--- decoder.c 2003/01/03 16:25:14 1.37.2.25 +++ decoder.c 2003/05/03 23:23:38 1.49.4.1 @@ -1,16 +1,9 @@ -/************************************************************************** +/***************************************************************************** * * XVID MPEG-4 VIDEO CODEC - * - Decoder main module - + * - Decoder 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 file is part of XviD, a free MPEG-4 video encoder/decoder * * 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 @@ -26,39 +19,11 @@ * 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 1.49.4.1 2003/05/03 23:23:38 Isibaar Exp $ * - * $Id: decoder.c,v 1.37.2.25 2003/01/03 16:25:14 suxen_drol Exp $ - * - *************************************************************************/ + ****************************************************************************/ +#include #include #include @@ -68,6 +33,7 @@ #include "xvid.h" #include "portab.h" +#include "global.h" #include "decoder.h" #include "bitstream/bitstream.h" @@ -91,25 +57,26 @@ #include "image/image.h" #include "image/colorspace.h" #include "utils/mem_align.h" +#include "image/postprocessing.h" int decoder_resize(DECODER * dec) { /* free existing */ - 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->tmp, dec->edged_width, dec->edged_height); image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); + image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); + if (dec->last_mbs) xvid_free(dec->last_mbs); if (dec->mbs) xvid_free(dec->mbs); /* realloc */ - dec->mb_width = (dec->width + 15) / 16; dec->mb_height = (dec->height + 15) / 16; @@ -127,8 +94,7 @@ return XVID_ERR_MEMORY; } - // add by chenm001 - // for support B-frame to reference last 2 frame + /* 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); @@ -152,6 +118,16 @@ return XVID_ERR_MEMORY; } + if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { + image_destroy(&dec->qtmp, 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); + image_destroy(&dec->tmp, 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); @@ -166,8 +142,7 @@ } memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); - // add by chenm001 - // for skip MB flag + /* For skip MB flag */ dec->last_mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); @@ -210,13 +185,16 @@ image_null(&dec->tmp); image_null(&dec->qtmp); + /* image based GMC */ + image_null(&dec->gmc); + + dec->mbs = NULL; dec->last_mbs = NULL; init_timer(); - // add by chenm001 - // for support B-frame to save reference frame's time + /* For B-frame support (used to save reference frame's time */ dec->frames = 0; dec->time = dec->time_base = dec->last_time_base = 0; dec->low_delay = 0; @@ -236,6 +214,10 @@ { xvid_free(dec->last_mbs); xvid_free(dec->mbs); + + /* image based GMC */ + image_destroy(&dec->gmc, 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->tmp, dec->edged_width, dec->edged_height); @@ -256,8 +238,7 @@ -// decode an intra macroblock - +/* decode an intra macroblock */ void decoder_mbintra(DECODER * dec, MACROBLOCK * pMB, @@ -292,7 +273,7 @@ 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++) { uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); @@ -315,7 +296,7 @@ dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; if (dc_size > 8) { - BitstreamSkip(bs, 1); // marker + BitstreamSkip(bs, 1); /* marker */ } block[i * 64 + 0] = dc_dif; @@ -327,7 +308,7 @@ } start_timer(); - if (cbp & (1 << (5 - i))) // coded + if (cbp & (1 << (5 - i))) /* coded */ { int direction = dec->alternate_vertical_scan ? 2 : pMB->acpred_directions[i]; @@ -384,18 +365,13 @@ - -#define SIGN(X) (((X)>0)?1:-1) -#define ABS(X) (((X)>0)?(X):-(X)) - -// 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 fcode, const uint32_t cbp, Bitstream * bs, const uint32_t quant, @@ -424,24 +400,19 @@ mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); } - }else{ + } else { 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); for (i = 0; i < 4; i++) mv[i] = pMB->mvs[i]; } - + if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { - uv_dx = mv[0].x; - uv_dy = mv[0].y; - - if (dec->quarterpel) - { - uv_dx /= 2; - uv_dy /= 2; - } + uv_dx = mv[0].x / (1 + dec->quarterpel); + uv_dy = mv[0].y / (1 + dec->quarterpel); + uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; @@ -461,11 +432,11 @@ if(dec->quarterpel) { interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, dec->qtmp.y + 128, 16*x_pos, 16*y_pos, - mv[0].x, mv[0].y, stride, rounding); + mv[0].x, mv[0].y, stride, rounding); } else { interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, - mv[0].x, mv[0].y, stride, rounding); + mv[0].x, mv[0].y, stride, rounding); } interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, @@ -497,49 +468,49 @@ { interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, mv[0].x, mv[0].y, stride, rounding); - interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos + 16, 32*y_pos, + interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, mv[1].x, mv[1].y, stride, rounding); - interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos + 16, + interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, mv[2].x, mv[2].y, stride, rounding); - interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos + 16, 32*y_pos + 16, + interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, mv[3].x, mv[3].y, stride, rounding); - interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, + interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, uv_dx, uv_dy, stride2, rounding); - interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, + interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, uv_dx, uv_dy, stride2, rounding); - // set_block(pY_Cur, stride, 32, 32, 127); + /* set_block(pY_Cur, stride, 32, 32, 127); */ } else { if(dec->quarterpel) { - interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, + interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, dec->qtmp.y + 128, 16*x_pos, 16*y_pos, mv[0].x, mv[0].y, stride, rounding); - interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, + interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, mv[1].x, mv[1].y, stride, rounding); - interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, + interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, mv[2].x, mv[2].y, stride, rounding); - interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, + interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, mv[3].x, mv[3].y, stride, rounding); } else { - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, + interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos, mv[0].x, mv[0].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, + interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, mv[1].x, mv[1].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, + interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8, mv[2].x, mv[2].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, + interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8, mv[3].x, mv[3].y, stride, rounding); } - interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, + 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, + interpolate8x8_switch(dec->cur.v, dec->refn[0].v , 8 * x_pos, 8 * y_pos, uv_dx, uv_dy, stride2, rounding); } stop_comp_timer(); @@ -548,9 +519,9 @@ for (i = 0; i < 6; i++) { int direction = dec->alternate_vertical_scan ? 2 : 0; - if (cbp & (1 << (5 - i))) // coded + 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], direction); @@ -609,6 +580,119 @@ stop_transfer_timer(); } +static __inline int gmc_sanitize(int value, int quarterpel, int fcode) +{ + int length = 1 << (fcode+4); + +/* if (quarterpel) value *= 2; */ + + if (value < -length) + return -length; + else if (value >= length) + return length-1; + else return value; +} + + +static void +decoder_mbgmc(DECODER * dec, + MACROBLOCK * const pMB, + const uint32_t x_pos, + const uint32_t y_pos, + const uint32_t fcode, + const uint32_t cbp, + Bitstream * bs, + const uint32_t quant, + const uint32_t rounding, + const int reduced_resolution) /* no reduced res support */ +{ + + DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); + DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); + + const uint32_t stride = dec->edged_width; + const uint32_t stride2 = stride / 2; + const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); + uint32_t i; + const uint32_t iQuant = pMB->quant; + uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); + uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); + uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); + + pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; + + start_timer(); + +/* this is where the calculations are done */ + + { + pMB->amv = generate_GMCimageMB(&dec->gmc_data, &dec->refn[0], x_pos, y_pos, + stride, stride2, dec->quarterpel, rounding, &dec->cur); + + pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); + pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); + } + pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; + +/* + transfer16x16_copy(pY_Cur, dec->gmc.y + (y_pos << 4)*stride + (x_pos << 4), stride); + transfer8x8_copy(pU_Cur, dec->gmc.u + (y_pos << 3)*stride2 + (x_pos << 3), stride2); + transfer8x8_copy(pV_Cur, dec->gmc.v + (y_pos << 3)*stride2 + (x_pos << 3), stride2); +*/ + + + stop_transfer_timer(); + + if (!cbp) return; + + for (i = 0; i < 6; i++) { + int direction = dec->alternate_vertical_scan ? 2 : 0; + + 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], direction); + 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(); + } + } + +/* interlace + GMC is this possible ??? */ +/* + 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, @@ -716,7 +800,7 @@ 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); + DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y); mv.x += pmv.x; mv.y += pmv.y; @@ -739,21 +823,9 @@ -static __inline int gmc_sanitize(int value, int quarterpel, int fcode) -{ - int length = 1 << (fcode+4); - if (quarterpel) value *= 2; - if (value < -length) - return -length; - else if (value >= length) - return length-1; - else return value; -} - - -/* for P_VOP set gmc_mv to NULL */ +/* for P_VOP set gmc_warp to NULL */ void decoder_pframe(DECODER * dec, Bitstream * bs, @@ -762,7 +834,7 @@ int quant, int fcode, int intra_dc_threshold, - VECTOR * gmc_mv) + const WARPPOINTS *const gmc_warp) { uint32_t x, y; @@ -782,6 +854,31 @@ dec->width, dec->height); stop_edges_timer(); + if (gmc_warp) + { + + /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ + if ( (dec->sprite_warping_accuracy != 3) || (dec->sprite_warping_points != 2) ) + { + fprintf(stderr,"Wrong GMC parameters acc=%d(-> 1/%d), %d!!!\n", + dec->sprite_warping_accuracy,(2<sprite_warping_accuracy), + dec->sprite_warping_points); + } + + generate_GMCparameters( dec->sprite_warping_points, + (2 << dec->sprite_warping_accuracy), gmc_warp, + dec->width, dec->height, &dec->gmc_data); + +/* image warping is done block-based in decoder_mbgmc(), now */ +/* + generate_GMCimage(&dec->gmc_data, &dec->refn[0], + mb_width, mb_height, + dec->edged_width, dec->edged_width/2, + fcode, dec->quarterpel, 0, + rounding, dec->mbs, &dec->gmc); +*/ + } + bound = 0; for (y = 0; y < mb_height; y++) { @@ -789,7 +886,7 @@ for (x = 0; x < mb_width; x++) { MACROBLOCK *mb; - // skip stuffing + /* skip stuffing */ while (BitstreamShowBits(bs, 10) == 1) BitstreamSkip(bs, 10); @@ -804,8 +901,8 @@ 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 + /* if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) */ /* not_coded */ + if (!(BitstreamGetBit(bs))) /* block _is_ coded */ { uint32_t mcbpc; uint32_t cbpc; @@ -813,7 +910,7 @@ uint32_t cbpy; uint32_t cbp; uint32_t intra; - int mcsel = 0; // mcsel: '0'=local motion, '1'=GMC + int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ cp_mb++; mcbpc = get_mcbpc_inter(bs); @@ -830,13 +927,13 @@ acpred_flag = BitstreamGetBit(bs); } - if (gmc_mv && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) + if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) { mcsel = BitstreamGetBit(bs); } cbpy = get_cbpy(bs, intra); - DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); + DPRINTF(DPRINTF_MB, "cbpy %i mcsel %i ", cbpy,mcsel); cbp = (cbpy << 2) | cbpc; @@ -871,15 +968,15 @@ } } } + + if (mcsel) { + decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, + rounding, reduced_resolution); + continue; - if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { - - if (mcsel) - { - mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode); - mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode); + } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { - } else if (dec->interlacing && mb->field_pred) { + 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], @@ -887,10 +984,7 @@ } 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; + mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; } } else if (mb->mode == MODE_INTER4V ) { @@ -898,7 +992,7 @@ 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 + } else /* MODE_INTRA, MODE_INTRA_Q */ { mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; @@ -909,16 +1003,26 @@ continue; } - decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, + decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, rounding, reduced_resolution); - } - else if (gmc_mv) /* not coded S_VOP macroblock */ + } + else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ { mb->mode = MODE_NOT_CODED_GMC; - mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode); - mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode); - decoder_mbinter(dec, mb, x, y, 0, 0, bs, quant, rounding, reduced_resolution); + + start_timer(); + + decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, quant, + rounding, reduced_resolution); + + 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; } else /* not coded P_VOP macroblock */ { @@ -926,7 +1030,7 @@ 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 + /* copy macroblock directly from ref to cur */ start_timer(); @@ -974,8 +1078,7 @@ } -// add by MinChen -// decode B-frame motion vector +/* decode B-frame motion vector */ void get_b_motion_vector(DECODER * dec, Bitstream * bs, @@ -1019,8 +1122,7 @@ } -// add by MinChen -// decode an B-frame forward & backward inter macroblock +/* decode an B-frame forward & backward inter macroblock */ void decoder_bf_mbinter(DECODER * dec, const MACROBLOCK * pMB, @@ -1104,9 +1206,9 @@ for (i = 0; i < 6; i++) { int direction = dec->alternate_vertical_scan ? 2 : 0; - if (cbp & (1 << (5 - i))) // coded + 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], direction); @@ -1147,8 +1249,7 @@ stop_transfer_timer(); } -// add by MinChen -// decode an B-frame direct & inter macroblock +/* decode an B-frame direct & inter macroblock */ void decoder_bf_interpolate_mbinter(DECODER * dec, IMAGE forward, @@ -1345,9 +1446,9 @@ for (i = 0; i < 6; i++) { int direction = dec->alternate_vertical_scan ? 2 : 0; - if (cbp & (1 << (5 - i))) // coded + 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], direction); @@ -1389,26 +1490,26 @@ } -// add by MinChen -// for decode B-frame dbquant +/* for decode B-frame dbquant */ int32_t __inline get_dbquant(Bitstream * bs) { - if (!BitstreamGetBit(bs)) // '0' + if (!BitstreamGetBit(bs)) /* '0' */ return (0); - else if (!BitstreamGetBit(bs)) // '10' + else if (!BitstreamGetBit(bs)) /* '10' */ return (-2); - else - return (2); // '11' + else /* '11' */ + return (2); } -// add by MinChen -// for decode B-frame mb_type -// bit ret_value -// 1 0 -// 01 1 -// 001 2 -// 0001 3 +/* + * For decode B-frame mb_type + * bit ret_value + * 1 0 + * 01 1 + * 001 2 + * 0001 3 + */ int32_t __inline get_mbtype(Bitstream * bs) { @@ -1457,7 +1558,7 @@ #endif for (y = 0; y < dec->mb_height; y++) { - // Initialize Pred Motion Vector + /* 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]; @@ -1467,11 +1568,14 @@ 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; - // skip if the co-located P_VOP macroblock is not coded - // note: gmc+not_coded isn't skipped + /* + * skip if the co-located P_VOP macroblock is not coded + * if not codec in co-located S_VOP macroblock is _not_ + * automatically skipped + */ if (last_mb->mode == MODE_NOT_CODED) { - //DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); + /* DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); */ mb->cbp = 0; #ifdef BFRAMES_DEC_DEBUG mb->mb_type = MODE_NOT_CODED; @@ -1479,19 +1583,21 @@ #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; + /* + 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' + if (!BitstreamGetBit(bs)) { /* modb=='0' */ const uint8_t modb2 = BitstreamGetBit(bs); mb->mb_type = get_mbtype(bs); - if (!modb2) { // modb=='00' + if (!modb2) { /* modb=='00' */ mb->cbp = BitstreamGetBits(bs, 6); } else { mb->cbp = 0; @@ -1512,7 +1618,7 @@ mb->quant = quant; mb->mode = MODE_INTER4V; - //DEBUG1("Switch bm_type=",mb->mb_type); + /* DEBUG1("Switch bm_type=",mb->mb_type); */ #ifdef BFRAMES_DEC_DEBUG BFRAME_DEBUG @@ -1541,7 +1647,7 @@ / TRD : mb->mvs[i].y - last_mb->mvs[i].y); } - //DEBUG("B-frame Direct!\n"); + /* DEBUG("B-frame Direct!\n"); */ } decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, bs); @@ -1559,7 +1665,7 @@ decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, bs); - //DEBUG("B-frame Bidir!\n"); + /* DEBUG("B-frame Bidir!\n"); */ break; case MODE_BACKWARD: @@ -1569,7 +1675,7 @@ mb->mode = MODE_INTER; decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); - //DEBUG("B-frame Backward!\n"); + /* DEBUG("B-frame Backward!\n"); */ break; case MODE_FORWARD: @@ -1579,15 +1685,15 @@ mb->mode = MODE_INTER; decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); - //DEBUG("B-frame Forward!\n"); + /* DEBUG("B-frame Forward!\n"); */ break; default: DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type); } - - } // end of FOR + } /* End of for */ } + #ifdef BFRAMES_DEC_DEBUG if (!first){ first=1; @@ -1597,7 +1703,7 @@ #endif } -// swap two MACROBLOCK array +/* swap two MACROBLOCK array */ void mb_swap(MACROBLOCK ** mb1, MACROBLOCK ** mb2) @@ -1618,9 +1724,9 @@ { /* note: image is stored to tmp */ image_copy(&dec->tmp, img, dec->edged_width, dec->height); - image_deblock_rrv(&dec->tmp, dec->edged_width, - mbs, dec->mb_width, dec->mb_height, dec->mb_width, - 8, frame->general); + image_deblock(&dec->tmp, dec->edged_width, + mbs, dec->mb_width, dec->mb_height, dec->mb_width, + frame->general); img = &dec->tmp; } @@ -1642,8 +1748,8 @@ uint32_t fcode_forward; uint32_t fcode_backward; uint32_t intra_dc_threshold; - VECTOR gmc_mv[5]; - uint32_t vop_type; + WARPPOINTS gmc_warp; + int vop_type; int success = 0; int output = 0; int seen_something = 0; @@ -1671,7 +1777,7 @@ { stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; stats->data.vop.time_base = (int)dec->time_base; - stats->data.vop.time_increment = 0; //XXX: todo + stats->data.vop.time_increment = 0; /* XXX: todo */ } emms(); @@ -1682,7 +1788,7 @@ BitstreamInit(&bs, frame->bitstream, frame->length); - // XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's + /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) { if (stats) @@ -1697,12 +1803,12 @@ repeat: vop_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, - &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, gmc_mv); + &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); - DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%i, time_pp=%i, time_bp=%i", + DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i", vop_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); - if (vop_type == - 1) + if (vop_type == -1) { if (success) goto done; emms(); @@ -1732,7 +1838,7 @@ goto repeat; } - dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 + dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ /* packed_mode: special-N_VOP treament */ @@ -1758,7 +1864,7 @@ break; case S_VOP : decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, - fcode_forward, intra_dc_threshold, gmc_mv); + fcode_forward, intra_dc_threshold, &gmc_warp); break; case N_VOP : image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); @@ -1862,7 +1968,7 @@ { stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; stats->data.vop.time_base = (int)dec->time_base; - stats->data.vop.time_increment = 0; //XXX: todo + stats->data.vop.time_increment = 0; /* XXX: todo */ } emms();