gstreamer/subprojects/gst-plugins-base/gst-libs/gst/audio/gstaudiostreamalign.c
Sebastian Dröge c4789e6de5 audio: Consider the expected timestamp for discont-wait handling
Otherwise if there is a huge gap it will only be considered a
discontinuity after another discont-time amount of buffers has passed.

Like this it will be immediately a discontinuity if the gap between the
expected and received time becomes bigger than the discont-time.

The last part of the test was actually testing for this behaviour and
expected the previous behaviour. Most other tests also had to be
adjusted because discont will now happen at slightly different times
than before.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5759>
2023-12-17 12:01:27 +00:00

477 lines
14 KiB
C

/* GStreamer
* Copyright (C) 2017 Sebastian Dröge <sebastian@centricular.com>
*
* gstaudiostreamalign.h:
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
* Boston, MA 02110-1301, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "gstaudiostreamalign.h"
/**
* SECTION:gstaudiostreamalign
* @title: GstAudioStreamAlign
* @short_description: Helper object for tracking audio stream alignment and discontinuities
*
* #GstAudioStreamAlign provides a helper object that helps tracking audio
* stream alignment and discontinuities, and detects discontinuities if
* possible.
*
* See gst_audio_stream_align_new() for a description of its parameters and
* gst_audio_stream_align_process() for the details of the processing.
*/
G_DEFINE_BOXED_TYPE (GstAudioStreamAlign, gst_audio_stream_align,
(GBoxedCopyFunc) gst_audio_stream_align_copy,
(GBoxedFreeFunc) gst_audio_stream_align_free);
struct _GstAudioStreamAlign
{
gint rate;
GstClockTime alignment_threshold;
GstClockTime discont_wait;
/* counter to keep track of timestamps */
guint64 next_offset;
GstClockTime timestamp_at_discont;
guint64 samples_since_discont;
/* Last time we noticed a discont */
GstClockTime discont_time;
};
/**
* gst_audio_stream_align_new:
* @rate: a sample rate
* @alignment_threshold: a alignment threshold in nanoseconds
* @discont_wait: discont wait in nanoseconds
*
* Allocate a new #GstAudioStreamAlign with the given configuration. All
* processing happens according to sample rate @rate, until
* gst_audio_stream_align_set_rate() is called with a new @rate.
* A negative rate can be used for reverse playback.
*
* @alignment_threshold gives the tolerance in nanoseconds after which a
* timestamp difference is considered a discontinuity. Once detected,
* @discont_wait nanoseconds have to pass without going below the threshold
* again until the output buffer is marked as a discontinuity. These can later
* be re-configured with gst_audio_stream_align_set_alignment_threshold() and
* gst_audio_stream_align_set_discont_wait().
*
* Returns: a new #GstAudioStreamAlign. free with gst_audio_stream_align_free().
*
* Since: 1.14
*/
GstAudioStreamAlign *
gst_audio_stream_align_new (gint rate, GstClockTime alignment_threshold,
GstClockTime discont_wait)
{
GstAudioStreamAlign *align;
g_return_val_if_fail (rate != 0, NULL);
g_return_val_if_fail (GST_CLOCK_TIME_IS_VALID (alignment_threshold), NULL);
g_return_val_if_fail (GST_CLOCK_TIME_IS_VALID (discont_wait), NULL);
align = g_new0 (GstAudioStreamAlign, 1);
align->rate = rate;
align->alignment_threshold = alignment_threshold;
align->discont_wait = discont_wait;
align->timestamp_at_discont = GST_CLOCK_TIME_NONE;
align->samples_since_discont = 0;
gst_audio_stream_align_mark_discont (align);
return align;
}
/**
* gst_audio_stream_align_copy:
* @align: a #GstAudioStreamAlign
*
* Copy a GstAudioStreamAlign structure.
*
* Returns: a new #GstAudioStreamAlign. free with gst_audio_stream_align_free.
*
* Since: 1.14
*/
GstAudioStreamAlign *
gst_audio_stream_align_copy (const GstAudioStreamAlign * align)
{
GstAudioStreamAlign *copy;
g_return_val_if_fail (align != NULL, NULL);
copy = g_new0 (GstAudioStreamAlign, 1);
*copy = *align;
return copy;
}
/**
* gst_audio_stream_align_free:
* @align: a #GstAudioStreamAlign
*
* Free a GstAudioStreamAlign structure previously allocated with gst_audio_stream_align_new()
* or gst_audio_stream_align_copy().
*
* Since: 1.14
*/
void
gst_audio_stream_align_free (GstAudioStreamAlign * align)
{
g_return_if_fail (align != NULL);
g_free (align);
}
/**
* gst_audio_stream_align_set_rate:
* @align: a #GstAudioStreamAlign
* @rate: a new sample rate
*
* Sets @rate as new sample rate for the following processing. If the sample
* rate differs this implicitly marks the next data as discontinuous.
*
* Since: 1.14
*/
void
gst_audio_stream_align_set_rate (GstAudioStreamAlign * align, gint rate)
{
g_return_if_fail (align != NULL);
g_return_if_fail (rate != 0);
if (align->rate == rate)
return;
align->rate = rate;
gst_audio_stream_align_mark_discont (align);
}
/**
* gst_audio_stream_align_get_rate:
* @align: a #GstAudioStreamAlign
*
* Gets the currently configured sample rate.
*
* Returns: The currently configured sample rate
*
* Since: 1.14
*/
gint
gst_audio_stream_align_get_rate (const GstAudioStreamAlign * align)
{
g_return_val_if_fail (align != NULL, 0);
return align->rate;
}
/**
* gst_audio_stream_align_set_alignment_threshold:
* @align: a #GstAudioStreamAlign
* @alignment_threshold: a new alignment threshold
*
* Sets @alignment_treshold as new alignment threshold for the following processing.
*
* Since: 1.14
*/
void
gst_audio_stream_align_set_alignment_threshold (GstAudioStreamAlign *
align, GstClockTime alignment_threshold)
{
g_return_if_fail (align != NULL);
g_return_if_fail (GST_CLOCK_TIME_IS_VALID (alignment_threshold));
align->alignment_threshold = alignment_threshold;
}
/**
* gst_audio_stream_align_get_alignment_threshold:
* @align: a #GstAudioStreamAlign
*
* Gets the currently configured alignment threshold.
*
* Returns: The currently configured alignment threshold
*
* Since: 1.14
*/
GstClockTime
gst_audio_stream_align_get_alignment_threshold (const GstAudioStreamAlign *
align)
{
g_return_val_if_fail (align != NULL, 0);
return align->alignment_threshold;
}
/**
* gst_audio_stream_align_set_discont_wait:
* @align: a #GstAudioStreamAlign
* @discont_wait: a new discont wait
*
* Sets @alignment_treshold as new discont wait for the following processing.
*
* Since: 1.14
*/
void
gst_audio_stream_align_set_discont_wait (GstAudioStreamAlign * align,
GstClockTime discont_wait)
{
g_return_if_fail (align != NULL);
g_return_if_fail (GST_CLOCK_TIME_IS_VALID (discont_wait));
align->discont_wait = discont_wait;
}
/**
* gst_audio_stream_align_get_discont_wait:
* @align: a #GstAudioStreamAlign
*
* Gets the currently configured discont wait.
*
* Returns: The currently configured discont wait
*
* Since: 1.14
*/
GstClockTime
gst_audio_stream_align_get_discont_wait (const GstAudioStreamAlign * align)
{
g_return_val_if_fail (align != NULL, 0);
return align->discont_wait;
}
/**
* gst_audio_stream_align_mark_discont:
* @align: a #GstAudioStreamAlign
*
* Marks the next buffer as discontinuous and resets timestamp tracking.
*
* Since: 1.14
*/
void
gst_audio_stream_align_mark_discont (GstAudioStreamAlign * align)
{
g_return_if_fail (align != NULL);
align->next_offset = -1;
align->discont_time = GST_CLOCK_TIME_NONE;
}
/**
* gst_audio_stream_align_get_timestamp_at_discont:
* @align: a #GstAudioStreamAlign
*
* Timestamp that was passed when a discontinuity was detected, i.e. the first
* timestamp after the discontinuity.
*
* Returns: The last timestamp at when a discontinuity was detected
*
* Since: 1.14
*/
GstClockTime
gst_audio_stream_align_get_timestamp_at_discont (const GstAudioStreamAlign *
align)
{
g_return_val_if_fail (align != NULL, GST_CLOCK_TIME_NONE);
return align->timestamp_at_discont;
}
/**
* gst_audio_stream_align_get_samples_since_discont:
* @align: a #GstAudioStreamAlign
*
* Returns the number of samples that were processed since the last
* discontinuity was detected.
*
* Returns: The number of samples processed since the last discontinuity.
*
* Since: 1.14
*/
guint64
gst_audio_stream_align_get_samples_since_discont (const GstAudioStreamAlign *
align)
{
g_return_val_if_fail (align != NULL, 0);
return align->samples_since_discont;
}
/**
* gst_audio_stream_align_process:
* @align: a #GstAudioStreamAlign
* @discont: if this data is considered to be discontinuous
* @timestamp: a #GstClockTime of the start of the data
* @n_samples: number of samples to process
* @out_timestamp: (out): output timestamp of the data
* @out_duration: (out): output duration of the data
* @out_sample_position: (out): output sample position of the start of the data
*
* Processes data with @timestamp and @n_samples, and returns the output
* timestamp, duration and sample position together with a boolean to signal
* whether a discontinuity was detected or not. All non-discontinuous data
* will have perfect timestamps and durations.
*
* A discontinuity is detected once the difference between the actual
* timestamp and the timestamp calculated from the sample count since the last
* discontinuity differs by more than the alignment threshold for a duration
* longer than discont wait.
*
* Note: In reverse playback, every buffer is considered discontinuous in the
* context of buffer flags because the last sample of the previous buffer is
* discontinuous with the first sample of the current one. However for this
* function they are only considered discontinuous in reverse playback if the
* first sample of the previous buffer is discontinuous with the last sample
* of the current one.
*
* Returns: %TRUE if a discontinuity was detected, %FALSE otherwise.
*
* Since: 1.14
*/
#define ABSDIFF(a, b) ((a) > (b) ? (a) - (b) : (b) - (a))
gboolean
gst_audio_stream_align_process (GstAudioStreamAlign * align,
gboolean discont, GstClockTime timestamp, guint n_samples,
GstClockTime * out_timestamp, GstClockTime * out_duration,
guint64 * out_sample_position)
{
GstClockTime start_time, end_time, duration;
guint64 start_offset, end_offset;
g_return_val_if_fail (align != NULL, FALSE);
start_time = timestamp;
start_offset =
gst_util_uint64_scale (start_time, ABS (align->rate), GST_SECOND);
end_offset = start_offset + n_samples;
end_time =
gst_util_uint64_scale_int (end_offset, GST_SECOND, ABS (align->rate));
duration = end_time - start_time;
if (align->next_offset == (guint64) - 1 || discont) {
discont = TRUE;
} else {
guint64 diff, max_sample_diff;
GstClockTime expected_time;
/* Check discont */
if (align->rate > 0) {
diff = ABSDIFF (start_offset, align->next_offset);
} else {
diff = ABSDIFF (end_offset, align->next_offset);
}
expected_time =
gst_util_uint64_scale (align->next_offset, GST_SECOND,
ABS (align->rate));
max_sample_diff =
gst_util_uint64_scale_int (align->alignment_threshold,
ABS (align->rate), GST_SECOND);
/* Discont! */
if (G_UNLIKELY (diff >= max_sample_diff)) {
if (align->discont_wait > 0) {
if (align->discont_time == GST_CLOCK_TIME_NONE) {
if (align->rate > 0
&& ABSDIFF (expected_time, start_time) >= align->discont_wait)
discont = TRUE;
else if (align->rate < 0
&& ABSDIFF (expected_time, end_time) >= align->discont_wait)
discont = TRUE;
else
align->discont_time = expected_time;
} else if ((align->rate > 0
&& ABSDIFF (start_time,
align->discont_time) >= align->discont_wait)
|| (align->rate < 0
&& ABSDIFF (end_time,
align->discont_time) >= align->discont_wait)) {
discont = TRUE;
align->discont_time = GST_CLOCK_TIME_NONE;
}
} else {
discont = TRUE;
}
} else if (G_UNLIKELY (align->discont_time != GST_CLOCK_TIME_NONE)) {
/* we have had a discont, but are now back on track! */
align->discont_time = GST_CLOCK_TIME_NONE;
}
}
if (discont) {
/* Have discont, need resync and use the capture timestamps */
if (align->next_offset != (guint64) - 1)
GST_INFO ("Have discont. Expected %"
G_GUINT64_FORMAT ", got %" G_GUINT64_FORMAT,
align->next_offset, start_offset);
align->next_offset = align->rate > 0 ? end_offset : start_offset;
align->timestamp_at_discont = start_time;
align->samples_since_discont = 0;
/* Got a discont and adjusted, reset the discont_time marker */
align->discont_time = GST_CLOCK_TIME_NONE;
} else {
/* No discont, just keep counting */
if (align->rate > 0) {
timestamp =
gst_util_uint64_scale (align->next_offset, GST_SECOND,
ABS (align->rate));
start_offset = align->next_offset;
align->next_offset += n_samples;
duration =
gst_util_uint64_scale (align->next_offset, GST_SECOND,
ABS (align->rate)) - timestamp;
} else {
guint64 old_offset = align->next_offset;
if (align->next_offset > n_samples)
align->next_offset -= n_samples;
else
align->next_offset = 0;
start_offset = align->next_offset;
timestamp =
gst_util_uint64_scale (align->next_offset, GST_SECOND,
ABS (align->rate));
duration =
gst_util_uint64_scale (old_offset, GST_SECOND,
ABS (align->rate)) - timestamp;
}
}
align->samples_since_discont += n_samples;
if (out_timestamp)
*out_timestamp = timestamp;
if (out_duration)
*out_duration = duration;
if (out_sample_position)
*out_sample_position = start_offset;
return discont;
}
#undef ABSDIFF