You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

869 lines
27 KiB

/*
* The simplest mpeg encoder (well, it was the simplest!)
* Copyright (c) 2000,2001 Fabrice Bellard
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
* 4MV & hq & B-frame encoding stuff by Michael Niedermayer <michaelni@gmx.at>
*
* 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 2.1 of the License, or (at your option) any later version.
*
* 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 FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* The simplest mpeg encoder (well, it was the simplest!).
*/
#include "libavutil/attributes.h"
#include "libavutil/avassert.h"
#include "libavutil/imgutils.h"
#include "libavutil/internal.h"
#include "libavutil/mem.h"
#include "avcodec.h"
#include "blockdsp.h"
#include "idctdsp.h"
#include "mathops.h"
#include "mpeg_er.h"
#include "mpegutils.h"
#include "mpegvideo.h"
#include "mpegvideodata.h"
#include "refstruct.h"
static void dct_unquantize_mpeg1_intra_c(MpegEncContext *s,
int16_t *block, int n, int qscale)
{
int i, level, nCoeffs;
const uint16_t *quant_matrix;
nCoeffs= s->block_last_index[n];
block[0] *= n < 4 ? s->y_dc_scale : s->c_dc_scale;
/* XXX: only MPEG-1 */
quant_matrix = s->intra_matrix;
for(i=1;i<=nCoeffs;i++) {
int j= s->intra_scantable.permutated[i];
level = block[j];
if (level) {
if (level < 0) {
level = -level;
level = (int)(level * qscale * quant_matrix[j]) >> 3;
level = (level - 1) | 1;
level = -level;
} else {
level = (int)(level * qscale * quant_matrix[j]) >> 3;
level = (level - 1) | 1;
}
block[j] = level;
}
}
}
static void dct_unquantize_mpeg1_inter_c(MpegEncContext *s,
int16_t *block, int n, int qscale)
{
int i, level, nCoeffs;
const uint16_t *quant_matrix;
nCoeffs= s->block_last_index[n];
quant_matrix = s->inter_matrix;
for(i=0; i<=nCoeffs; i++) {
int j= s->intra_scantable.permutated[i];
level = block[j];
if (level) {
if (level < 0) {
level = -level;
level = (((level << 1) + 1) * qscale *
((int) (quant_matrix[j]))) >> 4;
level = (level - 1) | 1;
level = -level;
} else {
level = (((level << 1) + 1) * qscale *
((int) (quant_matrix[j]))) >> 4;
level = (level - 1) | 1;
}
block[j] = level;
}
}
}
static void dct_unquantize_mpeg2_intra_c(MpegEncContext *s,
int16_t *block, int n, int qscale)
{
int i, level, nCoeffs;
const uint16_t *quant_matrix;
if (s->q_scale_type) qscale = ff_mpeg2_non_linear_qscale[qscale];
else qscale <<= 1;
if(s->alternate_scan) nCoeffs= 63;
else nCoeffs= s->block_last_index[n];
block[0] *= n < 4 ? s->y_dc_scale : s->c_dc_scale;
quant_matrix = s->intra_matrix;
for(i=1;i<=nCoeffs;i++) {
int j= s->intra_scantable.permutated[i];
level = block[j];
if (level) {
if (level < 0) {
level = -level;
level = (int)(level * qscale * quant_matrix[j]) >> 4;
level = -level;
} else {
level = (int)(level * qscale * quant_matrix[j]) >> 4;
}
block[j] = level;
}
}
}
static void dct_unquantize_mpeg2_intra_bitexact(MpegEncContext *s,
int16_t *block, int n, int qscale)
{
int i, level, nCoeffs;
const uint16_t *quant_matrix;
int sum=-1;
if (s->q_scale_type) qscale = ff_mpeg2_non_linear_qscale[qscale];
else qscale <<= 1;
if(s->alternate_scan) nCoeffs= 63;
else nCoeffs= s->block_last_index[n];
block[0] *= n < 4 ? s->y_dc_scale : s->c_dc_scale;
sum += block[0];
quant_matrix = s->intra_matrix;
for(i=1;i<=nCoeffs;i++) {
int j= s->intra_scantable.permutated[i];
level = block[j];
if (level) {
if (level < 0) {
level = -level;
level = (int)(level * qscale * quant_matrix[j]) >> 4;
level = -level;
} else {
level = (int)(level * qscale * quant_matrix[j]) >> 4;
}
block[j] = level;
sum+=level;
}
}
block[63]^=sum&1;
}
static void dct_unquantize_mpeg2_inter_c(MpegEncContext *s,
int16_t *block, int n, int qscale)
{
int i, level, nCoeffs;
const uint16_t *quant_matrix;
int sum=-1;
if (s->q_scale_type) qscale = ff_mpeg2_non_linear_qscale[qscale];
else qscale <<= 1;
if(s->alternate_scan) nCoeffs= 63;
else nCoeffs= s->block_last_index[n];
quant_matrix = s->inter_matrix;
for(i=0; i<=nCoeffs; i++) {
int j= s->intra_scantable.permutated[i];
level = block[j];
if (level) {
if (level < 0) {
level = -level;
level = (((level << 1) + 1) * qscale *
((int) (quant_matrix[j]))) >> 5;
level = -level;
} else {
level = (((level << 1) + 1) * qscale *
((int) (quant_matrix[j]))) >> 5;
}
block[j] = level;
sum+=level;
}
}
block[63]^=sum&1;
}
static void dct_unquantize_h263_intra_c(MpegEncContext *s,
int16_t *block, int n, int qscale)
{
int i, level, qmul, qadd;
int nCoeffs;
av_assert2(s->block_last_index[n]>=0 || s->h263_aic);
qmul = qscale << 1;
if (!s->h263_aic) {
block[0] *= n < 4 ? s->y_dc_scale : s->c_dc_scale;
qadd = (qscale - 1) | 1;
}else{
qadd = 0;
}
if(s->ac_pred)
nCoeffs=63;
else
nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];
for(i=1; i<=nCoeffs; i++) {
level = block[i];
if (level) {
if (level < 0) {
level = level * qmul - qadd;
} else {
level = level * qmul + qadd;
}
block[i] = level;
}
}
}
static void dct_unquantize_h263_inter_c(MpegEncContext *s,
int16_t *block, int n, int qscale)
{
int i, level, qmul, qadd;
int nCoeffs;
av_assert2(s->block_last_index[n]>=0);
qadd = (qscale - 1) | 1;
qmul = qscale << 1;
nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
for(i=0; i<=nCoeffs; i++) {
level = block[i];
if (level) {
if (level < 0) {
level = level * qmul - qadd;
} else {
level = level * qmul + qadd;
}
block[i] = level;
}
}
}
static void gray16(uint8_t *dst, const uint8_t *src, ptrdiff_t linesize, int h)
{
while(h--)
memset(dst + h*linesize, 128, 16);
}
static void gray8(uint8_t *dst, const uint8_t *src, ptrdiff_t linesize, int h)
{
while(h--)
memset(dst + h*linesize, 128, 8);
}
/* init common dct for both encoder and decoder */
static av_cold void dsp_init(MpegEncContext *s)
{
ff_blockdsp_init(&s->bdsp);
ff_hpeldsp_init(&s->hdsp, s->avctx->flags);
ff_videodsp_init(&s->vdsp, s->avctx->bits_per_raw_sample);
if (s->avctx->debug & FF_DEBUG_NOMC) {
int i;
for (i=0; i<4; i++) {
s->hdsp.avg_pixels_tab[0][i] = gray16;
s->hdsp.put_pixels_tab[0][i] = gray16;
s->hdsp.put_no_rnd_pixels_tab[0][i] = gray16;
s->hdsp.avg_pixels_tab[1][i] = gray8;
s->hdsp.put_pixels_tab[1][i] = gray8;
s->hdsp.put_no_rnd_pixels_tab[1][i] = gray8;
}
}
}
av_cold void ff_init_scantable(const uint8_t *permutation, ScanTable *st,
const uint8_t *src_scantable)
{
st->scantable = src_scantable;
for (int i = 0, end = -1; i < 64; i++) {
int j = src_scantable[i];
st->permutated[i] = permutation[j];
if (permutation[j] > end)
end = permutation[j];
st->raster_end[i] = end;
}
}
av_cold void ff_mpv_idct_init(MpegEncContext *s)
{
if (s->codec_id == AV_CODEC_ID_MPEG4)
s->idsp.mpeg4_studio_profile = s->studio_profile;
ff_idctdsp_init(&s->idsp, s->avctx);
/* load & permutate scantables
* note: only wmv uses different ones
*/
if (s->alternate_scan) {
ff_init_scantable(s->idsp.idct_permutation, &s->inter_scantable, ff_alternate_vertical_scan);
ff_init_scantable(s->idsp.idct_permutation, &s->intra_scantable, ff_alternate_vertical_scan);
} else {
ff_init_scantable(s->idsp.idct_permutation, &s->inter_scantable, ff_zigzag_direct);
ff_init_scantable(s->idsp.idct_permutation, &s->intra_scantable, ff_zigzag_direct);
}
ff_permute_scantable(s->permutated_intra_h_scantable, ff_alternate_horizontal_scan,
s->idsp.idct_permutation);
ff_permute_scantable(s->permutated_intra_v_scantable, ff_alternate_vertical_scan,
s->idsp.idct_permutation);
s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_c;
s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_c;
s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_c;
s->dct_unquantize_mpeg1_inter = dct_unquantize_mpeg1_inter_c;
s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_c;
if (s->avctx->flags & AV_CODEC_FLAG_BITEXACT)
s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_bitexact;
s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_c;
#if HAVE_INTRINSICS_NEON
ff_mpv_common_init_neon(s);
#endif
#if ARCH_ARM
ff_mpv_common_init_arm(s);
#elif ARCH_PPC
ff_mpv_common_init_ppc(s);
#elif ARCH_X86
ff_mpv_common_init_x86(s);
#elif ARCH_MIPS
ff_mpv_common_init_mips(s);
#endif
}
static int init_duplicate_context(MpegEncContext *s)
{
if (s->encoding) {
s->me.map = av_mallocz(2 * ME_MAP_SIZE * sizeof(*s->me.map));
if (!s->me.map)
return AVERROR(ENOMEM);
s->me.score_map = s->me.map + ME_MAP_SIZE;
if (s->noise_reduction) {
if (!FF_ALLOCZ_TYPED_ARRAY(s->dct_error_sum, 2))
return AVERROR(ENOMEM);
}
}
if (!FF_ALLOCZ_TYPED_ARRAY(s->blocks, 1 + s->encoding))
return AVERROR(ENOMEM);
s->block = s->blocks[0];
if (s->out_format == FMT_H263) {
int mb_height = s->msmpeg4_version == MSMP4_VC1 ?
FFALIGN(s->mb_height, 2) : s->mb_height;
int y_size = s->b8_stride * (2 * mb_height + 1);
int c_size = s->mb_stride * (mb_height + 1);
int yc_size = y_size + 2 * c_size;
/* ac values */
if (!FF_ALLOCZ_TYPED_ARRAY(s->ac_val_base, yc_size))
return AVERROR(ENOMEM);
s->ac_val[0] = s->ac_val_base + s->b8_stride + 1;
s->ac_val[1] = s->ac_val_base + y_size + s->mb_stride + 1;
s->ac_val[2] = s->ac_val[1] + c_size;
}
return 0;
}
int ff_mpv_init_duplicate_contexts(MpegEncContext *s)
{
int nb_slices = s->slice_context_count, ret;
/* We initialize the copies before the original so that
* fields allocated in init_duplicate_context are NULL after
* copying. This prevents double-frees upon allocation error. */
for (int i = 1; i < nb_slices; i++) {
s->thread_context[i] = av_memdup(s, sizeof(MpegEncContext));
if (!s->thread_context[i])
return AVERROR(ENOMEM);
if ((ret = init_duplicate_context(s->thread_context[i])) < 0)
return ret;
s->thread_context[i]->start_mb_y =
(s->mb_height * (i ) + nb_slices / 2) / nb_slices;
s->thread_context[i]->end_mb_y =
(s->mb_height * (i + 1) + nb_slices / 2) / nb_slices;
}
s->start_mb_y = 0;
s->end_mb_y = nb_slices > 1 ? (s->mb_height + nb_slices / 2) / nb_slices
: s->mb_height;
return init_duplicate_context(s);
}
static void free_duplicate_context(MpegEncContext *s)
{
if (!s)
return;
av_freep(&s->sc.edge_emu_buffer);
av_freep(&s->sc.scratchpad_buf);
s->me.temp = s->me.scratchpad =
s->sc.obmc_scratchpad = NULL;
s->sc.linesize = 0;
av_freep(&s->dct_error_sum);
av_freep(&s->me.map);
s->me.score_map = NULL;
av_freep(&s->blocks);
av_freep(&s->ac_val_base);
s->block = NULL;
}
static void free_duplicate_contexts(MpegEncContext *s)
{
for (int i = 1; i < s->slice_context_count; i++) {
free_duplicate_context(s->thread_context[i]);
av_freep(&s->thread_context[i]);
}
free_duplicate_context(s);
}
static void backup_duplicate_context(MpegEncContext *bak, MpegEncContext *src)
{
#define COPY(a) bak->a = src->a
COPY(sc);
COPY(me.map);
COPY(me.score_map);
COPY(blocks);
COPY(block);
COPY(start_mb_y);
COPY(end_mb_y);
COPY(me.map_generation);
COPY(dct_error_sum);
COPY(dct_count[0]);
COPY(dct_count[1]);
COPY(ac_val_base);
COPY(ac_val[0]);
COPY(ac_val[1]);
COPY(ac_val[2]);
#undef COPY
}
int ff_update_duplicate_context(MpegEncContext *dst, const MpegEncContext *src)
{
MpegEncContext bak;
int ret;
// FIXME copy only needed parts
backup_duplicate_context(&bak, dst);
memcpy(dst, src, sizeof(MpegEncContext));
backup_duplicate_context(dst, &bak);
ret = ff_mpv_framesize_alloc(dst->avctx, &dst->sc, dst->linesize);
if (ret < 0) {
av_log(dst->avctx, AV_LOG_ERROR, "failed to allocate context "
"scratch buffers.\n");
return ret;
}
return 0;
}
/**
* Set the given MpegEncContext to common defaults
* (same for encoding and decoding).
* The changed fields will not depend upon the
* prior state of the MpegEncContext.
*/
void ff_mpv_common_defaults(MpegEncContext *s)
{
s->y_dc_scale_table =
s->c_dc_scale_table = ff_mpeg1_dc_scale_table;
s->chroma_qscale_table = ff_default_chroma_qscale_table;
s->progressive_frame = 1;
s->progressive_sequence = 1;
s->picture_structure = PICT_FRAME;
s->picture_number = 0;
s->f_code = 1;
s->b_code = 1;
s->slice_context_count = 1;
}
static void free_buffer_pools(BufferPoolContext *pools)
{
ff_refstruct_pool_uninit(&pools->mbskip_table_pool);
ff_refstruct_pool_uninit(&pools->qscale_table_pool);
ff_refstruct_pool_uninit(&pools->mb_type_pool);
ff_refstruct_pool_uninit(&pools->motion_val_pool);
ff_refstruct_pool_uninit(&pools->ref_index_pool);
pools->alloc_mb_height = pools->alloc_mb_width = pools->alloc_mb_stride = 0;
}
int ff_mpv_init_context_frame(MpegEncContext *s)
{
BufferPoolContext *const pools = &s->buffer_pools;
int y_size, c_size, yc_size, i, mb_array_size, mv_table_size, x, y;
int mb_height;
if (s->codec_id == AV_CODEC_ID_MPEG2VIDEO && !s->progressive_sequence)
s->mb_height = (s->height + 31) / 32 * 2;
else
s->mb_height = (s->height + 15) / 16;
/* VC-1 can change from being progressive to interlaced on a per-frame
* basis. We therefore allocate certain buffers so big that they work
* in both instances. */
mb_height = s->msmpeg4_version == MSMP4_VC1 ?
FFALIGN(s->mb_height, 2) : s->mb_height;
s->mb_width = (s->width + 15) / 16;
s->mb_stride = s->mb_width + 1;
s->b8_stride = s->mb_width * 2 + 1;
mb_array_size = mb_height * s->mb_stride;
mv_table_size = (mb_height + 2) * s->mb_stride + 1;
/* set default edge pos, will be overridden
* in decode_header if needed */
s->h_edge_pos = s->mb_width * 16;
s->v_edge_pos = s->mb_height * 16;
s->mb_num = s->mb_width * s->mb_height;
s->block_wrap[0] =
s->block_wrap[1] =
s->block_wrap[2] =
s->block_wrap[3] = s->b8_stride;
s->block_wrap[4] =
s->block_wrap[5] = s->mb_stride;
y_size = s->b8_stride * (2 * mb_height + 1);
c_size = s->mb_stride * (mb_height + 1);
yc_size = y_size + 2 * c_size;
if (!FF_ALLOCZ_TYPED_ARRAY(s->mb_index2xy, s->mb_num + 1))
return AVERROR(ENOMEM);
for (y = 0; y < s->mb_height; y++)
for (x = 0; x < s->mb_width; x++)
s->mb_index2xy[x + y * s->mb_width] = x + y * s->mb_stride;
s->mb_index2xy[s->mb_height * s->mb_width] = (s->mb_height - 1) * s->mb_stride + s->mb_width; // FIXME really needed?
#define ALLOC_POOL(name, size, flags) do { \
pools->name ##_pool = ff_refstruct_pool_alloc((size), (flags)); \
if (!pools->name ##_pool) \
return AVERROR(ENOMEM); \
} while (0)
if (s->codec_id == AV_CODEC_ID_MPEG4 ||
(s->avctx->flags & AV_CODEC_FLAG_INTERLACED_ME)) {
/* interlaced direct mode decoding tables */
int16_t (*tmp)[2] = av_calloc(mv_table_size, 4 * sizeof(*tmp));
if (!tmp)
return AVERROR(ENOMEM);
s->p_field_mv_table_base = tmp;
tmp += s->mb_stride + 1;
for (int i = 0; i < 2; i++) {
for (int j = 0; j < 2; j++) {
s->p_field_mv_table[i][j] = tmp;
tmp += mv_table_size;
}
}
if (s->codec_id == AV_CODEC_ID_MPEG4) {
ALLOC_POOL(mbskip_table, mb_array_size + 2,
!s->encoding ? FF_REFSTRUCT_POOL_FLAG_ZERO_EVERY_TIME : 0);
if (!s->encoding) {
/* cbp, pred_dir */
if (!(s->cbp_table = av_mallocz(mb_array_size)) ||
!(s->pred_dir_table = av_mallocz(mb_array_size)))
return AVERROR(ENOMEM);
}
}
}
if (s->msmpeg4_version >= MSMP4_V3) {
s->coded_block_base = av_mallocz(y_size);
if (!s->coded_block_base)
return AVERROR(ENOMEM);
s->coded_block = s->coded_block_base + s->b8_stride + 1;
}
if (s->h263_pred || s->h263_plus || !s->encoding) {
/* dc values */
// MN: we need these for error resilience of intra-frames
if (!FF_ALLOCZ_TYPED_ARRAY(s->dc_val_base, yc_size))
return AVERROR(ENOMEM);
s->dc_val[0] = s->dc_val_base + s->b8_stride + 1;
s->dc_val[1] = s->dc_val_base + y_size + s->mb_stride + 1;
s->dc_val[2] = s->dc_val[1] + c_size;
for (i = 0; i < yc_size; i++)
s->dc_val_base[i] = 1024;
}
// Note the + 1 is for a quicker MPEG-4 slice_end detection
if (!(s->mbskip_table = av_mallocz(mb_array_size + 2)) ||
/* which mb is an intra block, init macroblock skip table */
!(s->mbintra_table = av_malloc(mb_array_size)))
return AVERROR(ENOMEM);
memset(s->mbintra_table, 1, mb_array_size);
ALLOC_POOL(qscale_table, mv_table_size, 0);
ALLOC_POOL(mb_type, mv_table_size * sizeof(uint32_t), 0);
if (s->out_format == FMT_H263 || s->encoding ||
(s->avctx->export_side_data & AV_CODEC_EXPORT_DATA_MVS)) {
const int b8_array_size = s->b8_stride * mb_height * 2;
int mv_size = 2 * (b8_array_size + 4) * sizeof(int16_t);
int ref_index_size = 4 * mb_array_size;
/* FIXME: The output of H.263 with OBMC depends upon
* the earlier content of the buffer; therefore we set
* the flags to always reset returned buffers here. */
ALLOC_POOL(motion_val, mv_size, FF_REFSTRUCT_POOL_FLAG_ZERO_EVERY_TIME);
ALLOC_POOL(ref_index, ref_index_size, 0);
}
#undef ALLOC_POOL
pools->alloc_mb_width = s->mb_width;
pools->alloc_mb_height = mb_height;
pools->alloc_mb_stride = s->mb_stride;
return !CONFIG_MPEGVIDEODEC || s->encoding ? 0 : ff_mpeg_er_init(s);
}
static void clear_context(MpegEncContext *s)
{
memset(&s->buffer_pools, 0, sizeof(s->buffer_pools));
memset(&s->next_pic, 0, sizeof(s->next_pic));
memset(&s->last_pic, 0, sizeof(s->last_pic));
memset(&s->cur_pic, 0, sizeof(s->cur_pic));
memset(s->thread_context, 0, sizeof(s->thread_context));
s->me.map = NULL;
s->me.score_map = NULL;
s->dct_error_sum = NULL;
s->block = NULL;
s->blocks = NULL;
s->ac_val_base = NULL;
s->ac_val[0] =
s->ac_val[1] =
s->ac_val[2] =NULL;
s->me.scratchpad = NULL;
s->me.temp = NULL;
memset(&s->sc, 0, sizeof(s->sc));
s->bitstream_buffer = NULL;
s->allocated_bitstream_buffer_size = 0;
s->p_field_mv_table_base = NULL;
for (int i = 0; i < 2; i++)
for (int j = 0; j < 2; j++)
s->p_field_mv_table[i][j] = NULL;
s->dc_val_base = NULL;
s->coded_block_base = NULL;
s->mbintra_table = NULL;
s->cbp_table = NULL;
s->pred_dir_table = NULL;
s->mbskip_table = NULL;
s->er.error_status_table = NULL;
s->er.er_temp_buffer = NULL;
s->mb_index2xy = NULL;
}
/**
* init common structure for both encoder and decoder.
* this assumes that some variables like width/height are already set
*/
av_cold int ff_mpv_common_init(MpegEncContext *s)
{
int nb_slices = (HAVE_THREADS &&
s->avctx->active_thread_type & FF_THREAD_SLICE) ?
s->avctx->thread_count : 1;
avcodec/mpegpicture: Make MPVPicture refcounted Up until now, an initialized MpegEncContext had an array of MPVPictures (way more than were ever needed) and the MPVPicture* contained in the MPVWorkPictures as well as the input_picture and reordered_input_picture arrays (for the encoder) pointed into this array. Several of the pointers could point to the same slot and because there was no reference counting involved, one had to check for aliasing before unreferencing. Furthermore, given that these pointers were not ownership pointers the pointers were often simply reset without unreferencing the slot (happened e.g. for the RV30 and RV40 decoders) or there were moved without resetting the src pointer (happened for the encoders where the entries in the input_picture and reordered_input_picture arrays were not reset). Instead actually releasing these pictures was performed by looping over the whole array and checking which one of the entries needed to be kept. Given that the array had way too many slots (36), this meant that more than 30 MPVPictures have been unnecessarily unreferenced in every ff_mpv_frame_start(); something similar happened for the encoder. This commit changes this by making the MPVPictures refcounted via the RefStruct API. The MPVPictures itself are part of a pool so that this does not entail constant allocations; instead, the amount of allocations actually goes down, because the earlier code used such a large array of MPVPictures (36 entries) and allocated an AVFrame for every one of these on every ff_mpv_common_init(). In fact, the pool is only freed when closing the codec, so that reinitializations don't lead to new allocations (this avoids having to sync the pool in update_thread_context). Making MPVPictures refcounted also has another key benefit: It makes it possible to directly share them across threads (when using frame-threaded decoding), eliminating ugly code with underlying av_frame_ref()'s; sharing these pictures can't fail any more. The pool is allocated in ff_mpv_decode_init() for decoders, which therefore can fail now. This and the fact that the pool is not unreferenced in ff_mpv_common_end() also necessitated to mark several mpegvideo-decoders with the FF_CODEC_CAP_INIT_CLEANUP flag. *: This also means that there is no good reason any more for ff_mpv_common_frame_size_change() to exist. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
int ret;
clear_context(s);
if (s->encoding && s->avctx->slices)
nb_slices = s->avctx->slices;
if (s->avctx->pix_fmt == AV_PIX_FMT_NONE) {
av_log(s->avctx, AV_LOG_ERROR,
"decoding to AV_PIX_FMT_NONE is not supported.\n");
return AVERROR(EINVAL);
}
if ((s->width || s->height) &&
av_image_check_size(s->width, s->height, 0, s->avctx))
return AVERROR(EINVAL);
dsp_init(s);
/* set chroma shifts */
ret = av_pix_fmt_get_chroma_sub_sample(s->avctx->pix_fmt,
&s->chroma_x_shift,
&s->chroma_y_shift);
if (ret)
return ret;
if ((ret = ff_mpv_init_context_frame(s)))
Revert "avcodec: add FF_CODEC_CAP_INIT_CLEANUP for all codecs which use ff_mpv_common_init()" This mostly reverts commit 4b2863ff01b1fe93d9a518523c9098d17a9d8c6f. Said commit removed the freeing code from ff_mpv_common_init(), ff_mpv_common_frame_size_change() and ff_mpeg_framesize_alloc() and instead added the FF_CODEC_CAP_INIT_CLEANUP to several codecs that use ff_mpv_common_init(). This introduced several bugs: a) Several decoders using ff_mpv_common_init() in their init function were forgotten: This affected FLV, Intel H.263, RealVideo 3.0 and V4.0 as well as VC-1/WMV3. b) ff_mpv_common_init() is not only called from the init function of codecs, it is also called from AVCodec.decode functions. If an error happens after an allocation has succeeded, it can lead to memleaks; furthermore, it is now possible for the MpegEncContext to be marked as initialized even when ff_mpv_common_init() returns an error and this can lead to segfaults because decoders that call ff_mpv_common_init() when decoding a frame can mistakenly think that the MpegEncContext has been properly initialized. This can e.g. happen with H.261 or MPEG-4. c) Removing code for freeing from ff_mpeg_framesize_alloc() (which can't be called from any init function) can lead to segfaults because the check for whether it needs to allocate consists of checking whether the first of the buffers allocated there has been allocated. This part has already been fixed in 76cea1d2ce3f23e8131c8664086a1daf873ed694. d) ff_mpv_common_frame_size_change() can also not be reached from any AVCodec.init function; yet the changes can e.g. lead to segfaults with decoders using ff_h263_decode_frame() upon allocation failure, because the MpegEncContext will upon return be flagged as both initialized and not in need of reinitialization (granted, the fact that ff_h263_decode_frame() clears context_reinit before the context has been reinited is a bug in itself). With the earlier version, the context would be cleaned upon failure and it would be attempted to initialize the context again in the next call to ff_h263_decode_frame(). While a) could be fixed by adding the missing FF_CODEC_CAP_INIT_CLEANUP, keeping the current approach would entail adding cleanup code to several other places because of b). Therefore ff_mpv_common_init() is again made to clean up after itself; the changes to the wmv2 decoder and the SVQ1 encoder have not been reverted: The former fixed a memleak, the latter allowed to remove cleanup code. Fixes: double free Fixes: ff_free_picture_tables.mp4 Fixes: ff_mpeg_update_thread_context.mp4 Fixes: decode_colskip.mp4 Fixes: memset.mp4 Reviewed-by: Michael Niedermayer <michael@niedermayer.cc> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
goto fail;
if (nb_slices > MAX_THREADS || (nb_slices > s->mb_height && s->mb_height)) {
int max_slices;
if (s->mb_height)
max_slices = FFMIN(MAX_THREADS, s->mb_height);
else
max_slices = MAX_THREADS;
av_log(s->avctx, AV_LOG_WARNING, "too many threads/slices (%d),"
" reducing to %d\n", nb_slices, max_slices);
nb_slices = max_slices;
}
s->context_initialized = 1;
memset(s->thread_context, 0, sizeof(s->thread_context));
s->thread_context[0] = s;
s->slice_context_count = nb_slices;
// if (s->width && s->height) {
ret = ff_mpv_init_duplicate_contexts(s);
if (ret < 0)
goto fail;
// }
return 0;
Revert "avcodec: add FF_CODEC_CAP_INIT_CLEANUP for all codecs which use ff_mpv_common_init()" This mostly reverts commit 4b2863ff01b1fe93d9a518523c9098d17a9d8c6f. Said commit removed the freeing code from ff_mpv_common_init(), ff_mpv_common_frame_size_change() and ff_mpeg_framesize_alloc() and instead added the FF_CODEC_CAP_INIT_CLEANUP to several codecs that use ff_mpv_common_init(). This introduced several bugs: a) Several decoders using ff_mpv_common_init() in their init function were forgotten: This affected FLV, Intel H.263, RealVideo 3.0 and V4.0 as well as VC-1/WMV3. b) ff_mpv_common_init() is not only called from the init function of codecs, it is also called from AVCodec.decode functions. If an error happens after an allocation has succeeded, it can lead to memleaks; furthermore, it is now possible for the MpegEncContext to be marked as initialized even when ff_mpv_common_init() returns an error and this can lead to segfaults because decoders that call ff_mpv_common_init() when decoding a frame can mistakenly think that the MpegEncContext has been properly initialized. This can e.g. happen with H.261 or MPEG-4. c) Removing code for freeing from ff_mpeg_framesize_alloc() (which can't be called from any init function) can lead to segfaults because the check for whether it needs to allocate consists of checking whether the first of the buffers allocated there has been allocated. This part has already been fixed in 76cea1d2ce3f23e8131c8664086a1daf873ed694. d) ff_mpv_common_frame_size_change() can also not be reached from any AVCodec.init function; yet the changes can e.g. lead to segfaults with decoders using ff_h263_decode_frame() upon allocation failure, because the MpegEncContext will upon return be flagged as both initialized and not in need of reinitialization (granted, the fact that ff_h263_decode_frame() clears context_reinit before the context has been reinited is a bug in itself). With the earlier version, the context would be cleaned upon failure and it would be attempted to initialize the context again in the next call to ff_h263_decode_frame(). While a) could be fixed by adding the missing FF_CODEC_CAP_INIT_CLEANUP, keeping the current approach would entail adding cleanup code to several other places because of b). Therefore ff_mpv_common_init() is again made to clean up after itself; the changes to the wmv2 decoder and the SVQ1 encoder have not been reverted: The former fixed a memleak, the latter allowed to remove cleanup code. Fixes: double free Fixes: ff_free_picture_tables.mp4 Fixes: ff_mpeg_update_thread_context.mp4 Fixes: decode_colskip.mp4 Fixes: memset.mp4 Reviewed-by: Michael Niedermayer <michael@niedermayer.cc> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
4 years ago
fail:
ff_mpv_common_end(s);
return ret;
}
void ff_mpv_free_context_frame(MpegEncContext *s)
{
free_duplicate_contexts(s);
free_buffer_pools(&s->buffer_pools);
av_freep(&s->p_field_mv_table_base);
for (int i = 0; i < 2; i++)
for (int j = 0; j < 2; j++)
s->p_field_mv_table[i][j] = NULL;
av_freep(&s->dc_val_base);
av_freep(&s->coded_block_base);
av_freep(&s->mbintra_table);
av_freep(&s->cbp_table);
av_freep(&s->pred_dir_table);
av_freep(&s->mbskip_table);
av_freep(&s->er.error_status_table);
av_freep(&s->er.er_temp_buffer);
av_freep(&s->mb_index2xy);
s->linesize = s->uvlinesize = 0;
}
void ff_mpv_common_end(MpegEncContext *s)
{
ff_mpv_free_context_frame(s);
if (s->slice_context_count > 1)
s->slice_context_count = 1;
av_freep(&s->bitstream_buffer);
s->allocated_bitstream_buffer_size = 0;
avcodec/mpegpicture: Make MPVPicture refcounted Up until now, an initialized MpegEncContext had an array of MPVPictures (way more than were ever needed) and the MPVPicture* contained in the MPVWorkPictures as well as the input_picture and reordered_input_picture arrays (for the encoder) pointed into this array. Several of the pointers could point to the same slot and because there was no reference counting involved, one had to check for aliasing before unreferencing. Furthermore, given that these pointers were not ownership pointers the pointers were often simply reset without unreferencing the slot (happened e.g. for the RV30 and RV40 decoders) or there were moved without resetting the src pointer (happened for the encoders where the entries in the input_picture and reordered_input_picture arrays were not reset). Instead actually releasing these pictures was performed by looping over the whole array and checking which one of the entries needed to be kept. Given that the array had way too many slots (36), this meant that more than 30 MPVPictures have been unnecessarily unreferenced in every ff_mpv_frame_start(); something similar happened for the encoder. This commit changes this by making the MPVPictures refcounted via the RefStruct API. The MPVPictures itself are part of a pool so that this does not entail constant allocations; instead, the amount of allocations actually goes down, because the earlier code used such a large array of MPVPictures (36 entries) and allocated an AVFrame for every one of these on every ff_mpv_common_init(). In fact, the pool is only freed when closing the codec, so that reinitializations don't lead to new allocations (this avoids having to sync the pool in update_thread_context). Making MPVPictures refcounted also has another key benefit: It makes it possible to directly share them across threads (when using frame-threaded decoding), eliminating ugly code with underlying av_frame_ref()'s; sharing these pictures can't fail any more. The pool is allocated in ff_mpv_decode_init() for decoders, which therefore can fail now. This and the fact that the pool is not unreferenced in ff_mpv_common_end() also necessitated to mark several mpegvideo-decoders with the FF_CODEC_CAP_INIT_CLEANUP flag. *: This also means that there is no good reason any more for ff_mpv_common_frame_size_change() to exist. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
1 year ago
ff_mpv_unref_picture(&s->last_pic);
ff_mpv_unref_picture(&s->cur_pic);
ff_mpv_unref_picture(&s->next_pic);
s->context_initialized = 0;
s->context_reinit = 0;
s->linesize = s->uvlinesize = 0;
}
/**
* Clean dc, ac for the current non-intra MB.
*/
void ff_clean_intra_table_entries(MpegEncContext *s)
{
int wrap = s->b8_stride;
int xy = s->block_index[0];
s->dc_val[0][xy ] =
s->dc_val[0][xy + 1 ] =
s->dc_val[0][xy + wrap] =
s->dc_val[0][xy + 1 + wrap] = 1024;
/* ac pred */
memset(s->ac_val[0][xy ], 0, 32 * sizeof(int16_t));
memset(s->ac_val[0][xy + wrap], 0, 32 * sizeof(int16_t));
/* chroma */
wrap = s->mb_stride;
xy = s->mb_x + s->mb_y * wrap;
s->dc_val[1][xy] =
s->dc_val[2][xy] = 1024;
/* ac pred */
memset(s->ac_val[1][xy], 0, 16 * sizeof(int16_t));
memset(s->ac_val[2][xy], 0, 16 * sizeof(int16_t));
s->mbintra_table[xy]= 0;
}
void ff_init_block_index(MpegEncContext *s){ //FIXME maybe rename
const int linesize = s->cur_pic.linesize[0]; //not s->linesize as this would be wrong for field pics
const int uvlinesize = s->cur_pic.linesize[1];
const int width_of_mb = (4 + (s->avctx->bits_per_raw_sample > 8)) - s->avctx->lowres;
const int height_of_mb = 4 - s->avctx->lowres;
s->block_index[0]= s->b8_stride*(s->mb_y*2 ) - 2 + s->mb_x*2;
s->block_index[1]= s->b8_stride*(s->mb_y*2 ) - 1 + s->mb_x*2;
s->block_index[2]= s->b8_stride*(s->mb_y*2 + 1) - 2 + s->mb_x*2;
s->block_index[3]= s->b8_stride*(s->mb_y*2 + 1) - 1 + s->mb_x*2;
s->block_index[4]= s->mb_stride*(s->mb_y + 1) + s->b8_stride*s->mb_height*2 + s->mb_x - 1;
s->block_index[5]= s->mb_stride*(s->mb_y + s->mb_height + 2) + s->b8_stride*s->mb_height*2 + s->mb_x - 1;
//block_index is not used by mpeg2, so it is not affected by chroma_format
s->dest[0] = s->cur_pic.data[0] + (int)((s->mb_x - 1U) << width_of_mb);
s->dest[1] = s->cur_pic.data[1] + (int)((s->mb_x - 1U) << (width_of_mb - s->chroma_x_shift));
s->dest[2] = s->cur_pic.data[2] + (int)((s->mb_x - 1U) << (width_of_mb - s->chroma_x_shift));
if (s->picture_structure == PICT_FRAME) {
s->dest[0] += s->mb_y * linesize << height_of_mb;
s->dest[1] += s->mb_y * uvlinesize << (height_of_mb - s->chroma_y_shift);
s->dest[2] += s->mb_y * uvlinesize << (height_of_mb - s->chroma_y_shift);
} else {
s->dest[0] += (s->mb_y>>1) * linesize << height_of_mb;
s->dest[1] += (s->mb_y>>1) * uvlinesize << (height_of_mb - s->chroma_y_shift);
s->dest[2] += (s->mb_y>>1) * uvlinesize << (height_of_mb - s->chroma_y_shift);
av_assert1((s->mb_y&1) == (s->picture_structure == PICT_BOTTOM_FIELD));
}
}
/**
* set qscale and update qscale dependent variables.
*/
void ff_set_qscale(MpegEncContext * s, int qscale)
{
if (qscale < 1)
qscale = 1;
else if (qscale > 31)
qscale = 31;
s->qscale = qscale;
s->chroma_qscale= s->chroma_qscale_table[qscale];
s->y_dc_scale= s->y_dc_scale_table[ qscale ];
s->c_dc_scale= s->c_dc_scale_table[ s->chroma_qscale ];
}