1 << (5 - i))) // coded { 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); 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); } stop_iquant_timer(); start_timer(); idct(&data[i * 64]); stop_idct_timer(); } 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); stop_transfer_timer(); } #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 }; // 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) { 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; uint32_t stride2 = stride / 2; uint32_t next_block = stride * 8; uint32_t i; uint32_t iQuant = pMB->quant; uint8_t *pY_Cur, *pU_Cur, *pV_Cur; int uv_dx, uv_dy; pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); 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) { 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 { 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)); 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)); } 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); stop_comp_timer(); for (i = 0; i < 6; i++) { if (cbp & (1 << (5 - i))) // coded { memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear start_timer(); 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); } stop_iquant_timer(); start_timer(); idct(&data[i * 64]); stop_idct_timer(); } } if (dec->interlacing && pMB->field_dct) { next_block = stride; stride *= 2; } start_timer(); if (cbp & 32) transfer_16to8add(pY_Cur, &data[0 * 64], stride); if (cbp & 16) transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); if (cbp & 8) transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); if (cbp & 4) transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); if (cbp & 2) transfer_16to8add(pU_Cur, &data[4 * 64], stride2); if (cbp & 1) transfer_16to8add(pV_Cur, &data[5 * 64], stride2); stop_transfer_timer(); } void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) { uint32_t bound; uint32_t x, y; 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); cbpy = get_cbpy(bs, 1); cbp = (cbpy << 2) | cbpc; if (mb->mode == MODE_INTRA_Q) { quant += dquant_table[BitstreamGetBits(bs, 2)]; if (quant > 31) { quant = 31; } 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) { mb->field_dct = BitstreamGetBit(bs); DEBUG1("deci: field_dct: ", mb->field_dct); } 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, const int bound) { int scale_fac = 1 << (fcode - 1); int high = (32 * scale_fac) - 1; int low = ((-32) * scale_fac); int range = (64 * scale_fac); VECTOR pmv; int mv_x, mv_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); 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) { mv_x -= range; } if (mv_y < low) { mv_y += range; } else if (mv_y > high) { mv_y -= range; } mv->x = mv_x; mv->y = mv_y; } 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); stop_edges_timer(); 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 (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded if (!(BitstreamGetBit(bs))) // not_coded { uint32_t mcbpc; uint32_t cbpc; uint32_t acpred_flag; uint32_t cbpy; 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); } 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) { int dquant = dquant_table[BitstreamGetBits(bs, 2)]; DPRINTF(DPRINTF_MB, "dquant %i", dquant); quant += dquant; if (quant > 31) { quant = 31; } 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 (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { mb->field_pred = BitstreamGetBit(bs); DEBUG1("decp: field_pred: ", mb->field_pred); if (mb->field_pred) { mb->field_for_top = BitstreamGetBit(bs); DEBUG1("decp: field_for_top: ", mb->field_for_top); mb->field_for_bot = BitstreamGetBit(bs); DEBUG1("decp: field_for_bot: ", 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, 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 if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */ ) { 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, bound); continue; } decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); } else // not coded { //DEBUG2("P-frame MB at (X,Y)=",x,y); 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(); 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); } } // add by MinChen // decode B-frame motion vector void get_b_motion_vector(DECODER * dec, Bitstream * bs, int x, int y, VECTOR * mv, int fcode, const VECTOR pmv) { int scale_fac = 1 << (fcode - 1); int high = (32 * scale_fac) - 1; int low = ((-32) * scale_fac); int range = (64 * scale_fac); int mv_x, mv_y; int pmv_x, pmv_y; pmv_x = pmv.x; pmv_y = pmv.y; mv_x = get_mv(bs, fcode); mv_y = get_mv(bs, fcode); mv_x += pmv_x; mv_y += pmv_y; if (mv_x < low) { mv_x += range; } else if (mv_x > high) { mv_x -= range; } if (mv_y < low) { mv_y += range; } else if (mv_y > high) { mv_y -= range; } mv->x = mv_x; mv->y = mv_y; } // add by MinChen // decode an B-frame forward & backward inter macroblock void decoder_bf_mbinter(DECODER * dec, const MACROBLOCK * pMB, const uint32_t x_pos, const uint32_t y_pos, const uint32_t cbp, Bitstream * bs, const uint32_t quant, const uint8_t ref) { 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; uint32_t stride2 = stride / 2; uint32_t next_block = stride * 8; uint32_t i; uint32_t iQuant = pMB->quant; uint8_t *pY_Cur, *pU_Cur, *pV_Cur; int uv_dx, uv_dy; pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); 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)) { 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 { 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)); 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)); } start_timer(); interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos, pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, 16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, 16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, uv_dx, uv_dy, stride2, 0); interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, uv_dx, uv_dy, stride2, 0); stop_comp_timer(); for (i = 0; i < 6; i++) { if (cbp & (1 << (5 - i))) // coded { memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear start_timer(); 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); } stop_iquant_timer(); start_timer(); idct(&data[i * 64]); stop_idct_timer(); } } if (dec->interlacing && pMB->field_dct) { next_block = stride; stride *= 2; } start_timer(); if (cbp & 32) transfer_16to8add(pY_Cur, &data[0 * 64], stride); if (cbp & 16) transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); if (cbp & 8) transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); if (cbp & 4) transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); if (cbp & 2) transfer_16to8add(pU_Cur, &data[4 * 64], stride2); if (cbp & 1) transfer_16to8add(pV_Cur, &data[5 * 64], stride2); stop_transfer_timer(); } // add by MinChen // decode an B-frame direct & inter macroblock void decoder_bf_interpolate_mbinter(DECODER * dec, IMAGE forward, IMAGE backward, const MACROBLOCK * pMB, const uint32_t x_pos, const uint32_t y_pos, Bitstream * bs) { 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; uint32_t stride2 = stride / 2; uint32_t next_block = stride * 8; uint32_t iQuant = pMB->quant; int uv_dx, uv_dy; int b_uv_dx, b_uv_dy; uint32_t i; uint8_t *pY_Cur, *pU_Cur, *pV_Cur; const uint32_t cbp = pMB->cbp; pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); 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)) { 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; b_uv_dx = pMB->b_mvs[0].x; b_uv_dy = pMB->b_mvs[0].y; b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; } 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)); 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)); sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; b_uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; b_uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); } start_timer(); interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, uv_dy, stride2, 0); interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, uv_dy, stride2, 0); interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, 16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, 16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, b_uv_dx, b_uv_dy, stride2, 0); interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, b_uv_dx, b_uv_dy, stride2, 0); interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos, stride); interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, stride); interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, stride); interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos + 8, stride); interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, stride2); interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos, stride2); stop_comp_timer(); for (i = 0; i < 6; i++) { if (cbp & (1 << (5 - i))) // coded { memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear start_timer(); 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); } stop_iquant_timer(); start_timer(); idct(&data[i * 64]); stop_idct_timer(); } } if (dec->interlacing && pMB->field_dct) { next_block = stride; stride *= 2; } start_timer(); if (cbp & 32) transfer_16to8add(pY_Cur, &data[0 * 64], stride); if (cbp & 16) transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); if (cbp & 8) transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); if (cbp & 4) transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); if (cbp & 2) transfer_16to8add(pU_Cur, &data[4 * 64], stride2); if (cbp & 1) transfer_16to8add(pV_Cur, &data[5 * 64], stride2); stop_transfer_timer(); } // add by MinChen // for decode B-frame dbquant int32_t __inline get_dbquant(Bitstream * bs) { if (!BitstreamGetBit(bs)) // '0' return (0); else if (!BitstreamGetBit(bs)) // '10' return (-2); else return (2); // '11' } // add by MinChen // for decode B-frame mb_type // bit ret_value // 1 0 // 01 1 // 001 2 // 0001 3 int32_t __inline get_mbtype(Bitstream * bs) { int32_t mb_type; for (mb_type = 0; mb_type <= 3; mb_type++) { if (BitstreamGetBit(bs)) break; } if (mb_type <= 3) return (mb_type); else return (-1); } void decoder_bframe(DECODER * dec, Bitstream * bs, int quant, int fcode_forward, int fcode_backward) { uint32_t x, y; VECTOR mv; const VECTOR zeromv = {0,0}; #ifdef BFRAMES_DEC_DEBUG FILE *fp; static char first=0; #define BFRAME_DEBUG if (!first && fp){ \ fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ } #endif start_timer(); image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); stop_edges_timer(); #ifdef BFRAMES_DEC_DEBUG if (!first){ fp=fopen("C:\\XVIDDBG.TXT","w"); } #endif for (y = 0; y < dec->mb_height; y++) { // Initialize Pred Motion Vector dec->p_fmv = dec->p_bmv = zeromv; for (x = 0; x < dec->mb_width; x++) { MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; mv = mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; // the last P_VOP is skip macroblock ? if (last_mb->mode == MODE_NOT_CODED) { //DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); mb->cbp = 0; #ifdef BFRAMES_DEC_DEBUG mb->mb_type = MODE_NOT_CODED; BFRAME_DEBUG #endif mb->mb_type = MODE_FORWARD; mb->quant = last_mb->quant; //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; decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); continue; } if (!BitstreamGetBit(bs)) { // modb=='0' const uint8_t modb2 = BitstreamGetBit(bs); mb->mb_type = get_mbtype(bs); if (!modb2) { // modb=='00' mb->cbp = BitstreamGetBits(bs, 6); } else { mb->cbp = 0; } if (mb->mb_type && mb->cbp) { quant += get_dbquant(bs); if (quant > 31) { quant = 31; } else if (quant < 1) { quant = 1; } } } else { mb->mb_type = MODE_DIRECT_NONE_MV; mb->cbp = 0; } mb->quant = quant; mb->mode = MODE_INTER4V; //DEBUG1("Switch bm_type=",mb->mb_type); #ifdef BFRAMES_DEC_DEBUG BFRAME_DEBUG #endif switch (mb->mb_type) { case MODE_DIRECT: get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); case MODE_DIRECT_NONE_MV: { // Because this file is a C file not C++ so I use '{' to define var const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; int i; for (i = 0; i < 4; i++) { mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x); mb->b_mvs[i].x = (int32_t) ((mv.x == 0) ? ((TRB - TRD) * last_mb->mvs[i].x) / TRD : mb->mvs[i].x - last_mb->mvs[i].x); mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y); mb->b_mvs[i].y = (int32_t) ((mv.y == 0) ? ((TRB - TRD) * last_mb->mvs[i].y) / TRD : mb->mvs[i].y - last_mb->mvs[i].y); } //DEBUG("B-frame Direct!\n"); } decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, bs); break; case MODE_INTERPOLATE: get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, dec->p_fmv); dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], fcode_backward, dec->p_bmv); dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0]; decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, bs); //DEBUG("B-frame Bidir!\n"); break; case MODE_BACKWARD: get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, dec->p_bmv); dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; mb->mode = MODE_INTER; decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); //DEBUG("B-frame Backward!\n"); break; case MODE_FORWARD: get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, dec->p_fmv); dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; mb->mode = MODE_INTER; decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); //DEBUG("B-frame Forward!\n"); break; default: //DEBUG1("Not support B-frame mb_type =", mb->mb_type); ; } } // end of FOR } #ifdef BFRAMES_DEC_DEBUG if (!first){ first=1; if (fp) fclose(fp); } #endif } // 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_forward; uint32_t fcode_backward; uint32_t intra_dc_threshold; uint32_t vop_type; start_global_timer(); dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; BitstreamInit(&bs, frame->bitstream, frame->length); // 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); #ifdef BFRAMES_DEC DEBUG1("P_VOP Time=", dec->time); #endif break; case I_VOP: decoder_iframe(dec, &bs, quant, intra_dc_threshold); #ifdef BFRAMES_DEC DEBUG1("I_VOP Time=", dec->time); #endif break; case B_VOP: #ifdef BFRAMES_DEC if (dec->time_pp > dec->time_bp) { DEBUG1("B_VOP Time=", dec->time); decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); } else { DEBUG("broken B-frame!"); } #else image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); #endif break; case N_VOP: // vop not coded // 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: return XVID_ERR_FAIL; } #ifdef BFRAMES_DEC_DEBUG if (frame->length != BitstreamPos(&bs) / 8){ DEBUG2("InLen/UseLen",frame->length, BitstreamPos(&bs) / 8); } #endif frame->length = BitstreamPos(&bs) / 8; #ifdef BFRAMES_DEC // test if no B_VOP if (dec->low_delay) { #endif image_output(&dec->cur, dec->width, dec->height, dec->edged_width, frame->image, frame->stride, frame->colorspace); #ifdef BFRAMES_DEC } else { if (dec->frames >= 0) { start_timer(); if ((vop_type == I_VOP || vop_type == P_VOP)) { image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, frame->image, frame->stride, frame->colorspace); } else if (vop_type == B_VOP) { image_output(&dec->cur, dec->width, dec->height, dec->edged_width, frame->image, frame->stride, frame->colorspace); } stop_conv_timer(); } } #endif 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; }