static inline int get_frame_length
(
libavsmash_audio_decode_handler_t *adhp,
uint32_t frame_number,
uint32_t *frame_length,
libavsmash_summary_t **sp
)
{
lsmash_sample_t sample;
if( lsmash_get_sample_info_from_media_timeline( adhp->root, adhp->track_ID, frame_number, &sample ) )
return -1;
*sp = &adhp->config.entries[ sample.index - 1 ];
libavsmash_summary_t *s = *sp;
if( s->extended.frame_length == 0 )
{
/* variable frame length
* Guess the frame length from sample duration. */
if( lsmash_get_sample_delta_from_media_timeline( adhp->root, adhp->track_ID, frame_number, frame_length ) )
return -1;
*frame_length *= s->extended.upsampling;
}
else
/* constant frame length */
*frame_length = s->extended.frame_length;
return 0;
}
static uint32_t libavsmash_vfr2cfr
(
libavsmash_video_decode_handler_t *vdhp,
libavsmash_video_output_handler_t *vohp,
uint32_t sample_number
)
{
/* Convert VFR to CFR. */
double target_pts = (double)((uint64_t)(sample_number - 1) * vohp->cfr_den) / vohp->cfr_num;
double current_pts = DBL_MAX;
lsmash_sample_t sample;
if( vdhp->last_sample_number <= vdhp->sample_count )
{
uint32_t last_decoding_sample_number = get_decoding_sample_number( vdhp->order_converter, vdhp->last_sample_number );
if( lsmash_get_sample_info_from_media_timeline( vdhp->root, vdhp->track_id, last_decoding_sample_number, &sample ) < 0 )
return 0;
current_pts = (double)(sample.cts - vdhp->min_cts) / vdhp->media_timescale;
if( target_pts == current_pts )
return vdhp->last_sample_number;
}
if( target_pts < current_pts )
{
uint32_t composition_sample_number;
for( composition_sample_number = vdhp->last_sample_number - 1;
composition_sample_number;
composition_sample_number-- )
{
uint32_t decoding_sample_number = get_decoding_sample_number( vdhp->order_converter, composition_sample_number );
if( lsmash_get_sample_info_from_media_timeline( vdhp->root, vdhp->track_id, decoding_sample_number, &sample ) < 0 )
return 0;
current_pts = (double)(sample.cts - vdhp->min_cts) / vdhp->media_timescale;
if( current_pts <= target_pts )
{
sample_number = composition_sample_number;
break;
}
}
if( composition_sample_number == 0 )
return 0;
}
else
{
uint32_t composition_sample_number;
for( composition_sample_number = vdhp->last_sample_number + 1;
composition_sample_number <= vdhp->sample_count;
composition_sample_number++ )
{
uint32_t decoding_sample_number = get_decoding_sample_number( vdhp->order_converter, composition_sample_number );
if( lsmash_get_sample_info_from_media_timeline( vdhp->root, vdhp->track_id, decoding_sample_number, &sample ) < 0 )
return 0;
current_pts = (double)(sample.cts - vdhp->min_cts) / vdhp->media_timescale;
if( current_pts > target_pts )
{
sample_number = composition_sample_number - 1;
break;
}
}
if( composition_sample_number > vdhp->sample_count )
sample_number = vdhp->sample_count;
}
return sample_number;
}
static int get_requested_picture
(
libavsmash_video_decode_handler_t *vdhp,
AVFrame *picture,
uint32_t sample_number
)
{
#define MAX_ERROR_COUNT 3 /* arbitrary */
codec_configuration_t *config = &vdhp->config;
uint32_t config_index;
if( sample_number < vdhp->first_valid_frame_number || vdhp->sample_count == 1 )
{
/* Get the index of the decoder configuration. */
lsmash_sample_t sample;
uint32_t decoding_sample_number = get_decoding_sample_number( vdhp->order_converter, vdhp->first_valid_frame_number );
if( lsmash_get_sample_info_from_media_timeline( vdhp->root, vdhp->track_id, decoding_sample_number, &sample ) < 0 )
goto video_fail;
config_index = sample.index;
/* Copy the first valid video frame data. */
av_frame_unref( picture );
if( av_frame_ref( picture, vdhp->first_valid_frame ) < 0 )
goto video_fail;
/* Force seeking at the next access for valid video frame. */
vdhp->last_sample_number = vdhp->sample_count + 1;
goto return_frame;
}
uint32_t start_number; /* number of sample, for normal decoding, where decoding starts excluding decoding delay */
uint32_t rap_number; /* number of sample, for seeking, where decoding starts excluding decoding delay */
int seek_mode = vdhp->seek_mode;
int roll_recovery = 0;
if( sample_number > vdhp->last_sample_number
&& sample_number <= vdhp->last_sample_number + vdhp->forward_seek_threshold )
{
start_number = vdhp->last_sample_number + 1 + config->delay_count;
rap_number = vdhp->last_rap_number;
}
else
{
roll_recovery = find_random_accessible_point( vdhp, sample_number, 0, &rap_number );
if( rap_number == vdhp->last_rap_number && sample_number > vdhp->last_sample_number )
{
roll_recovery = 0;
start_number = vdhp->last_sample_number + 1 + config->delay_count;
}
else
{
/* Require starting to decode from random accessible sample. */
vdhp->last_rap_number = rap_number;
start_number = seek_video( vdhp, picture, sample_number, rap_number, roll_recovery || seek_mode != SEEK_MODE_NORMAL );
}
}
/* Get the desired picture. */
int error_count = 0;
while( start_number == 0 /* Failed to seek. */
|| config->update_pending /* Need to update the decoder configuration to decode pictures. */
|| get_picture( vdhp, picture, start_number, sample_number + config->delay_count ) < 0 )
{
if( config->update_pending )
{
roll_recovery = find_random_accessible_point( vdhp, sample_number, 0, &rap_number );
vdhp->last_rap_number = rap_number;
}
else
{
/* Failed to get the desired picture. */
if( config->error || seek_mode == SEEK_MODE_AGGRESSIVE )
goto video_fail;
if( ++error_count > MAX_ERROR_COUNT || rap_number <= 1 )
{
if( seek_mode == SEEK_MODE_UNSAFE )
goto video_fail;
/* Retry to decode from the same random accessible sample with error ignorance. */
seek_mode = SEEK_MODE_AGGRESSIVE;
}
else
{
/* Retry to decode from more past random accessible sample. */
roll_recovery = find_random_accessible_point( vdhp, sample_number, rap_number - 1, &rap_number );
if( vdhp->last_rap_number == rap_number )
goto video_fail;
vdhp->last_rap_number = rap_number;
}
}
start_number = seek_video( vdhp, picture, sample_number, rap_number, roll_recovery || seek_mode != SEEK_MODE_NORMAL );
}
vdhp->last_sample_number = sample_number;
config_index = config->index;
return_frame:;
/* Don't exceed the maximum presentation size specified for each sequence. */
extended_summary_t *extended = &config->entries[ config_index - 1 ].extended;
if( config->ctx->width > extended->width )
config->ctx->width = extended->width;
if( config->ctx->height > extended->height )
config->ctx->height = extended->height;
return 0;
video_fail:
/* fatal error of decoding */
lw_log_show( &config->lh, LW_LOG_WARNING, "Couldn't read video frame." );
return -1;
#undef MAX_ERROR_COUNT
}
static int find_random_accessible_point
(
libavsmash_video_decode_handler_t *vdhp,
uint32_t composition_sample_number,
uint32_t decoding_sample_number,
uint32_t *rap_number
)
{
if( decoding_sample_number == 0 )
decoding_sample_number = get_decoding_sample_number( vdhp->order_converter, composition_sample_number );
lsmash_random_access_flag ra_flags;
uint32_t distance; /* distance from the closest random accessible point to the previous. */
uint32_t number_of_leadings;
if( lsmash_get_closest_random_accessible_point_detail_from_media_timeline( vdhp->root, vdhp->track_id,
decoding_sample_number, rap_number,
&ra_flags, &number_of_leadings, &distance ) < 0 )
*rap_number = 1;
int roll_recovery = !!(ra_flags & ISOM_SAMPLE_RANDOM_ACCESS_FLAG_GDR);
int is_leading = number_of_leadings && (decoding_sample_number - *rap_number <= number_of_leadings);
if( (roll_recovery || is_leading) && *rap_number > distance )
*rap_number -= distance;
/* Check whether random accessible point has the same decoder configuration or not. */
decoding_sample_number = get_decoding_sample_number( vdhp->order_converter, composition_sample_number );
do
{
lsmash_sample_t sample;
lsmash_sample_t rap_sample;
if( lsmash_get_sample_info_from_media_timeline( vdhp->root, vdhp->track_id, decoding_sample_number, &sample ) < 0
|| lsmash_get_sample_info_from_media_timeline( vdhp->root, vdhp->track_id, *rap_number, &rap_sample ) < 0 )
{
/* Fatal error. */
*rap_number = vdhp->last_rap_number;
return 0;
}
if( sample.index == rap_sample.index )
break;
uint32_t sample_index = sample.index;
for( uint32_t i = decoding_sample_number - 1; i; i-- )
{
if( lsmash_get_sample_info_from_media_timeline( vdhp->root, vdhp->track_id, i, &sample ) < 0 )
{
/* Fatal error. */
*rap_number = vdhp->last_rap_number;
return 0;
}
if( sample.index != sample_index )
{
if( distance )
{
*rap_number += distance;
distance = 0;
continue;
}
else
*rap_number = i + 1;
}
}
break;
} while( 1 );
return roll_recovery;
}
uint64_t libavsmash_count_overall_pcm_samples
(
libavsmash_audio_decode_handler_t *adhp,
int output_sample_rate,
uint64_t *skip_decoded_samples /* converted to upsampled in this function */
)
{
codec_configuration_t *config = &adhp->config;
extended_summary_t *es = NULL;
/* Here, the decoder upsampling is defined only when libavcodec doesn't return upsampled length as AVCodecContext.frame_size. */
int current_sample_rate = 0; /* after the decoder upsampling */
uint32_t current_index = 0;
uint32_t current_frame_length = 0; /* before the decoder upsampling */
uint64_t sequence_pcm_count = 0; /* before the decoder upsampling */
uint64_t prior_sequences_pcm_count = 0; /* before the decoder upsampling */
uint64_t overall_pcm_count = 0; /* after the decoder and the resampler upsampling */
uint64_t orig_skip_decoded_samples = *skip_decoded_samples; /* before the decoder upsampling */
/* Count the number of output PCM audio samples in each sequence. */
*skip_decoded_samples = 0;
for( uint32_t i = 1; i <= adhp->frame_count; i++ )
{
/* Get configuration index. */
lsmash_sample_t sample;
if( lsmash_get_sample_info_from_media_timeline( adhp->root, adhp->track_ID, i, &sample ) )
continue;
if( current_index != sample.index )
{
es = &config->entries[ sample.index - 1 ].extended;
current_index = sample.index;
}
else if( !es )
continue;
/* Get audio frame length. */
uint32_t frame_length;
if( es->frame_length )
frame_length = es->frame_length;
else if( lsmash_get_sample_delta_from_media_timeline( adhp->root, adhp->track_ID, i, &frame_length ) )
continue;
/* */
if( (current_sample_rate != es->sample_rate && es->sample_rate > 0)
|| current_frame_length != frame_length )
{
/* Encountered a new sequence. */
if( current_sample_rate > 0 )
{
/* Add the number of decoded PCM audio samples which shall be skipped. */
if( orig_skip_decoded_samples > prior_sequences_pcm_count )
{
if( orig_skip_decoded_samples >= prior_sequences_pcm_count + sequence_pcm_count )
/* All decoded PCM audio samples in the previous sequence shall be skipped. */
*skip_decoded_samples += sequence_pcm_count * es->upsampling;
else
/* 0 < orig_skip_decoded_samples - prior_sequences_pcm_count < sequence_pcm_count
* Partial decoded PCM audio samples in the previous sequence are not skipped. */
*skip_decoded_samples += (orig_skip_decoded_samples - prior_sequences_pcm_count) * es->upsampling;
}
prior_sequences_pcm_count += sequence_pcm_count;
/* Add the number of output PCM audio samples in the previous sequence. */
overall_pcm_count += count_sequence_output_pcm_samples( sequence_pcm_count * es->upsampling,
*skip_decoded_samples,
current_sample_rate,
output_sample_rate );
sequence_pcm_count = 0;
}
current_sample_rate = es->sample_rate > 0 ? es->sample_rate : config->ctx->sample_rate;
current_frame_length = frame_length;
}
sequence_pcm_count += frame_length;
}
if( !es || (sequence_pcm_count == 0 && overall_pcm_count == 0) )
return 0;
/* Count the number of output PCM audio samples in the last sequence. */
if( orig_skip_decoded_samples > prior_sequences_pcm_count )
*skip_decoded_samples += (orig_skip_decoded_samples - prior_sequences_pcm_count) * es->upsampling;
current_sample_rate = es->sample_rate > 0 ? es->sample_rate : config->ctx->sample_rate;
overall_pcm_count += count_sequence_output_pcm_samples( sequence_pcm_count * es->upsampling,
*skip_decoded_samples,
current_sample_rate,
output_sample_rate );
/* Return the number of output PCM audio samples. */
return overall_pcm_count;
}
