X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fdv.c;h=3983381640cf0a991f9b953bc561aabb4e4cce85;hb=6f7b915a0d69ea33a6777c35f720cf31f122ed02;hp=4d79b2cc5333ba209bb8c6f686baedaa20cf5530;hpb=d2d230a7569154306a1625ca37dbfa4c36627ec6;p=ffmpeg diff --git a/libavcodec/dv.c b/libavcodec/dv.c index 4d79b2cc533..3983381640c 100644 --- a/libavcodec/dv.c +++ b/libavcodec/dv.c @@ -6,21 +6,26 @@ * DV encoder * Copyright (c) 2003 Roman Shaposhnik. * + * 50 Mbps (DVCPRO50) support + * Copyright (c) 2006 Daniel Maas + * * Many thanks to Dan Dennedy for providing wealth * of DV technical info. * - * This library is free software; you can redistribute it and/or + * This file is part of FFmpeg. + * + * FFmpeg 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 of the License, or (at your option) any later version. + * version 2.1 of the License, or (at your option) any later version. * - * This library is distributed in the hope that it will be useful, + * FFmpeg 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 this library; if not, write to the Free Software + * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -28,9 +33,10 @@ * @file dv.c * DV codec. */ +#define ALT_BITSTREAM_READER #include "avcodec.h" #include "dsputil.h" -#include "mpegvideo.h" +#include "bitstream.h" #include "simple_idct.h" #include "dvdata.h" @@ -51,8 +57,12 @@ typedef struct DVVideoContext { void (*idct_put[2])(uint8_t *dest, int line_size, DCTELEM *block); } DVVideoContext; -/* MultiThreading - applies to entire DV codec, not just the avcontext */ -uint8_t** dv_anchor; +/* MultiThreading - dv_anchor applies to entire DV codec, not just the avcontext */ +/* one element is needed for each video segment in a DV frame */ +/* at most there are 2 DIF channels * 12 DIF sequences * 27 video segments (PAL 50Mbps) */ +#define DV_ANCHOR_SIZE (2*12*27) + +static void* dv_anchor[DV_ANCHOR_SIZE]; #define TEX_VLC_BITS 9 @@ -65,12 +75,12 @@ uint8_t** dv_anchor; #endif /* XXX: also include quantization */ -static RL_VLC_ELEM *dv_rl_vlc; +static RL_VLC_ELEM dv_rl_vlc[1184]; /* VLC encoding lookup table */ static struct dv_vlc_pair { uint32_t vlc; uint8_t size; -} (*dv_vlc_map)[DV_VLC_MAP_LEV_SIZE] = NULL; +} dv_vlc_map[DV_VLC_MAP_RUN_SIZE][DV_VLC_MAP_LEV_SIZE]; static void dv_build_unquantize_tables(DVVideoContext *s, uint8_t* perm) { @@ -97,7 +107,7 @@ static void dv_build_unquantize_tables(DVVideoContext *s, uint8_t* perm) } } -static int dvvideo_init(AVCodecContext *avctx) +static av_cold int dvvideo_init(AVCodecContext *avctx) { DVVideoContext *s = avctx->priv_data; DSPContext dsp; @@ -113,16 +123,8 @@ static int dvvideo_init(AVCodecContext *avctx) done = 1; - dv_vlc_map = av_mallocz_static(DV_VLC_MAP_LEV_SIZE*DV_VLC_MAP_RUN_SIZE*sizeof(struct dv_vlc_pair)); - if (!dv_vlc_map) - return -ENOMEM; - /* dv_anchor lets each thread know its Id */ - dv_anchor = av_malloc(12*27*sizeof(void*)); - if (!dv_anchor) { - return -ENOMEM; - } - for (i=0; i<12*27; i++) + for (i=0; ifdct[1] = dsp.fdct248; - s->idct_put[1] = simple_idct248_put; // FIXME: need to add it to DSP + s->idct_put[1] = ff_simple_idct248_put; // FIXME: need to add it to DSP if(avctx->lowres){ for (i=0; i<64; i++){ int j= ff_zigzag248_direct[i]; @@ -238,9 +237,6 @@ static int dvvideo_init(AVCodecContext *avctx) /* XXX: do it only for constant case */ dv_build_unquantize_tables(s, dsp.idct_permutation); - /* FIXME: I really don't think this should be here */ - if (dv_codec_profile(avctx)) - avctx->pix_fmt = dv_codec_profile(avctx)->pix_fmt; avctx->coded_frame = &s->picture; s->avctx= avctx; @@ -253,6 +249,7 @@ static int dvvideo_init(AVCodecContext *avctx) typedef struct BlockInfo { const uint8_t *shift_table; const uint8_t *scan_table; + const int *iweight_table; uint8_t pos; /* position in block */ uint8_t dct_mode; uint8_t partial_bit_count; @@ -269,11 +266,6 @@ static const int vs_total_ac_bits = (100 * 4 + 68*2) * 5; /* see dv_88_areas and dv_248_areas for details */ static const int mb_area_start[5] = { 1, 6, 21, 43, 64 }; -#ifndef ALT_BITSTREAM_READER -#warning only works with ALT_BITSTREAM_READER -static int re_index; //Hack to make it compile -#endif - static inline int get_bits_left(GetBitContext *s) { return s->size_in_bits - get_bits_count(s); @@ -295,6 +287,7 @@ static void dv_decode_ac(GetBitContext *gb, BlockInfo *mb, DCTELEM *block) int last_index = get_bits_size(gb); const uint8_t *scan_table = mb->scan_table; const uint8_t *shift_table = mb->shift_table; + const int *iweight_table = mb->iweight_table; int pos = mb->pos; int partial_bit_count = mb->partial_bit_count; int level, pos1, run, vlc_len, index; @@ -342,9 +335,13 @@ static void dv_decode_ac(GetBitContext *gb, BlockInfo *mb, DCTELEM *block) if (pos >= 64) break; - assert(level); pos1 = scan_table[pos]; - block[pos1] = level << shift_table[pos1]; + level <<= shift_table[pos1]; + + /* unweigh, round, and shift down */ + level = (level*iweight_table[pos] + (1 << (dv_iweight_bits-1))) >> dv_iweight_bits; + + block[pos1] = level; UPDATE_CACHE(re, gb); } @@ -366,7 +363,7 @@ static inline void bit_copy(PutBitContext *pb, GetBitContext *gb) /* mb_x and mb_y are in units of 8 pixels */ static inline void dv_decode_video_segment(DVVideoContext *s, - uint8_t *buf_ptr1, + const uint8_t *buf_ptr1, const uint16_t *mb_pos_ptr) { int quant, dc, dct_mode, class1, j; @@ -375,11 +372,11 @@ static inline void dv_decode_video_segment(DVVideoContext *s, int c_offset; uint8_t *y_ptr; void (*idct_put)(uint8_t *dest, int line_size, DCTELEM *block); - uint8_t *buf_ptr; + const uint8_t *buf_ptr; PutBitContext pb, vs_pb; GetBitContext gb; BlockInfo mb_data[5 * 6], *mb, *mb1; - DECLARE_ALIGNED_8(DCTELEM, sblock[5*6][64]); + DECLARE_ALIGNED_16(DCTELEM, sblock[5*6][64]); DECLARE_ALIGNED_8(uint8_t, mb_bit_buffer[80 + 4]); /* allow some slack */ DECLARE_ALIGNED_8(uint8_t, vs_bit_buffer[5 * 80 + 4]); /* allow some slack */ const int log2_blocksize= 3-s->avctx->lowres; @@ -410,6 +407,7 @@ static inline void dv_decode_video_segment(DVVideoContext *s, dct_mode = get_bits1(&gb); mb->dct_mode = dct_mode; mb->scan_table = s->dv_zigzag[dct_mode]; + mb->iweight_table = dct_mode ? dv_iweight_248 : dv_iweight_88; class1 = get_bits(&gb, 2); mb->shift_table = s->dv_idct_shift[class1 == 3][dct_mode] [quant + dv_quant_offset[class1]]; @@ -488,45 +486,63 @@ static inline void dv_decode_video_segment(DVVideoContext *s, v = *mb_pos_ptr++; mb_x = v & 0xff; mb_y = v >> 8; - y_ptr = s->picture.data[0] + ((mb_y * s->picture.linesize[0] + mb_x)<sys->pix_fmt == PIX_FMT_YUV411P) + if (s->sys->pix_fmt == PIX_FMT_YUV422P) { + y_ptr = s->picture.data[0] + ((mb_y * s->picture.linesize[0] + (mb_x>>1))<picture.linesize[1] + (mb_x >> 2))<> 1) * s->picture.linesize[1] + (mb_x >> 1))<picture.data[0] + ((mb_y * s->picture.linesize[0] + mb_x)<sys->pix_fmt == PIX_FMT_YUV411P) + c_offset = ((mb_y * s->picture.linesize[1] + (mb_x >> 2))<> 1) * s->picture.linesize[1] + (mb_x >> 1))<idct_put[mb->dct_mode && log2_blocksize==3]; - if (j < 4) { - if (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x < (704 / 8)) { - /* NOTE: at end of line, the macroblock is handled as 420 */ - idct_put(y_ptr + (j<picture.linesize[0], block); - } else { - idct_put(y_ptr + (((j & 1) + (j >> 1) * s->picture.linesize[0])<sys->pix_fmt == PIX_FMT_YUV422P) { /* 4:2:2 */ + if (j == 0 || j == 2) { + /* Y0 Y1 */ + idct_put(y_ptr + ((j >> 1)<picture.linesize[0], block); + } else if(j > 3) { + /* Cr Cb */ + idct_put(s->picture.data[6 - j] + c_offset, + s->picture.linesize[6 - j], block); } - } else { - if (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x >= (704 / 8)) { - uint64_t aligned_pixels[64/8]; - uint8_t *pixels= (uint8_t*)aligned_pixels; - uint8_t *c_ptr, *c_ptr1, *ptr, *ptr1; - int x, y, linesize; - /* NOTE: at end of line, the macroblock is handled as 420 */ - idct_put(pixels, 8, block); - linesize = s->picture.linesize[6 - j]; - c_ptr = s->picture.data[6 - j] + c_offset; - ptr = pixels; - for(y = 0;y < (1<sys->pix_fmt == PIX_FMT_YUV411P && mb_x < (704 / 8)) { + /* NOTE: at end of line, the macroblock is handled as 420 */ + idct_put(y_ptr + (j<picture.linesize[0], block); + } else { + idct_put(y_ptr + (((j & 1) + (j >> 1) * s->picture.linesize[0])<picture.linesize[0], block); } } else { - /* don't ask me why they inverted Cb and Cr ! */ - idct_put(s->picture.data[6 - j] + c_offset, - s->picture.linesize[6 - j], block); + if (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x >= (704 / 8)) { + uint64_t aligned_pixels[64/8]; + uint8_t *pixels= (uint8_t*)aligned_pixels; + uint8_t *c_ptr, *c_ptr1, *ptr, *ptr1; + int x, y, linesize; + /* NOTE: at end of line, the macroblock is handled as 420 */ + idct_put(pixels, 8, block); + linesize = s->picture.linesize[6 - j]; + c_ptr = s->picture.data[6 - j] + c_offset; + ptr = pixels; + for(y = 0;y < (1<picture.data[6 - j] + c_offset, + s->picture.linesize[6 - j], block); + } } } block += 64; @@ -537,7 +553,7 @@ static inline void dv_decode_video_segment(DVVideoContext *s, #ifdef DV_CODEC_TINY_TARGET /* Converts run and level (where level != 0) pair into vlc, returning bit size */ -static always_inline int dv_rl2vlc(int run, int level, int sign, uint32_t* vlc) +static av_always_inline int dv_rl2vlc(int run, int level, int sign, uint32_t* vlc) { int size; if (run < DV_VLC_MAP_RUN_SIZE && level < DV_VLC_MAP_LEV_SIZE) { @@ -562,7 +578,7 @@ static always_inline int dv_rl2vlc(int run, int level, int sign, uint32_t* vlc) return size; } -static always_inline int dv_rl2vlc_size(int run, int level) +static av_always_inline int dv_rl2vlc_size(int run, int level) { int size; @@ -578,13 +594,13 @@ static always_inline int dv_rl2vlc_size(int run, int level) return size; } #else -static always_inline int dv_rl2vlc(int run, int l, int sign, uint32_t* vlc) +static av_always_inline int dv_rl2vlc(int run, int l, int sign, uint32_t* vlc) { *vlc = dv_vlc_map[run][l].vlc | sign; return dv_vlc_map[run][l].size; } -static always_inline int dv_rl2vlc_size(int run, int l) +static av_always_inline int dv_rl2vlc_size(int run, int l) { return dv_vlc_map[run][l].size; } @@ -604,7 +620,7 @@ typedef struct EncBlockInfo { uint32_t partial_bit_buffer; /* we can't use uint16_t here */ } EncBlockInfo; -static always_inline PutBitContext* dv_encode_ac(EncBlockInfo* bi, PutBitContext* pb_pool, +static av_always_inline PutBitContext* dv_encode_ac(EncBlockInfo* bi, PutBitContext* pb_pool, PutBitContext* pb_end) { int prev; @@ -647,12 +663,26 @@ static always_inline PutBitContext* dv_encode_ac(EncBlockInfo* bi, PutBitContext return pb; } -static always_inline void dv_set_class_number(DCTELEM* blk, EncBlockInfo* bi, - const uint8_t* zigzag_scan, int bias) +static av_always_inline void dv_set_class_number(DCTELEM* blk, EncBlockInfo* bi, + const uint8_t* zigzag_scan, const int *weight, int bias) { int i, area; + /* We offer two different methods for class number assignment: the + method suggested in SMPTE 314M Table 22, and an improved + method. The SMPTE method is very conservative; it assigns class + 3 (i.e. severe quantization) to any block where the largest AC + component is greater than 36. ffmpeg's DV encoder tracks AC bit + consumption precisely, so there is no need to bias most blocks + towards strongly lossy compression. Instead, we assign class 2 + to most blocks, and use class 3 only when strictly necessary + (for blocks whose largest AC component exceeds 255). */ + +#if 0 /* SMPTE spec method */ static const int classes[] = {12, 24, 36, 0xffff}; - int max=12; +#else /* improved ffmpeg method */ + static const int classes[] = {-1, -1, 255, 0xffff}; +#endif + int max=classes[0]; int prev=0; bi->mb[0] = blk[0]; @@ -665,7 +695,11 @@ static always_inline void dv_set_class_number(DCTELEM* blk, EncBlockInfo* bi, if (level+15 > 30U) { bi->sign[i] = (level>>31)&1; - bi->mb[i] = level= ABS(level)>>4; + /* weigh it and and shift down into range, adding for rounding */ + /* the extra division by a factor of 2^4 reverses the 8x expansion of the DCT + AND the 2x doubling of the weights */ + level = (FFABS(level) * weight[i] + (1<<(dv_weight_bits+3))) >> (dv_weight_bits+4); + bi->mb[i] = level; if(level>max) max= level; bi->bit_size[area] += dv_rl2vlc_size(i - prev - 1, level); bi->next[prev]= i; @@ -701,7 +735,7 @@ static always_inline void dv_set_class_number(DCTELEM* blk, EncBlockInfo* bi, //FIXME replace this by dsputil #define SC(x, y) ((s[x] - s[y]) ^ ((s[x] - s[y]) >> 7)) -static always_inline int dv_guess_dct_mode(DCTELEM *blk) { +static av_always_inline int dv_guess_dct_mode(DCTELEM *blk) { DCTELEM *s; int score88 = 0; int score248 = 0; @@ -731,6 +765,7 @@ static inline void dv_guess_qnos(EncBlockInfo* blks, int* qnos) int i, j, k, a, prev, a2; EncBlockInfo* b; + size[0] = size[1] = size[2] = size[3] = size[4] = 1<<24; do { b = blks; for (i=0; i<5; i++) { @@ -745,6 +780,7 @@ static inline void dv_guess_qnos(EncBlockInfo* blks, int* qnos) b->bit_size[a] = 1; // 4 areas 4 bits for EOB :) b->area_q[a]++; prev= b->prev[a]; + assert(b->next[prev] >= mb_area_start[a+1] || b->mb[prev]); for (k= b->next[prev] ; knext[k]) { b->mb[k] >>= 1; if (b->mb[k]) { @@ -752,11 +788,14 @@ static inline void dv_guess_qnos(EncBlockInfo* blks, int* qnos) prev= k; } else { if(b->next[k] >= mb_area_start[a+1] && b->next[k]<64){ - for(a2=a+1; b->next[k] >= mb_area_start[a2+1]; a2++); + for(a2=a+1; b->next[k] >= mb_area_start[a2+1]; a2++) + b->prev[a2] = prev; assert(a2<4); assert(b->mb[b->next[k]]); b->bit_size[a2] += dv_rl2vlc_size(b->next[k] - prev - 1, b->mb[b->next[k]]) -dv_rl2vlc_size(b->next[k] - k - 1, b->mb[b->next[k]]); + assert(b->prev[a2]==k && (a2+1 >= 4 || b->prev[a2+1]!=k)); + b->prev[a2] = prev; } b->next[prev] = b->next[k]; } @@ -766,15 +805,15 @@ static inline void dv_guess_qnos(EncBlockInfo* blks, int* qnos) size[i] += b->bit_size[a]; } } + if(vs_total_ac_bits >= size[0] + size[1] + size[2] + size[3] + size[4]) + return; } - if(vs_total_ac_bits >= size[0] + size[1] + size[2] + size[3] + size[4]) - return; } while (qnos[0]|qnos[1]|qnos[2]|qnos[3]|qnos[4]); for(a=2; a==2 || vs_total_ac_bits < size[0]; a+=a){ b = blks; - size[0] = 0; + size[0] = 5*6*4; //EOB for (j=0; j<6*5; j++, b++) { prev= b->prev[0]; for (k= b->next[prev]; k<64; k= b->next[k]) { @@ -789,11 +828,6 @@ static inline void dv_guess_qnos(EncBlockInfo* blks, int* qnos) } } -/* - * This is a very rough initial implementaion. The performance is - * horrible and the weighting is missing. But it's missing from the - * decoding step also -- so at least we're on the same page with decoder ;-) - */ static inline void dv_encode_video_segment(DVVideoContext *s, uint8_t *dif, const uint16_t *mb_pos_ptr) @@ -804,7 +838,7 @@ static inline void dv_encode_video_segment(DVVideoContext *s, uint8_t* data; uint8_t* ptr; int do_edge_wrap; - DECLARE_ALIGNED_8(DCTELEM, block[64]); + DECLARE_ALIGNED_16(DCTELEM, block[64]); EncBlockInfo enc_blks[5*6]; PutBitContext pbs[5*6]; PutBitContext* pb; @@ -812,7 +846,7 @@ static inline void dv_encode_video_segment(DVVideoContext *s, int vs_bit_size = 0; int qnos[5]; - assert((((int)block) & 7) == 0); + assert((((int)block) & 15) == 0); enc_blk = &enc_blks[0]; pb = &pbs[0]; @@ -820,28 +854,52 @@ static inline void dv_encode_video_segment(DVVideoContext *s, v = *mb_pos_ptr++; mb_x = v & 0xff; mb_y = v >> 8; - y_ptr = s->picture.data[0] + (mb_y * s->picture.linesize[0] * 8) + (mb_x * 8); - c_offset = (s->sys->pix_fmt == PIX_FMT_YUV411P) ? - ((mb_y * s->picture.linesize[1] * 8) + ((mb_x >> 2) * 8)) : - (((mb_y >> 1) * s->picture.linesize[1] * 8) + ((mb_x >> 1) * 8)); + if (s->sys->pix_fmt == PIX_FMT_YUV422P) { + y_ptr = s->picture.data[0] + (mb_y * s->picture.linesize[0] * 8) + (mb_x * 4); + } else { /* 4:1:1 */ + y_ptr = s->picture.data[0] + (mb_y * s->picture.linesize[0] * 8) + (mb_x * 8); + } + if (s->sys->pix_fmt == PIX_FMT_YUV420P) { + c_offset = (((mb_y >> 1) * s->picture.linesize[1] * 8) + ((mb_x >> 1) * 8)); + } else { /* 4:2:2 or 4:1:1 */ + c_offset = ((mb_y * s->picture.linesize[1] * 8) + ((mb_x >> 2) * 8)); + } do_edge_wrap = 0; qnos[mb_index] = 15; /* No quantization */ ptr = dif + mb_index*80 + 4; for(j = 0;j < 6; j++) { - if (j < 4) { /* Four Y blocks */ - /* NOTE: at end of line, the macroblock is handled as 420 */ - if (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x < (704 / 8)) { - data = y_ptr + (j * 8); + int dummy = 0; + if (s->sys->pix_fmt == PIX_FMT_YUV422P) { /* 4:2:2 */ + if (j == 0 || j == 2) { + /* Y0 Y1 */ + data = y_ptr + ((j>>1) * 8); + linesize = s->picture.linesize[0]; + } else if (j > 3) { + /* Cr Cb */ + data = s->picture.data[6 - j] + c_offset; + linesize = s->picture.linesize[6 - j]; } else { - data = y_ptr + ((j & 1) * 8) + ((j >> 1) * 8 * s->picture.linesize[0]); + /* j=1 and j=3 are "dummy" blocks, used for AC data only */ + data = 0; + linesize = 0; + dummy = 1; + } + } else { /* 4:1:1 or 4:2:0 */ + if (j < 4) { /* Four Y blocks */ + /* NOTE: at end of line, the macroblock is handled as 420 */ + if (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x < (704 / 8)) { + data = y_ptr + (j * 8); + } else { + data = y_ptr + ((j & 1) * 8) + ((j >> 1) * 8 * s->picture.linesize[0]); + } + linesize = s->picture.linesize[0]; + } else { /* Cr and Cb blocks */ + /* don't ask Fabrice why they inverted Cb and Cr ! */ + data = s->picture.data[6 - j] + c_offset; + linesize = s->picture.linesize[6 - j]; + if (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x >= (704 / 8)) + do_edge_wrap = 1; } - linesize = s->picture.linesize[0]; - } else { /* Cr and Cb blocks */ - /* don't ask Fabrice why they inverted Cb and Cr ! */ - data = s->picture.data[6 - j] + c_offset; - linesize = s->picture.linesize[6 - j]; - if (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x >= (704 / 8)) - do_edge_wrap = 1; } /* Everything is set up -- now just copy data -> DCT block */ @@ -856,7 +914,8 @@ static inline void dv_encode_video_segment(DVVideoContext *s, b += 8; } } else { /* Simple copy: 8x8 -> 8x8 */ - s->get_pixels(block, data, linesize); + if (!dummy) + s->get_pixels(block, data, linesize); } if(s->avctx->flags & CODEC_FLAG_INTERLACED_DCT) @@ -868,10 +927,18 @@ static inline void dv_encode_video_segment(DVVideoContext *s, enc_blk->partial_bit_buffer = 0; enc_blk->cur_ac = 0; - s->fdct[enc_blk->dct_mode](block); + if (dummy) { + /* We rely on the fact that encoding all zeros leads to an immediate EOB, + which is precisely what the spec calls for in the "dummy" blocks. */ + memset(block, 0, sizeof(block)); + } else { + s->fdct[enc_blk->dct_mode](block); + } dv_set_class_number(block, enc_blk, - enc_blk->dct_mode ? ff_zigzag248_direct : ff_zigzag_direct, j/4); + enc_blk->dct_mode ? ff_zigzag248_direct : ff_zigzag_direct, + enc_blk->dct_mode ? dv_weight_248 : dv_weight_88, + j/4); init_put_bits(pb, ptr, block_sizes[j]/8); put_bits(pb, 9, (uint16_t)(((enc_blk->mb[0] >> 3) - 1024 + 2) >> 2)); @@ -923,25 +990,48 @@ static int dv_decode_mt(AVCodecContext *avctx, void* sl) { DVVideoContext *s = avctx->priv_data; int slice = (size_t)sl; - dv_decode_video_segment(s, &s->buf[((slice/27)*6+(slice/3)+slice*5+7)*80], + + /* which DIF channel is this? */ + int chan = slice / (s->sys->difseg_size * 27); + + /* slice within the DIF channel */ + int chan_slice = slice % (s->sys->difseg_size * 27); + + /* byte offset of this channel's data */ + int chan_offset = chan * s->sys->difseg_size * 150 * 80; + + dv_decode_video_segment(s, &s->buf[((chan_slice/27)*6+(chan_slice/3)+chan_slice*5+7)*80 + chan_offset], &s->sys->video_place[slice*5]); return 0; } +#ifdef CONFIG_ENCODERS static int dv_encode_mt(AVCodecContext *avctx, void* sl) { DVVideoContext *s = avctx->priv_data; int slice = (size_t)sl; - dv_encode_video_segment(s, &s->buf[((slice/27)*6+(slice/3)+slice*5+7)*80], + + /* which DIF channel is this? */ + int chan = slice / (s->sys->difseg_size * 27); + + /* slice within the DIF channel */ + int chan_slice = slice % (s->sys->difseg_size * 27); + + /* byte offset of this channel's data */ + int chan_offset = chan * s->sys->difseg_size * 150 * 80; + + dv_encode_video_segment(s, &s->buf[((chan_slice/27)*6+(chan_slice/3)+chan_slice*5+7)*80 + chan_offset], &s->sys->video_place[slice*5]); return 0; } +#endif +#ifdef CONFIG_DECODERS /* NOTE: exactly one frame must be given (120000 bytes for NTSC, - 144000 bytes for PAL) */ + 144000 bytes for PAL - or twice those for 50Mbps) */ static int dvvideo_decode_frame(AVCodecContext *avctx, void *data, int *data_size, - uint8_t *buf, int buf_size) + const uint8_t *buf, int buf_size) { DVVideoContext *s = avctx->priv_data; @@ -966,7 +1056,7 @@ static int dvvideo_decode_frame(AVCodecContext *avctx, s->buf = buf; avctx->execute(avctx, dv_decode_mt, (void**)&dv_anchor[0], NULL, - s->sys->difseg_size * 27); + s->sys->n_difchan * s->sys->difseg_size * 27); emms_c(); @@ -976,7 +1066,132 @@ static int dvvideo_decode_frame(AVCodecContext *avctx, return s->sys->frame_size; } +#endif + + +static inline int dv_write_pack(enum dv_pack_type pack_id, DVVideoContext *c, uint8_t* buf) +{ + /* + * Here's what SMPTE314M says about these two: + * (page 6) APTn, AP1n, AP2n, AP3n: These data shall be identical + * as track application IDs (APTn = 001, AP1n = + * 001, AP2n = 001, AP3n = 001), if the source signal + * comes from a digital VCR. If the signal source is + * unknown, all bits for these data shall be set to 1. + * (page 12) STYPE: STYPE defines a signal type of video signal + * 00000b = 4:1:1 compression + * 00100b = 4:2:2 compression + * XXXXXX = Reserved + * Now, I've got two problems with these statements: + * 1. it looks like APT == 111b should be a safe bet, but it isn't. + * It seems that for PAL as defined in IEC 61834 we have to set + * APT to 000 and for SMPTE314M to 001. + * 2. It is not at all clear what STYPE is used for 4:2:0 PAL + * compression scheme (if any). + */ + int apt = (c->sys->pix_fmt == PIX_FMT_YUV420P ? 0 : 1); + int stype = (c->sys->pix_fmt == PIX_FMT_YUV422P ? 4 : 0); + + uint8_t aspect = 0; + if((int)(av_q2d(c->avctx->sample_aspect_ratio) * c->avctx->width / c->avctx->height * 10) == 17) /* 16:9 */ + aspect = 0x02; + + buf[0] = (uint8_t)pack_id; + switch (pack_id) { + case dv_header525: /* I can't imagine why these two weren't defined as real */ + case dv_header625: /* packs in SMPTE314M -- they definitely look like ones */ + buf[1] = 0xf8 | /* reserved -- always 1 */ + (apt & 0x07); /* APT: Track application ID */ + buf[2] = (0 << 7) | /* TF1: audio data is 0 - valid; 1 - invalid */ + (0x0f << 3) | /* reserved -- always 1 */ + (apt & 0x07); /* AP1: Audio application ID */ + buf[3] = (0 << 7) | /* TF2: video data is 0 - valid; 1 - invalid */ + (0x0f << 3) | /* reserved -- always 1 */ + (apt & 0x07); /* AP2: Video application ID */ + buf[4] = (0 << 7) | /* TF3: subcode(SSYB) is 0 - valid; 1 - invalid */ + (0x0f << 3) | /* reserved -- always 1 */ + (apt & 0x07); /* AP3: Subcode application ID */ + break; + case dv_video_source: + buf[1] = 0xff; /* reserved -- always 1 */ + buf[2] = (1 << 7) | /* B/W: 0 - b/w, 1 - color */ + (1 << 6) | /* following CLF is valid - 0, invalid - 1 */ + (3 << 4) | /* CLF: color frames id (see ITU-R BT.470-4) */ + 0xf; /* reserved -- always 1 */ + buf[3] = (3 << 6) | /* reserved -- always 1 */ + (c->sys->dsf << 5) | /* system: 60fields/50fields */ + stype; /* signal type video compression */ + buf[4] = 0xff; /* VISC: 0xff -- no information */ + break; + case dv_video_control: + buf[1] = (0 << 6) | /* Copy generation management (CGMS) 0 -- free */ + 0x3f; /* reserved -- always 1 */ + buf[2] = 0xc8 | /* reserved -- always b11001xxx */ + aspect; + buf[3] = (1 << 7) | /* Frame/field flag 1 -- frame, 0 -- field */ + (1 << 6) | /* First/second field flag 0 -- field 2, 1 -- field 1 */ + (1 << 5) | /* Frame change flag 0 -- same picture as before, 1 -- different */ + (1 << 4) | /* 1 - interlaced, 0 - noninterlaced */ + 0xc; /* reserved -- always b1100 */ + buf[4] = 0xff; /* reserved -- always 1 */ + break; + default: + buf[1] = buf[2] = buf[3] = buf[4] = 0xff; + } + return 5; +} + +static void dv_format_frame(DVVideoContext* c, uint8_t* buf) +{ + int chan, i, j, k; + + for (chan = 0; chan < c->sys->n_difchan; chan++) { + for (i = 0; i < c->sys->difseg_size; i++) { + memset(buf, 0xff, 80 * 6); /* First 6 DIF blocks are for control data */ + + /* DV header: 1DIF */ + buf += dv_write_dif_id(dv_sect_header, chan, i, 0, buf); + buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525), c, buf); + buf += 72; /* unused bytes */ + + /* DV subcode: 2DIFs */ + for (j = 0; j < 2; j++) { + buf += dv_write_dif_id(dv_sect_subcode, chan, i, j, buf); + for (k = 0; k < 6; k++) + buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size/2), buf) + 5; + buf += 29; /* unused bytes */ + } + + /* DV VAUX: 3DIFS */ + for (j = 0; j < 3; j++) { + buf += dv_write_dif_id(dv_sect_vaux, chan, i, j, buf); + buf += dv_write_pack(dv_video_source, c, buf); + buf += dv_write_pack(dv_video_control, c, buf); + buf += 7*5; + buf += dv_write_pack(dv_video_source, c, buf); + buf += dv_write_pack(dv_video_control, c, buf); + buf += 4*5 + 2; /* unused bytes */ + } + + /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */ + for (j = 0; j < 135; j++) { + if (j%15 == 0) { + memset(buf, 0xff, 80); + buf += dv_write_dif_id(dv_sect_audio, chan, i, j/15, buf); + buf += 77; /* audio control & shuffled PCM audio */ + } + buf += dv_write_dif_id(dv_sect_video, chan, i, j, buf); + buf += 77; /* 1 video macro block: 1 bytes control + 4 * 14 bytes Y 8x8 data + 10 bytes Cr 8x8 data + 10 bytes Cb 8x8 data */ + } + } + } +} + +#ifdef CONFIG_ENCODERS static int dvvideo_encode_frame(AVCodecContext *c, uint8_t *buf, int buf_size, void *data) { @@ -995,14 +1210,22 @@ static int dvvideo_encode_frame(AVCodecContext *c, uint8_t *buf, int buf_size, s->buf = buf; c->execute(c, dv_encode_mt, (void**)&dv_anchor[0], NULL, - s->sys->difseg_size * 27); + s->sys->n_difchan * s->sys->difseg_size * 27); emms_c(); + + dv_format_frame(s, buf); + return s->sys->frame_size; } +#endif static int dvvideo_close(AVCodecContext *c) { + DVVideoContext *s = c->priv_data; + + if(s->picture.data[0]) + c->release_buffer(c, &s->picture); return 0; } @@ -1016,13 +1239,11 @@ AVCodec dvvideo_encoder = { sizeof(DVVideoContext), dvvideo_init, dvvideo_encode_frame, - dvvideo_close, - NULL, - CODEC_CAP_DR1, - NULL + .pix_fmts = (enum PixelFormat[]) {PIX_FMT_YUV411P, PIX_FMT_YUV422P, PIX_FMT_YUV420P, -1}, }; #endif // CONFIG_DVVIDEO_ENCODER +#ifdef CONFIG_DVVIDEO_DECODER AVCodec dvvideo_decoder = { "dvvideo", CODEC_TYPE_VIDEO, @@ -1035,3 +1256,4 @@ AVCodec dvvideo_decoder = { CODEC_CAP_DR1, NULL }; +#endif