/*****************************************************************************
* aout_FifoMoveDates : Move forwards or backwards all dates in the FIFO
*****************************************************************************/
void aout_FifoMoveDates( aout_fifo_t * p_fifo, mtime_t difference )
{
aout_buffer_t * p_buffer;
date_Move( &p_fifo->end_date, difference );
p_buffer = p_fifo->p_first;
while ( p_buffer != NULL )
{
p_buffer->i_pts += difference;
p_buffer = p_buffer->p_next;
}
}
/**
* Move forwards or backwards all dates in the FIFO.
*/
static void aout_FifoMoveDates( aout_fifo_t *fifo, mtime_t difference )
{
if( date_Get( &fifo->end_date ) == VLC_TS_INVALID )
{
assert( fifo->p_first == NULL );
return;
}
date_Move( &fifo->end_date, difference );
for( block_t *block = fifo->p_first; block != NULL; block = block->p_next )
block->i_pts += difference;
}
/*****************************************************************************
* aout_FifoMoveDates : Move forwards or backwards all dates in the FIFO
*****************************************************************************/
void aout_FifoMoveDates( aout_instance_t * p_aout, aout_fifo_t * p_fifo,
mtime_t difference )
{
aout_buffer_t * p_buffer;
(void)p_aout;
AOUT_ASSERT_FIFO_LOCKED;
date_Move( &p_fifo->end_date, difference );
p_buffer = p_fifo->p_first;
while ( p_buffer != NULL )
{
p_buffer->i_pts += difference;
p_buffer = p_buffer->p_next;
}
}
/**
* Notifies the audio input of the drift from the requested audio
* playback timestamp (@ref block_t.i_pts) to the anticipated playback time
* as reported by the audio output hardware.
* Depending on the drift amplitude, the input core may ignore the drift
* trigger upsampling or downsampling, or even discard samples.
* Future VLC versions may instead adjust the input decoding speed.
*
* The audio output plugin is responsible for estimating the ideal current
* playback time defined as follows:
* ideal time = buffer timestamp - (output latency + pending buffer duration)
*
* Practically, this is the PTS (block_t.i_pts) of the current buffer minus
* the latency reported by the output programming interface.
* Computing the estimated drift directly would probably be more intuitive.
* However the use of an absolute time value does not introduce extra
* measurement errors due to the CPU scheduling jitter and clock resolution.
* Furthermore, the ideal while it is an abstract value, is easy for most
* audio output plugins to compute.
* The following definition is equivalent but depends on the clock time:
* ideal time = real time + drift
* @note If aout_LatencyReport() is never called, the core will assume that
* there is no drift.
*
* @param ideal estimated ideal time as defined above.
*/
void aout_TimeReport (audio_output_t *aout, mtime_t ideal)
{
mtime_t delta = mdate() - ideal /* = -drift */;
aout_assert_locked (aout);
if (delta < -AOUT_MAX_PTS_ADVANCE || +AOUT_MAX_PTS_DELAY < delta)
{
aout_owner_t *owner = aout_owner (aout);
msg_Warn (aout, "not synchronized (%"PRId64" us), resampling",
delta);
if (date_Get (&owner->sync.date) != VLC_TS_INVALID)
date_Move (&owner->sync.date, delta);
}
}
