--- decoder.c 2004/03/22 22:36:23 1.51 +++ decoder.c 2004/08/10 21:58:55 1.65 @@ -4,7 +4,7 @@ * - Decoder Module - * * Copyright(C) 2002 MinChen - * 2002-2003 Peter Ross + * 2002-2004 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 @@ -20,7 +20,7 @@ * along with this program ; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * - * $Id: decoder.c,v 1.51 2004/03/22 22:36:23 edgomez Exp $ + * $Id: decoder.c,v 1.65 2004/08/10 21:58:55 edgomez Exp $ * ****************************************************************************/ @@ -48,6 +48,7 @@ #include "image/interpolate8x8.h" #include "image/reduced.h" #include "image/font.h" +#include "image/qpel.h" #include "bitstream/mbcoding.h" #include "prediction/mbprediction.h" @@ -222,6 +223,7 @@ dec->time = dec->time_base = dec->last_time_base = 0; dec->low_delay = 0; dec->packed_mode = 0; + dec->time_inc_resolution = 1; /* until VOL header says otherwise */ dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); @@ -338,7 +340,7 @@ stop_coding_timer(); start_timer(); - add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); + add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version); stop_prediction_timer(); start_timer(); @@ -389,74 +391,78 @@ uint8_t * pY_Cur, uint8_t * pU_Cur, uint8_t * pV_Cur, - const int reduced_resolution, + int reduced_resolution, const MACROBLOCK * pMB) { - DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); - DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); + DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE); int stride = dec->edged_width; int next_block = stride * (reduced_resolution ? 16 : 8); - const int stride2 = stride/2; int i; const uint32_t iQuant = pMB->quant; const int direction = dec->alternate_vertical_scan ? 2 : 0; - const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; + typedef void (*get_inter_block_function_t)( + Bitstream * bs, + int16_t * block, + int direction, + const int quant, + const uint16_t *matrix); + typedef void (*add_residual_function_t)( + uint8_t *predicted_block, + const int16_t *residual, + int stride); + + const get_inter_block_function_t get_inter_block = (dec->quant_type == 0) + ? (get_inter_block_function_t)get_inter_block_h263 + : (get_inter_block_function_t)get_inter_block_mpeg; + + const add_residual_function_t add_residual = (reduced_resolution) + ? (add_residual_function_t)add_upsampled_8x8_16to8 + : (add_residual_function_t)transfer_16to8add; - for (i = 0; i < 6; i++) { + uint8_t *dst[6]; + int strides[6]; + - if (cbp & (1 << (5 - i))) { /* coded */ + if (dec->interlacing && pMB->field_dct) { + next_block = stride; + stride *= 2; + } - memset(block, 0, 64 * sizeof(int16_t)); /* clear */ + reduced_resolution = !!reduced_resolution; + dst[0] = pY_Cur; + dst[2] = pY_Cur + next_block; + dst[1] = dst[0] + (8<mpeg_quant_matrices)); stop_coding_timer(); + /* iDCT */ start_timer(); - dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); - stop_iquant_timer(); + idct(&data[0]); + stop_idct_timer(); + /* Add this residual to the predicted block */ start_timer(); - idct(&data[i * 64]); - stop_idct_timer(); + add_residual(dst[i], &data[0], strides[i]); + stop_transfer_timer(); } } - - if (dec->interlacing && pMB->field_dct) { - next_block = stride; - stride *= 2; - } - - start_timer(); - if (reduced_resolution) { - if (cbp & 32) - add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); - if (cbp & 16) - add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); - if (cbp & 8) - add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); - if (cbp & 4) - add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); - if (cbp & 2) - add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); - if (cbp & 1) - add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); - } else { - 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(); } /* decode an inter macroblock */ @@ -496,6 +502,33 @@ mv[i] = pMB->mvs[i]; } + for (i = 0; i < 4; i++) { + /* clip to valid range */ + int border = (int)(dec->mb_width - x_pos) << (5 + dec->quarterpel); + if (mv[i].x > border) { + DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); + mv[i].x = border; + } else { + border = (-(int)x_pos-1) << (5 + dec->quarterpel); + if (mv[i].x < border) { + DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); + mv[i].x = border; + } + } + + border = (int)(dec->mb_height - y_pos) << (5 + dec->quarterpel); + if (mv[i].y > border) { + DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); + mv[i].y = border; + } else { + border = (-(int)y_pos-1) << (5 + dec->quarterpel); + if (mv[i].y < border) { + DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); + mv[i].y = border; + } + } + } + start_timer(); if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ @@ -761,20 +794,6 @@ mv.y -= range; } - /* clip to valid range */ - - if (mv.x > ((int)(dec->mb_width - x) << (5 + dec->quarterpel)) ) - mv.x = (int)(dec->mb_width - x) << (5 + dec->quarterpel); - - else if (mv.x < (int)(-x-1) << (5 + dec->quarterpel)) - mv.x = (int)(-x-1) << (5 + dec->quarterpel); - - if (mv.y > ((int)(dec->mb_height - y) << (5 + dec->quarterpel)) ) - mv.y = (int)(dec->mb_height - y) << (5 + dec->quarterpel); - - else if (mv.y < ((int)(-y-1)) << (5 + dec->quarterpel) ) - mv.y = (int)(-y-1) << (5 + dec->quarterpel); - ret_mv->x = mv.x; ret_mv->y = mv.y; } @@ -801,10 +820,13 @@ mb_height = (dec->height + 31) / 32; } - start_timer(); - image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, - dec->width, dec->height, dec->bs_version); - stop_edges_timer(); + if (!dec->is_edged[0]) { + start_timer(); + image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, + dec->width, dec->height, dec->bs_version); + dec->is_edged[0] = 1; + stop_edges_timer(); + } if (gmc_warp) { /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ @@ -986,20 +1008,6 @@ else if (mv_y > high) mv_y -= range; - - /* clip to valid range */ - if (mv_x > ((int)(dec->mb_width - x) << (5 + dec->quarterpel)) ) - mv_x = (int)(dec->mb_width - x) << (5 + dec->quarterpel); - - else if (mv_x < (int)(-x-1) << (5 + dec->quarterpel)) - mv_x = (int)(-x-1) << (5 + dec->quarterpel); - - if (mv_y > ((int)(dec->mb_height - y) << (5 + dec->quarterpel)) ) - mv_y = (int)(dec->mb_height - y) << (5 + dec->quarterpel); - - else if (mv_y < ((int)(-y-1)) << (5 + dec->quarterpel) ) - mv_y = (int)(-y-1) << (5 + dec->quarterpel); - mv->x = mv_x; mv->y = mv_y; } @@ -1029,7 +1037,6 @@ if (!direct) { uv_dx = pMB->mvs[0].x; uv_dy = pMB->mvs[0].y; - b_uv_dx = pMB->b_mvs[0].x; b_uv_dy = pMB->b_mvs[0].y; @@ -1042,21 +1049,20 @@ uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; - b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; } else { - if(dec->quarterpel) { - uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); - uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); - b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); - b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); - } else { - uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; - uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; - b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; - b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; + uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; + uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; + b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; + b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; + + if (dec->quarterpel) { + uv_dx /= 2; + uv_dy /= 2; + b_uv_dx /= 2; + b_uv_dy /= 2; } uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; @@ -1104,68 +1110,38 @@ if(dec->quarterpel) { if(!direct) { - interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, - dec->qtmp.y + 128, 16*x_pos, 16*y_pos, - pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); + interpolate16x16_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, + dec->qtmp.y + 128, 16*x_pos, 16*y_pos, + pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); } else { - interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, - dec->qtmp.y + 128, 16*x_pos, 16*y_pos, - pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); - interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, - dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, - pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); - interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, - dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, - pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); - interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, - dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, - pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); + interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, + dec->qtmp.y + 128, 16*x_pos, 16*y_pos, + pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); + interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, + dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, + pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); + interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, + dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, + pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); + interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, + dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, + pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); } } else { - interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, - pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); - interpolate8x8_switch(dec->tmp.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->tmp.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->tmp.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->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, - b_uv_dx, b_uv_dy, stride2, 0); - interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, - b_uv_dx, b_uv_dy, stride2, 0); - - interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, - dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, - dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, - stride, 1, 8); - - interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, - dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, - dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, - stride, 1, 8); - - interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, - dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, - dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, - stride, 1, 8); - - interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, - dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, - dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, - stride, 1, 8); - - interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, - dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, - dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, - stride2, 1, 8); - - interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, - dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, - dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, - stride2, 1, 8); + interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos, + pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); + interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, + 16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); + interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, + 16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); + interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, + 16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); + } + + interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos, + b_uv_dx, b_uv_dy, stride2, 0); + interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos, + b_uv_dx, b_uv_dy, stride2, 0); stop_comp_timer(); @@ -1215,15 +1191,23 @@ uint32_t x, y; VECTOR mv; const VECTOR zeromv = {0,0}; - const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; int i; - start_timer(); - image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, - dec->width, dec->height, dec->bs_version); - image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, - dec->width, dec->height, dec->bs_version); - stop_edges_timer(); + if (!dec->is_edged[0]) { + start_timer(); + image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, + dec->width, dec->height, dec->bs_version); + dec->is_edged[0] = 1; + stop_edges_timer(); + } + + if (!dec->is_edged[1]) { + start_timer(); + image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, + dec->width, dec->height, dec->bs_version); + dec->is_edged[1] = 1; + stop_edges_timer(); + } for (y = 0; y < dec->mb_height; y++) { /* Initialize Pred Motion Vector */ @@ -1310,14 +1294,15 @@ case MODE_DIRECT_NONE_MV: 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); + mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x; + mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y; + + mb->b_mvs[i].x = (mv.x) + ? mb->mvs[i].x - last_mb->mvs[i].x + : last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp; + mb->b_mvs[i].y = (mv.y) + ? mb->mvs[i].y - last_mb->mvs[i].y + : last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp; } decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], @@ -1361,16 +1346,19 @@ xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type, int quant) { + const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0; + if (dec->cartoon_mode) frame->general &= ~XVID_FILMEFFECT; - if (frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) && mbs != NULL) /* post process */ + if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0) + && mbs != NULL) /* post process */ { /* note: image is stored to tmp */ image_copy(&dec->tmp, img, dec->edged_width, dec->height); image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, mbs, dec->mb_width, dec->mb_height, dec->mb_width, - frame->general, dec->frames, (coding_type == B_VOP)); + frame->general, brightness, dec->frames, (coding_type == B_VOP)); img = &dec->tmp; } @@ -1490,6 +1478,11 @@ goto repeat; } + if(dec->frames == 0 && coding_type != I_VOP) { + /* 1st frame is not an i-vop */ + goto repeat; + } + 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 */ @@ -1539,7 +1532,9 @@ } image_swap(&dec->refn[0], &dec->refn[1]); + dec->is_edged[1] = dec->is_edged[0]; image_swap(&dec->cur, &dec->refn[0]); + dec->is_edged[0] = 0; SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); dec->last_reduced_resolution = reduced_resolution; dec->last_coding_type = coding_type; @@ -1551,7 +1546,7 @@ if (dec->low_delay) { DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); - dec->low_delay = 1; + dec->low_delay = 0; } if (dec->frames < 2) {