--- decoder.c 2002/07/21 23:34:07 1.33 +++ decoder.c 2002/10/05 21:34:21 1.37.2.2 @@ -55,7 +55,7 @@ * 22.12.2001 lock based interpolation * 01.12.2001 inital version; (c)2001 peter ross * - * $Id: decoder.c,v 1.33 2002/07/21 23:34:07 chl Exp $ + * $Id: decoder.c,v 1.37.2.2 2002/10/05 21:34:21 Isibaar Exp $ * *************************************************************************/ @@ -136,6 +136,15 @@ return XVID_ERR_MEMORY; } + if (image_create(&dec->refh, 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->refn[2], dec->edged_width, dec->edged_height); + xvid_free(dec); + return XVID_ERR_MEMORY; + } + dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); @@ -144,6 +153,7 @@ 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->refh, dec->edged_width, dec->edged_height); xvid_free(dec); return XVID_ERR_MEMORY; } @@ -161,6 +171,7 @@ 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->refh, dec->edged_width, dec->edged_height); xvid_free(dec); return XVID_ERR_MEMORY; } @@ -186,6 +197,7 @@ 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->refh, dec->edged_width, dec->edged_height); image_destroy(&dec->cur, dec->edged_width, dec->edged_height); xvid_free(dec); @@ -268,8 +280,10 @@ start_timer(); if (cbp & (1 << (5 - i))) // coded { - get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], - start_coeff); + int direction = dec->alternate_vertical_scan ? + 2 : pMB->acpred_directions[i]; + + get_intra_block(bs, &block[i * 64], direction, start_coeff); } stop_coding_timer(); @@ -348,36 +362,61 @@ uv_dx = pMB->mvs[0].x; uv_dy = pMB->mvs[0].y; + if (dec->quarterpel) + { + uv_dx = (uv_dx >> 1) | (uv_dx & 1); + uv_dy = (uv_dy >> 1) | (uv_dy & 1); + } + 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)); + + if (dec->quarterpel) + { + sum /= 2; + } + + uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; - uv_dy = - (sum == - 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + - (ABS(sum) / 16) * 2)); + + if (dec->quarterpel) + { + sum /= 2; + } + + uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); } start_timer(); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, 16 * y_pos, - pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos + 8, - 16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, - rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, - 16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, - rounding); - interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos + 8, - 16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, - rounding); + 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); + } + 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, @@ -385,12 +424,14 @@ stop_comp_timer(); 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]); + get_inter_block(bs, &block[i * 64], direction); stop_coding_timer(); start_timer(); @@ -563,7 +604,7 @@ start_timer(); image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, - dec->width, dec->height, dec->interlacing); + dec->width, dec->height); stop_edges_timer(); bound = 0; @@ -631,8 +672,10 @@ mb->quant = quant; if (dec->interlacing) { - mb->field_dct = BitstreamGetBit(bs); - DEBUG1("decp: field_dct: ", mb->field_dct); + if (cbp || intra) { + 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); @@ -661,8 +704,8 @@ 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 */ ) { + } else if (mb->mode == MODE_INTER4V ) { + get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); @@ -682,7 +725,7 @@ rounding); } else // not coded { - //DEBUG2("P-frame MB at (X,Y)=",x,y); + 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; @@ -852,12 +895,14 @@ stop_comp_timer(); 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]); + get_inter_block(bs, &block[i * 64], direction); stop_coding_timer(); start_timer(); @@ -1003,27 +1048,47 @@ 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); + interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, + dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, + dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos, + stride, 0); + + 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->refn[2].y + (16 * y_pos * stride) + 16 * x_pos + 8, + stride, 0); + + 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->refn[2].y + (16 * (y_pos + 8) * stride) + 16 * x_pos, + stride, 0); + + 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->refn[2].y + (16 * (y_pos + 8) * stride) + 16 * x_pos + 8, + stride, 0); + + interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride) + 8 * x_pos, + dec->cur.u + (8 * y_pos * stride) + 8 * x_pos, + dec->refn[2].u + (8 * y_pos * stride) + 8 * x_pos, + stride2, 0); + + interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride) + 8 * x_pos, + dec->cur.v + (8 * y_pos * stride) + 8 * x_pos, + dec->refn[2].v + (8 * y_pos * stride) + 8 * x_pos, + stride2, 0); + stop_comp_timer(); 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]); + get_inter_block(bs, &block[i * 64], direction); stop_coding_timer(); start_timer(); @@ -1118,9 +1183,9 @@ start_timer(); image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, - dec->width, dec->height, dec->interlacing); + dec->width, dec->height); image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, - dec->width, dec->height, dec->interlacing); + dec->width, dec->height); stop_edges_timer(); #ifdef BFRAMES_DEC_DEBUG @@ -1255,17 +1320,8 @@ default: DEBUG1("Not support B-frame mb_type =", mb->mb_type); - ; } - if ( (x==19) && (y==8) ) - { - fprintf(stderr,"D %d %d %d %d %d %d \n",0, mb->mb_type, - mb->mvs[0].x, mb->mvs[0].y,mb->b_mvs[0].x, mb->b_mvs[0].y ); - } - - - } // end of FOR } #ifdef BFRAMES_DEC_DEBUG @@ -1364,14 +1420,14 @@ #ifdef BFRAMES_DEC // test if no B_VOP - if (dec->low_delay) { + if (dec->low_delay || dec->frames == 0) { #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) { + if (dec->frames >= 1) { start_timer(); if ((vop_type == I_VOP || vop_type == P_VOP)) { image_output(&dec->refn[0], dec->width, dec->height,