mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-12-20 23:36:38 +00:00
495 lines
14 KiB
C
495 lines
14 KiB
C
/*
|
|
* Copyright (C) 2019 Collabora Ltd.
|
|
* Author: Xavier Claessens <xavier.claessens@collabora.com>
|
|
*
|
|
* This library 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
|
|
* version 2.1 of the License.
|
|
*
|
|
* 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
|
|
* Lesser General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
* License along with this library; if not, write to the Free Software
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
*
|
|
*/
|
|
|
|
/**
|
|
* SECTION:mlaudiosink
|
|
* @short_description: Audio sink for Magic Leap platform
|
|
* @see_also: #GstAudioSink
|
|
*
|
|
* An audio sink element for LuminOS, the Magic Leap platform. There are 2 modes
|
|
* supported: normal and spatial. By default the audio is output directly to the
|
|
* stereo speakers, but in spatial mode the audio will be localised in the 3D
|
|
* environment. The user ears the sound as coming from a point in space, from a
|
|
* given distance and direction.
|
|
*
|
|
* To enable the spatial mode, the application needs to set a sync bus
|
|
* handler, using gst_bus_set_sync_handler(), to catch messages of type
|
|
* %GST_MESSAGE_ELEMENT named "gst.mlaudiosink.need-app" and
|
|
* "gst.mlaudiosink.need-audio-node". The need-app message will be posted first,
|
|
* application must then set the #GstMLAudioSink::app property with the pointer
|
|
* to application's lumin::BaseApp C++ object. That property can also be set on
|
|
* element creation in which case the need-app message won't be posted. After
|
|
* that, and if #GstMLAudioSink::app has been set, the need-audio-node message
|
|
* is posted from lumin::BaseApp's main thread. The application must then create
|
|
* a lumin::AudioNode C++ object, using lumin::Prism::createAudioNode(), and set
|
|
* the #GstMLAudioSink::audio-node property. Note that it is important that the
|
|
* lumin::AudioNode object must be created from within that message handler,
|
|
* and in the caller's thread, this is a limitation/bug of the platform
|
|
* (atleast until version 0.97).
|
|
*
|
|
* Here is an example of bus message handler to enable spatial sound:
|
|
* ```C
|
|
* static GstBusSyncReply
|
|
* bus_sync_handler_cb (GstBus * bus, GstMessage * msg, gpointer user_data)
|
|
* {
|
|
* MyApplication * self = user_data;
|
|
*
|
|
* if (GST_MESSAGE_TYPE (msg) == GST_MESSAGE_ELEMENT) {
|
|
* if (gst_message_has_name (msg, "gst.mlaudiosink.need-app")) {
|
|
* g_object_set (G_OBJECT (msg->src), "app", &self->app, NULL);
|
|
* } else if (gst_message_has_name (msg, "gst.mlaudiosink.need-audio-node")) {
|
|
* self->audio_node = self->prism->createAudioNode ();
|
|
* self->audio_node->setSpatialSoundEnable (true);
|
|
* self->ui_node->addChild(self->audio_node);
|
|
* g_object_set (G_OBJECT (msg->src), "audio-node", self->audio_node, NULL);
|
|
* }
|
|
* }
|
|
* return GST_BUS_PASS;
|
|
* }
|
|
* ```
|
|
*
|
|
* Since: 1.18
|
|
*/
|
|
|
|
#ifdef HAVE_CONFIG_H
|
|
#include "config.h"
|
|
#endif
|
|
|
|
#include "mlaudiosink.h"
|
|
#include "mlaudiowrapper.h"
|
|
|
|
GST_DEBUG_CATEGORY_EXTERN (mgl_debug);
|
|
#define GST_CAT_DEFAULT mgl_debug
|
|
|
|
static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
|
|
GST_PAD_SINK,
|
|
GST_PAD_ALWAYS,
|
|
GST_STATIC_CAPS ("audio/x-raw, "
|
|
"format = (string) { S16LE }, "
|
|
"channels = (int) [ 1, 2 ], "
|
|
"rate = (int) [ 16000, 48000 ], " "layout = (string) interleaved"));
|
|
|
|
/* HACK: After calling MLAudioStopSound() there is no way to know when it will
|
|
* actually stop calling buffer_cb(). If the sink is disposed first, it would
|
|
* crash. Keep here a set of active sinks. */
|
|
static GHashTable *active_sinks;
|
|
static GMutex active_sinks_mutex;
|
|
|
|
struct _GstMLAudioSink
|
|
{
|
|
GstAudioSink parent;
|
|
|
|
gpointer audio_node;
|
|
gpointer app;
|
|
|
|
GstMLAudioWrapper *wrapper;
|
|
MLAudioBufferFormat format;
|
|
uint32_t recommended_buffer_size;
|
|
MLAudioBuffer buffer;
|
|
guint buffer_offset;
|
|
gboolean has_buffer;
|
|
gboolean paused;
|
|
gboolean stopped;
|
|
|
|
GMutex mutex;
|
|
GCond cond;
|
|
};
|
|
|
|
G_DEFINE_TYPE (GstMLAudioSink, gst_ml_audio_sink, GST_TYPE_AUDIO_SINK);
|
|
GST_ELEMENT_REGISTER_DEFINE_WITH_CODE (mlaudiosink, "mlaudiosink",
|
|
GST_RANK_PRIMARY + 10, GST_TYPE_ML_AUDIO_SINK,
|
|
GST_DEBUG_CATEGORY_INIT (mgl_debug, "magicleap", 0, "Magic Leap elements"));
|
|
|
|
enum
|
|
{
|
|
PROP_0,
|
|
PROP_AUDIO_NODE,
|
|
PROP_APP,
|
|
};
|
|
|
|
static void
|
|
gst_ml_audio_sink_init (GstMLAudioSink * self)
|
|
{
|
|
g_mutex_init (&self->mutex);
|
|
g_cond_init (&self->cond);
|
|
}
|
|
|
|
static void
|
|
gst_ml_audio_sink_dispose (GObject * object)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (object);
|
|
|
|
g_mutex_clear (&self->mutex);
|
|
g_cond_clear (&self->cond);
|
|
|
|
G_OBJECT_CLASS (gst_ml_audio_sink_parent_class)->dispose (object);
|
|
}
|
|
|
|
static void
|
|
gst_ml_audio_sink_set_property (GObject * object, guint prop_id,
|
|
const GValue * value, GParamSpec * pspec)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (object);
|
|
|
|
switch (prop_id) {
|
|
case PROP_AUDIO_NODE:
|
|
self->audio_node = g_value_get_pointer (value);
|
|
break;
|
|
case PROP_APP:
|
|
self->app = g_value_get_pointer (value);
|
|
break;
|
|
default:
|
|
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
|
|
break;
|
|
}
|
|
}
|
|
|
|
static void
|
|
gst_ml_audio_sink_get_property (GObject * object, guint prop_id, GValue * value,
|
|
GParamSpec * pspec)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (object);
|
|
|
|
switch (prop_id) {
|
|
case PROP_AUDIO_NODE:
|
|
g_value_set_pointer (value, self->audio_node);
|
|
break;
|
|
case PROP_APP:
|
|
g_value_set_pointer (value, self->app);
|
|
break;
|
|
default:
|
|
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
|
|
break;
|
|
}
|
|
}
|
|
|
|
static GstCaps *
|
|
gst_ml_audio_sink_getcaps (GstBaseSink * bsink, GstCaps * filter)
|
|
{
|
|
GstCaps *caps;
|
|
|
|
caps = gst_static_caps_get (&sink_template.static_caps);
|
|
|
|
if (filter) {
|
|
gst_caps_replace (&caps,
|
|
gst_caps_intersect_full (filter, caps, GST_CAPS_INTERSECT_FIRST));
|
|
}
|
|
|
|
return caps;
|
|
}
|
|
|
|
static gboolean
|
|
gst_ml_audio_sink_open (GstAudioSink * sink)
|
|
{
|
|
/* Nothing to do in open/close */
|
|
return TRUE;
|
|
}
|
|
|
|
static void
|
|
buffer_cb (MLHandle handle, gpointer user_data)
|
|
{
|
|
GstMLAudioSink *self = user_data;
|
|
|
|
g_mutex_lock (&active_sinks_mutex);
|
|
if (!g_hash_table_contains (active_sinks, self))
|
|
goto out;
|
|
|
|
gst_ml_audio_wrapper_set_handle (self->wrapper, handle);
|
|
|
|
g_mutex_lock (&self->mutex);
|
|
g_cond_signal (&self->cond);
|
|
g_mutex_unlock (&self->mutex);
|
|
|
|
out:
|
|
g_mutex_unlock (&active_sinks_mutex);
|
|
}
|
|
|
|
/* Must be called with self->mutex locked */
|
|
static gboolean
|
|
wait_for_buffer (GstMLAudioSink * self)
|
|
{
|
|
gboolean ret = TRUE;
|
|
|
|
while (!self->has_buffer && !self->stopped) {
|
|
MLResult result;
|
|
|
|
result = gst_ml_audio_wrapper_get_buffer (self->wrapper, &self->buffer);
|
|
if (result == MLResult_Ok) {
|
|
self->has_buffer = TRUE;
|
|
self->buffer_offset = 0;
|
|
} else if (result == MLAudioResult_BufferNotReady) {
|
|
g_cond_wait (&self->cond, &self->mutex);
|
|
} else {
|
|
GST_ERROR_OBJECT (self, "Failed to get output buffer: %d", result);
|
|
ret = FALSE;
|
|
break;
|
|
}
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
static gboolean
|
|
create_sound_cb (GstMLAudioWrapper * wrapper, gpointer user_data)
|
|
{
|
|
GstMLAudioSink *self = user_data;
|
|
MLResult result;
|
|
|
|
if (self->app) {
|
|
gst_element_post_message (GST_ELEMENT (self),
|
|
gst_message_new_element (GST_OBJECT (self),
|
|
gst_structure_new_empty ("gst.mlaudiosink.need-audio-node")));
|
|
}
|
|
|
|
gst_ml_audio_wrapper_set_node (self->wrapper, self->audio_node);
|
|
|
|
result = gst_ml_audio_wrapper_create_sound (self->wrapper, &self->format,
|
|
self->recommended_buffer_size, buffer_cb, self);
|
|
if (result != MLResult_Ok) {
|
|
GST_ERROR_OBJECT (self, "Failed to create output stream: %d", result);
|
|
return FALSE;
|
|
}
|
|
|
|
return TRUE;
|
|
}
|
|
|
|
static gboolean
|
|
gst_ml_audio_sink_prepare (GstAudioSink * sink, GstAudioRingBufferSpec * spec)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (sink);
|
|
float max_pitch = 1.0f;
|
|
uint32_t min_size;
|
|
MLResult result;
|
|
|
|
result =
|
|
MLAudioGetOutputStreamDefaults (GST_AUDIO_INFO_CHANNELS (&spec->info),
|
|
GST_AUDIO_INFO_RATE (&spec->info), max_pitch, &self->format,
|
|
&self->recommended_buffer_size, &min_size);
|
|
if (result != MLResult_Ok) {
|
|
GST_ERROR_OBJECT (self, "Failed to get output stream defaults: %d", result);
|
|
return FALSE;
|
|
}
|
|
|
|
if (!self->app) {
|
|
gst_element_post_message (GST_ELEMENT (self),
|
|
gst_message_new_element (GST_OBJECT (self),
|
|
gst_structure_new_empty ("gst.mlaudiosink.need-app")));
|
|
}
|
|
|
|
self->wrapper = gst_ml_audio_wrapper_new (self->app);
|
|
self->has_buffer = FALSE;
|
|
self->stopped = FALSE;
|
|
self->paused = FALSE;
|
|
|
|
g_mutex_lock (&active_sinks_mutex);
|
|
g_hash_table_add (active_sinks, self);
|
|
g_mutex_unlock (&active_sinks_mutex);
|
|
|
|
/* createAudioNode() and createSoundWithOutputStream() must both be called in
|
|
* application's main thread, and in a single main loop iteration. */
|
|
if (!gst_ml_audio_wrapper_invoke_sync (self->wrapper, create_sound_cb, self))
|
|
return FALSE;
|
|
|
|
return TRUE;
|
|
}
|
|
|
|
static void
|
|
release_current_buffer (GstMLAudioSink * self)
|
|
{
|
|
if (self->has_buffer) {
|
|
memset (self->buffer.ptr + self->buffer_offset, 0,
|
|
self->buffer.size - self->buffer_offset);
|
|
gst_ml_audio_wrapper_release_buffer (self->wrapper);
|
|
self->has_buffer = false;
|
|
}
|
|
}
|
|
|
|
static gboolean
|
|
gst_ml_audio_sink_unprepare (GstAudioSink * sink)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (sink);
|
|
|
|
g_mutex_lock (&active_sinks_mutex);
|
|
g_hash_table_remove (active_sinks, self);
|
|
release_current_buffer (self);
|
|
g_clear_pointer (&self->wrapper, gst_ml_audio_wrapper_free);
|
|
g_mutex_unlock (&active_sinks_mutex);
|
|
|
|
return TRUE;
|
|
}
|
|
|
|
static gboolean
|
|
gst_ml_audio_sink_close (GstAudioSink * sink)
|
|
{
|
|
/* Nothing to do in open/close */
|
|
return TRUE;
|
|
}
|
|
|
|
static gint
|
|
gst_ml_audio_sink_write (GstAudioSink * sink, gpointer data, guint length)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (sink);
|
|
guint8 *input = data;
|
|
gint written = 0;
|
|
|
|
g_mutex_lock (&self->mutex);
|
|
|
|
while (length > 0) {
|
|
MLResult result;
|
|
guint to_write;
|
|
|
|
if (!wait_for_buffer (self)) {
|
|
written = -1;
|
|
break;
|
|
}
|
|
|
|
if (self->stopped) {
|
|
/* Pretend we have written the full buffer (drop data) and return
|
|
* immediately. */
|
|
release_current_buffer (self);
|
|
gst_ml_audio_wrapper_stop_sound (self->wrapper);
|
|
written = length;
|
|
break;
|
|
}
|
|
|
|
to_write = MIN (length, self->buffer.size - self->buffer_offset);
|
|
memcpy (self->buffer.ptr + self->buffer_offset, input + written, to_write);
|
|
self->buffer_offset += to_write;
|
|
if (self->buffer_offset == self->buffer.size) {
|
|
result = gst_ml_audio_wrapper_release_buffer (self->wrapper);
|
|
if (result != MLResult_Ok) {
|
|
GST_ERROR_OBJECT (self, "Failed to release buffer: %d", result);
|
|
written = -1;
|
|
break;
|
|
}
|
|
self->has_buffer = FALSE;
|
|
}
|
|
|
|
length -= to_write;
|
|
written += to_write;
|
|
}
|
|
|
|
if (self->paused) {
|
|
/* Pause was requested and we finished writing current buffer.
|
|
* See https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/665
|
|
*/
|
|
gst_ml_audio_wrapper_pause_sound (self->wrapper);
|
|
}
|
|
|
|
g_mutex_unlock (&self->mutex);
|
|
|
|
return written;
|
|
}
|
|
|
|
static guint
|
|
gst_ml_audio_sink_delay (GstAudioSink * sink)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (sink);
|
|
MLResult result;
|
|
float latency_ms;
|
|
|
|
result = gst_ml_audio_wrapper_get_latency (self->wrapper, &latency_ms);
|
|
if (result != MLResult_Ok) {
|
|
GST_ERROR_OBJECT (self, "Failed to get latency: %d", result);
|
|
return 0;
|
|
}
|
|
|
|
return latency_ms * self->format.samples_per_second / 1000;
|
|
}
|
|
|
|
static void
|
|
gst_ml_audio_sink_pause (GstAudioSink * sink)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (sink);
|
|
|
|
g_mutex_lock (&self->mutex);
|
|
self->paused = TRUE;
|
|
g_cond_signal (&self->cond);
|
|
g_mutex_unlock (&self->mutex);
|
|
}
|
|
|
|
static void
|
|
gst_ml_audio_sink_resume (GstAudioSink * sink)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (sink);
|
|
|
|
g_mutex_lock (&self->mutex);
|
|
self->paused = FALSE;
|
|
gst_ml_audio_wrapper_resume_sound (self->wrapper);
|
|
g_mutex_unlock (&self->mutex);
|
|
}
|
|
|
|
static void
|
|
gst_ml_audio_sink_stop (GstAudioSink * sink)
|
|
{
|
|
GstMLAudioSink *self = GST_ML_AUDIO_SINK (sink);
|
|
|
|
g_mutex_lock (&self->mutex);
|
|
self->stopped = TRUE;
|
|
g_cond_signal (&self->cond);
|
|
g_mutex_unlock (&self->mutex);
|
|
}
|
|
|
|
static void
|
|
gst_ml_audio_sink_class_init (GstMLAudioSinkClass * klass)
|
|
{
|
|
GObjectClass *gobject_class = G_OBJECT_CLASS (klass);
|
|
GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
|
|
GstBaseSinkClass *basesink_class = GST_BASE_SINK_CLASS (klass);
|
|
GstAudioSinkClass *audiosink_class = GST_AUDIO_SINK_CLASS (klass);
|
|
|
|
active_sinks = g_hash_table_new (NULL, NULL);
|
|
g_mutex_init (&active_sinks_mutex);
|
|
|
|
gobject_class->dispose = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_dispose);
|
|
gobject_class->set_property =
|
|
GST_DEBUG_FUNCPTR (gst_ml_audio_sink_set_property);
|
|
gobject_class->get_property =
|
|
GST_DEBUG_FUNCPTR (gst_ml_audio_sink_get_property);
|
|
|
|
g_object_class_install_property (gobject_class,
|
|
PROP_AUDIO_NODE, g_param_spec_pointer ("audio-node",
|
|
"A pointer to a lumin::AudioNode object",
|
|
"Enable spatial sound", G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
|
|
|
|
g_object_class_install_property (gobject_class,
|
|
PROP_APP, g_param_spec_pointer ("app",
|
|
"A pointer to a lumin::BaseApp object",
|
|
"Enable spatial sound", G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
|
|
|
|
gst_element_class_set_static_metadata (element_class,
|
|
"Magic Leap Audio Sink",
|
|
"Sink/Audio", "Plays audio on a Magic Leap device",
|
|
"Xavier Claessens <xavier.claessens@collabora.com>");
|
|
|
|
gst_element_class_add_static_pad_template (element_class, &sink_template);
|
|
|
|
basesink_class->get_caps = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_getcaps);
|
|
|
|
audiosink_class->open = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_open);
|
|
audiosink_class->prepare = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_prepare);
|
|
audiosink_class->unprepare = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_unprepare);
|
|
audiosink_class->close = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_close);
|
|
audiosink_class->write = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_write);
|
|
audiosink_class->delay = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_delay);
|
|
audiosink_class->pause = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_pause);
|
|
audiosink_class->resume = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_resume);
|
|
audiosink_class->stop = GST_DEBUG_FUNCPTR (gst_ml_audio_sink_stop);
|
|
}
|