X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fvc1.c;h=229c8850600a9a1ed1ff94d96808614fe3bc1ac7;hb=6708cfc4d3a079527409b2ba2fc75997a7975104;hp=6c478d1c41c4f3aa3459bb88a1e95063baadb432;hpb=f87d2632b3495246abea6c49d8c66e61f71ae618;p=ffmpeg diff --git a/libavcodec/vc1.c b/libavcodec/vc1.c index 6c478d1c41c..229c8850600 100644 --- a/libavcodec/vc1.c +++ b/libavcodec/vc1.c @@ -21,10 +21,11 @@ */ /** - * @file vc1.c + * @file libavcodec/vc1.c * VC-1 and WMV3 decoder * */ +#include "internal.h" #include "dsputil.h" #include "avcodec.h" #include "mpegvideo.h" @@ -34,6 +35,8 @@ #include "msmpeg4data.h" #include "unary.h" #include "simple_idct.h" +#include "mathops.h" +#include "vdpau_internal.h" #undef NDEBUG #include @@ -44,13 +47,6 @@ static const uint16_t table_mb_intra[64][2]; -static inline int decode210(GetBitContext *gb){ - if (get_bits1(gb)) - return 0; - else - return 2 - get_bits1(gb); -} - /** * Init VC-1 specific tables and VC1Context members * @param v The VC1Context to initialize @@ -121,12 +117,12 @@ static int vc1_init_common(VC1Context *v) /***********************************************************************/ /** - * @defgroup bitplane VC9 Bitplane decoding + * @defgroup vc1bitplane VC-1 Bitplane decoding * @see 8.7, p56 * @{ */ -/** @addtogroup bitplane +/** * Imode types * @{ */ @@ -182,7 +178,8 @@ static void decode_colskip(uint8_t* plane, int width, int height, int stride, Ge } /** Decode a bitplane's bits - * @param bp Bitplane where to store the decode bits + * @param data bitplane where to store the decode bits + * @param[out] raw_flag pointer to the flag indicating that this bitplane is not coded explicitly * @param v VC-1 context for bit reading and logging * @return Status * @todo FIXME: Optimize @@ -313,6 +310,91 @@ static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v) /** @} */ //Bitplane group +#define FILTSIGN(a) ((a) >= 0 ? 1 : -1) +/** + * VC-1 in-loop deblocking filter for one line + * @param src source block type + * @param stride block stride + * @param pq block quantizer + * @return whether other 3 pairs should be filtered or not + * @see 8.6 + */ +static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){ + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3; + int a0_sign = a0 >> 31; /* Store sign */ + a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */ + if(a0 < pq){ + int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3); + int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3); + if(a1 < a0 || a2 < a0){ + int clip = src[-1*stride] - src[ 0*stride]; + int clip_sign = clip >> 31; + clip = ((clip ^ clip_sign) - clip_sign)>>1; + if(clip){ + int a3 = FFMIN(a1, a2); + int d = 5 * (a3 - a0); + int d_sign = (d >> 31); + d = ((d ^ d_sign) - d_sign) >> 3; + d_sign ^= a0_sign; + + if( d_sign ^ clip_sign ) + d = 0; + else{ + d = FFMIN(d, clip); + d = (d ^ d_sign) - d_sign; /* Restore sign */ + src[-1*stride] = cm[src[-1*stride] - d]; + src[ 0*stride] = cm[src[ 0*stride] + d]; + } + return 1; + } + } + } + return 0; +} + +/** + * VC-1 in-loop deblocking filter + * @param src source block type + * @param step distance between horizontally adjacent elements + * @param stride distance between vertically adjacent elements + * @param len edge length to filter (4 or 8 pixels) + * @param pq block quantizer + * @see 8.6 + */ +static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq) +{ + int i; + int filt3; + + for(i = 0; i < len; i += 4){ + filt3 = vc1_filter_line(src + 2*step, stride, pq); + if(filt3){ + vc1_filter_line(src + 0*step, stride, pq); + vc1_filter_line(src + 1*step, stride, pq); + vc1_filter_line(src + 3*step, stride, pq); + } + src += step * 4; + } +} + +static void vc1_loop_filter_iblk(MpegEncContext *s, int pq) +{ + int i, j; + if(!s->first_slice_line) + vc1_loop_filter(s->dest[0], 1, s->linesize, 16, pq); + vc1_loop_filter(s->dest[0] + 8*s->linesize, 1, s->linesize, 16, pq); + for(i = !s->mb_x*8; i < 16; i += 8) + vc1_loop_filter(s->dest[0] + i, s->linesize, 1, 16, pq); + for(j = 0; j < 2; j++){ + if(!s->first_slice_line) + vc1_loop_filter(s->dest[j+1], 1, s->uvlinesize, 8, pq); + if(s->mb_x) + vc1_loop_filter(s->dest[j+1], s->uvlinesize, 1, 8, pq); + } +} + /***********************************************************************/ /** VOP Dquant decoding * @param v VC-1 Context @@ -407,7 +489,7 @@ static void vc1_mc_1mv(VC1Context *v, int dir) my = s->mv[dir][0][1]; // store motion vectors for further use in B frames - if(s->pict_type == P_TYPE) { + if(s->pict_type == FF_P_TYPE) { s->current_picture.motion_val[1][s->block_index[0]][0] = mx; s->current_picture.motion_val[1][s->block_index[0]][1] = my; } @@ -784,10 +866,14 @@ static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb) if (v->profile == PROFILE_ADVANCED) { + v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz; + v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz; return decode_sequence_header_adv(v, gb); } else { + v->zz_8x4 = wmv2_scantableA; + v->zz_4x8 = wmv2_scantableB; v->res_sm = get_bits(gb, 2); //reserved if (v->res_sm) { @@ -807,13 +893,18 @@ static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb) av_log(avctx, AV_LOG_ERROR, "LOOPFILTER shell not be enabled in simple profile\n"); } + if(v->s.avctx->skip_loop_filter >= AVDISCARD_ALL) + v->s.loop_filter = 0; v->res_x8 = get_bits1(gb); //reserved v->multires = get_bits1(gb); v->res_fasttx = get_bits1(gb); if (!v->res_fasttx) { - v->s.dsp.vc1_inv_trans_8x8 = simple_idct; + v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct; + v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add; + v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add; + v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add; } v->fastuvmc = get_bits1(gb); //common @@ -932,8 +1023,8 @@ static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb) if(get_bits1(gb)) { //Display Info - decoding is not affected by it int w, h, ar = 0; av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n"); - v->s.avctx->width = v->s.width = w = get_bits(gb, 14) + 1; - v->s.avctx->height = v->s.height = h = get_bits(gb, 14) + 1; + v->s.avctx->coded_width = w = get_bits(gb, 14) + 1; + v->s.avctx->coded_height = h = get_bits(gb, 14) + 1; av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h); if(get_bits1(gb)) ar = get_bits(gb, 4); @@ -944,6 +1035,7 @@ static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb) h = get_bits(gb, 8); v->s.avctx->sample_aspect_ratio = (AVRational){w, h}; } + av_log(v->s.avctx, AV_LOG_DEBUG, "Aspect: %i:%i\n", v->s.avctx->sample_aspect_ratio.num, v->s.avctx->sample_aspect_ratio.den); if(get_bits1(gb)){ //framerate stuff if(get_bits1(gb)) { @@ -984,13 +1076,13 @@ static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb) static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb) { VC1Context *v = avctx->priv_data; - int i, blink, clentry, refdist; + int i; av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32)); - blink = get_bits1(gb); // broken link - clentry = get_bits1(gb); // closed entry + v->broken_link = get_bits1(gb); + v->closed_entry = get_bits1(gb); v->panscanflag = get_bits1(gb); - refdist = get_bits1(gb); // refdist flag + v->refdist_flag = get_bits1(gb); v->s.loop_filter = get_bits1(gb); v->fastuvmc = get_bits1(gb); v->extended_mv = get_bits1(gb); @@ -1011,20 +1103,20 @@ static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb) } if(v->extended_mv) v->extended_dmv = get_bits1(gb); - if(get_bits1(gb)) { + if((v->range_mapy_flag = get_bits1(gb))) { av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n"); - skip_bits(gb, 3); // Y range, ignored for now + v->range_mapy = get_bits(gb, 3); } - if(get_bits1(gb)) { + if((v->range_mapuv_flag = get_bits1(gb))) { av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n"); - skip_bits(gb, 3); // UV range, ignored for now + v->range_mapuv = get_bits(gb, 3); } av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n" "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n" "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n" "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n", - blink, clentry, v->panscanflag, refdist, v->s.loop_filter, + v->broken_link, v->closed_entry, v->panscanflag, v->refdist_flag, v->s.loop_filter, v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode); return 0; @@ -1041,30 +1133,31 @@ static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb) v->s.pict_type = get_bits1(gb); if (v->s.avctx->max_b_frames) { if (!v->s.pict_type) { - if (get_bits1(gb)) v->s.pict_type = I_TYPE; - else v->s.pict_type = B_TYPE; - } else v->s.pict_type = P_TYPE; - } else v->s.pict_type = v->s.pict_type ? P_TYPE : I_TYPE; + if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE; + else v->s.pict_type = FF_B_TYPE; + } else v->s.pict_type = FF_P_TYPE; + } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE; v->bi_type = 0; - if(v->s.pict_type == B_TYPE) { - v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1); - v->bfraction = ff_vc1_bfraction_lut[v->bfraction]; + if(v->s.pict_type == FF_B_TYPE) { + v->bfraction_lut_index = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1); + v->bfraction = ff_vc1_bfraction_lut[v->bfraction_lut_index]; if(v->bfraction == 0) { - v->s.pict_type = BI_TYPE; + v->s.pict_type = FF_BI_TYPE; } } - if(v->s.pict_type == I_TYPE || v->s.pict_type == BI_TYPE) + if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) skip_bits(gb, 7); // skip buffer fullness /* calculate RND */ - if(v->s.pict_type == I_TYPE || v->s.pict_type == BI_TYPE) + if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) v->rnd = 1; - if(v->s.pict_type == P_TYPE) + if(v->s.pict_type == FF_P_TYPE) v->rnd ^= 1; /* Quantizer stuff */ pqindex = get_bits(gb, 5); + if(!pqindex) return -1; if (v->quantizer_mode == QUANT_FRAME_IMPLICIT) v->pq = ff_vc1_pquant_table[0][pqindex]; else @@ -1086,23 +1179,18 @@ static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb) v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11 v->range_x = 1 << (v->k_x - 1); v->range_y = 1 << (v->k_y - 1); - if (v->profile == PROFILE_ADVANCED) - { - if (v->postprocflag) v->postproc = get_bits1(gb); - } - else - if (v->multires && v->s.pict_type != B_TYPE) v->respic = get_bits(gb, 2); + if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2); - if(v->res_x8 && (v->s.pict_type == I_TYPE || v->s.pict_type == BI_TYPE)){ + if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){ v->x8_type = get_bits1(gb); }else v->x8_type = 0; //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n", -// (v->s.pict_type == P_TYPE) ? 'P' : ((v->s.pict_type == I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm); +// (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm); - if(v->s.pict_type == I_TYPE || v->s.pict_type == P_TYPE) v->use_ic = 0; + if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0; switch(v->s.pict_type) { - case P_TYPE: + case FF_P_TYPE: if (v->pq < 5) v->tt_index = 0; else if(v->pq < 13) v->tt_index = 1; else v->tt_index = 2; @@ -1185,7 +1273,7 @@ static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb) v->ttfrm = TT_8X8; } break; - case B_TYPE: + case FF_B_TYPE: if (v->pq < 5) v->tt_index = 0; else if(v->pq < 13) v->tt_index = 1; else v->tt_index = 2; @@ -1232,7 +1320,7 @@ static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb) { /* AC Syntax */ v->c_ac_table_index = decode012(gb); - if (v->s.pict_type == I_TYPE || v->s.pict_type == BI_TYPE) + if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) { v->y_ac_table_index = decode012(gb); } @@ -1240,8 +1328,8 @@ static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb) v->s.dc_table_index = get_bits1(gb); } - if(v->s.pict_type == BI_TYPE) { - v->s.pict_type = B_TYPE; + if(v->s.pict_type == FF_BI_TYPE) { + v->s.pict_type = FF_B_TYPE; v->bi_type = 1; } return 0; @@ -1260,19 +1348,19 @@ static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb) } switch(get_unary(gb, 0, 4)) { case 0: - v->s.pict_type = P_TYPE; + v->s.pict_type = FF_P_TYPE; break; case 1: - v->s.pict_type = B_TYPE; + v->s.pict_type = FF_B_TYPE; break; case 2: - v->s.pict_type = I_TYPE; + v->s.pict_type = FF_I_TYPE; break; case 3: - v->s.pict_type = BI_TYPE; + v->s.pict_type = FF_BI_TYPE; break; case 4: - v->s.pict_type = P_TYPE; // skipped pic + v->s.pict_type = FF_P_TYPE; // skipped pic v->p_frame_skipped = 1; return 0; } @@ -1293,14 +1381,15 @@ static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb) if(v->interlace) v->uvsamp = get_bits1(gb); if(v->finterpflag) v->interpfrm = get_bits1(gb); - if(v->s.pict_type == B_TYPE) { - v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1); - v->bfraction = ff_vc1_bfraction_lut[v->bfraction]; + if(v->s.pict_type == FF_B_TYPE) { + v->bfraction_lut_index = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1); + v->bfraction = ff_vc1_bfraction_lut[v->bfraction_lut_index]; if(v->bfraction == 0) { - v->s.pict_type = BI_TYPE; /* XXX: should not happen here */ + v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */ } } pqindex = get_bits(gb, 5); + if(!pqindex) return -1; v->pqindex = pqindex; if (v->quantizer_mode == QUANT_FRAME_IMPLICIT) v->pq = ff_vc1_pquant_table[0][pqindex]; @@ -1317,12 +1406,14 @@ static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb) else v->halfpq = 0; if (v->quantizer_mode == QUANT_FRAME_EXPLICIT) v->pquantizer = get_bits1(gb); + if(v->postprocflag) + v->postproc = get_bits(gb, 2); - if(v->s.pict_type == I_TYPE || v->s.pict_type == P_TYPE) v->use_ic = 0; + if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0; switch(v->s.pict_type) { - case I_TYPE: - case BI_TYPE: + case FF_I_TYPE: + case FF_BI_TYPE: status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v); if (status < 0) return -1; av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: " @@ -1338,9 +1429,7 @@ static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb) } } break; - case P_TYPE: - if(v->postprocflag) - v->postproc = get_bits1(gb); + case FF_P_TYPE: if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3); else v->mvrange = 0; v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13 @@ -1429,9 +1518,7 @@ static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb) v->ttfrm = TT_8X8; } break; - case B_TYPE: - if(v->postprocflag) - v->postproc = get_bits1(gb); + case FF_B_TYPE: if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3); else v->mvrange = 0; v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13 @@ -1483,20 +1570,20 @@ static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb) /* AC Syntax */ v->c_ac_table_index = decode012(gb); - if (v->s.pict_type == I_TYPE || v->s.pict_type == BI_TYPE) + if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) { v->y_ac_table_index = decode012(gb); } /* DC Syntax */ v->s.dc_table_index = get_bits1(gb); - if ((v->s.pict_type == I_TYPE || v->s.pict_type == BI_TYPE) && v->dquant) { + if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) { av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n"); vop_dquant_decoding(v); } v->bi_type = 0; - if(v->s.pict_type == BI_TYPE) { - v->s.pict_type = B_TYPE; + if(v->s.pict_type == FF_BI_TYPE) { + v->s.pict_type = FF_B_TYPE; v->bi_type = 1; } return 0; @@ -1504,7 +1591,7 @@ static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb) /***********************************************************************/ /** - * @defgroup block VC-1 Block-level functions + * @defgroup vc1block VC-1 Block-level functions * @see 7.1.4, p91 and 8.1.1.7, p(1)04 * @{ */ @@ -2090,6 +2177,8 @@ static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int /** Get predicted DC value for I-frames only * prediction dir: left=0, top=1 * @param s MpegEncContext + * @param overlap flag indicating that overlap filtering is used + * @param pq integer part of picture quantizer * @param[in] n block index in the current MB * @param dc_val_ptr Pointer to DC predictor * @param dir_ptr Prediction direction for use in AC prediction @@ -2150,7 +2239,11 @@ static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n, /** Get predicted DC value * prediction dir: left=0, top=1 * @param s MpegEncContext + * @param overlap flag indicating that overlap filtering is used + * @param pq integer part of picture quantizer * @param[in] n block index in the current MB + * @param a_avail flag indicating top block availability + * @param c_avail flag indicating left block availability * @param dc_val_ptr Pointer to DC predictor * @param dir_ptr Prediction direction for use in AC prediction */ @@ -2221,9 +2314,10 @@ static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n, return pred; } +/** @} */ // Block group /** - * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles + * @defgroup vc1_std_mb VC1 Macroblock-level functions in Simple/Main Profiles * @see 7.1.4, p91 and 8.1.1.7, p(1)04 * @{ */ @@ -2260,6 +2354,7 @@ static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **code * @param last Last coefficient * @param skip How much zero coefficients to skip * @param value Decoded AC coefficient value + * @param codingset set of VLC to decode data * @see 8.1.3.4 */ static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset) @@ -2323,6 +2418,7 @@ static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, /** Decode intra block in intra frames - should be faster than decode_intra_block * @param v VC1Context * @param block block to decode + * @param[in] n subblock index * @param coded are AC coeffs present or not * @param codingset set of VLC to decode data */ @@ -2396,11 +2492,11 @@ static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded if(v->s.ac_pred) { if(!dc_pred_dir) - zz_table = ff_vc1_horizontal_zz; + zz_table = wmv1_scantable[2]; else - zz_table = ff_vc1_vertical_zz; + zz_table = wmv1_scantable[3]; } else - zz_table = ff_vc1_normal_zz; + zz_table = wmv1_scantable[1]; ac_val = s->ac_val[0][0] + s->block_index[n] * 16; ac_val2 = ac_val; @@ -2488,8 +2584,10 @@ not_coded: /** Decode intra block in intra frames - should be faster than decode_intra_block * @param v VC1Context * @param block block to decode + * @param[in] n subblock number * @param coded are AC coeffs present or not * @param codingset set of VLC to decode data + * @param mquant quantizer value for this macroblock */ static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant) { @@ -2579,11 +2677,11 @@ static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int c if(v->s.ac_pred) { if(!dc_pred_dir) - zz_table = ff_vc1_horizontal_zz; + zz_table = wmv1_scantable[2]; else - zz_table = ff_vc1_vertical_zz; + zz_table = wmv1_scantable[3]; } else - zz_table = ff_vc1_normal_zz; + zz_table = wmv1_scantable[1]; while (!last) { vc1_decode_ac_coeff(v, &last, &skip, &value, codingset); @@ -2684,6 +2782,7 @@ static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int c /** Decode intra block in inter frames - more generic version than vc1_decode_i_block * @param v VC1Context * @param block block to decode + * @param[in] n subblock index * @param coded are AC coeffs present or not * @param mquant block quantizer * @param codingset set of VLC to decode data @@ -2784,7 +2883,7 @@ static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int c const int8_t *zz_table; int k; - zz_table = ff_vc1_simple_progressive_8x8_zz; + zz_table = wmv1_scantable[0]; while (!last) { vc1_decode_ac_coeff(v, &last, &skip, &value, codingset); @@ -2884,7 +2983,8 @@ static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int c /** Decode P block */ -static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block) +static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block, + uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left) { MpegEncContext *s = &v->s; GetBitContext *gb = &s->gb; @@ -2892,6 +2992,7 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan int subblkpat = 0; int scale, off, idx, last, skip, value; int ttblk = ttmb & 7; + int pat = 0; if(ttmb == -1) { ttblk = ff_vc1_ttblk_to_tt[v->tt_index][get_vlc2(gb, ff_vc1_ttblk_vlc[v->tt_index].table, VC1_TTBLK_VLC_BITS, 1)]; @@ -2918,6 +3019,7 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan } switch(ttblk) { case TT_8X8: + pat = 0xF; i = 0; last = 0; while (!last) { @@ -2925,14 +3027,22 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan i += skip; if(i > 63) break; - idx = ff_vc1_simple_progressive_8x8_zz[i++]; + idx = wmv1_scantable[0][i++]; block[idx] = value * scale; if(!v->pquantizer) block[idx] += (block[idx] < 0) ? -mquant : mquant; } - s->dsp.vc1_inv_trans_8x8(block); + if(!skip_block){ + s->dsp.vc1_inv_trans_8x8(block); + s->dsp.add_pixels_clamped(block, dst, linesize); + if(apply_filter && cbp_top & 0xC) + vc1_loop_filter(dst, 1, linesize, 8, mquant); + if(apply_filter && cbp_left & 0xA) + vc1_loop_filter(dst, linesize, 1, 8, mquant); + } break; case TT_4X4: + pat = ~subblkpat & 0xF; for(j = 0; j < 4; j++) { last = subblkpat & (1 << (3 - j)); i = 0; @@ -2947,11 +3057,17 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan if(!v->pquantizer) block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant; } - if(!(subblkpat & (1 << (3 - j)))) - s->dsp.vc1_inv_trans_4x4(block, j); + if(!(subblkpat & (1 << (3 - j))) && !skip_block){ + s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off); + if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2))))) + vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, 1, linesize, 4, mquant); + if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1))))) + vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, linesize, 1, 4, mquant); + } } break; case TT_8X4: + pat = ~((subblkpat & 2)*6 + (subblkpat & 1)*3) & 0xF; for(j = 0; j < 2; j++) { last = subblkpat & (1 << (1 - j)); i = 0; @@ -2961,19 +3077,22 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan i += skip; if(i > 31) break; - if(v->profile < PROFILE_ADVANCED) - idx = ff_vc1_simple_progressive_8x4_zz[i++]; - else - idx = ff_vc1_adv_progressive_8x4_zz[i++]; - block[idx + off] = value * scale; + idx = v->zz_8x4[i++]+off; + block[idx] = value * scale; if(!v->pquantizer) - block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant; + block[idx] += (block[idx] < 0) ? -mquant : mquant; + } + if(!(subblkpat & (1 << (1 - j))) && !skip_block){ + s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off); + if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC)) + vc1_loop_filter(dst + j*4*linesize, 1, linesize, 8, mquant); + if(apply_filter && cbp_left & (2 << j)) + vc1_loop_filter(dst + j*4*linesize, linesize, 1, 4, mquant); } - if(!(subblkpat & (1 << (1 - j)))) - s->dsp.vc1_inv_trans_8x4(block, j); } break; case TT_4X8: + pat = ~(subblkpat*5) & 0xF; for(j = 0; j < 2; j++) { last = subblkpat & (1 << (1 - j)); i = 0; @@ -2983,22 +3102,28 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan i += skip; if(i > 31) break; - if(v->profile < PROFILE_ADVANCED) - idx = ff_vc1_simple_progressive_4x8_zz[i++]; - else - idx = ff_vc1_adv_progressive_4x8_zz[i++]; - block[idx + off] = value * scale; + idx = v->zz_4x8[i++]+off; + block[idx] = value * scale; if(!v->pquantizer) - block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant; + block[idx] += (block[idx] < 0) ? -mquant : mquant; + } + if(!(subblkpat & (1 << (1 - j))) && !skip_block){ + s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off); + if(apply_filter && cbp_top & (2 << j)) + vc1_loop_filter(dst + j*4, 1, linesize, 4, mquant); + if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA)) + vc1_loop_filter(dst + j*4, linesize, 1, 8, mquant); } - if(!(subblkpat & (1 << (1 - j)))) - s->dsp.vc1_inv_trans_4x8(block, j); } break; } - return 0; + return pat; } +/** @} */ // Macroblock group + +static const int size_table [6] = { 0, 2, 3, 4, 5, 8 }; +static const int offset_table[6] = { 0, 1, 3, 7, 15, 31 }; /** Decode one P-frame MB (in Simple/Main profile) */ @@ -3011,17 +3136,16 @@ static int vc1_decode_p_mb(VC1Context *v) int cbp; /* cbp decoding stuff */ int mqdiff, mquant; /* MB quantization */ int ttmb = v->ttfrm; /* MB Transform type */ - int status; - static const int size_table[6] = { 0, 2, 3, 4, 5, 8 }, - offset_table[6] = { 0, 1, 3, 7, 15, 31 }; int mb_has_coeffs = 1; /* last_flag */ int dmv_x, dmv_y; /* Differential MV components */ - int index, index1; /* LUT indices */ + int index, index1; /* LUT indexes */ int val, sign; /* temp values */ int first_block = 1; int dst_idx, off; int skipped, fourmv; + int block_cbp = 0, pat; + int apply_loop_filter; mquant = v->pq; /* Loosy initialization */ @@ -3036,6 +3160,7 @@ static int vc1_decode_p_mb(VC1Context *v) s->dsp.clear_blocks(s->block[0]); + apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY); if (!fourmv) /* 1MV mode */ { if (!skipped) @@ -3100,12 +3225,41 @@ static int vc1_decode_p_mb(VC1Context *v) if(v->a_avail) s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); } + if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){ + int left_cbp, top_cbp; + if(i & 4){ + left_cbp = v->cbp[s->mb_x - 1] >> (i * 4); + top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4); + }else{ + left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4)); + top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4)); + } + if(left_cbp & 0xC) + vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant); + if(top_cbp & 0xA) + vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant); + } + block_cbp |= 0xF << (i << 2); } else if(val) { - vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block); + int left_cbp = 0, top_cbp = 0, filter = 0; + if(apply_loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){ + filter = 1; + if(i & 4){ + left_cbp = v->cbp[s->mb_x - 1] >> (i * 4); + top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4); + }else{ + left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4)); + top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4)); + } + if(left_cbp & 0xC) + vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant); + if(top_cbp & 0xA) + vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant); + } + pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp); + block_cbp |= pat << (i << 2); if(!v->ttmbf && ttmb < 8) ttmb = -1; first_block = 0; - if((i<4) || !(s->flags & CODEC_FLAG_GRAY)) - s->dsp.add_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize); } } } @@ -3202,15 +3356,44 @@ static int vc1_decode_p_mb(VC1Context *v) if(v->a_avail) s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); } + if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){ + int left_cbp, top_cbp; + if(i & 4){ + left_cbp = v->cbp[s->mb_x - 1] >> (i * 4); + top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4); + }else{ + left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4)); + top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4)); + } + if(left_cbp & 0xC) + vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant); + if(top_cbp & 0xA) + vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant); + } + block_cbp |= 0xF << (i << 2); } else if(is_coded[i]) { - status = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block); + int left_cbp = 0, top_cbp = 0, filter = 0; + if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){ + filter = 1; + if(i & 4){ + left_cbp = v->cbp[s->mb_x - 1] >> (i * 4); + top_cbp = v->cbp[s->mb_x - s->mb_stride] >> (i * 4); + }else{ + left_cbp = (i & 1) ? (cbp >> ((i-1)*4)) : (v->cbp[s->mb_x - 1] >> ((i+1)*4)); + top_cbp = (i & 2) ? (cbp >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4)); + } + if(left_cbp & 0xC) + vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant); + if(top_cbp & 0xA) + vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant); + } + pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp); + block_cbp |= pat << (i << 2); if(!v->ttmbf && ttmb < 8) ttmb = -1; first_block = 0; - if((i<4) || !(s->flags & CODEC_FLAG_GRAY)) - s->dsp.add_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize); } } - return status; + return 0; } else //Skipped MB { @@ -3230,6 +3413,7 @@ static int vc1_decode_p_mb(VC1Context *v) return 0; } } + v->cbp[s->mb_x] = block_cbp; /* Should never happen */ return -1; @@ -3246,11 +3430,8 @@ static void vc1_decode_b_mb(VC1Context *v) int cbp = 0; /* cbp decoding stuff */ int mqdiff, mquant; /* MB quantization */ int ttmb = v->ttfrm; /* MB Transform type */ - - static const int size_table[6] = { 0, 2, 3, 4, 5, 8 }, - offset_table[6] = { 0, 1, 3, 7, 15, 31 }; int mb_has_coeffs = 0; /* last_flag */ - int index, index1; /* LUT indices */ + int index, index1; /* LUT indexes */ int val, sign; /* temp values */ int first_block = 1; int dst_idx, off; @@ -3377,11 +3558,9 @@ static void vc1_decode_b_mb(VC1Context *v) if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1; s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); } else if(val) { - vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block); + vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), 0, 0, 0); if(!v->ttmbf && ttmb < 8) ttmb = -1; first_block = 0; - if((i<4) || !(s->flags & CODEC_FLAG_GRAY)) - s->dsp.add_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize); } } } @@ -3485,6 +3664,7 @@ static void vc1_decode_i_blocks(VC1Context *v) s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize); s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize); } + if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]); if(get_bits_count(&s->gb) > v->bits) { ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)); @@ -3615,6 +3795,7 @@ static void vc1_decode_i_blocks_adv(VC1Context *v) s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize); s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize); } + if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]); if(get_bits_count(&s->gb) > v->bits) { ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)); @@ -3658,6 +3839,7 @@ static void vc1_decode_p_blocks(VC1Context *v) } s->first_slice_line = 1; + memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride); for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) { for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) { ff_init_block_index(s); @@ -3671,6 +3853,7 @@ static void vc1_decode_p_blocks(VC1Context *v) return; } } + memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride); ff_draw_horiz_band(s, s->mb_y * 16, 16); s->first_slice_line = 0; } @@ -3719,6 +3902,7 @@ static void vc1_decode_b_blocks(VC1Context *v) av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits,s->mb_x,s->mb_y); return; } + if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[s->mb_x + s->mb_y *s->mb_stride]); } ff_draw_horiz_band(s, s->mb_y * 16, 16); s->first_slice_line = 0; @@ -3742,7 +3926,7 @@ static void vc1_decode_skip_blocks(VC1Context *v) ff_draw_horiz_band(s, s->mb_y * 16, 16); s->first_slice_line = 0; } - s->pict_type = P_TYPE; + s->pict_type = FF_P_TYPE; } static void vc1_decode_blocks(VC1Context *v) @@ -3754,19 +3938,19 @@ static void vc1_decode_blocks(VC1Context *v) }else{ switch(v->s.pict_type) { - case I_TYPE: + case FF_I_TYPE: if(v->profile == PROFILE_ADVANCED) vc1_decode_i_blocks_adv(v); else vc1_decode_i_blocks(v); break; - case P_TYPE: + case FF_P_TYPE: if(v->p_frame_skipped) vc1_decode_skip_blocks(v); else vc1_decode_p_blocks(v); break; - case B_TYPE: + case FF_B_TYPE: if(v->bi_type){ if(v->profile == PROFILE_ADVANCED) vc1_decode_i_blocks_adv(v); @@ -3782,7 +3966,7 @@ static void vc1_decode_blocks(VC1Context *v) /** Find VC-1 marker in buffer * @return position where next marker starts or end of buffer if no marker found */ -static av_always_inline uint8_t* find_next_marker(uint8_t *src, uint8_t *end) +static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end) { uint32_t mrk = 0xFFFFFFFF; @@ -3795,7 +3979,7 @@ static av_always_inline uint8_t* find_next_marker(uint8_t *src, uint8_t *end) return end; } -static av_always_inline int vc1_unescape_buffer(uint8_t *src, int size, uint8_t *dst) +static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst) { int dsize = 0, i; @@ -3818,7 +4002,7 @@ static av_always_inline int vc1_unescape_buffer(uint8_t *src, int size, uint8_t * @todo TODO: Handle VC-1 IDUs (Transport level?) * @todo TODO: Decypher remaining bits in extra_data */ -static int vc1_decode_init(AVCodecContext *avctx) +static av_cold int vc1_decode_init(AVCodecContext *avctx) { VC1Context *v = avctx->priv_data; MpegEncContext *s = &v->s; @@ -3826,9 +4010,10 @@ static int vc1_decode_init(AVCodecContext *avctx) if (!avctx->extradata_size || !avctx->extradata) return -1; if (!(avctx->flags & CODEC_FLAG_GRAY)) - avctx->pix_fmt = PIX_FMT_YUV420P; + avctx->pix_fmt = avctx->get_format(avctx, avctx->codec->pix_fmts); else avctx->pix_fmt = PIX_FMT_GRAY8; + avctx->hwaccel = ff_find_hwaccel(avctx->codec->id, avctx->pix_fmt); v->s.avctx = avctx; avctx->flags |= CODEC_FLAG_EMU_EDGE; v->s.flags |= CODEC_FLAG_EMU_EDGE; @@ -3868,10 +4053,12 @@ static int vc1_decode_init(AVCodecContext *avctx) av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count); } } else { // VC1/WVC1 - uint8_t *start = avctx->extradata, *end = avctx->extradata + avctx->extradata_size; - uint8_t *next; int size, buf2_size; + const uint8_t *start = avctx->extradata; + uint8_t *end = avctx->extradata + avctx->extradata_size; + const uint8_t *next; + int size, buf2_size; uint8_t *buf2 = NULL; - int seq_inited = 0, ep_inited = 0; + int seq_initialized = 0, ep_initialized = 0; if(avctx->extradata_size < 16) { av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size); @@ -3893,19 +4080,19 @@ static int vc1_decode_init(AVCodecContext *avctx) av_free(buf2); return -1; } - seq_inited = 1; + seq_initialized = 1; break; case VC1_CODE_ENTRYPOINT: if(decode_entry_point(avctx, &gb) < 0){ av_free(buf2); return -1; } - ep_inited = 1; + ep_initialized = 1; break; } } av_free(buf2); - if(!seq_inited || !ep_inited){ + if(!seq_initialized || !ep_initialized){ av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n"); return -1; } @@ -3922,6 +4109,9 @@ static int vc1_decode_init(AVCodecContext *avctx) v->acpred_plane = av_malloc(s->mb_stride * s->mb_height); v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height); + v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride); + v->cbp = v->cbp_base + s->mb_stride; + /* allocate block type info in that way so it could be used with s->block_index[] */ v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2); v->mb_type[0] = v->mb_type_base + s->b8_stride + 1; @@ -3947,12 +4137,13 @@ static int vc1_decode_init(AVCodecContext *avctx) */ static int vc1_decode_frame(AVCodecContext *avctx, void *data, int *data_size, - uint8_t *buf, int buf_size) + const uint8_t *buf, int buf_size) { VC1Context *v = avctx->priv_data; MpegEncContext *s = &v->s; AVFrame *pict = data; uint8_t *buf2 = NULL; + const uint8_t *buf_start = buf; /* no supplementary picture */ if (buf_size == 0) { @@ -3974,13 +4165,20 @@ static int vc1_decode_frame(AVCodecContext *avctx, s->current_picture_ptr= &s->picture[i]; } + if (s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU){ + if (v->profile < PROFILE_ADVANCED) + avctx->pix_fmt = PIX_FMT_VDPAU_WMV3; + else + avctx->pix_fmt = PIX_FMT_VDPAU_VC1; + } + //for advanced profile we may need to parse and unescape data if (avctx->codec_id == CODEC_ID_VC1) { int buf_size2 = 0; buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE); if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */ - uint8_t *start, *end, *next; + const uint8_t *start, *end, *next; int size; next = buf; @@ -3990,6 +4188,9 @@ static int vc1_decode_frame(AVCodecContext *avctx, if(size <= 0) continue; switch(AV_RB32(start)){ case VC1_CODE_FRAME: + if (avctx->hwaccel || + s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) + buf_start = start; buf_size2 = vc1_unescape_buffer(start + 4, size, buf2); break; case VC1_CODE_ENTRYPOINT: /* it should be before frame data */ @@ -4004,11 +4205,12 @@ static int vc1_decode_frame(AVCodecContext *avctx, } } }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */ - uint8_t *divider; + const uint8_t *divider; divider = find_next_marker(buf, buf + buf_size); if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){ av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n"); + av_free(buf2); return -1; } @@ -4034,24 +4236,24 @@ static int vc1_decode_frame(AVCodecContext *avctx, } } - if(s->pict_type != I_TYPE && !v->res_rtm_flag){ + if(s->pict_type != FF_I_TYPE && !v->res_rtm_flag){ av_free(buf2); return -1; } // for hurry_up==5 s->current_picture.pict_type= s->pict_type; - s->current_picture.key_frame= s->pict_type == I_TYPE; + s->current_picture.key_frame= s->pict_type == FF_I_TYPE; /* skip B-frames if we don't have reference frames */ - if(s->last_picture_ptr==NULL && (s->pict_type==B_TYPE || s->dropable)){ + if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){ av_free(buf2); return -1;//buf_size; } /* skip b frames if we are in a hurry */ - if(avctx->hurry_up && s->pict_type==B_TYPE) return -1;//buf_size; - if( (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==B_TYPE) - || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=I_TYPE) + if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size; + if( (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE) + || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE) || avctx->skip_frame >= AVDISCARD_ALL) { av_free(buf2); return buf_size; @@ -4063,7 +4265,7 @@ static int vc1_decode_frame(AVCodecContext *avctx, } if(s->next_p_frame_damaged){ - if(s->pict_type==B_TYPE) + if(s->pict_type==FF_B_TYPE) return buf_size; else s->next_p_frame_damaged=0; @@ -4077,20 +4279,32 @@ static int vc1_decode_frame(AVCodecContext *avctx, s->me.qpel_put= s->dsp.put_qpel_pixels_tab; s->me.qpel_avg= s->dsp.avg_qpel_pixels_tab; - ff_er_frame_start(s); + if ((CONFIG_VC1_VDPAU_DECODER || CONFIG_WMV3_VDPAU_DECODER) + &&s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) + ff_vdpau_vc1_decode_picture(s, buf_start, (buf + buf_size) - buf_start); + else if (avctx->hwaccel) { + if (avctx->hwaccel->start_frame(avctx, buf, buf_size) < 0) + return -1; + if (avctx->hwaccel->decode_slice(avctx, buf_start, (buf + buf_size) - buf_start) < 0) + return -1; + if (avctx->hwaccel->end_frame(avctx) < 0) + return -1; + } else { + ff_er_frame_start(s); - v->bits = buf_size * 8; - vc1_decode_blocks(v); + v->bits = buf_size * 8; + vc1_decode_blocks(v); //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8); // if(get_bits_count(&s->gb) > buf_size * 8) // return -1; - ff_er_frame_end(s); + ff_er_frame_end(s); + } MPV_frame_end(s); assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type); assert(s->current_picture.pict_type == s->pict_type); - if (s->pict_type == B_TYPE || s->low_delay) { + if (s->pict_type == FF_B_TYPE || s->low_delay) { *pict= *(AVFrame*)s->current_picture_ptr; } else if (s->last_picture_ptr != NULL) { *pict= *(AVFrame*)s->last_picture_ptr; @@ -4113,7 +4327,7 @@ assert(s->current_picture.pict_type == s->pict_type); /** Close a VC1/WMV3 decoder * @warning Initial try at using MpegEncContext stuff */ -static int vc1_decode_end(AVCodecContext *avctx) +static av_cold int vc1_decode_end(AVCodecContext *avctx) { VC1Context *v = avctx->priv_data; @@ -4125,6 +4339,8 @@ static int vc1_decode_end(AVCodecContext *avctx) av_freep(&v->acpred_plane); av_freep(&v->over_flags_plane); av_freep(&v->mb_type_base); + av_freep(&v->cbp_base); + ff_intrax8_common_end(&v->x8); return 0; } @@ -4139,7 +4355,9 @@ AVCodec vc1_decoder = { vc1_decode_end, vc1_decode_frame, CODEC_CAP_DELAY, - NULL + NULL, + .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"), + .pix_fmts = ff_hwaccel_pixfmt_list_420 }; AVCodec wmv3_decoder = { @@ -4152,5 +4370,41 @@ AVCodec wmv3_decoder = { vc1_decode_end, vc1_decode_frame, CODEC_CAP_DELAY, - NULL + NULL, + .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"), + .pix_fmts = ff_hwaccel_pixfmt_list_420 }; + +#if CONFIG_WMV3_VDPAU_DECODER +AVCodec wmv3_vdpau_decoder = { + "wmv3_vdpau", + CODEC_TYPE_VIDEO, + CODEC_ID_WMV3, + sizeof(VC1Context), + vc1_decode_init, + NULL, + vc1_decode_end, + vc1_decode_frame, + CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU, + NULL, + .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 VDPAU"), + .pix_fmts = (enum PixelFormat[]){PIX_FMT_VDPAU_WMV3, PIX_FMT_NONE} +}; +#endif + +#if CONFIG_VC1_VDPAU_DECODER +AVCodec vc1_vdpau_decoder = { + "vc1_vdpau", + CODEC_TYPE_VIDEO, + CODEC_ID_VC1, + sizeof(VC1Context), + vc1_decode_init, + NULL, + vc1_decode_end, + vc1_decode_frame, + CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU, + NULL, + .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1 VDPAU"), + .pix_fmts = (enum PixelFormat[]){PIX_FMT_VDPAU_VC1, PIX_FMT_NONE} +}; +#endif