--- decoder.c 2002/11/20 22:28:25 1.37.2.14 +++ decoder.c 2002/12/13 11:26:41 1.37.2.20 @@ -55,7 +55,7 @@ * 22.12.2001 lock based interpolation * 01.12.2001 inital version; (c)2001 peter ross * - * $Id: decoder.c,v 1.37.2.14 2002/11/20 22:28:25 Isibaar Exp $ + * $Id: decoder.c,v 1.37.2.20 2002/12/13 11:26:41 suxen_drol Exp $ * *************************************************************************/ @@ -79,6 +79,8 @@ #include "dct/fdct.h" #include "utils/mem_transfer.h" #include "image/interpolate8x8.h" +#include "image/reduced.h" +#include "image/font.h" #include "bitstream/mbcoding.h" #include "prediction/mbprediction.h" @@ -215,9 +217,10 @@ // add by chenm001 // for support B-frame to save reference frame's time - dec->frames = -1; + dec->frames = 0; dec->time = dec->time_base = dec->last_time_base = 0; dec->low_delay = 0; + dec->packed_mode = 0; dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); @@ -265,7 +268,8 @@ Bitstream * bs, const uint32_t quant, const uint32_t intra_dc_threshold, - const unsigned int bound) + const unsigned int bound, + const int reduced_resolution) { DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); @@ -278,9 +282,15 @@ 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); + if (reduced_resolution) { + pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); + pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); + pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); + }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); + } memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear @@ -341,6 +351,7 @@ start_timer(); idct(&data[i * 64]); stop_idct_timer(); + } if (dec->interlacing && pMB->field_dct) { @@ -349,12 +360,24 @@ } 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); + + if (reduced_resolution) + { + next_block*=2; + copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); + copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); + copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); + copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); + copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); + copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); + }else{ + 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(); } @@ -376,7 +399,8 @@ const uint32_t cbp, Bitstream * bs, const uint32_t quant, - const uint32_t rounding) + const uint32_t rounding, + const int reduced_resolution) { DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); @@ -384,19 +408,33 @@ uint32_t stride = dec->edged_width; uint32_t stride2 = stride / 2; - uint32_t next_block = stride * 8; + uint32_t next_block = stride * (reduced_resolution ? 16 : 8); uint32_t i; uint32_t iQuant = pMB->quant; uint8_t *pY_Cur, *pU_Cur, *pV_Cur; + int uv_dx, uv_dy; + VECTOR mv[4]; /* local copy of mvs */ - 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 (reduced_resolution) { + pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); + pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); + pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); + for (i = 0; i < 4; i++) { + mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); + mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); + } + }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 = pMB->mvs[0].x; - uv_dy = pMB->mvs[0].y; + uv_dx = mv[0].x; + uv_dy = mv[0].y; if (dec->quarterpel) { @@ -408,75 +446,102 @@ uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; start_timer(); - if(dec->quarterpel) { - interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, - dec->refh.y + 128, 16*x_pos, 16*y_pos, - pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); - } - else { - 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); + if (reduced_resolution) + { + interpolate32x32_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.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, + uv_dx, uv_dy, stride2, rounding); + } + else + { + if(dec->quarterpel) { + interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, + dec->refh.y + 128, 16*x_pos, 16*y_pos, + 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); + } - 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.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(); - } else { + } else { /* MODE_INTER4V */ int sum; if(dec->quarterpel) - sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); + sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); else - sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; + sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; if(dec->quarterpel) - sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); + sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); else - sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; + sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; start_timer(); - if(dec->quarterpel) { - interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, - dec->refh.y + 128, 16*x_pos, 16*y_pos, - pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); - interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, - dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, - pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); - interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, - dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, - pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); - interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, - dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, - pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); - } - else { - 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); - } + if (reduced_resolution) + { + 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, + mv[1].x, mv[1].y, stride, rounding); + 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, + mv[3].x, mv[3].y, stride, rounding); + 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, + uv_dx, uv_dy, stride2, rounding); + + // set_block(pY_Cur, stride, 32, 32, 127); + } + else + { + if(dec->quarterpel) { + interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, + dec->refh.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->refh.y, dec->refh.y + 64, + dec->refh.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->refh.y, dec->refh.y + 64, + dec->refh.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->refh.y, dec->refh.y + 64, + dec->refh.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, + mv[0].x, mv[0].y, stride, rounding); + 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, + 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, + mv[3].x, mv[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.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(); } @@ -511,18 +576,36 @@ } 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); + 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(); } @@ -541,8 +624,8 @@ if (reduced_resolution) { - mb_width /= 2; - mb_height /= 2; + mb_width = (dec->width + 31) / 32; + mb_height = (dec->height + 31) / 32; } bound = 0; @@ -563,8 +646,8 @@ { bound = read_video_packet_header(bs, dec, 0, &quant, NULL, NULL, &intra_dc_threshold); - x = bound % dec->mb_width; - y = bound / dec->mb_width; + x = bound % mb_width; + y = bound / mb_width; } mb = &dec->mbs[y * dec->mb_width + x]; @@ -599,11 +682,11 @@ } decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, - intra_dc_threshold, bound); + intra_dc_threshold, bound, reduced_resolution); } if(dec->out_frm) - output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width); + output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); } } @@ -615,7 +698,7 @@ int x, int y, int k, - VECTOR * mv, + VECTOR * ret_mv, int fcode, const int bound) { @@ -626,32 +709,32 @@ int range = (64 * scale_fac); VECTOR pmv; - int mv_x, mv_y; + VECTOR mv; 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 = 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)", mv.x, mv.y, pmv.x, pmv.y); - mv_x += pmv.x; - mv_y += 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.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; + if (mv.y < low) { + mv.y += range; + } else if (mv.y > high) { + mv.y -= range; } - mv->x = mv_x; - mv->y = mv_y; + ret_mv->x = mv.x; + ret_mv->y = mv.y; } @@ -685,6 +768,14 @@ uint32_t x, y; uint32_t bound; int cp_mb, st_mb; + int mb_width = dec->mb_width; + int mb_height = dec->mb_height; + + if (reduced_resolution) + { + mb_width = (dec->width + 31) / 32; + mb_height = (dec->height + 31) / 32; + } start_timer(); image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, @@ -693,9 +784,9 @@ bound = 0; - for (y = 0; y < dec->mb_height; y++) { + for (y = 0; y < mb_height; y++) { cp_mb = st_mb = 0; - for (x = 0; x < dec->mb_width; x++) { + for (x = 0; x < mb_width; x++) { MACROBLOCK *mb; // skip stuffing @@ -706,8 +797,8 @@ { bound = read_video_packet_header(bs, dec, fcode - 1, &quant, &fcode, NULL, &intra_dc_threshold); - x = bound % dec->mb_width; - y = bound / dec->mb_width; + x = bound % mb_width; + y = bound / mb_width; } mb = &dec->mbs[y * dec->mb_width + x]; @@ -814,12 +905,12 @@ 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); + intra_dc_threshold, bound, reduced_resolution); continue; } decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, - rounding); + rounding, reduced_resolution); } else if (gmc_mv) /* not coded S_VOP macroblock */ @@ -827,50 +918,46 @@ mb->mode = MODE_NOT_CODED; 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); + decoder_mbinter(dec, mb, x, y, 0, 0, bs, quant, rounding, reduced_resolution); } else /* not coded P_VOP macroblock */ { 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); + if (reduced_resolution) + { + transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), + dec->refn[0].y + (32*y)*dec->edged_width + (32*x), + dec->edged_width); + + transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), + dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), + dec->edged_width/2); + + transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), + dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), + dec->edged_width/2); + } + else + { + transfer16x16_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.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(); @@ -1519,6 +1606,7 @@ *mb2 = temp; } + int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) @@ -1534,9 +1622,13 @@ VECTOR gmc_mv[5]; uint32_t vop_type; int success = 0; + int output = 0; + int seen_something = 0; start_global_timer(); + dec->low_delay_default = (frame->general & XVID_DEC_LOWDELAY); + dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; BitstreamInit(&bs, frame->bitstream, frame->length); @@ -1556,17 +1648,20 @@ start: // add by chenm001 // for support B-frame to reference last 2 frame - dec->frames++; xxx: vop_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, gmc_mv); - DPRINTF(DPRINTF_HEADER, "vop_type=%i", vop_type); + DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%i, time_pp=%i, time_bp=%i", + vop_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); - if (vop_type == -1 && success) - goto done; + if (vop_type == - 1) + { + if (success) goto done; + return XVID_ERR_FAIL; + } if (vop_type == -2 || vop_type == -3) { @@ -1592,111 +1687,138 @@ 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, reduced_resolution, quant, - fcode_forward, intra_dc_threshold, NULL); -#ifdef BFRAMES_DEC - DEBUG1("P_VOP Time=", dec->time); -#endif - break; - case I_VOP: - decoder_iframe(dec, &bs, reduced_resolution, 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!"); + /* packed_mode: special-N_VOP treament */ + if (dec->packed_mode && vop_type == N_VOP) + { + if (dec->low_delay_default) + { + image_output(&dec->refn[0], dec->width, dec->height, + dec->edged_width, frame->image, frame->stride, + frame->colorspace, dec->interlacing); + output = 1; } -#else - image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); -#endif - break; - - case S_VOP : - decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, - fcode_forward, intra_dc_threshold, gmc_mv); - 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); -#ifdef BFRAMES_DEC - DEBUG1("N_VOP Time=", dec->time); -#endif - break; - - default: - if (stats) - stats->notify = 0; - - emms(); - return XVID_ERR_FAIL; + /* ignore otherwise */ } + else if (vop_type != B_VOP) + { + switch(vop_type) + { + case I_VOP : + decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); + break; + case P_VOP : + decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, + fcode_forward, intra_dc_threshold, NULL); + break; + case S_VOP : + decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, + fcode_forward, intra_dc_threshold, gmc_mv); + break; + case N_VOP : + image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); + break; + } - BitstreamByteAlign(&bs); - -#ifdef BFRAMES_DEC - // test if no B_VOP - if (dec->low_delay || dec->frames == 0 || ((dec->packed_mode) && !(frame->length > BitstreamPos(&bs) / 8))) { -#endif - image_output(&dec->cur, dec->width, dec->height, dec->edged_width, - frame->image, frame->stride, frame->colorspace, dec->interlacing); + if (reduced_resolution) + { + image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, + (dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width); + } -#ifdef BFRAMES_DEC - } else { - if (dec->frames >= 1 && !(dec->packed_mode)) { - start_timer(); - if ((vop_type == I_VOP || vop_type == P_VOP || vop_type == S_VOP)) { - image_output(&dec->refn[0], dec->width, dec->height, + /* note: for packed_mode, output is performed when the special-N_VOP is decoded */ + if (!(dec->low_delay_default && dec->packed_mode)) + { + if (dec->low_delay) + { + image_output(&dec->cur, dec->width, dec->height, dec->edged_width, frame->image, frame->stride, frame->colorspace, dec->interlacing); - } else if (vop_type == B_VOP) { - image_output(&dec->cur, dec->width, dec->height, + output = 1; + } + else if (dec->frames > 0) /* is the reference frame valid? */ + { + image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, frame->image, frame->stride, frame->colorspace, dec->interlacing); + output = 1; } - stop_conv_timer(); } - } -#endif - if (vop_type == I_VOP || vop_type == P_VOP || vop_type == S_VOP) { image_swap(&dec->refn[0], &dec->refn[1]); image_swap(&dec->cur, &dec->refn[0]); + mb_swap(&dec->mbs, &dec->last_mbs); + + dec->frames++; + seen_something = 1; - // 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); + }else{ /* B_VOP */ + + if (dec->low_delay) + { + DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream"); + dec->low_delay = 1; + } + + if (dec->time_pp <= dec->time_bp) { + /* this occurs when dx50_bvop_compatibility==0 sequences are + decoded in vfw. */ + image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, + "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); + }else{ + decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); + } + + image_output(&dec->cur, dec->width, dec->height, + dec->edged_width, frame->image, frame->stride, + frame->colorspace, dec->interlacing); + output = 1; + dec->frames++; } - success = 1; + emms(); + + BitstreamByteAlign(&bs); - if (frame->length > BitstreamPos(&bs) / 8) // multiple vops packed together + /* low_delay_default mode: repeat in packed_mode */ + if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) + { + success = 1; goto start; + } done : + /* low_delay_default mode: if we've gotten here without outputting anything, + then output the recently decoded frame, or print an error message */ + if (dec->low_delay_default && output == 0) + { + if (dec->packed_mode && seen_something) + { + /* output the recently decoded frame */ + image_output(&dec->refn[0], dec->width, dec->height, + dec->edged_width, frame->image, frame->stride, + frame->colorspace, dec->interlacing); + } + else + { + image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, + "warning: nothing to output"); + image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, + "bframe decoder lag"); + + image_output(&dec->cur, dec->width, dec->height, + dec->edged_width, frame->image, frame->stride, + frame->colorspace, dec->interlacing); + } + output = 1; + } + frame->length = BitstreamPos(&bs) / 8; if (stats) { - stats->notify = XVID_DEC_VOP; + 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 }