/*
* DV decoder
- * Copyright (c) 2002 Fabrice Bellard.
- * Copyright (c) 2004 Roman Shaposhnik.
+ * Copyright (c) 2002 Fabrice Bellard
+ * Copyright (c) 2004 Roman Shaposhnik
*
* DV encoder
- * Copyright (c) 2003 Roman Shaposhnik.
+ * Copyright (c) 2003 Roman Shaposhnik
*
* 50 Mbps (DVCPRO50) support
* Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
* Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
* of DV technical info.
*
- * 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 dv.c
+ * @file
* DV codec.
*/
#define ALT_BITSTREAM_READER
+#include "libavutil/pixdesc.h"
#include "avcodec.h"
#include "dsputil.h"
-#include "bitstream.h"
+#include "get_bits.h"
+#include "put_bits.h"
#include "simple_idct.h"
#include "dvdata.h"
+#include "dv_tablegen.h"
//#undef NDEBUG
//#include <assert.h>
uint8_t *buf;
uint8_t dv_zigzag[2][64];
- uint32_t dv_idct_factor[2][2][22][64];
- uint32_t dv100_idct_factor[4][4][16][64];
void (*get_pixels)(DCTELEM *block, const uint8_t *pixels, int line_size);
void (*fdct[2])(DCTELEM *block);
void (*idct_put[2])(uint8_t *dest, int line_size, DCTELEM *block);
+ me_cmp_func ildct_cmp;
} DVVideoContext;
#define TEX_VLC_BITS 9
-#if ENABLE_SMALL
-#define DV_VLC_MAP_RUN_SIZE 15
-#define DV_VLC_MAP_LEV_SIZE 23
-#else
-#define DV_VLC_MAP_RUN_SIZE 64
-#define DV_VLC_MAP_LEV_SIZE 512 //FIXME sign was removed so this should be /2 but needs check
-#endif
-
/* XXX: also include quantization */
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_RUN_SIZE][DV_VLC_MAP_LEV_SIZE];
static inline int dv_work_pool_size(const DVprofile *d)
{
return size;
}
-static int dv_init_dynamic_tables(const DVprofile *d)
+static inline void dv_calc_mb_coordinates(const DVprofile *d, int chan, int seq, int slot,
+ uint16_t *tbl)
{
- int j,i,c,s,p,k;
-
- if (d->work_chunks[dv_work_pool_size(d)-1].buf_offset)
- return 0;
-
- p = i = 0;
- for (c=0; c<d->n_difchan; c++) {
- for (s=0; s<d->difseg_size; s++) {
- p += 6;
- for (j=0; j<27; j++) {
- p += !(j%3);
- if (!(DV_PROFILE_IS_1080i50(d) && c != 0 && s == 11) &&
- !(DV_PROFILE_IS_720p50(d) && s > 9)) {
- for (k=0; k<5; k++)
- d->work_chunks[i].mb_coordinates[k] = d->video_place[(c*d->difseg_size+s)*27*5 + j*5 + k];
- d->work_chunks[i++].buf_offset = p;
- }
- p += 5;
- }
+ static const uint8_t off[] = { 2, 6, 8, 0, 4 };
+ static const uint8_t shuf1[] = { 36, 18, 54, 0, 72 };
+ static const uint8_t shuf2[] = { 24, 12, 36, 0, 48 };
+ static const uint8_t shuf3[] = { 18, 9, 27, 0, 36 };
+
+ static const uint8_t l_start[] = {0, 4, 9, 13, 18, 22, 27, 31, 36, 40};
+ static const uint8_t l_start_shuffled[] = { 9, 4, 13, 0, 18 };
+
+ static const uint8_t serpent1[] = {0, 1, 2, 2, 1, 0,
+ 0, 1, 2, 2, 1, 0,
+ 0, 1, 2, 2, 1, 0,
+ 0, 1, 2, 2, 1, 0,
+ 0, 1, 2};
+ static const uint8_t serpent2[] = {0, 1, 2, 3, 4, 5, 5, 4, 3, 2, 1, 0,
+ 0, 1, 2, 3, 4, 5, 5, 4, 3, 2, 1, 0,
+ 0, 1, 2, 3, 4, 5};
+
+ static const uint8_t remap[][2] = {{ 0, 0}, { 0, 0}, { 0, 0}, { 0, 0}, /* dummy */
+ { 0, 0}, { 0, 1}, { 0, 2}, { 0, 3}, {10, 0},
+ {10, 1}, {10, 2}, {10, 3}, {20, 0}, {20, 1},
+ {20, 2}, {20, 3}, {30, 0}, {30, 1}, {30, 2},
+ {30, 3}, {40, 0}, {40, 1}, {40, 2}, {40, 3},
+ {50, 0}, {50, 1}, {50, 2}, {50, 3}, {60, 0},
+ {60, 1}, {60, 2}, {60, 3}, {70, 0}, {70, 1},
+ {70, 2}, {70, 3}, { 0,64}, { 0,65}, { 0,66},
+ {10,64}, {10,65}, {10,66}, {20,64}, {20,65},
+ {20,66}, {30,64}, {30,65}, {30,66}, {40,64},
+ {40,65}, {40,66}, {50,64}, {50,65}, {50,66},
+ {60,64}, {60,65}, {60,66}, {70,64}, {70,65},
+ {70,66}, { 0,67}, {20,67}, {40,67}, {60,67}};
+
+ int i, k, m;
+ int x, y, blk;
+
+ for (m=0; m<5; m++) {
+ switch (d->width) {
+ case 1440:
+ blk = (chan*11+seq)*27+slot;
+
+ if (chan == 0 && seq == 11) {
+ x = m*27+slot;
+ if (x<90) {
+ y = 0;
+ } else {
+ x = (x - 90)*2;
+ y = 67;
+ }
+ } else {
+ i = (4*chan + blk + off[m])%11;
+ k = (blk/11)%27;
+
+ x = shuf1[m] + (chan&1)*9 + k%9;
+ y = (i*3+k/9)*2 + (chan>>1) + 1;
+ }
+ tbl[m] = (x<<1)|(y<<9);
+ break;
+ case 1280:
+ blk = (chan*10+seq)*27+slot;
+
+ i = (4*chan + (seq/5) + 2*blk + off[m])%10;
+ k = (blk/5)%27;
+
+ x = shuf1[m]+(chan&1)*9 + k%9;
+ y = (i*3+k/9)*2 + (chan>>1) + 4;
+
+ if (x >= 80) {
+ x = remap[y][0]+((x-80)<<(y>59));
+ y = remap[y][1];
+ }
+ tbl[m] = (x<<1)|(y<<9);
+ break;
+ case 960:
+ blk = (chan*10+seq)*27+slot;
+
+ i = (4*chan + (seq/5) + 2*blk + off[m])%10;
+ k = (blk/5)%27 + (i&1)*3;
+
+ x = shuf2[m] + k%6 + 6*(chan&1);
+ y = l_start[i] + k/6 + 45*(chan>>1);
+ tbl[m] = (x<<1)|(y<<9);
+ break;
+ case 720:
+ switch (d->pix_fmt) {
+ case PIX_FMT_YUV422P:
+ x = shuf3[m] + slot/3;
+ y = serpent1[slot] +
+ ((((seq + off[m]) % d->difseg_size)<<1) + chan)*3;
+ tbl[m] = (x<<1)|(y<<8);
+ break;
+ case PIX_FMT_YUV420P:
+ x = shuf3[m] + slot/3;
+ y = serpent1[slot] +
+ ((seq + off[m]) % d->difseg_size)*3;
+ tbl[m] = (x<<1)|(y<<9);
+ break;
+ case PIX_FMT_YUV411P:
+ i = (seq + off[m]) % d->difseg_size;
+ k = slot + ((m==1||m==2)?3:0);
+
+ x = l_start_shuffled[m] + k/6;
+ y = serpent2[k] + i*6;
+ if (x>21)
+ y = y*2 - i*6;
+ tbl[m] = (x<<2)|(y<<8);
+ break;
+ }
+ default:
+ break;
}
}
- return 0;
}
-static void dv_build_unquantize_tables(DVVideoContext *s, uint8_t* perm)
+static int dv_init_dynamic_tables(const DVprofile *d)
{
- int i, q, a;
-
- /* NOTE: max left shift is 6 */
- for (q = 0; q < 22; q++) {
- /* 88DCT */
- i = 1;
- for (a = 0; a < 4; a++) {
- for (; i < dv_quant_areas[a]; i++) {
- /* 88 table */
- s->dv_idct_factor[0][0][q][i] = dv_iweight_88[i] << (dv_quant_shifts[q][a] + 1);
- s->dv_idct_factor[1][0][q][i] = s->dv_idct_factor[0][0][q][i] << 1;
-
- /* 248 table */
- s->dv_idct_factor[0][1][q][i] = dv_iweight_248[i] << (dv_quant_shifts[q][a] + 1);
- s->dv_idct_factor[1][1][q][i] = s->dv_idct_factor[0][1][q][i] << 1;
+ int j,i,c,s,p;
+ uint32_t *factor1, *factor2;
+ const int *iweight1, *iweight2;
+
+ if (!d->work_chunks[dv_work_pool_size(d)-1].buf_offset) {
+ p = i = 0;
+ for (c=0; c<d->n_difchan; c++) {
+ for (s=0; s<d->difseg_size; s++) {
+ p += 6;
+ for (j=0; j<27; j++) {
+ p += !(j%3);
+ if (!(DV_PROFILE_IS_1080i50(d) && c != 0 && s == 11) &&
+ !(DV_PROFILE_IS_720p50(d) && s > 9)) {
+ dv_calc_mb_coordinates(d, c, s, j, &d->work_chunks[i].mb_coordinates[0]);
+ d->work_chunks[i++].buf_offset = p;
+ }
+ p += 5;
+ }
}
}
}
- for (a = 0; a < 4; a++) {
- for (q = 0; q < 16; q++) {
- for (i = 1; i < 64; i++) {
- s->dv100_idct_factor[0][a][q][i] = (dv100_qstep[q] << (a + 9)) * dv_iweight_1080_y[i];
- s->dv100_idct_factor[1][a][q][i] = (dv100_qstep[q] << (a + 9)) * dv_iweight_1080_c[i];
- s->dv100_idct_factor[2][a][q][i] = (dv100_qstep[q] << (a + 9)) * dv_iweight_720_y[i];
- s->dv100_idct_factor[3][a][q][i] = (dv100_qstep[q] << (a + 9)) * dv_iweight_720_c[i];
+ if (!d->idct_factor[DV_PROFILE_IS_HD(d)?8191:5631]) {
+ factor1 = &d->idct_factor[0];
+ factor2 = &d->idct_factor[DV_PROFILE_IS_HD(d)?4096:2816];
+ if (d->height == 720) {
+ iweight1 = &dv_iweight_720_y[0];
+ iweight2 = &dv_iweight_720_c[0];
+ } else {
+ iweight1 = &dv_iweight_1080_y[0];
+ iweight2 = &dv_iweight_1080_c[0];
+ }
+ if (DV_PROFILE_IS_HD(d)) {
+ for (c = 0; c < 4; c++) {
+ for (s = 0; s < 16; s++) {
+ for (i = 0; i < 64; i++) {
+ *factor1++ = (dv100_qstep[s] << (c + 9)) * iweight1[i];
+ *factor2++ = (dv100_qstep[s] << (c + 9)) * iweight2[i];
+ }
+ }
+ }
+ } else {
+ iweight1 = &dv_iweight_88[0];
+ for (j = 0; j < 2; j++, iweight1 = &dv_iweight_248[0]) {
+ for (s = 0; s < 22; s++) {
+ for (i = c = 0; c < 4; c++) {
+ for (; i < dv_quant_areas[c]; i++) {
+ *factor1 = iweight1[i] << (dv_quant_shifts[s][c] + 1);
+ *factor2++ = (*factor1++) << 1;
+ }
+ }
+ }
}
}
}
+
+ return 0;
}
static av_cold int dvvideo_init(AVCodecContext *avctx)
}
free_vlc(&dv_vlc);
- for (i = 0; i < NB_DV_VLC - 1; i++) {
- if (dv_vlc_run[i] >= DV_VLC_MAP_RUN_SIZE)
- continue;
-#if ENABLE_SMALL
- if (dv_vlc_level[i] >= DV_VLC_MAP_LEV_SIZE)
- continue;
-#endif
-
- if (dv_vlc_map[dv_vlc_run[i]][dv_vlc_level[i]].size != 0)
- continue;
-
- dv_vlc_map[dv_vlc_run[i]][dv_vlc_level[i]].vlc =
- dv_vlc_bits[i] << (!!dv_vlc_level[i]);
- dv_vlc_map[dv_vlc_run[i]][dv_vlc_level[i]].size =
- dv_vlc_len[i] + (!!dv_vlc_level[i]);
- }
- for (i = 0; i < DV_VLC_MAP_RUN_SIZE; i++) {
-#if ENABLE_SMALL
- for (j = 1; j < DV_VLC_MAP_LEV_SIZE; j++) {
- if (dv_vlc_map[i][j].size == 0) {
- dv_vlc_map[i][j].vlc = dv_vlc_map[0][j].vlc |
- (dv_vlc_map[i-1][0].vlc << (dv_vlc_map[0][j].size));
- dv_vlc_map[i][j].size = dv_vlc_map[i-1][0].size +
- dv_vlc_map[0][j].size;
- }
- }
-#else
- for (j = 1; j < DV_VLC_MAP_LEV_SIZE/2; j++) {
- if (dv_vlc_map[i][j].size == 0) {
- dv_vlc_map[i][j].vlc = dv_vlc_map[0][j].vlc |
- (dv_vlc_map[i-1][0].vlc << (dv_vlc_map[0][j].size));
- dv_vlc_map[i][j].size = dv_vlc_map[i-1][0].size +
- dv_vlc_map[0][j].size;
- }
- dv_vlc_map[i][((uint16_t)(-j))&0x1ff].vlc =
- dv_vlc_map[i][j].vlc | 1;
- dv_vlc_map[i][((uint16_t)(-j))&0x1ff].size =
- dv_vlc_map[i][j].size;
- }
-#endif
- }
+ dv_vlc_map_tableinit();
}
/* Generic DSP setup */
dsputil_init(&dsp, avctx);
+ ff_set_cmp(&dsp, dsp.ildct_cmp, avctx->ildct_cmp);
s->get_pixels = dsp.get_pixels;
+ s->ildct_cmp = dsp.ildct_cmp[5];
/* 88DCT setup */
s->fdct[0] = dsp.fdct;
}else
memcpy(s->dv_zigzag[1], ff_zigzag248_direct, 64);
- /* XXX: do it only for constant case */
- dv_build_unquantize_tables(s, dsp.idct_permutation);
-
avctx->coded_frame = &s->picture;
s->avctx = avctx;
+ avctx->chroma_sample_location = AVCHROMA_LOC_TOPLEFT;
return 0;
}
-// #define VLC_DEBUG
-// #define printf(...) av_log(NULL, AV_LOG_ERROR, __VA_ARGS__)
+static av_cold int dvvideo_init_encoder(AVCodecContext *avctx)
+{
+ if (!avpriv_dv_codec_profile(avctx)) {
+ av_log(avctx, AV_LOG_ERROR, "Found no DV profile for %ix%i %s video\n",
+ avctx->width, avctx->height, av_get_pix_fmt_name(avctx->pix_fmt));
+ return -1;
+ }
+
+ return dvvideo_init(avctx);
+}
typedef struct BlockInfo {
const uint32_t *factor_table;
/* bit budget for AC only in 5 MBs */
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 };
-static inline int get_bits_left(GetBitContext *s)
-{
- return s->size_in_bits - get_bits_count(s);
-}
-
static inline int put_bits_left(PutBitContext* s)
{
return (s->buf_end - s->buf) * 8 - put_bits_count(s);
}
-/* decode ac coefficients */
+/* decode AC coefficients */
static void dv_decode_ac(GetBitContext *gb, BlockInfo *mb, DCTELEM *block)
{
int last_index = gb->size_in_bits;
OPEN_READER(re, gb);
UPDATE_CACHE(re, gb);
- /* if we must parse a partial vlc, we do it here */
+ /* if we must parse a partial VLC, we do it here */
if (partial_bit_count > 0) {
re_cache = ((unsigned)re_cache >> partial_bit_count) |
(mb->partial_bit_buffer << (sizeof(re_cache) * 8 - partial_bit_count));
/* get the AC coefficients until last_index is reached */
for (;;) {
-#ifdef VLC_DEBUG
- printf("%2d: bits=%04x index=%d\n", pos, SHOW_UBITS(re, gb, 16), re_index);
-#endif
+ av_dlog(NULL, "%2d: bits=%04x index=%d\n", pos, SHOW_UBITS(re, gb, 16),
+ re_index);
/* our own optimized GET_RL_VLC */
index = NEG_USR32(re_cache, TEX_VLC_BITS);
vlc_len = dv_rl_vlc[index].len;
}
re_index += vlc_len;
-#ifdef VLC_DEBUG
- printf("run=%d level=%d\n", run, level);
-#endif
+ av_dlog(NULL, "run=%d level=%d\n", run, level);
pos += run;
if (pos >= 64)
break;
}
/* mb_x and mb_y are in units of 8 pixels */
-static int dv_decode_video_segment(AVCodecContext *avctx, DVwork_chunk *work_chunk)
+static int dv_decode_video_segment(AVCodecContext *avctx, void *arg)
{
DVVideoContext *s = avctx->priv_data;
+ DVwork_chunk *work_chunk = arg;
int quant, dc, dct_mode, class1, j;
int mb_index, mb_x, mb_y, last_index;
int y_stride, linesize;
PutBitContext pb, vs_pb;
GetBitContext gb;
BlockInfo mb_data[5 * DV_MAX_BPM], *mb, *mb1;
- DECLARE_ALIGNED_16(DCTELEM, sblock[5*DV_MAX_BPM][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 */
+ LOCAL_ALIGNED_16(DCTELEM, sblock, [5*DV_MAX_BPM], [64]);
+ LOCAL_ALIGNED_16(uint8_t, mb_bit_buffer, [ 80 + FF_INPUT_BUFFER_PADDING_SIZE]); /* allow some slack */
+ LOCAL_ALIGNED_16(uint8_t, vs_bit_buffer, [5*80 + FF_INPUT_BUFFER_PADDING_SIZE]); /* allow some slack */
const int log2_blocksize = 3-s->avctx->lowres;
int is_field_mode[5];
assert((((int)mb_bit_buffer) & 7) == 0);
assert((((int)vs_bit_buffer) & 7) == 0);
- memset(sblock, 0, sizeof(sblock));
+ memset(sblock, 0, 5*DV_MAX_BPM*sizeof(*sblock));
- /* pass 1 : read DC and AC coefficients in blocks */
+ /* pass 1: read DC and AC coefficients in blocks */
buf_ptr = &s->buf[work_chunk->buf_offset*80];
block1 = &sblock[0][0];
mb1 = mb_data;
last_index = s->sys->block_sizes[j];
init_get_bits(&gb, buf_ptr, last_index);
- /* get the dc */
+ /* get the DC */
dc = get_sbits(&gb, 9);
dct_mode = get_bits1(&gb);
class1 = get_bits(&gb, 2);
if (DV_PROFILE_IS_HD(s->sys)) {
mb->idct_put = s->idct_put[0];
mb->scan_table = s->dv_zigzag[0];
- mb->factor_table = s->dv100_idct_factor[((s->sys->height == 720) << 1) | (j >= 4)][class1][quant];
+ mb->factor_table = &s->sys->idct_factor[(j >= 4)*4*16*64 + class1*16*64 + quant*64];
is_field_mode[mb_index] |= !j && dct_mode;
} else {
mb->idct_put = s->idct_put[dct_mode && log2_blocksize == 3];
mb->scan_table = s->dv_zigzag[dct_mode];
- mb->factor_table = s->dv_idct_factor[class1 == 3][dct_mode]
- [quant + dv_quant_offset[class1]];
+ mb->factor_table = &s->sys->idct_factor[(class1 == 3)*2*22*64 + dct_mode*22*64 +
+ (quant + dv_quant_offset[class1])*64];
}
dc = dc << 2;
/* convert to unsigned because 128 is not added in the
mb->pos = 0;
mb->partial_bit_count = 0;
-#ifdef VLC_DEBUG
- printf("MB block: %d, %d ", mb_index, j);
-#endif
+ av_dlog(avctx, "MB block: %d, %d ", mb_index, j);
dv_decode_ac(&gb, mb, block);
- /* write the remaining bits in a new buffer only if the
+ /* write the remaining bits in a new buffer only if the
block is finished */
if (mb->pos >= 64)
bit_copy(&pb, &gb);
mb++;
}
- /* pass 2 : we can do it just after */
-#ifdef VLC_DEBUG
- printf("***pass 2 size=%d MB#=%d\n", put_bits_count(&pb), mb_index);
-#endif
+ /* pass 2: we can do it just after */
+ av_dlog(avctx, "***pass 2 size=%d MB#=%d\n", put_bits_count(&pb), mb_index);
block = block1;
mb = mb1;
init_get_bits(&gb, mb_bit_buffer, put_bits_count(&pb));
+ put_bits32(&pb, 0); // padding must be zeroed
flush_put_bits(&pb);
for (j = 0; j < s->sys->bpm; j++, block += 64, mb++) {
if (mb->pos < 64 && get_bits_left(&gb) > 0) {
bit_copy(&vs_pb, &gb);
}
- /* we need a pass other the whole video segment */
-#ifdef VLC_DEBUG
- printf("***pass 3 size=%d\n", put_bits_count(&vs_pb));
-#endif
+ /* we need a pass over the whole video segment */
+ av_dlog(avctx, "***pass 3 size=%d\n", put_bits_count(&vs_pb));
block = &sblock[0][0];
mb = mb_data;
init_get_bits(&gb, vs_bit_buffer, put_bits_count(&vs_pb));
+ put_bits32(&vs_pb, 0); // padding must be zeroed
flush_put_bits(&vs_pb);
for (mb_index = 0; mb_index < 5; mb_index++) {
for (j = 0; j < s->sys->bpm; j++) {
if (mb->pos < 64) {
-#ifdef VLC_DEBUG
- printf("start %d:%d\n", mb_index, j);
-#endif
+ av_dlog(avctx, "start %d:%d\n", mb_index, j);
dv_decode_ac(&gb, mb, block);
}
if (mb->pos >= 64 && mb->pos < 127)
- av_log(NULL, AV_LOG_ERROR, "AC EOB marker is absent pos=%d\n", mb->pos);
+ av_log(avctx, AV_LOG_ERROR, "AC EOB marker is absent pos=%d\n", mb->pos);
block += 64;
mb++;
}
return 0;
}
-#if ENABLE_SMALL
-/* Converts run and level (where level != 0) pair into vlc, returning bit size */
+#if CONFIG_SMALL
+/* Converts run and level (where level != 0) pair into VLC, returning bit size */
static av_always_inline int dv_rl2vlc(int run, int level, int sign, uint32_t* vlc)
{
int size;
return pb;
}
-static av_always_inline void dv_set_class_number(DCTELEM* blk, EncBlockInfo* bi,
- const uint8_t* zigzag_scan,
- const int *weight, int bias)
+static av_always_inline int dv_guess_dct_mode(DVVideoContext *s, uint8_t *data, int linesize) {
+ if (s->avctx->flags & CODEC_FLAG_INTERLACED_DCT) {
+ int ps = s->ildct_cmp(NULL, data, NULL, linesize, 8) - 400;
+ if (ps > 0) {
+ int is = s->ildct_cmp(NULL, data , NULL, linesize<<1, 4) +
+ s->ildct_cmp(NULL, data + linesize, NULL, linesize<<1, 4);
+ return (ps > is);
+ }
+ }
+
+ return 0;
+}
+
+static av_always_inline int dv_init_enc_block(EncBlockInfo* bi, uint8_t *data, int linesize, DVVideoContext *s, int bias)
{
+ const int *weight;
+ const uint8_t* zigzag_scan;
+ LOCAL_ALIGNED_16(DCTELEM, blk, [64]);
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
+ component is greater than 36. Libav'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
#if 0 /* SMPTE spec method */
static const int classes[] = {12, 24, 36, 0xffff};
-#else /* improved FFmpeg method */
+#else /* improved Libav method */
static const int classes[] = {-1, -1, 255, 0xffff};
#endif
int max = classes[0];
int prev = 0;
+ assert((((int)blk) & 15) == 0);
+
+ bi->area_q[0] = bi->area_q[1] = bi->area_q[2] = bi->area_q[3] = 0;
+ bi->partial_bit_count = 0;
+ bi->partial_bit_buffer = 0;
+ bi->cur_ac = 0;
+ if (data) {
+ bi->dct_mode = dv_guess_dct_mode(s, data, linesize);
+ s->get_pixels(blk, data, linesize);
+ s->fdct[bi->dct_mode](blk);
+ } else {
+ /* 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(blk, 0, 64*sizeof(*blk));
+ bi->dct_mode = 0;
+ }
bi->mb[0] = blk[0];
+ zigzag_scan = bi->dct_mode ? ff_zigzag248_direct : ff_zigzag_direct;
+ weight = bi->dct_mode ? dv_weight_248 : dv_weight_88;
+
for (area = 0; area < 4; area++) {
bi->prev[area] = prev;
bi->bit_size[area] = 1; // 4 areas 4 bits for EOB :)
if (level + 15 > 30U) {
bi->sign[i] = (level >> 31) & 1;
- /* weigh it and and shift down into range, adding for rounding */
+ /* weight 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->next[prev]= i;
}
-}
-
-//FIXME replace this by dsputil
-#define SC(x, y) ((s[x] - s[y]) ^ ((s[x] - s[y]) >> 7))
-static av_always_inline int dv_guess_dct_mode(DCTELEM *blk) {
- DCTELEM *s;
- int score88 = 0;
- int score248 = 0;
- int i;
-
- /* Compute 8-8 score (small values give a better chance for 8-8 DCT) */
- s = blk;
- for (i = 0; i < 7; i++) {
- score88 += SC(0, 8) + SC(1, 9) + SC(2, 10) + SC(3, 11) +
- SC(4, 12) + SC(5,13) + SC(6, 14) + SC(7, 15);
- s += 8;
- }
- /* Compute 2-4-8 score (small values give a better chance for 2-4-8 DCT) */
- s = blk;
- for (i = 0; i < 6; i++) {
- score248 += SC(0, 16) + SC(1,17) + SC(2, 18) + SC(3, 19) +
- SC(4, 20) + SC(5,21) + SC(6, 22) + SC(7, 23);
- s += 8;
- }
- return (score88 - score248 > -10);
+ return bi->bit_size[0] + bi->bit_size[1] + bi->bit_size[2] + bi->bit_size[3];
}
static inline void dv_guess_qnos(EncBlockInfo* blks, int* qnos)
}
}
-static int dv_encode_video_segment(AVCodecContext *avctx, DVwork_chunk *work_chunk)
+static int dv_encode_video_segment(AVCodecContext *avctx, void *arg)
{
DVVideoContext *s = avctx->priv_data;
+ DVwork_chunk *work_chunk = arg;
int mb_index, i, j;
- int mb_x, mb_y, c_offset, linesize;
+ int mb_x, mb_y, c_offset, linesize, y_stride;
uint8_t* y_ptr;
- uint8_t* data;
- uint8_t* ptr;
uint8_t* dif;
- int do_edge_wrap;
- DECLARE_ALIGNED_16(DCTELEM, block[64]);
- EncBlockInfo enc_blks[5*6];
- PutBitContext pbs[5*6];
+ LOCAL_ALIGNED_8(uint8_t, scratch, [64]);
+ EncBlockInfo enc_blks[5*DV_MAX_BPM];
+ PutBitContext pbs[5*DV_MAX_BPM];
PutBitContext* pb;
EncBlockInfo* enc_blk;
int vs_bit_size = 0;
- int qnos[5];
-
- assert((((int)block) & 15) == 0);
+ int qnos[5] = {15, 15, 15, 15, 15}; /* No quantization */
+ int* qnosp = &qnos[0];
dif = &s->buf[work_chunk->buf_offset*80];
enc_blk = &enc_blks[0];
- pb = &pbs[0];
for (mb_index = 0; mb_index < 5; mb_index++) {
dv_calculate_mb_xy(s, work_chunk, mb_index, &mb_x, &mb_y);
+
+ /* initializing luminance blocks */
+ if ((s->sys->pix_fmt == PIX_FMT_YUV420P) ||
+ (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x >= (704 / 8)) ||
+ (s->sys->height >= 720 && mb_y != 134)) {
+ y_stride = s->picture.linesize[0] << 3;
+ } else {
+ y_stride = 16;
+ }
y_ptr = s->picture.data[0] + ((mb_y * s->picture.linesize[0] + mb_x) << 3);
+ linesize = s->picture.linesize[0];
+
+ if (s->sys->video_stype == 4) { /* SD 422 */
+ vs_bit_size +=
+ dv_init_enc_block(enc_blk+0, y_ptr , linesize, s, 0) +
+ dv_init_enc_block(enc_blk+1, NULL , linesize, s, 0) +
+ dv_init_enc_block(enc_blk+2, y_ptr + 8 , linesize, s, 0) +
+ dv_init_enc_block(enc_blk+3, NULL , linesize, s, 0);
+ } else {
+ vs_bit_size +=
+ dv_init_enc_block(enc_blk+0, y_ptr , linesize, s, 0) +
+ dv_init_enc_block(enc_blk+1, y_ptr + 8 , linesize, s, 0) +
+ dv_init_enc_block(enc_blk+2, y_ptr + y_stride, linesize, s, 0) +
+ dv_init_enc_block(enc_blk+3, y_ptr + 8 + y_stride, linesize, s, 0);
+ }
+ enc_blk += 4;
+
+ /* initializing chrominance blocks */
c_offset = (((mb_y >> (s->sys->pix_fmt == PIX_FMT_YUV420P)) * s->picture.linesize[1] +
(mb_x >> ((s->sys->pix_fmt == PIX_FMT_YUV411P) ? 2 : 1))) << 3);
- do_edge_wrap = 0;
- qnos[mb_index] = 15; /* No quantization */
- ptr = dif + mb_index*80 + 4;
- for (j = 0; j < 6; j++) {
- 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 {
- /* 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;
- }
- }
-
- /* Everything is set up -- now just copy data -> DCT block */
- if (do_edge_wrap) { /* Edge wrap copy: 4x16 -> 8x8 */
+ for (j = 2; j; j--) {
+ uint8_t *c_ptr = s->picture.data[j] + c_offset;
+ linesize = s->picture.linesize[j];
+ y_stride = (mb_y == 134) ? 8 : (s->picture.linesize[j] << 3);
+ if (s->sys->pix_fmt == PIX_FMT_YUV411P && mb_x >= (704 / 8)) {
uint8_t* d;
- DCTELEM *b = block;
+ uint8_t* b = scratch;
for (i = 0; i < 8; i++) {
- d = data + 8 * linesize;
- b[0] = data[0]; b[1] = data[1]; b[2] = data[2]; b[3] = data[3];
- b[4] = d[0]; b[5] = d[1]; b[6] = d[2]; b[7] = d[3];
- data += linesize;
- b += 8;
+ d = c_ptr + (linesize << 3);
+ b[0] = c_ptr[0]; b[1] = c_ptr[1]; b[2] = c_ptr[2]; b[3] = c_ptr[3];
+ b[4] = d[0]; b[5] = d[1]; b[6] = d[2]; b[7] = d[3];
+ c_ptr += linesize;
+ b += 8;
}
- } else { /* Simple copy: 8x8 -> 8x8 */
- if (!dummy)
- s->get_pixels(block, data, linesize);
+ c_ptr = scratch;
+ linesize = 8;
}
- if (s->avctx->flags & CODEC_FLAG_INTERLACED_DCT)
- enc_blk->dct_mode = dv_guess_dct_mode(block);
- else
- enc_blk->dct_mode = 0;
- enc_blk->area_q[0] = enc_blk->area_q[1] = enc_blk->area_q[2] = enc_blk->area_q[3] = 0;
- enc_blk->partial_bit_count = 0;
- enc_blk->partial_bit_buffer = 0;
- enc_blk->cur_ac = 0;
-
- 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);
+ vs_bit_size += dv_init_enc_block( enc_blk++, c_ptr , linesize, s, 1);
+ if (s->sys->bpm == 8) {
+ vs_bit_size += dv_init_enc_block(enc_blk++, c_ptr + y_stride, linesize, s, 1);
}
-
- dv_set_class_number(block, enc_blk,
- 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, s->sys->block_sizes[j]/8);
- put_bits(pb, 9, (uint16_t)(((enc_blk->mb[0] >> 3) - 1024 + 2) >> 2));
- put_bits(pb, 1, enc_blk->dct_mode);
- put_bits(pb, 2, enc_blk->cno);
-
- vs_bit_size += enc_blk->bit_size[0] + enc_blk->bit_size[1] +
- enc_blk->bit_size[2] + enc_blk->bit_size[3];
- ++enc_blk;
- ++pb;
- ptr += s->sys->block_sizes[j]/8;
}
}
if (vs_total_ac_bits < vs_bit_size)
- dv_guess_qnos(&enc_blks[0], &qnos[0]);
+ dv_guess_qnos(&enc_blks[0], qnosp);
- for (i = 0; i < 5; i++) {
- dif[i*80 + 3] = qnos[i];
- }
+ /* DIF encoding process */
+ for (j=0; j<5*s->sys->bpm;) {
+ int start_mb = j;
+
+ dif[3] = *qnosp++;
+ dif += 4;
- /* First pass over individual cells only */
- for (j = 0; j < 5 * 6; j++)
- dv_encode_ac(&enc_blks[j], &pbs[j], &pbs[j+1]);
+ /* First pass over individual cells only */
+ for (i=0; i<s->sys->bpm; i++, j++) {
+ int sz = s->sys->block_sizes[i]>>3;
+
+ init_put_bits(&pbs[j], dif, sz);
+ put_sbits(&pbs[j], 9, ((enc_blks[j].mb[0] >> 3) - 1024 + 2) >> 2);
+ put_bits(&pbs[j], 1, enc_blks[j].dct_mode);
+ put_bits(&pbs[j], 2, enc_blks[j].cno);
+
+ dv_encode_ac(&enc_blks[j], &pbs[j], &pbs[j+1]);
+ dif += sz;
+ }
- /* Second pass over each MB space */
- for (j = 0; j < 5 * 6; j += 6) {
- pb = &pbs[j];
- for (i = 0; i < 6; i++) {
- if (enc_blks[i+j].partial_bit_count)
- pb = dv_encode_ac(&enc_blks[i+j], pb, &pbs[j+6]);
+ /* Second pass over each MB space */
+ pb = &pbs[start_mb];
+ for (i=0; i<s->sys->bpm; i++) {
+ if (enc_blks[start_mb+i].partial_bit_count)
+ pb = dv_encode_ac(&enc_blks[start_mb+i], pb, &pbs[start_mb+s->sys->bpm]);
}
}
/* Third and final pass over the whole video segment space */
pb = &pbs[0];
- for (j = 0; j < 5 * 6; j++) {
+ for (j=0; j<5*s->sys->bpm; j++) {
if (enc_blks[j].partial_bit_count)
- pb = dv_encode_ac(&enc_blks[j], pb, &pbs[6*5]);
+ pb = dv_encode_ac(&enc_blks[j], pb, &pbs[s->sys->bpm*5]);
if (enc_blks[j].partial_bit_count)
- av_log(NULL, AV_LOG_ERROR, "ac bitstream overflow\n");
+ av_log(avctx, AV_LOG_ERROR, "ac bitstream overflow\n");
}
- for (j = 0; j < 5 * 6; j++)
+ for (j=0; j<5*s->sys->bpm; j++) {
+ int pos;
+ int size = pbs[j].size_in_bits >> 3;
flush_put_bits(&pbs[j]);
+ pos = put_bits_count(&pbs[j]) >> 3;
+ if (pos > size) {
+ av_log(avctx, AV_LOG_ERROR, "bitstream written beyond buffer size\n");
+ return -1;
+ }
+ memset(pbs[j].buf + pos, 0xff, size - pos);
+ }
return 0;
}
-#ifdef CONFIG_DVVIDEO_DECODER
+#if CONFIG_DVVIDEO_DECODER
/* NOTE: exactly one frame must be given (120000 bytes for NTSC,
144000 bytes for PAL - or twice those for 50Mbps) */
static int dvvideo_decode_frame(AVCodecContext *avctx,
void *data, int *data_size,
- const uint8_t *buf, int buf_size)
+ AVPacket *avpkt)
{
+ const uint8_t *buf = avpkt->data;
+ int buf_size = avpkt->size;
DVVideoContext *s = avctx->priv_data;
+ const uint8_t* vsc_pack;
+ int apt, is16_9;
- s->sys = dv_frame_profile(buf);
- if (!s->sys || buf_size < s->sys->frame_size || dv_init_dynamic_tables(s->sys))
+ s->sys = avpriv_dv_frame_profile(s->sys, buf, buf_size);
+ if (!s->sys || buf_size < s->sys->frame_size || dv_init_dynamic_tables(s->sys)) {
+ av_log(avctx, AV_LOG_ERROR, "could not find dv frame profile\n");
return -1; /* NOTE: we only accept several full frames */
+ }
if (s->picture.data[0])
avctx->release_buffer(avctx, &s->picture);
s->picture.reference = 0;
s->picture.key_frame = 1;
- s->picture.pict_type = FF_I_TYPE;
+ s->picture.pict_type = AV_PICTURE_TYPE_I;
avctx->pix_fmt = s->sys->pix_fmt;
avctx->time_base = s->sys->time_base;
avcodec_set_dimensions(avctx, s->sys->width, s->sys->height);
*data_size = sizeof(AVFrame);
*(AVFrame*)data = s->picture;
+ /* Determine the codec's sample_aspect ratio from the packet */
+ vsc_pack = buf + 80*5 + 48 + 5;
+ if ( *vsc_pack == dv_video_control ) {
+ apt = buf[4] & 0x07;
+ is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 || (!apt && (vsc_pack[2] & 0x07) == 0x07)));
+ avctx->sample_aspect_ratio = s->sys->sar[is16_9];
+ }
+
return s->sys->frame_size;
}
#endif /* CONFIG_DVVIDEO_DECODER */
* 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 */
+ 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;
0xf; /* reserved -- always 1 */
buf[3] = (3 << 6) | /* reserved -- always 1 */
(c->sys->dsf << 5) | /* system: 60fields/50fields */
- stype; /* signal type video compression */
+ c->sys->video_stype; /* signal type video compression */
buf[4] = 0xff; /* VISC: 0xff -- no information */
break;
case dv_video_control:
return 5;
}
-#ifdef CONFIG_DVVIDEO_ENCODER
+#if CONFIG_DVVIDEO_ENCODER
static void dv_format_frame(DVVideoContext* c, uint8_t* buf)
{
int chan, i, j, k;
{
DVVideoContext *s = c->priv_data;
- s->sys = dv_codec_profile(c);
+ s->sys = avpriv_dv_codec_profile(c);
if (!s->sys || buf_size < s->sys->frame_size || dv_init_dynamic_tables(s->sys))
return -1;
c->pix_fmt = s->sys->pix_fmt;
s->picture = *((AVFrame *)data);
s->picture.key_frame = 1;
- s->picture.pict_type = FF_I_TYPE;
+ s->picture.pict_type = AV_PICTURE_TYPE_I;
s->buf = buf;
c->execute(c, dv_encode_video_segment, s->sys->work_chunks, NULL,
}
-#ifdef CONFIG_DVVIDEO_ENCODER
-AVCodec dvvideo_encoder = {
- "dvvideo",
- CODEC_TYPE_VIDEO,
- CODEC_ID_DVVIDEO,
- sizeof(DVVideoContext),
- dvvideo_init,
- dvvideo_encode_frame,
- .pix_fmts = (enum PixelFormat[]) {PIX_FMT_YUV411P, PIX_FMT_YUV422P, PIX_FMT_YUV420P, PIX_FMT_NONE},
+#if CONFIG_DVVIDEO_ENCODER
+AVCodec ff_dvvideo_encoder = {
+ .name = "dvvideo",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .id = CODEC_ID_DVVIDEO,
+ .priv_data_size = sizeof(DVVideoContext),
+ .init = dvvideo_init_encoder,
+ .encode = dvvideo_encode_frame,
+ .capabilities = CODEC_CAP_SLICE_THREADS,
+ .pix_fmts = (const enum PixelFormat[]) {PIX_FMT_YUV411P, PIX_FMT_YUV422P, PIX_FMT_YUV420P, PIX_FMT_NONE},
.long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
};
#endif // CONFIG_DVVIDEO_ENCODER
-#ifdef CONFIG_DVVIDEO_DECODER
-AVCodec dvvideo_decoder = {
- "dvvideo",
- CODEC_TYPE_VIDEO,
- CODEC_ID_DVVIDEO,
- sizeof(DVVideoContext),
- dvvideo_init,
- NULL,
- dvvideo_close,
- dvvideo_decode_frame,
- CODEC_CAP_DR1,
- NULL,
+#if CONFIG_DVVIDEO_DECODER
+AVCodec ff_dvvideo_decoder = {
+ .name = "dvvideo",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .id = CODEC_ID_DVVIDEO,
+ .priv_data_size = sizeof(DVVideoContext),
+ .init = dvvideo_init,
+ .close = dvvideo_close,
+ .decode = dvvideo_decode_frame,
+ .capabilities = CODEC_CAP_DR1 | CODEC_CAP_SLICE_THREADS,
+ .max_lowres = 3,
.long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
};
#endif