X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fh264_cavlc.c;h=fdb7ab57098bc8480ab1be22cdbfe881dd96de32;hb=f44005b610abc5b1aca5b3191b64a0a0ba65a03f;hp=0afabfc6e899b85c3f48a96ccc7f481b6d1cc7a0;hpb=81afcf1fae1f04573d65f3f4ecb0e0b3c60fb1c0;p=ffmpeg diff --git a/libavcodec/h264_cavlc.c b/libavcodec/h264_cavlc.c index 0afabfc6e89..fdb7ab57098 100644 --- a/libavcodec/h264_cavlc.c +++ b/libavcodec/h264_cavlc.c @@ -2,25 +2,25 @@ * H.26L/H.264/AVC/JVT/14496-10/... cavlc bitstream decoding * Copyright (c) 2003 Michael Niedermayer * - * This file is part of FFmpeg. + * This file is part of Libav. * - * FFmpeg is free software; you can redistribute it and/or + * Libav is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * - * FFmpeg is distributed in the hope that it will be useful, + * Libav is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with FFmpeg; if not, write to the Free Software + * License along with Libav; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ /** - * @file libavcodec/h264_cavlc.c + * @file * H.264 / AVC / MPEG4 part10 cavlc bitstream decoding. * @author Michael Niedermayer */ @@ -62,6 +62,30 @@ static const uint8_t chroma_dc_coeff_token_bits[4*5]={ 2, 3, 2, 0, }; +static const uint8_t chroma422_dc_coeff_token_len[4*9]={ + 1, 0, 0, 0, + 7, 2, 0, 0, + 7, 7, 3, 0, + 9, 7, 7, 5, + 9, 9, 7, 6, + 10, 10, 9, 7, + 11, 11, 10, 7, + 12, 12, 11, 10, + 13, 12, 12, 11, +}; + +static const uint8_t chroma422_dc_coeff_token_bits[4*9]={ + 1, 0, 0, 0, + 15, 1, 0, 0, + 14, 13, 1, 0, + 7, 12, 11, 1, + 6, 5, 10, 1, + 7, 6, 4, 9, + 7, 6, 5, 8, + 7, 6, 5, 4, + 7, 5, 4, 4, +}; + static const uint8_t coeff_token_len[4][4*17]={ { 1, 0, 0, 0, @@ -172,6 +196,26 @@ static const uint8_t chroma_dc_total_zeros_bits[3][4]= { { 1, 0, 0, 0,}, }; +static const uint8_t chroma422_dc_total_zeros_len[7][8]= { + { 1, 3, 3, 4, 4, 4, 5, 5 }, + { 3, 2, 3, 3, 3, 3, 3 }, + { 3, 3, 2, 2, 3, 3 }, + { 3, 2, 2, 2, 3 }, + { 2, 2, 2, 2 }, + { 2, 2, 1 }, + { 1, 1 }, +}; + +static const uint8_t chroma422_dc_total_zeros_bits[7][8]= { + { 1, 2, 3, 2, 3, 1, 1, 0 }, + { 0, 1, 1, 4, 5, 6, 7 }, + { 0, 1, 1, 2, 6, 7 }, + { 6, 0, 1, 2, 7 }, + { 0, 1, 2, 3 }, + { 0, 1, 1 }, + { 0, 1 }, +}; + static const uint8_t run_len[7][16]={ {1,1}, {1,2,2}, @@ -200,6 +244,10 @@ static VLC chroma_dc_coeff_token_vlc; static VLC_TYPE chroma_dc_coeff_token_vlc_table[256][2]; static const int chroma_dc_coeff_token_vlc_table_size = 256; +static VLC chroma422_dc_coeff_token_vlc; +static VLC_TYPE chroma422_dc_coeff_token_vlc_table[8192][2]; +static const int chroma422_dc_coeff_token_vlc_table_size = 8192; + static VLC total_zeros_vlc[15]; static VLC_TYPE total_zeros_vlc_tables[15][512][2]; static const int total_zeros_vlc_tables_size = 512; @@ -208,6 +256,10 @@ static VLC chroma_dc_total_zeros_vlc[3]; static VLC_TYPE chroma_dc_total_zeros_vlc_tables[3][8][2]; static const int chroma_dc_total_zeros_vlc_tables_size = 8; +static VLC chroma422_dc_total_zeros_vlc[7]; +static VLC_TYPE chroma422_dc_total_zeros_vlc_tables[7][32][2]; +static const int chroma422_dc_total_zeros_vlc_tables_size = 32; + static VLC run_vlc[6]; static VLC_TYPE run_vlc_tables[6][8][2]; static const int run_vlc_tables_size = 8; @@ -219,9 +271,17 @@ static const int run7_vlc_table_size = 96; #define LEVEL_TAB_BITS 8 static int8_t cavlc_level_tab[7][1<>(LEVEL_TAB_BITS-prefix-1-suffix_length)) - (1<>1) ^ mask) - mask; if(prefix + 1 + suffix_length <= LEVEL_TAB_BITS){ + int level_code = (prefix << suffix_length) + + (i >> (av_log2(i) - suffix_length)) - (1 << suffix_length); + int mask = -(level_code&1); + level_code = (((2 + level_code) >> 1) ^ mask) - mask; cavlc_level_tab[suffix_length][i][0]= level_code; cavlc_level_tab[suffix_length][i][1]= prefix + 1 + suffix_length; }else if(prefix + 1 <= LEVEL_TAB_BITS){ @@ -277,6 +338,13 @@ av_cold void ff_h264_decode_init_vlc(void){ &chroma_dc_coeff_token_bits[0], 1, 1, INIT_VLC_USE_NEW_STATIC); + chroma422_dc_coeff_token_vlc.table = chroma422_dc_coeff_token_vlc_table; + chroma422_dc_coeff_token_vlc.table_allocated = chroma422_dc_coeff_token_vlc_table_size; + init_vlc(&chroma422_dc_coeff_token_vlc, CHROMA422_DC_COEFF_TOKEN_VLC_BITS, 4*9, + &chroma422_dc_coeff_token_len [0], 1, 1, + &chroma422_dc_coeff_token_bits[0], 1, 1, + INIT_VLC_USE_NEW_STATIC); + offset = 0; for(i=0; i<4; i++){ coeff_token_vlc[i].table = coeff_token_vlc_tables+offset; @@ -303,6 +371,17 @@ av_cold void ff_h264_decode_init_vlc(void){ &chroma_dc_total_zeros_bits[i][0], 1, 1, INIT_VLC_USE_NEW_STATIC); } + + for(i=0; i<7; i++){ + chroma422_dc_total_zeros_vlc[i].table = chroma422_dc_total_zeros_vlc_tables[i]; + chroma422_dc_total_zeros_vlc[i].table_allocated = chroma422_dc_total_zeros_vlc_tables_size; + init_vlc(&chroma422_dc_total_zeros_vlc[i], + CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 8, + &chroma422_dc_total_zeros_len [i][0], 1, 1, + &chroma422_dc_total_zeros_bits[i][0], 1, 1, + INIT_VLC_USE_NEW_STATIC); + } + for(i=0; i<15; i++){ total_zeros_vlc[i].table = total_zeros_vlc_tables[i]; total_zeros_vlc[i].table_allocated = total_zeros_vlc_tables_size; @@ -357,7 +436,7 @@ static inline int get_level_prefix(GetBitContext *gb){ } /** - * decodes a residual block. + * Decode a residual block. * @param n block index * @param scantable scantable * @param max_coeff number of coefficients in the block @@ -367,25 +446,28 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in MpegEncContext * const s = &h->s; static const int coeff_token_table_index[17]= {0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3}; int level[16]; - int zeros_left, coeff_num, coeff_token, total_coeff, i, j, trailing_ones, run_before; + int zeros_left, coeff_token, total_coeff, i, trailing_ones, run_before; //FIXME put trailing_onex into the context - if(n == CHROMA_DC_BLOCK_INDEX){ - coeff_token= get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1); + if(max_coeff <= 8){ + if (max_coeff == 4) + coeff_token = get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1); + else + coeff_token = get_vlc2(gb, chroma422_dc_coeff_token_vlc.table, CHROMA422_DC_COEFF_TOKEN_VLC_BITS, 1); total_coeff= coeff_token>>2; }else{ - if(n == LUMA_DC_BLOCK_INDEX){ - total_coeff= pred_non_zero_count(h, 0); + if(n >= LUMA_DC_BLOCK_INDEX){ + total_coeff= pred_non_zero_count(h, (n - LUMA_DC_BLOCK_INDEX)*16); coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2); total_coeff= coeff_token>>2; }else{ total_coeff= pred_non_zero_count(h, n); coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2); total_coeff= coeff_token>>2; - h->non_zero_count_cache[ scan8[n] ]= total_coeff; } } + h->non_zero_count_cache[ scan8[n] ]= total_coeff; //FIXME set last_non_zero? @@ -431,8 +513,13 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in level_code= prefix + get_bits(gb, 4); //part }else{ level_code= 30 + get_bits(gb, prefix-3); //part - if(prefix>=16) + if(prefix>=16){ + if(prefix > 25+3){ + av_log(h->s.avctx, AV_LOG_ERROR, "Invalid level prefix\n"); + return -1; + } level_code += (1<<(prefix-3))-4096; + } } if(trailing_ones < 3) level_code += 2; @@ -477,46 +564,56 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in if(total_coeff == max_coeff) zeros_left=0; else{ - if(n == CHROMA_DC_BLOCK_INDEX) - zeros_left= get_vlc2(gb, (chroma_dc_total_zeros_vlc-1)[ total_coeff ].table, CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1); - else + if (max_coeff <= 8) { + if (max_coeff == 4) + zeros_left = get_vlc2(gb, (chroma_dc_total_zeros_vlc-1)[total_coeff].table, + CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1); + else + zeros_left = get_vlc2(gb, (chroma422_dc_total_zeros_vlc-1)[total_coeff].table, + CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 1); + } else { zeros_left= get_vlc2(gb, (total_zeros_vlc-1)[ total_coeff ].table, TOTAL_ZEROS_VLC_BITS, 1); + } } - coeff_num = zeros_left + total_coeff - 1; - j = scantable[coeff_num]; - if(n > 24){ - block[j] = level[0]; - for(i=1;i>6; - for(i=1;i= LUMA_DC_BLOCK_INDEX){ \ + ((type*)block)[*scantable] = level[0]; \ + for(i=1;i 0;i++) { \ + if(zeros_left < 7) \ + run_before= get_vlc2(gb, (run_vlc-1)[zeros_left].table, RUN_VLC_BITS, 1); \ + else \ + run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \ + zeros_left -= run_before; \ + scantable -= 1 + run_before; \ + ((type*)block)[*scantable]= level[i]; \ + } \ + for(;i>6; \ + for(i=1;i 0;i++) { \ + if(zeros_left < 7) \ + run_before= get_vlc2(gb, (run_vlc-1)[zeros_left].table, RUN_VLC_BITS, 1); \ + else \ + run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \ + zeros_left -= run_before; \ + scantable -= 1 + run_before; \ + ((type*)block)[*scantable]= ((int)(level[i] * qmul[*scantable] + 32))>>6; \ + } \ + for(;i>6; \ + } \ + } - block[j]= (level[i] * qmul[j] + 32)>>6; - } + if (h->pixel_shift) { + STORE_BLOCK(int32_t) + } else { + STORE_BLOCK(int16_t) } if(zeros_left<0){ @@ -527,19 +624,88 @@ static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, in return 0; } +static av_always_inline int decode_luma_residual(H264Context *h, GetBitContext *gb, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p){ + int i4x4, i8x8; + MpegEncContext * const s = &h->s; + int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1]; + if(IS_INTRA16x16(mb_type)){ + AV_ZERO128(h->mb_luma_dc[p]+0); + AV_ZERO128(h->mb_luma_dc[p]+8); + AV_ZERO128(h->mb_luma_dc[p]+16); + AV_ZERO128(h->mb_luma_dc[p]+24); + if( decode_residual(h, h->intra_gb_ptr, h->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX+p, scan, NULL, 16) < 0){ + return -1; //FIXME continue if partitioned and other return -1 too + } + + assert((cbp&15) == 0 || (cbp&15) == 15); + + if(cbp&15){ + for(i8x8=0; i8x8<4; i8x8++){ + for(i4x4=0; i4x4<4; i4x4++){ + const int index= i4x4 + 4*i8x8 + p*16; + if( decode_residual(h, h->intra_gb_ptr, h->mb + (16*index << pixel_shift), + index, scan + 1, h->dequant4_coeff[p][qscale], 15) < 0 ){ + return -1; + } + } + } + return 0xf; + }else{ + fill_rectangle(&h->non_zero_count_cache[scan8[p*16]], 4, 4, 8, 0, 1); + return 0; + } + }else{ + int cqm = (IS_INTRA( mb_type ) ? 0:3)+p; + /* For CAVLC 4:4:4, we need to keep track of the luma 8x8 CBP for deblocking nnz purposes. */ + int new_cbp = 0; + for(i8x8=0; i8x8<4; i8x8++){ + if(cbp & (1<mb[64*i8x8+256*p << pixel_shift]; + uint8_t *nnz; + for(i4x4=0; i4x4<4; i4x4++){ + const int index= i4x4 + 4*i8x8 + p*16; + if( decode_residual(h, gb, buf, index, scan8x8+16*i4x4, + h->dequant8_coeff[cqm][qscale], 16) < 0 ) + return -1; + } + nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ]; + nnz[0] += nnz[1] + nnz[8] + nnz[9]; + new_cbp |= !!nnz[0] << i8x8; + }else{ + for(i4x4=0; i4x4<4; i4x4++){ + const int index= i4x4 + 4*i8x8 + p*16; + if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, + scan, h->dequant4_coeff[cqm][qscale], 16) < 0 ){ + return -1; + } + new_cbp |= h->non_zero_count_cache[ scan8[index] ] << i8x8; + } + } + }else{ + uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ]; + nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0; + } + } + return new_cbp; + } +} + int ff_h264_decode_mb_cavlc(H264Context *h){ MpegEncContext * const s = &h->s; int mb_xy; int partition_count; unsigned int mb_type, cbp; int dct8x8_allowed= h->pps.transform_8x8_mode; + int decode_chroma = h->sps.chroma_format_idc == 1 || h->sps.chroma_format_idc == 2; + const int pixel_shift = h->pixel_shift; mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride; tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y); cbp = 0; /* avoid warning. FIXME: find a solution without slowing down the code */ - if(h->slice_type_nos != FF_I_TYPE){ + if(h->slice_type_nos != AV_PICTURE_TYPE_I){ if(s->mb_skip_run==-1) s->mb_skip_run= get_ue_golomb(&s->gb); @@ -547,8 +713,6 @@ int ff_h264_decode_mb_cavlc(H264Context *h){ if(FRAME_MBAFF && (s->mb_y&1) == 0){ if(s->mb_skip_run==0) h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&s->gb); - else - predict_field_decoding_flag(h); } decode_mb_skip(h); return 0; @@ -562,7 +726,7 @@ int ff_h264_decode_mb_cavlc(H264Context *h){ h->prev_mb_skipped= 0; mb_type= get_ue_golomb(&s->gb); - if(h->slice_type_nos == FF_B_TYPE){ + if(h->slice_type_nos == AV_PICTURE_TYPE_B){ if(mb_type < 23){ partition_count= b_mb_type_info[mb_type].partition_count; mb_type= b_mb_type_info[mb_type].type; @@ -570,7 +734,7 @@ int ff_h264_decode_mb_cavlc(H264Context *h){ mb_type -= 23; goto decode_intra_mb; } - }else if(h->slice_type_nos == FF_P_TYPE){ + }else if(h->slice_type_nos == AV_PICTURE_TYPE_P){ if(mb_type < 5){ partition_count= p_mb_type_info[mb_type].partition_count; mb_type= p_mb_type_info[mb_type].type; @@ -579,12 +743,12 @@ int ff_h264_decode_mb_cavlc(H264Context *h){ goto decode_intra_mb; } }else{ - assert(h->slice_type_nos == FF_I_TYPE); - if(h->slice_type == FF_SI_TYPE && mb_type) + assert(h->slice_type_nos == AV_PICTURE_TYPE_I); + if(h->slice_type == AV_PICTURE_TYPE_SI && mb_type) mb_type--; decode_intra_mb: if(mb_type > 25){ - av_log(h->s.avctx, AV_LOG_ERROR, "mb_type %d in %c slice too large at %d %d\n", mb_type, av_get_pict_type_char(h->slice_type), s->mb_x, s->mb_y); + av_log(h->s.avctx, AV_LOG_ERROR, "mb_type %d in %c slice too large at %d %d\n", mb_type, av_get_picture_type_char(h->slice_type), s->mb_x, s->mb_y); return -1; } partition_count=0; @@ -600,21 +764,23 @@ decode_intra_mb: if(IS_INTRA_PCM(mb_type)){ unsigned int x; + static const uint16_t mb_sizes[4] = {256,384,512,768}; + const int mb_size = mb_sizes[h->sps.chroma_format_idc]*h->sps.bit_depth_luma >> 3; // We assume these blocks are very rare so we do not optimize it. align_get_bits(&s->gb); // The pixels are stored in the same order as levels in h->mb array. - for(x=0; x < (CHROMA ? 384 : 256); x++){ + for(x=0; x < mb_size; x++){ ((uint8_t*)h->mb)[x]= get_bits(&s->gb, 8); } // In deblocking, the quantizer is 0 - s->current_picture.qscale_table[mb_xy]= 0; + s->current_picture.f.qscale_table[mb_xy] = 0; // All coeffs are present - memset(h->non_zero_count[mb_xy], 16, 32); + memset(h->non_zero_count[mb_xy], 16, 48); - s->current_picture.mb_type[mb_xy]= mb_type; + s->current_picture.f.mb_type[mb_xy] = mb_type; return 0; } @@ -623,6 +789,7 @@ decode_intra_mb: h->ref_count[1] <<= 1; } + fill_decode_neighbors(h, mb_type); fill_decode_caches(h, mb_type); //mb_pred @@ -651,7 +818,7 @@ decode_intra_mb: else h->intra4x4_pred_mode_cache[ scan8[i] ] = mode; } - ff_h264_write_back_intra_pred_mode(h); + write_back_intra_pred_mode(h); if( ff_h264_check_intra4x4_pred_mode(h) < 0) return -1; }else{ @@ -659,16 +826,18 @@ decode_intra_mb: if(h->intra16x16_pred_mode < 0) return -1; } - if(CHROMA){ + if(decode_chroma){ pred_mode= ff_h264_check_intra_pred_mode(h, get_ue_golomb_31(&s->gb)); if(pred_mode < 0) return -1; h->chroma_pred_mode= pred_mode; + } else { + h->chroma_pred_mode = DC_128_PRED8x8; } }else if(partition_count==4){ int i, j, sub_partition_count[4], list, ref[2][4]; - if(h->slice_type_nos == FF_B_TYPE){ + if(h->slice_type_nos == AV_PICTURE_TYPE_B){ for(i=0; i<4; i++){ h->sub_mb_type[i]= get_ue_golomb_31(&s->gb); if(h->sub_mb_type[i] >=13){ @@ -686,7 +855,7 @@ decode_intra_mb: h->ref_cache[1][scan8[12]] = PART_NOT_AVAILABLE; } }else{ - assert(h->slice_type_nos == FF_P_TYPE); //FIXME SP correct ? + assert(h->slice_type_nos == AV_PICTURE_TYPE_P); //FIXME SP correct ? for(i=0; i<4; i++){ h->sub_mb_type[i]= get_ue_golomb_31(&s->gb); if(h->sub_mb_type[i] >=4){ @@ -790,22 +959,18 @@ decode_intra_mb: return -1; } } - }else - val= LIST_NOT_USED&0xFF; fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, val, 1); + } } for(list=0; listlist_count; list++){ - unsigned int val; if(IS_DIR(mb_type, 0, list)){ pred_motion(h, 0, 4, list, h->ref_cache[list][ scan8[0] ], &mx, &my); mx += get_se_golomb(&s->gb); my += get_se_golomb(&s->gb); tprintf(s->avctx, "final mv:%d %d\n", mx, my); - val= pack16to32(mx,my); - }else - val=0; - fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, val, 4); + fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4); + } } } else if(IS_16X8(mb_type)){ @@ -889,53 +1054,55 @@ decode_intra_mb: if(!IS_INTRA16x16(mb_type)){ cbp= get_ue_golomb(&s->gb); - if(cbp > 47){ - av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y); - return -1; - } - if(CHROMA){ + if(decode_chroma){ + if(cbp > 47){ + av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y); + return -1; + } if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp[cbp]; else cbp= golomb_to_inter_cbp [cbp]; }else{ + if(cbp > 15){ + av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y); + return -1; + } if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp_gray[cbp]; else cbp= golomb_to_inter_cbp_gray[cbp]; } } if(dct8x8_allowed && (cbp&15) && !IS_INTRA(mb_type)){ - if(get_bits1(&s->gb)){ - mb_type |= MB_TYPE_8x8DCT; - } + mb_type |= MB_TYPE_8x8DCT*get_bits1(&s->gb); } h->cbp= h->cbp_table[mb_xy]= cbp; - s->current_picture.mb_type[mb_xy]= mb_type; + s->current_picture.f.mb_type[mb_xy] = mb_type; if(cbp || IS_INTRA16x16(mb_type)){ - int i8x8, i4x4, chroma_idx; + int i4x4, i8x8, chroma_idx; int dquant; + int ret; GetBitContext *gb= IS_INTRA(mb_type) ? h->intra_gb_ptr : h->inter_gb_ptr; - const uint8_t *scan, *scan8x8, *dc_scan; + const uint8_t *scan, *scan8x8; + const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8); if(IS_INTERLACED(mb_type)){ scan8x8= s->qscale ? h->field_scan8x8_cavlc : h->field_scan8x8_cavlc_q0; scan= s->qscale ? h->field_scan : h->field_scan_q0; - dc_scan= luma_dc_field_scan; }else{ scan8x8= s->qscale ? h->zigzag_scan8x8_cavlc : h->zigzag_scan8x8_cavlc_q0; scan= s->qscale ? h->zigzag_scan : h->zigzag_scan_q0; - dc_scan= luma_dc_zigzag_scan; } dquant= get_se_golomb(&s->gb); s->qscale += dquant; - if(((unsigned)s->qscale) > 51){ - if(s->qscale<0) s->qscale+= 52; - else s->qscale-= 52; - if(((unsigned)s->qscale) > 51){ + if(((unsigned)s->qscale) > max_qp){ + if(s->qscale<0) s->qscale+= max_qp+1; + else s->qscale-= max_qp+1; + if(((unsigned)s->qscale) > max_qp){ av_log(h->s.avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, s->mb_x, s->mb_y); return -1; } @@ -943,83 +1110,74 @@ decode_intra_mb: h->chroma_qp[0]= get_chroma_qp(h, 0, s->qscale); h->chroma_qp[1]= get_chroma_qp(h, 1, s->qscale); - if(IS_INTRA16x16(mb_type)){ - if( decode_residual(h, h->intra_gb_ptr, h->mb, LUMA_DC_BLOCK_INDEX, dc_scan, h->dequant4_coeff[0][s->qscale], 16) < 0){ - return -1; //FIXME continue if partitioned and other return -1 too - } - assert((cbp&15) == 0 || (cbp&15) == 15); - - if(cbp&15){ - for(i8x8=0; i8x8<4; i8x8++){ - for(i4x4=0; i4x4<4; i4x4++){ - const int index= i4x4 + 4*i8x8; - if( decode_residual(h, h->intra_gb_ptr, h->mb + 16*index, index, scan + 1, h->dequant4_coeff[0][s->qscale], 15) < 0 ){ - return -1; - } + if( (ret = decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 0)) < 0 ){ + return -1; + } + h->cbp_table[mb_xy] |= ret << 12; + if(CHROMA444){ + if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 1) < 0 ){ + return -1; + } + if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 2) < 0 ){ + return -1; + } + } else if (CHROMA422) { + if(cbp&0x30){ + for(chroma_idx=0; chroma_idx<2; chroma_idx++) + if (decode_residual(h, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift), + CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma422_dc_scan, + NULL, 8) < 0) { + return -1; } - } - }else{ - fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1); } - }else{ - for(i8x8=0; i8x8<4; i8x8++){ - if(cbp & (1<mb[64*i8x8]; - uint8_t *nnz; - for(i4x4=0; i4x4<4; i4x4++){ - if( decode_residual(h, gb, buf, i4x4+4*i8x8, scan8x8+16*i4x4, - h->dequant8_coeff[IS_INTRA( mb_type ) ? 0:1][s->qscale], 16) <0 ) - return -1; - } - nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ]; - nnz[0] += nnz[1] + nnz[8] + nnz[9]; - }else{ - for(i4x4=0; i4x4<4; i4x4++){ - const int index= i4x4 + 4*i8x8; - if( decode_residual(h, gb, h->mb + 16*index, index, scan, h->dequant4_coeff[IS_INTRA( mb_type ) ? 0:3][s->qscale], 16) <0 ){ + if(cbp&0x20){ + for(chroma_idx=0; chroma_idx<2; chroma_idx++){ + const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]]; + DCTELEM *mb = h->mb + (16*(16 + 16*chroma_idx) << pixel_shift); + for (i8x8 = 0; i8x8 < 2; i8x8++) { + for (i4x4 = 0; i4x4 < 4; i4x4++) { + const int index = 16 + 16*chroma_idx + 8*i8x8 + i4x4; + if (decode_residual(h, gb, mb, index, scan + 1, qmul, 15) < 0) return -1; - } + mb += 16 << pixel_shift; } } - }else{ - uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ]; - nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0; } + }else{ + fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1); + fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1); } - } - - if(cbp&0x30){ - for(chroma_idx=0; chroma_idx<2; chroma_idx++) - if( decode_residual(h, gb, h->mb + 256 + 16*4*chroma_idx, CHROMA_DC_BLOCK_INDEX, chroma_dc_scan, NULL, 4) < 0){ - return -1; - } - } - - if(cbp&0x20){ - for(chroma_idx=0; chroma_idx<2; chroma_idx++){ - const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]]; - for(i4x4=0; i4x4<4; i4x4++){ - const int index= 16 + 4*chroma_idx + i4x4; - if( decode_residual(h, gb, h->mb + 16*index, index, scan + 1, qmul, 15) < 0){ + } else /* yuv420 */ { + if(cbp&0x30){ + for(chroma_idx=0; chroma_idx<2; chroma_idx++) + if( decode_residual(h, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift), CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma_dc_scan, NULL, 4) < 0){ return -1; } + } + + if(cbp&0x20){ + for(chroma_idx=0; chroma_idx<2; chroma_idx++){ + const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]]; + for(i4x4=0; i4x4<4; i4x4++){ + const int index= 16 + 16*chroma_idx + i4x4; + if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){ + return -1; + } + } } + }else{ + fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1); + fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1); } - }else{ - uint8_t * const nnz= &h->non_zero_count_cache[0]; - nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] = - nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0; } }else{ - uint8_t * const nnz= &h->non_zero_count_cache[0]; - fill_rectangle(&nnz[scan8[0]], 4, 4, 8, 0, 1); - nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] = - nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0; + fill_rectangle(&h->non_zero_count_cache[scan8[ 0]], 4, 4, 8, 0, 1); + fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1); + fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1); } - s->current_picture.qscale_table[mb_xy]= s->qscale; + s->current_picture.f.qscale_table[mb_xy] = s->qscale; write_back_non_zero_count(h); if(MB_MBAFF){ @@ -1029,4 +1187,3 @@ decode_intra_mb: return 0; } -