59e11934d8
Affects the Math class, a good chunk of the audio code, and a lot of other miscellaneous classes, too. - `var2str` -> `var_to_str` - `str2var` -> `str_to_var` - `bytes2var` -> `bytes_to_var` - `bytes2var_with_objects` -> `bytes_to_var_with_objects` - `var2bytes` -> `var_to_bytes` - `var2bytes_with_objects` -> `var_to_bytes_with_objects` - `linear2db` -> `linear_to_db` - `db2linear` -> `db_to_linear` - `deg2rad` -> `deg_to_rad` - `rad2deg` -> `rad_to_deg` - `dict2inst` -> `dict_to_inst` - `inst2dict` -> `inst_to_dict`
1885 lines
58 KiB
C++
1885 lines
58 KiB
C++
/*************************************************************************/
|
|
/* audio_server.cpp */
|
|
/*************************************************************************/
|
|
/* This file is part of: */
|
|
/* GODOT ENGINE */
|
|
/* https://godotengine.org */
|
|
/*************************************************************************/
|
|
/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */
|
|
/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */
|
|
/* */
|
|
/* Permission is hereby granted, free of charge, to any person obtaining */
|
|
/* a copy of this software and associated documentation files (the */
|
|
/* "Software"), to deal in the Software without restriction, including */
|
|
/* without limitation the rights to use, copy, modify, merge, publish, */
|
|
/* distribute, sublicense, and/or sell copies of the Software, and to */
|
|
/* permit persons to whom the Software is furnished to do so, subject to */
|
|
/* the following conditions: */
|
|
/* */
|
|
/* The above copyright notice and this permission notice shall be */
|
|
/* included in all copies or substantial portions of the Software. */
|
|
/* */
|
|
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
|
|
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
|
|
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
|
|
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
|
|
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
|
|
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
|
|
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
|
|
/*************************************************************************/
|
|
|
|
#include "audio_server.h"
|
|
|
|
#include "core/config/project_settings.h"
|
|
#include "core/debugger/engine_debugger.h"
|
|
#include "core/error/error_macros.h"
|
|
#include "core/io/file_access.h"
|
|
#include "core/io/resource_loader.h"
|
|
#include "core/math/audio_frame.h"
|
|
#include "core/os/os.h"
|
|
#include "core/string/string_name.h"
|
|
#include "core/templates/pair.h"
|
|
#include "scene/resources/audio_stream_wav.h"
|
|
#include "servers/audio/audio_driver_dummy.h"
|
|
#include "servers/audio/effects/audio_effect_compressor.h"
|
|
|
|
#include <cstring>
|
|
|
|
#ifdef TOOLS_ENABLED
|
|
#define MARK_EDITED set_edited(true);
|
|
#else
|
|
#define MARK_EDITED
|
|
#endif
|
|
|
|
AudioDriver *AudioDriver::singleton = nullptr;
|
|
AudioDriver *AudioDriver::get_singleton() {
|
|
return singleton;
|
|
}
|
|
|
|
void AudioDriver::set_singleton() {
|
|
singleton = this;
|
|
}
|
|
|
|
void AudioDriver::audio_server_process(int p_frames, int32_t *p_buffer, bool p_update_mix_time) {
|
|
if (p_update_mix_time) {
|
|
update_mix_time(p_frames);
|
|
}
|
|
|
|
if (AudioServer::get_singleton()) {
|
|
AudioServer::get_singleton()->_driver_process(p_frames, p_buffer);
|
|
}
|
|
}
|
|
|
|
void AudioDriver::update_mix_time(int p_frames) {
|
|
_last_mix_frames = p_frames;
|
|
if (OS::get_singleton()) {
|
|
_last_mix_time = OS::get_singleton()->get_ticks_usec();
|
|
}
|
|
}
|
|
|
|
double AudioDriver::get_time_since_last_mix() {
|
|
lock();
|
|
uint64_t last_mix_time = _last_mix_time;
|
|
unlock();
|
|
return (OS::get_singleton()->get_ticks_usec() - last_mix_time) / 1000000.0;
|
|
}
|
|
|
|
double AudioDriver::get_time_to_next_mix() {
|
|
lock();
|
|
uint64_t last_mix_time = _last_mix_time;
|
|
uint64_t last_mix_frames = _last_mix_frames;
|
|
unlock();
|
|
double total = (OS::get_singleton()->get_ticks_usec() - last_mix_time) / 1000000.0;
|
|
double mix_buffer = last_mix_frames / (double)get_mix_rate();
|
|
return mix_buffer - total;
|
|
}
|
|
|
|
void AudioDriver::input_buffer_init(int driver_buffer_frames) {
|
|
const int input_buffer_channels = 2;
|
|
input_buffer.resize(driver_buffer_frames * input_buffer_channels * 4);
|
|
input_position = 0;
|
|
input_size = 0;
|
|
}
|
|
|
|
void AudioDriver::input_buffer_write(int32_t sample) {
|
|
if ((int)input_position < input_buffer.size()) {
|
|
input_buffer.write[input_position++] = sample;
|
|
if ((int)input_position >= input_buffer.size()) {
|
|
input_position = 0;
|
|
}
|
|
if ((int)input_size < input_buffer.size()) {
|
|
input_size++;
|
|
}
|
|
} else {
|
|
WARN_PRINT("input_buffer_write: Invalid input_position=" + itos(input_position) + " input_buffer.size()=" + itos(input_buffer.size()));
|
|
}
|
|
}
|
|
|
|
AudioDriver::SpeakerMode AudioDriver::get_speaker_mode_by_total_channels(int p_channels) const {
|
|
switch (p_channels) {
|
|
case 4:
|
|
return SPEAKER_SURROUND_31;
|
|
case 6:
|
|
return SPEAKER_SURROUND_51;
|
|
case 8:
|
|
return SPEAKER_SURROUND_71;
|
|
}
|
|
|
|
// Default to STEREO
|
|
return SPEAKER_MODE_STEREO;
|
|
}
|
|
|
|
int AudioDriver::get_total_channels_by_speaker_mode(AudioDriver::SpeakerMode p_mode) const {
|
|
switch (p_mode) {
|
|
case SPEAKER_MODE_STEREO:
|
|
return 2;
|
|
case SPEAKER_SURROUND_31:
|
|
return 4;
|
|
case SPEAKER_SURROUND_51:
|
|
return 6;
|
|
case SPEAKER_SURROUND_71:
|
|
return 8;
|
|
}
|
|
|
|
ERR_FAIL_V(2);
|
|
}
|
|
|
|
PackedStringArray AudioDriver::get_device_list() {
|
|
PackedStringArray list;
|
|
|
|
list.push_back("Default");
|
|
|
|
return list;
|
|
}
|
|
|
|
String AudioDriver::get_device() {
|
|
return "Default";
|
|
}
|
|
|
|
PackedStringArray AudioDriver::capture_get_device_list() {
|
|
PackedStringArray list;
|
|
|
|
list.push_back("Default");
|
|
|
|
return list;
|
|
}
|
|
|
|
AudioDriverDummy AudioDriverManager::dummy_driver;
|
|
AudioDriver *AudioDriverManager::drivers[MAX_DRIVERS] = {
|
|
&AudioDriverManager::dummy_driver,
|
|
};
|
|
int AudioDriverManager::driver_count = 1;
|
|
|
|
void AudioDriverManager::add_driver(AudioDriver *p_driver) {
|
|
ERR_FAIL_COND(driver_count >= MAX_DRIVERS);
|
|
drivers[driver_count - 1] = p_driver;
|
|
|
|
// Last driver is always our dummy driver
|
|
drivers[driver_count++] = &AudioDriverManager::dummy_driver;
|
|
}
|
|
|
|
int AudioDriverManager::get_driver_count() {
|
|
return driver_count;
|
|
}
|
|
|
|
void AudioDriverManager::initialize(int p_driver) {
|
|
GLOBAL_DEF_RST("audio/driver/enable_input", false);
|
|
GLOBAL_DEF_RST("audio/driver/mix_rate", DEFAULT_MIX_RATE);
|
|
GLOBAL_DEF_RST("audio/driver/mix_rate.web", 0); // Safer default output_latency for web (use browser default).
|
|
GLOBAL_DEF_RST("audio/driver/output_latency", DEFAULT_OUTPUT_LATENCY);
|
|
GLOBAL_DEF_RST("audio/driver/output_latency.web", 50); // Safer default output_latency for web.
|
|
|
|
int failed_driver = -1;
|
|
|
|
// Check if there is a selected driver
|
|
if (p_driver >= 0 && p_driver < driver_count) {
|
|
if (drivers[p_driver]->init() == OK) {
|
|
drivers[p_driver]->set_singleton();
|
|
return;
|
|
} else {
|
|
failed_driver = p_driver;
|
|
}
|
|
}
|
|
|
|
// No selected driver, try them all in order
|
|
for (int i = 0; i < driver_count; i++) {
|
|
// Don't re-init the driver if it failed above
|
|
if (i == failed_driver) {
|
|
continue;
|
|
}
|
|
|
|
if (drivers[i]->init() == OK) {
|
|
drivers[i]->set_singleton();
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (driver_count > 1 && String(AudioDriver::get_singleton()->get_name()) == "Dummy") {
|
|
WARN_PRINT("All audio drivers failed, falling back to the dummy driver.");
|
|
}
|
|
}
|
|
|
|
AudioDriver *AudioDriverManager::get_driver(int p_driver) {
|
|
ERR_FAIL_INDEX_V(p_driver, driver_count, nullptr);
|
|
return drivers[p_driver];
|
|
}
|
|
|
|
//////////////////////////////////////////////
|
|
//////////////////////////////////////////////
|
|
//////////////////////////////////////////////
|
|
//////////////////////////////////////////////
|
|
|
|
void AudioServer::_driver_process(int p_frames, int32_t *p_buffer) {
|
|
mix_count++;
|
|
int todo = p_frames;
|
|
|
|
#ifdef DEBUG_ENABLED
|
|
uint64_t prof_ticks = OS::get_singleton()->get_ticks_usec();
|
|
#endif
|
|
|
|
if (channel_count != get_channel_count()) {
|
|
// Amount of channels changed due to a device change
|
|
// reinitialize the buses channels and buffers
|
|
init_channels_and_buffers();
|
|
}
|
|
|
|
ERR_FAIL_COND_MSG(buses.is_empty() && todo, "AudioServer bus count is less than 1.");
|
|
while (todo) {
|
|
if (to_mix == 0) {
|
|
_mix_step();
|
|
}
|
|
|
|
int to_copy = MIN(to_mix, todo);
|
|
|
|
Bus *master = buses[0];
|
|
|
|
int from = buffer_size - to_mix;
|
|
int from_buf = p_frames - todo;
|
|
|
|
//master master, send to output
|
|
int cs = master->channels.size();
|
|
for (int k = 0; k < cs; k++) {
|
|
if (master->channels[k].active) {
|
|
const AudioFrame *buf = master->channels[k].buffer.ptr();
|
|
|
|
for (int j = 0; j < to_copy; j++) {
|
|
float l = CLAMP(buf[from + j].l, -1.0, 1.0);
|
|
int32_t vl = l * ((1 << 20) - 1);
|
|
int32_t vl2 = (vl < 0 ? -1 : 1) * (ABS(vl) << 11);
|
|
p_buffer[(from_buf + j) * (cs * 2) + k * 2 + 0] = vl2;
|
|
|
|
float r = CLAMP(buf[from + j].r, -1.0, 1.0);
|
|
int32_t vr = r * ((1 << 20) - 1);
|
|
int32_t vr2 = (vr < 0 ? -1 : 1) * (ABS(vr) << 11);
|
|
p_buffer[(from_buf + j) * (cs * 2) + k * 2 + 1] = vr2;
|
|
}
|
|
|
|
} else {
|
|
for (int j = 0; j < to_copy; j++) {
|
|
p_buffer[(from_buf + j) * (cs * 2) + k * 2 + 0] = 0;
|
|
p_buffer[(from_buf + j) * (cs * 2) + k * 2 + 1] = 0;
|
|
}
|
|
}
|
|
}
|
|
|
|
todo -= to_copy;
|
|
to_mix -= to_copy;
|
|
}
|
|
|
|
#ifdef DEBUG_ENABLED
|
|
prof_time += OS::get_singleton()->get_ticks_usec() - prof_ticks;
|
|
#endif
|
|
}
|
|
|
|
void AudioServer::_mix_step() {
|
|
bool solo_mode = false;
|
|
|
|
for (int i = 0; i < buses.size(); i++) {
|
|
Bus *bus = buses[i];
|
|
bus->index_cache = i; //might be moved around by editor, so..
|
|
for (int k = 0; k < bus->channels.size(); k++) {
|
|
bus->channels.write[k].used = false;
|
|
}
|
|
|
|
if (bus->solo) {
|
|
//solo chain
|
|
solo_mode = true;
|
|
bus->soloed = true;
|
|
do {
|
|
if (bus != buses[0]) {
|
|
//everything has a send save for master bus
|
|
if (!bus_map.has(bus->send)) {
|
|
bus = buses[0]; //send to master
|
|
} else {
|
|
int prev_index_cache = bus->index_cache;
|
|
bus = bus_map[bus->send];
|
|
if (prev_index_cache >= bus->index_cache) { //invalid, send to master
|
|
bus = buses[0];
|
|
}
|
|
}
|
|
|
|
bus->soloed = true;
|
|
} else {
|
|
bus = nullptr;
|
|
}
|
|
|
|
} while (bus);
|
|
} else {
|
|
bus->soloed = false;
|
|
}
|
|
}
|
|
for (CallbackItem *ci : mix_callback_list) {
|
|
ci->callback(ci->userdata);
|
|
}
|
|
|
|
for (AudioStreamPlaybackListNode *playback : playback_list) {
|
|
// Paused streams are no-ops. Don't even mix audio from the stream playback.
|
|
if (playback->state.load() == AudioStreamPlaybackListNode::PAUSED) {
|
|
continue;
|
|
}
|
|
|
|
bool fading_out = playback->state.load() == AudioStreamPlaybackListNode::FADE_OUT_TO_DELETION || playback->state.load() == AudioStreamPlaybackListNode::FADE_OUT_TO_PAUSE;
|
|
|
|
AudioFrame *buf = mix_buffer.ptrw();
|
|
|
|
// Copy the lookeahead buffer into the mix buffer.
|
|
for (int i = 0; i < LOOKAHEAD_BUFFER_SIZE; i++) {
|
|
buf[i] = playback->lookahead[i];
|
|
}
|
|
|
|
// Mix the audio stream
|
|
unsigned int mixed_frames = playback->stream_playback->mix(&buf[LOOKAHEAD_BUFFER_SIZE], playback->pitch_scale.get(), buffer_size);
|
|
|
|
if (tag_used_audio_streams && playback->stream_playback->is_playing()) {
|
|
playback->stream_playback->tag_used_streams();
|
|
}
|
|
|
|
if (mixed_frames != buffer_size) {
|
|
// We know we have at least the size of our lookahead buffer for fade-out purposes.
|
|
|
|
float fadeout_base = 0.94;
|
|
float fadeout_coefficient = 1;
|
|
static_assert(LOOKAHEAD_BUFFER_SIZE == 64, "Update fadeout_base and comment here if you change LOOKAHEAD_BUFFER_SIZE.");
|
|
// 0.94 ^ 64 = 0.01906. There might still be a pop but it'll be way better than if we didn't do this.
|
|
for (unsigned int idx = mixed_frames; idx < buffer_size; idx++) {
|
|
fadeout_coefficient *= fadeout_base;
|
|
buf[idx] *= fadeout_coefficient;
|
|
}
|
|
AudioStreamPlaybackListNode::PlaybackState new_state;
|
|
new_state = AudioStreamPlaybackListNode::AWAITING_DELETION;
|
|
playback->state.store(new_state);
|
|
} else {
|
|
// Move the last little bit of what we just mixed into our lookahead buffer.
|
|
for (int i = 0; i < LOOKAHEAD_BUFFER_SIZE; i++) {
|
|
playback->lookahead[i] = buf[buffer_size + i];
|
|
}
|
|
}
|
|
|
|
AudioStreamPlaybackBusDetails *ptr = playback->bus_details.load();
|
|
ERR_FAIL_COND(ptr == nullptr);
|
|
// By putting null into the bus details pointers, we're taking ownership of their memory for the duration of this mix.
|
|
AudioStreamPlaybackBusDetails bus_details = *ptr;
|
|
|
|
// Mix to any active buses.
|
|
for (int idx = 0; idx < MAX_BUSES_PER_PLAYBACK; idx++) {
|
|
if (!bus_details.bus_active[idx]) {
|
|
continue;
|
|
}
|
|
int bus_idx = thread_find_bus_index(bus_details.bus[idx]);
|
|
|
|
int prev_bus_idx = -1;
|
|
for (int search_idx = 0; search_idx < MAX_BUSES_PER_PLAYBACK; search_idx++) {
|
|
if (!playback->prev_bus_details->bus_active[search_idx]) {
|
|
continue;
|
|
}
|
|
if (playback->prev_bus_details->bus[search_idx].hash() == bus_details.bus[idx].hash()) {
|
|
prev_bus_idx = search_idx;
|
|
}
|
|
}
|
|
|
|
for (int channel_idx = 0; channel_idx < channel_count; channel_idx++) {
|
|
AudioFrame *channel_buf = thread_get_channel_mix_buffer(bus_idx, channel_idx);
|
|
if (fading_out) {
|
|
bus_details.volume[idx][channel_idx] = AudioFrame(0, 0);
|
|
}
|
|
AudioFrame channel_vol = bus_details.volume[idx][channel_idx];
|
|
|
|
AudioFrame prev_channel_vol = AudioFrame(0, 0);
|
|
if (prev_bus_idx != -1) {
|
|
prev_channel_vol = playback->prev_bus_details->volume[prev_bus_idx][channel_idx];
|
|
}
|
|
_mix_step_for_channel(channel_buf, buf, prev_channel_vol, channel_vol, playback->attenuation_filter_cutoff_hz.get(), playback->highshelf_gain.get(), &playback->filter_process[channel_idx * 2], &playback->filter_process[channel_idx * 2 + 1]);
|
|
}
|
|
}
|
|
|
|
// Now go through and fade-out any buses that were being played to previously that we missed by going through current data.
|
|
for (int idx = 0; idx < MAX_BUSES_PER_PLAYBACK; idx++) {
|
|
if (!playback->prev_bus_details->bus_active[idx]) {
|
|
continue;
|
|
}
|
|
int bus_idx = thread_find_bus_index(playback->prev_bus_details->bus[idx]);
|
|
|
|
int current_bus_idx = -1;
|
|
for (int search_idx = 0; search_idx < MAX_BUSES_PER_PLAYBACK; search_idx++) {
|
|
if (bus_details.bus[search_idx] == playback->prev_bus_details->bus[idx]) {
|
|
current_bus_idx = search_idx;
|
|
}
|
|
}
|
|
if (current_bus_idx != -1) {
|
|
// If we found a corresponding bus in the current bus assignments, we've already mixed to this bus.
|
|
continue;
|
|
}
|
|
|
|
for (int channel_idx = 0; channel_idx < channel_count; channel_idx++) {
|
|
AudioFrame *channel_buf = thread_get_channel_mix_buffer(bus_idx, channel_idx);
|
|
AudioFrame prev_channel_vol = playback->prev_bus_details->volume[idx][channel_idx];
|
|
// Fade out to silence
|
|
_mix_step_for_channel(channel_buf, buf, prev_channel_vol, AudioFrame(0, 0), playback->attenuation_filter_cutoff_hz.get(), playback->highshelf_gain.get(), &playback->filter_process[channel_idx * 2], &playback->filter_process[channel_idx * 2 + 1]);
|
|
}
|
|
}
|
|
|
|
// Copy the bus details we mixed with to the previous bus details to maintain volume ramps.
|
|
std::copy(std::begin(bus_details.bus_active), std::end(bus_details.bus_active), std::begin(playback->prev_bus_details->bus_active));
|
|
std::copy(std::begin(bus_details.bus), std::end(bus_details.bus), std::begin(playback->prev_bus_details->bus));
|
|
for (int bus_idx = 0; bus_idx < MAX_BUSES_PER_PLAYBACK; bus_idx++) {
|
|
std::copy(std::begin(bus_details.volume[bus_idx]), std::end(bus_details.volume[bus_idx]), std::begin(playback->prev_bus_details->volume[bus_idx]));
|
|
}
|
|
|
|
switch (playback->state.load()) {
|
|
case AudioStreamPlaybackListNode::AWAITING_DELETION:
|
|
case AudioStreamPlaybackListNode::FADE_OUT_TO_DELETION:
|
|
playback_list.erase(playback, [](AudioStreamPlaybackListNode *p) {
|
|
if (p->prev_bus_details) {
|
|
delete p->prev_bus_details;
|
|
}
|
|
if (p->bus_details) {
|
|
delete p->bus_details;
|
|
}
|
|
p->stream_playback.unref();
|
|
delete p;
|
|
});
|
|
break;
|
|
case AudioStreamPlaybackListNode::FADE_OUT_TO_PAUSE: {
|
|
// Pause the stream.
|
|
AudioStreamPlaybackListNode::PlaybackState old_state, new_state;
|
|
do {
|
|
old_state = playback->state.load();
|
|
new_state = AudioStreamPlaybackListNode::PAUSED;
|
|
} while (!playback->state.compare_exchange_strong(/* expected= */ old_state, new_state));
|
|
} break;
|
|
case AudioStreamPlaybackListNode::PLAYING:
|
|
case AudioStreamPlaybackListNode::PAUSED:
|
|
// No-op!
|
|
break;
|
|
}
|
|
}
|
|
|
|
for (int i = buses.size() - 1; i >= 0; i--) {
|
|
//go bus by bus
|
|
Bus *bus = buses[i];
|
|
|
|
for (int k = 0; k < bus->channels.size(); k++) {
|
|
if (bus->channels[k].active && !bus->channels[k].used) {
|
|
//buffer was not used, but it's still active, so it must be cleaned
|
|
AudioFrame *buf = bus->channels.write[k].buffer.ptrw();
|
|
|
|
for (uint32_t j = 0; j < buffer_size; j++) {
|
|
buf[j] = AudioFrame(0, 0);
|
|
}
|
|
}
|
|
}
|
|
|
|
//process effects
|
|
if (!bus->bypass) {
|
|
for (int j = 0; j < bus->effects.size(); j++) {
|
|
if (!bus->effects[j].enabled) {
|
|
continue;
|
|
}
|
|
|
|
#ifdef DEBUG_ENABLED
|
|
uint64_t ticks = OS::get_singleton()->get_ticks_usec();
|
|
#endif
|
|
|
|
for (int k = 0; k < bus->channels.size(); k++) {
|
|
if (!(bus->channels[k].active || bus->channels[k].effect_instances[j]->process_silence())) {
|
|
continue;
|
|
}
|
|
bus->channels.write[k].effect_instances.write[j]->process(bus->channels[k].buffer.ptr(), temp_buffer.write[k].ptrw(), buffer_size);
|
|
}
|
|
|
|
//swap buffers, so internal buffer always has the right data
|
|
for (int k = 0; k < bus->channels.size(); k++) {
|
|
if (!(buses[i]->channels[k].active || bus->channels[k].effect_instances[j]->process_silence())) {
|
|
continue;
|
|
}
|
|
SWAP(bus->channels.write[k].buffer, temp_buffer.write[k]);
|
|
}
|
|
|
|
#ifdef DEBUG_ENABLED
|
|
bus->effects.write[j].prof_time += OS::get_singleton()->get_ticks_usec() - ticks;
|
|
#endif
|
|
}
|
|
}
|
|
|
|
//process send
|
|
|
|
Bus *send = nullptr;
|
|
|
|
if (i > 0) {
|
|
//everything has a send save for master bus
|
|
if (!bus_map.has(bus->send)) {
|
|
send = buses[0];
|
|
} else {
|
|
send = bus_map[bus->send];
|
|
if (send->index_cache >= bus->index_cache) { //invalid, send to master
|
|
send = buses[0];
|
|
}
|
|
}
|
|
}
|
|
|
|
for (int k = 0; k < bus->channels.size(); k++) {
|
|
if (!bus->channels[k].active) {
|
|
bus->channels.write[k].peak_volume = AudioFrame(AUDIO_MIN_PEAK_DB, AUDIO_MIN_PEAK_DB);
|
|
continue;
|
|
}
|
|
|
|
AudioFrame *buf = bus->channels.write[k].buffer.ptrw();
|
|
|
|
AudioFrame peak = AudioFrame(0, 0);
|
|
|
|
float volume = Math::db_to_linear(bus->volume_db);
|
|
|
|
if (solo_mode) {
|
|
if (!bus->soloed) {
|
|
volume = 0.0;
|
|
}
|
|
} else {
|
|
if (bus->mute) {
|
|
volume = 0.0;
|
|
}
|
|
}
|
|
|
|
//apply volume and compute peak
|
|
for (uint32_t j = 0; j < buffer_size; j++) {
|
|
buf[j] *= volume;
|
|
|
|
float l = ABS(buf[j].l);
|
|
if (l > peak.l) {
|
|
peak.l = l;
|
|
}
|
|
float r = ABS(buf[j].r);
|
|
if (r > peak.r) {
|
|
peak.r = r;
|
|
}
|
|
}
|
|
|
|
bus->channels.write[k].peak_volume = AudioFrame(Math::linear_to_db(peak.l + AUDIO_PEAK_OFFSET), Math::linear_to_db(peak.r + AUDIO_PEAK_OFFSET));
|
|
|
|
if (!bus->channels[k].used) {
|
|
//see if any audio is contained, because channel was not used
|
|
|
|
if (MAX(peak.r, peak.l) > Math::db_to_linear(channel_disable_threshold_db)) {
|
|
bus->channels.write[k].last_mix_with_audio = mix_frames;
|
|
} else if (mix_frames - bus->channels[k].last_mix_with_audio > channel_disable_frames) {
|
|
bus->channels.write[k].active = false;
|
|
continue; //went inactive, don't mix.
|
|
}
|
|
}
|
|
|
|
if (send) {
|
|
//if not master bus, send
|
|
AudioFrame *target_buf = thread_get_channel_mix_buffer(send->index_cache, k);
|
|
|
|
for (uint32_t j = 0; j < buffer_size; j++) {
|
|
target_buf[j] += buf[j];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
mix_frames += buffer_size;
|
|
to_mix = buffer_size;
|
|
}
|
|
|
|
void AudioServer::_mix_step_for_channel(AudioFrame *p_out_buf, AudioFrame *p_source_buf, AudioFrame p_vol_start, AudioFrame p_vol_final, float p_attenuation_filter_cutoff_hz, float p_highshelf_gain, AudioFilterSW::Processor *p_processor_l, AudioFilterSW::Processor *p_processor_r) {
|
|
if (p_highshelf_gain != 0) {
|
|
AudioFilterSW filter;
|
|
filter.set_mode(AudioFilterSW::HIGHSHELF);
|
|
filter.set_sampling_rate(AudioServer::get_singleton()->get_mix_rate());
|
|
filter.set_cutoff(p_attenuation_filter_cutoff_hz);
|
|
filter.set_resonance(1);
|
|
filter.set_stages(1);
|
|
filter.set_gain(p_highshelf_gain);
|
|
|
|
ERR_FAIL_COND(p_processor_l == nullptr);
|
|
ERR_FAIL_COND(p_processor_r == nullptr);
|
|
|
|
bool is_just_started = p_vol_start.l == 0 && p_vol_start.r == 0;
|
|
p_processor_l->set_filter(&filter, /* clear_history= */ is_just_started);
|
|
p_processor_l->update_coeffs(buffer_size);
|
|
p_processor_r->set_filter(&filter, /* clear_history= */ is_just_started);
|
|
p_processor_r->update_coeffs(buffer_size);
|
|
|
|
for (unsigned int frame_idx = 0; frame_idx < buffer_size; frame_idx++) {
|
|
// Make this buffer size invariant if buffer_size ever becomes a project setting.
|
|
float lerp_param = (float)frame_idx / buffer_size;
|
|
AudioFrame vol = p_vol_final * lerp_param + (1 - lerp_param) * p_vol_start;
|
|
AudioFrame mixed = vol * p_source_buf[frame_idx];
|
|
p_processor_l->process_one_interp(mixed.l);
|
|
p_processor_r->process_one_interp(mixed.r);
|
|
p_out_buf[frame_idx] += mixed;
|
|
}
|
|
|
|
} else {
|
|
for (unsigned int frame_idx = 0; frame_idx < buffer_size; frame_idx++) {
|
|
// Make this buffer size invariant if buffer_size ever becomes a project setting.
|
|
float lerp_param = (float)frame_idx / buffer_size;
|
|
p_out_buf[frame_idx] += (p_vol_final * lerp_param + (1 - lerp_param) * p_vol_start) * p_source_buf[frame_idx];
|
|
}
|
|
}
|
|
}
|
|
|
|
AudioServer::AudioStreamPlaybackListNode *AudioServer::_find_playback_list_node(Ref<AudioStreamPlayback> p_playback) {
|
|
for (AudioStreamPlaybackListNode *playback_list_node : playback_list) {
|
|
if (playback_list_node->stream_playback == p_playback) {
|
|
return playback_list_node;
|
|
}
|
|
}
|
|
return nullptr;
|
|
}
|
|
|
|
bool AudioServer::thread_has_channel_mix_buffer(int p_bus, int p_buffer) const {
|
|
if (p_bus < 0 || p_bus >= buses.size()) {
|
|
return false;
|
|
}
|
|
if (p_buffer < 0 || p_buffer >= buses[p_bus]->channels.size()) {
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
AudioFrame *AudioServer::thread_get_channel_mix_buffer(int p_bus, int p_buffer) {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), nullptr);
|
|
ERR_FAIL_INDEX_V(p_buffer, buses[p_bus]->channels.size(), nullptr);
|
|
|
|
AudioFrame *data = buses.write[p_bus]->channels.write[p_buffer].buffer.ptrw();
|
|
|
|
if (!buses[p_bus]->channels[p_buffer].used) {
|
|
buses.write[p_bus]->channels.write[p_buffer].used = true;
|
|
buses.write[p_bus]->channels.write[p_buffer].active = true;
|
|
buses.write[p_bus]->channels.write[p_buffer].last_mix_with_audio = mix_frames;
|
|
for (uint32_t i = 0; i < buffer_size; i++) {
|
|
data[i] = AudioFrame(0, 0);
|
|
}
|
|
}
|
|
|
|
return data;
|
|
}
|
|
|
|
int AudioServer::thread_get_mix_buffer_size() const {
|
|
return buffer_size;
|
|
}
|
|
|
|
int AudioServer::thread_find_bus_index(const StringName &p_name) {
|
|
if (bus_map.has(p_name)) {
|
|
return bus_map[p_name]->index_cache;
|
|
} else {
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
void AudioServer::set_bus_count(int p_count) {
|
|
ERR_FAIL_COND(p_count < 1);
|
|
ERR_FAIL_INDEX(p_count, 256);
|
|
|
|
MARK_EDITED
|
|
|
|
lock();
|
|
int cb = buses.size();
|
|
|
|
if (p_count < buses.size()) {
|
|
for (int i = p_count; i < buses.size(); i++) {
|
|
bus_map.erase(buses[i]->name);
|
|
memdelete(buses[i]);
|
|
}
|
|
}
|
|
|
|
buses.resize(p_count);
|
|
|
|
for (int i = cb; i < buses.size(); i++) {
|
|
String attempt = "New Bus";
|
|
int attempts = 1;
|
|
while (true) {
|
|
bool name_free = true;
|
|
for (int j = 0; j < i; j++) {
|
|
if (buses[j]->name == attempt) {
|
|
name_free = false;
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (!name_free) {
|
|
attempts++;
|
|
attempt = "New Bus " + itos(attempts);
|
|
} else {
|
|
break;
|
|
}
|
|
}
|
|
|
|
buses.write[i] = memnew(Bus);
|
|
buses.write[i]->channels.resize(channel_count);
|
|
for (int j = 0; j < channel_count; j++) {
|
|
buses.write[i]->channels.write[j].buffer.resize(buffer_size);
|
|
}
|
|
buses[i]->name = attempt;
|
|
buses[i]->solo = false;
|
|
buses[i]->mute = false;
|
|
buses[i]->bypass = false;
|
|
buses[i]->volume_db = 0;
|
|
if (i > 0) {
|
|
buses[i]->send = "Master";
|
|
}
|
|
|
|
bus_map[attempt] = buses[i];
|
|
}
|
|
|
|
unlock();
|
|
|
|
emit_signal(SNAME("bus_layout_changed"));
|
|
}
|
|
|
|
void AudioServer::remove_bus(int p_index) {
|
|
ERR_FAIL_INDEX(p_index, buses.size());
|
|
ERR_FAIL_COND(p_index == 0);
|
|
|
|
MARK_EDITED
|
|
|
|
lock();
|
|
bus_map.erase(buses[p_index]->name);
|
|
memdelete(buses[p_index]);
|
|
buses.remove_at(p_index);
|
|
unlock();
|
|
|
|
emit_signal(SNAME("bus_layout_changed"));
|
|
}
|
|
|
|
void AudioServer::add_bus(int p_at_pos) {
|
|
MARK_EDITED
|
|
|
|
if (p_at_pos >= buses.size()) {
|
|
p_at_pos = -1;
|
|
} else if (p_at_pos == 0) {
|
|
if (buses.size() > 1) {
|
|
p_at_pos = 1;
|
|
} else {
|
|
p_at_pos = -1;
|
|
}
|
|
}
|
|
|
|
String attempt = "New Bus";
|
|
int attempts = 1;
|
|
while (true) {
|
|
bool name_free = true;
|
|
for (int j = 0; j < buses.size(); j++) {
|
|
if (buses[j]->name == attempt) {
|
|
name_free = false;
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (!name_free) {
|
|
attempts++;
|
|
attempt = "New Bus " + itos(attempts);
|
|
} else {
|
|
break;
|
|
}
|
|
}
|
|
|
|
Bus *bus = memnew(Bus);
|
|
bus->channels.resize(channel_count);
|
|
for (int j = 0; j < channel_count; j++) {
|
|
bus->channels.write[j].buffer.resize(buffer_size);
|
|
}
|
|
bus->name = attempt;
|
|
bus->solo = false;
|
|
bus->mute = false;
|
|
bus->bypass = false;
|
|
bus->volume_db = 0;
|
|
|
|
bus_map[attempt] = bus;
|
|
|
|
if (p_at_pos == -1) {
|
|
buses.push_back(bus);
|
|
} else {
|
|
buses.insert(p_at_pos, bus);
|
|
}
|
|
|
|
emit_signal(SNAME("bus_layout_changed"));
|
|
}
|
|
|
|
void AudioServer::move_bus(int p_bus, int p_to_pos) {
|
|
ERR_FAIL_COND(p_bus < 1 || p_bus >= buses.size());
|
|
ERR_FAIL_COND(p_to_pos != -1 && (p_to_pos < 1 || p_to_pos > buses.size()));
|
|
|
|
MARK_EDITED
|
|
|
|
if (p_bus == p_to_pos) {
|
|
return;
|
|
}
|
|
|
|
Bus *bus = buses[p_bus];
|
|
buses.remove_at(p_bus);
|
|
|
|
if (p_to_pos == -1) {
|
|
buses.push_back(bus);
|
|
} else if (p_to_pos < p_bus) {
|
|
buses.insert(p_to_pos, bus);
|
|
} else {
|
|
buses.insert(p_to_pos - 1, bus);
|
|
}
|
|
|
|
emit_signal(SNAME("bus_layout_changed"));
|
|
}
|
|
|
|
int AudioServer::get_bus_count() const {
|
|
return buses.size();
|
|
}
|
|
|
|
void AudioServer::set_bus_name(int p_bus, const String &p_name) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
if (p_bus == 0 && p_name != "Master") {
|
|
return; //bus 0 is always master
|
|
}
|
|
|
|
MARK_EDITED
|
|
|
|
lock();
|
|
|
|
if (buses[p_bus]->name == p_name) {
|
|
unlock();
|
|
return;
|
|
}
|
|
|
|
String attempt = p_name;
|
|
int attempts = 1;
|
|
|
|
while (true) {
|
|
bool name_free = true;
|
|
for (int i = 0; i < buses.size(); i++) {
|
|
if (buses[i]->name == attempt) {
|
|
name_free = false;
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (name_free) {
|
|
break;
|
|
}
|
|
|
|
attempts++;
|
|
attempt = p_name + " " + itos(attempts);
|
|
}
|
|
bus_map.erase(buses[p_bus]->name);
|
|
buses[p_bus]->name = attempt;
|
|
bus_map[attempt] = buses[p_bus];
|
|
unlock();
|
|
|
|
emit_signal(SNAME("bus_layout_changed"));
|
|
}
|
|
|
|
String AudioServer::get_bus_name(int p_bus) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), String());
|
|
return buses[p_bus]->name;
|
|
}
|
|
|
|
int AudioServer::get_bus_index(const StringName &p_bus_name) const {
|
|
for (int i = 0; i < buses.size(); ++i) {
|
|
if (buses[i]->name == p_bus_name) {
|
|
return i;
|
|
}
|
|
}
|
|
return -1;
|
|
}
|
|
|
|
void AudioServer::set_bus_volume_db(int p_bus, float p_volume_db) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
|
|
MARK_EDITED
|
|
|
|
buses[p_bus]->volume_db = p_volume_db;
|
|
}
|
|
|
|
float AudioServer::get_bus_volume_db(int p_bus) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), 0);
|
|
return buses[p_bus]->volume_db;
|
|
}
|
|
|
|
int AudioServer::get_bus_channels(int p_bus) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), 0);
|
|
return buses[p_bus]->channels.size();
|
|
}
|
|
|
|
void AudioServer::set_bus_send(int p_bus, const StringName &p_send) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
|
|
MARK_EDITED
|
|
|
|
buses[p_bus]->send = p_send;
|
|
}
|
|
|
|
StringName AudioServer::get_bus_send(int p_bus) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), StringName());
|
|
return buses[p_bus]->send;
|
|
}
|
|
|
|
void AudioServer::set_bus_solo(int p_bus, bool p_enable) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
|
|
MARK_EDITED
|
|
|
|
buses[p_bus]->solo = p_enable;
|
|
}
|
|
|
|
bool AudioServer::is_bus_solo(int p_bus) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), false);
|
|
|
|
return buses[p_bus]->solo;
|
|
}
|
|
|
|
void AudioServer::set_bus_mute(int p_bus, bool p_enable) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
|
|
MARK_EDITED
|
|
|
|
buses[p_bus]->mute = p_enable;
|
|
}
|
|
|
|
bool AudioServer::is_bus_mute(int p_bus) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), false);
|
|
|
|
return buses[p_bus]->mute;
|
|
}
|
|
|
|
void AudioServer::set_bus_bypass_effects(int p_bus, bool p_enable) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
|
|
MARK_EDITED
|
|
|
|
buses[p_bus]->bypass = p_enable;
|
|
}
|
|
|
|
bool AudioServer::is_bus_bypassing_effects(int p_bus) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), false);
|
|
|
|
return buses[p_bus]->bypass;
|
|
}
|
|
|
|
void AudioServer::_update_bus_effects(int p_bus) {
|
|
for (int i = 0; i < buses[p_bus]->channels.size(); i++) {
|
|
buses.write[p_bus]->channels.write[i].effect_instances.resize(buses[p_bus]->effects.size());
|
|
for (int j = 0; j < buses[p_bus]->effects.size(); j++) {
|
|
Ref<AudioEffectInstance> fx = buses.write[p_bus]->effects.write[j].effect->instantiate();
|
|
if (Object::cast_to<AudioEffectCompressorInstance>(*fx)) {
|
|
Object::cast_to<AudioEffectCompressorInstance>(*fx)->set_current_channel(i);
|
|
}
|
|
buses.write[p_bus]->channels.write[i].effect_instances.write[j] = fx;
|
|
}
|
|
}
|
|
}
|
|
|
|
void AudioServer::add_bus_effect(int p_bus, const Ref<AudioEffect> &p_effect, int p_at_pos) {
|
|
ERR_FAIL_COND(p_effect.is_null());
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
|
|
MARK_EDITED
|
|
|
|
lock();
|
|
|
|
Bus::Effect fx;
|
|
fx.effect = p_effect;
|
|
//fx.instance=p_effect->instantiate();
|
|
fx.enabled = true;
|
|
#ifdef DEBUG_ENABLED
|
|
fx.prof_time = 0;
|
|
#endif
|
|
|
|
if (p_at_pos >= buses[p_bus]->effects.size() || p_at_pos < 0) {
|
|
buses[p_bus]->effects.push_back(fx);
|
|
} else {
|
|
buses[p_bus]->effects.insert(p_at_pos, fx);
|
|
}
|
|
|
|
_update_bus_effects(p_bus);
|
|
|
|
unlock();
|
|
}
|
|
|
|
void AudioServer::remove_bus_effect(int p_bus, int p_effect) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
|
|
MARK_EDITED
|
|
|
|
lock();
|
|
|
|
buses[p_bus]->effects.remove_at(p_effect);
|
|
_update_bus_effects(p_bus);
|
|
|
|
unlock();
|
|
}
|
|
|
|
int AudioServer::get_bus_effect_count(int p_bus) {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), 0);
|
|
|
|
return buses[p_bus]->effects.size();
|
|
}
|
|
|
|
Ref<AudioEffectInstance> AudioServer::get_bus_effect_instance(int p_bus, int p_effect, int p_channel) {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), Ref<AudioEffectInstance>());
|
|
ERR_FAIL_INDEX_V(p_effect, buses[p_bus]->effects.size(), Ref<AudioEffectInstance>());
|
|
ERR_FAIL_INDEX_V(p_channel, buses[p_bus]->channels.size(), Ref<AudioEffectInstance>());
|
|
|
|
return buses[p_bus]->channels[p_channel].effect_instances[p_effect];
|
|
}
|
|
|
|
Ref<AudioEffect> AudioServer::get_bus_effect(int p_bus, int p_effect) {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), Ref<AudioEffect>());
|
|
ERR_FAIL_INDEX_V(p_effect, buses[p_bus]->effects.size(), Ref<AudioEffect>());
|
|
|
|
return buses[p_bus]->effects[p_effect].effect;
|
|
}
|
|
|
|
void AudioServer::swap_bus_effects(int p_bus, int p_effect, int p_by_effect) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
ERR_FAIL_INDEX(p_effect, buses[p_bus]->effects.size());
|
|
ERR_FAIL_INDEX(p_by_effect, buses[p_bus]->effects.size());
|
|
|
|
MARK_EDITED
|
|
|
|
lock();
|
|
SWAP(buses.write[p_bus]->effects.write[p_effect], buses.write[p_bus]->effects.write[p_by_effect]);
|
|
_update_bus_effects(p_bus);
|
|
unlock();
|
|
}
|
|
|
|
void AudioServer::set_bus_effect_enabled(int p_bus, int p_effect, bool p_enabled) {
|
|
ERR_FAIL_INDEX(p_bus, buses.size());
|
|
ERR_FAIL_INDEX(p_effect, buses[p_bus]->effects.size());
|
|
|
|
MARK_EDITED
|
|
|
|
buses.write[p_bus]->effects.write[p_effect].enabled = p_enabled;
|
|
}
|
|
|
|
bool AudioServer::is_bus_effect_enabled(int p_bus, int p_effect) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), false);
|
|
ERR_FAIL_INDEX_V(p_effect, buses[p_bus]->effects.size(), false);
|
|
return buses[p_bus]->effects[p_effect].enabled;
|
|
}
|
|
|
|
float AudioServer::get_bus_peak_volume_left_db(int p_bus, int p_channel) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), 0);
|
|
ERR_FAIL_INDEX_V(p_channel, buses[p_bus]->channels.size(), 0);
|
|
|
|
return buses[p_bus]->channels[p_channel].peak_volume.l;
|
|
}
|
|
|
|
float AudioServer::get_bus_peak_volume_right_db(int p_bus, int p_channel) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), 0);
|
|
ERR_FAIL_INDEX_V(p_channel, buses[p_bus]->channels.size(), 0);
|
|
|
|
return buses[p_bus]->channels[p_channel].peak_volume.r;
|
|
}
|
|
|
|
bool AudioServer::is_bus_channel_active(int p_bus, int p_channel) const {
|
|
ERR_FAIL_INDEX_V(p_bus, buses.size(), false);
|
|
ERR_FAIL_INDEX_V(p_channel, buses[p_bus]->channels.size(), false);
|
|
|
|
return buses[p_bus]->channels[p_channel].active;
|
|
}
|
|
|
|
void AudioServer::set_playback_speed_scale(float p_scale) {
|
|
ERR_FAIL_COND(p_scale <= 0);
|
|
|
|
playback_speed_scale = p_scale;
|
|
}
|
|
|
|
float AudioServer::get_playback_speed_scale() const {
|
|
return playback_speed_scale;
|
|
}
|
|
|
|
void AudioServer::start_playback_stream(Ref<AudioStreamPlayback> p_playback, StringName p_bus, Vector<AudioFrame> p_volume_db_vector, float p_start_time, float p_pitch_scale) {
|
|
ERR_FAIL_COND(p_playback.is_null());
|
|
|
|
HashMap<StringName, Vector<AudioFrame>> map;
|
|
map[p_bus] = p_volume_db_vector;
|
|
|
|
start_playback_stream(p_playback, map, p_start_time, p_pitch_scale);
|
|
}
|
|
|
|
void AudioServer::start_playback_stream(Ref<AudioStreamPlayback> p_playback, HashMap<StringName, Vector<AudioFrame>> p_bus_volumes, float p_start_time, float p_pitch_scale, float p_highshelf_gain, float p_attenuation_cutoff_hz) {
|
|
ERR_FAIL_COND(p_playback.is_null());
|
|
|
|
AudioStreamPlaybackListNode *playback_node = new AudioStreamPlaybackListNode();
|
|
playback_node->stream_playback = p_playback;
|
|
playback_node->stream_playback->start(p_start_time);
|
|
|
|
AudioStreamPlaybackBusDetails *new_bus_details = new AudioStreamPlaybackBusDetails();
|
|
int idx = 0;
|
|
for (KeyValue<StringName, Vector<AudioFrame>> pair : p_bus_volumes) {
|
|
if (pair.value.size() < channel_count || pair.value.size() != MAX_CHANNELS_PER_BUS) {
|
|
delete new_bus_details;
|
|
ERR_FAIL();
|
|
}
|
|
|
|
new_bus_details->bus_active[idx] = true;
|
|
new_bus_details->bus[idx] = pair.key;
|
|
for (int channel_idx = 0; channel_idx < MAX_CHANNELS_PER_BUS; channel_idx++) {
|
|
new_bus_details->volume[idx][channel_idx] = pair.value[channel_idx];
|
|
}
|
|
}
|
|
playback_node->bus_details = new_bus_details;
|
|
playback_node->prev_bus_details = new AudioStreamPlaybackBusDetails();
|
|
|
|
playback_node->pitch_scale.set(p_pitch_scale);
|
|
playback_node->highshelf_gain.set(p_highshelf_gain);
|
|
playback_node->attenuation_filter_cutoff_hz.set(p_attenuation_cutoff_hz);
|
|
|
|
memset(playback_node->prev_bus_details->volume, 0, sizeof(playback_node->prev_bus_details->volume));
|
|
|
|
for (AudioFrame &frame : playback_node->lookahead) {
|
|
frame = AudioFrame(0, 0);
|
|
}
|
|
|
|
playback_node->state.store(AudioStreamPlaybackListNode::PLAYING);
|
|
|
|
playback_list.insert(playback_node);
|
|
}
|
|
|
|
void AudioServer::stop_playback_stream(Ref<AudioStreamPlayback> p_playback) {
|
|
ERR_FAIL_COND(p_playback.is_null());
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return;
|
|
}
|
|
|
|
AudioStreamPlaybackListNode::PlaybackState new_state, old_state;
|
|
do {
|
|
old_state = playback_node->state.load();
|
|
if (old_state == AudioStreamPlaybackListNode::AWAITING_DELETION) {
|
|
break; // Don't fade out again.
|
|
}
|
|
new_state = AudioStreamPlaybackListNode::FADE_OUT_TO_DELETION;
|
|
|
|
} while (!playback_node->state.compare_exchange_strong(old_state, new_state));
|
|
}
|
|
|
|
void AudioServer::set_playback_bus_exclusive(Ref<AudioStreamPlayback> p_playback, StringName p_bus, Vector<AudioFrame> p_volumes) {
|
|
ERR_FAIL_COND(p_volumes.size() != MAX_CHANNELS_PER_BUS);
|
|
|
|
HashMap<StringName, Vector<AudioFrame>> map;
|
|
map[p_bus] = p_volumes;
|
|
|
|
set_playback_bus_volumes_linear(p_playback, map);
|
|
}
|
|
|
|
void AudioServer::set_playback_bus_volumes_linear(Ref<AudioStreamPlayback> p_playback, HashMap<StringName, Vector<AudioFrame>> p_bus_volumes) {
|
|
ERR_FAIL_COND(p_bus_volumes.size() > MAX_BUSES_PER_PLAYBACK);
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return;
|
|
}
|
|
AudioStreamPlaybackBusDetails *old_bus_details, *new_bus_details = new AudioStreamPlaybackBusDetails();
|
|
|
|
int idx = 0;
|
|
for (KeyValue<StringName, Vector<AudioFrame>> pair : p_bus_volumes) {
|
|
if (idx >= MAX_BUSES_PER_PLAYBACK) {
|
|
break;
|
|
}
|
|
ERR_FAIL_COND(pair.value.size() < channel_count);
|
|
ERR_FAIL_COND(pair.value.size() != MAX_CHANNELS_PER_BUS);
|
|
|
|
new_bus_details->bus_active[idx] = true;
|
|
new_bus_details->bus[idx] = pair.key;
|
|
for (int channel_idx = 0; channel_idx < MAX_CHANNELS_PER_BUS; channel_idx++) {
|
|
new_bus_details->volume[idx][channel_idx] = pair.value[channel_idx];
|
|
}
|
|
idx++;
|
|
}
|
|
|
|
do {
|
|
old_bus_details = playback_node->bus_details.load();
|
|
} while (!playback_node->bus_details.compare_exchange_strong(old_bus_details, new_bus_details));
|
|
|
|
bus_details_graveyard.insert(old_bus_details);
|
|
}
|
|
|
|
void AudioServer::set_playback_all_bus_volumes_linear(Ref<AudioStreamPlayback> p_playback, Vector<AudioFrame> p_volumes) {
|
|
ERR_FAIL_COND(p_playback.is_null());
|
|
ERR_FAIL_COND(p_volumes.size() != MAX_CHANNELS_PER_BUS);
|
|
|
|
HashMap<StringName, Vector<AudioFrame>> map;
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return;
|
|
}
|
|
for (int bus_idx = 0; bus_idx < MAX_BUSES_PER_PLAYBACK; bus_idx++) {
|
|
if (playback_node->bus_details.load()->bus_active[bus_idx]) {
|
|
map[playback_node->bus_details.load()->bus[bus_idx]] = p_volumes;
|
|
}
|
|
}
|
|
|
|
set_playback_bus_volumes_linear(p_playback, map);
|
|
}
|
|
|
|
void AudioServer::set_playback_pitch_scale(Ref<AudioStreamPlayback> p_playback, float p_pitch_scale) {
|
|
ERR_FAIL_COND(p_playback.is_null());
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return;
|
|
}
|
|
|
|
playback_node->pitch_scale.set(p_pitch_scale);
|
|
}
|
|
|
|
void AudioServer::set_playback_paused(Ref<AudioStreamPlayback> p_playback, bool p_paused) {
|
|
ERR_FAIL_COND(p_playback.is_null());
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return;
|
|
}
|
|
|
|
AudioStreamPlaybackListNode::PlaybackState new_state, old_state;
|
|
do {
|
|
old_state = playback_node->state.load();
|
|
new_state = p_paused ? AudioStreamPlaybackListNode::FADE_OUT_TO_PAUSE : AudioStreamPlaybackListNode::PLAYING;
|
|
if (!p_paused && old_state == AudioStreamPlaybackListNode::PLAYING) {
|
|
return; // No-op.
|
|
}
|
|
if (p_paused && (old_state == AudioStreamPlaybackListNode::PAUSED || old_state == AudioStreamPlaybackListNode::FADE_OUT_TO_PAUSE)) {
|
|
return; // No-op.
|
|
}
|
|
|
|
} while (!playback_node->state.compare_exchange_strong(old_state, new_state));
|
|
}
|
|
|
|
void AudioServer::set_playback_highshelf_params(Ref<AudioStreamPlayback> p_playback, float p_gain, float p_attenuation_cutoff_hz) {
|
|
ERR_FAIL_COND(p_playback.is_null());
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return;
|
|
}
|
|
|
|
playback_node->attenuation_filter_cutoff_hz.set(p_attenuation_cutoff_hz);
|
|
playback_node->highshelf_gain.set(p_gain);
|
|
}
|
|
|
|
bool AudioServer::is_playback_active(Ref<AudioStreamPlayback> p_playback) {
|
|
ERR_FAIL_COND_V(p_playback.is_null(), false);
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return false;
|
|
}
|
|
|
|
return playback_node->state.load() == AudioStreamPlaybackListNode::PLAYING;
|
|
}
|
|
|
|
float AudioServer::get_playback_position(Ref<AudioStreamPlayback> p_playback) {
|
|
ERR_FAIL_COND_V(p_playback.is_null(), 0);
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return 0;
|
|
}
|
|
|
|
return playback_node->stream_playback->get_playback_position();
|
|
}
|
|
|
|
bool AudioServer::is_playback_paused(Ref<AudioStreamPlayback> p_playback) {
|
|
ERR_FAIL_COND_V(p_playback.is_null(), false);
|
|
|
|
AudioStreamPlaybackListNode *playback_node = _find_playback_list_node(p_playback);
|
|
if (!playback_node) {
|
|
return false;
|
|
}
|
|
|
|
return playback_node->state.load() == AudioStreamPlaybackListNode::PAUSED || playback_node->state.load() == AudioStreamPlaybackListNode::FADE_OUT_TO_PAUSE;
|
|
}
|
|
|
|
uint64_t AudioServer::get_mix_count() const {
|
|
return mix_count;
|
|
}
|
|
|
|
uint64_t AudioServer::get_mixed_frames() const {
|
|
return mix_frames;
|
|
}
|
|
|
|
void AudioServer::notify_listener_changed() {
|
|
for (CallbackItem *ci : listener_changed_callback_list) {
|
|
ci->callback(ci->userdata);
|
|
}
|
|
}
|
|
|
|
void AudioServer::init_channels_and_buffers() {
|
|
channel_count = get_channel_count();
|
|
temp_buffer.resize(channel_count);
|
|
mix_buffer.resize(buffer_size + LOOKAHEAD_BUFFER_SIZE);
|
|
|
|
for (int i = 0; i < temp_buffer.size(); i++) {
|
|
temp_buffer.write[i].resize(buffer_size);
|
|
}
|
|
|
|
for (int i = 0; i < buses.size(); i++) {
|
|
buses[i]->channels.resize(channel_count);
|
|
for (int j = 0; j < channel_count; j++) {
|
|
buses.write[i]->channels.write[j].buffer.resize(buffer_size);
|
|
}
|
|
_update_bus_effects(i);
|
|
}
|
|
}
|
|
|
|
void AudioServer::init() {
|
|
channel_disable_threshold_db = GLOBAL_DEF_RST("audio/buses/channel_disable_threshold_db", -60.0);
|
|
channel_disable_frames = float(GLOBAL_DEF_RST("audio/buses/channel_disable_time", 2.0)) * get_mix_rate();
|
|
ProjectSettings::get_singleton()->set_custom_property_info("audio/buses/channel_disable_time", PropertyInfo(Variant::FLOAT, "audio/buses/channel_disable_time", PROPERTY_HINT_RANGE, "0,5,0.01,or_greater"));
|
|
buffer_size = 512; //hardcoded for now
|
|
|
|
init_channels_and_buffers();
|
|
|
|
mix_count = 0;
|
|
set_bus_count(1);
|
|
set_bus_name(0, "Master");
|
|
|
|
if (AudioDriver::get_singleton()) {
|
|
AudioDriver::get_singleton()->start();
|
|
}
|
|
|
|
#ifdef TOOLS_ENABLED
|
|
set_edited(false); //avoid editors from thinking this was edited
|
|
#endif
|
|
|
|
GLOBAL_DEF_RST("audio/video/video_delay_compensation_ms", 0);
|
|
}
|
|
|
|
void AudioServer::update() {
|
|
#ifdef DEBUG_ENABLED
|
|
if (EngineDebugger::is_profiling("servers")) {
|
|
// Driver time includes server time + effects times
|
|
// Server time includes effects times
|
|
uint64_t driver_time = AudioDriver::get_singleton()->get_profiling_time();
|
|
uint64_t server_time = prof_time;
|
|
|
|
// Subtract the server time from the driver time
|
|
if (driver_time > server_time) {
|
|
driver_time -= server_time;
|
|
}
|
|
|
|
Array values;
|
|
|
|
for (int i = buses.size() - 1; i >= 0; i--) {
|
|
Bus *bus = buses[i];
|
|
if (bus->bypass) {
|
|
continue;
|
|
}
|
|
|
|
for (int j = 0; j < bus->effects.size(); j++) {
|
|
if (!bus->effects[j].enabled) {
|
|
continue;
|
|
}
|
|
|
|
values.push_back(String(bus->name) + bus->effects[j].effect->get_name());
|
|
values.push_back(USEC_TO_SEC(bus->effects[j].prof_time));
|
|
|
|
// Subtract the effect time from the driver and server times
|
|
if (driver_time > bus->effects[j].prof_time) {
|
|
driver_time -= bus->effects[j].prof_time;
|
|
}
|
|
if (server_time > bus->effects[j].prof_time) {
|
|
server_time -= bus->effects[j].prof_time;
|
|
}
|
|
}
|
|
}
|
|
|
|
values.push_back("audio_server");
|
|
values.push_back(USEC_TO_SEC(server_time));
|
|
values.push_back("audio_driver");
|
|
values.push_back(USEC_TO_SEC(driver_time));
|
|
|
|
values.push_front("audio_thread");
|
|
EngineDebugger::profiler_add_frame_data("servers", values);
|
|
}
|
|
|
|
// Reset profiling times
|
|
for (int i = buses.size() - 1; i >= 0; i--) {
|
|
Bus *bus = buses[i];
|
|
if (bus->bypass) {
|
|
continue;
|
|
}
|
|
|
|
for (int j = 0; j < bus->effects.size(); j++) {
|
|
if (!bus->effects[j].enabled) {
|
|
continue;
|
|
}
|
|
|
|
bus->effects.write[j].prof_time = 0;
|
|
}
|
|
}
|
|
|
|
AudioDriver::get_singleton()->reset_profiling_time();
|
|
prof_time = 0;
|
|
#endif
|
|
|
|
for (CallbackItem *ci : update_callback_list) {
|
|
ci->callback(ci->userdata);
|
|
}
|
|
mix_callback_list.maybe_cleanup();
|
|
update_callback_list.maybe_cleanup();
|
|
listener_changed_callback_list.maybe_cleanup();
|
|
playback_list.maybe_cleanup();
|
|
for (AudioStreamPlaybackBusDetails *bus_details : bus_details_graveyard_frame_old) {
|
|
bus_details_graveyard_frame_old.erase(bus_details, [](AudioStreamPlaybackBusDetails *d) { delete d; });
|
|
}
|
|
for (AudioStreamPlaybackBusDetails *bus_details : bus_details_graveyard) {
|
|
bus_details_graveyard_frame_old.insert(bus_details);
|
|
bus_details_graveyard.erase(bus_details);
|
|
}
|
|
bus_details_graveyard.maybe_cleanup();
|
|
bus_details_graveyard_frame_old.maybe_cleanup();
|
|
}
|
|
|
|
void AudioServer::load_default_bus_layout() {
|
|
String layout_path = ProjectSettings::get_singleton()->get("audio/buses/default_bus_layout");
|
|
|
|
if (ResourceLoader::exists(layout_path)) {
|
|
Ref<AudioBusLayout> default_layout = ResourceLoader::load(layout_path);
|
|
if (default_layout.is_valid()) {
|
|
set_bus_layout(default_layout);
|
|
}
|
|
}
|
|
}
|
|
|
|
void AudioServer::finish() {
|
|
for (int i = 0; i < AudioDriverManager::get_driver_count(); i++) {
|
|
AudioDriverManager::get_driver(i)->finish();
|
|
}
|
|
|
|
for (int i = 0; i < buses.size(); i++) {
|
|
memdelete(buses[i]);
|
|
}
|
|
|
|
buses.clear();
|
|
}
|
|
|
|
/* MISC config */
|
|
|
|
void AudioServer::lock() {
|
|
AudioDriver::get_singleton()->lock();
|
|
}
|
|
|
|
void AudioServer::unlock() {
|
|
AudioDriver::get_singleton()->unlock();
|
|
}
|
|
|
|
AudioServer::SpeakerMode AudioServer::get_speaker_mode() const {
|
|
return (AudioServer::SpeakerMode)AudioDriver::get_singleton()->get_speaker_mode();
|
|
}
|
|
|
|
float AudioServer::get_mix_rate() const {
|
|
return AudioDriver::get_singleton()->get_mix_rate();
|
|
}
|
|
|
|
float AudioServer::read_output_peak_db() const {
|
|
return 0;
|
|
}
|
|
|
|
AudioServer *AudioServer::get_singleton() {
|
|
return singleton;
|
|
}
|
|
|
|
double AudioServer::get_output_latency() const {
|
|
return AudioDriver::get_singleton()->get_latency();
|
|
}
|
|
|
|
double AudioServer::get_time_to_next_mix() const {
|
|
return AudioDriver::get_singleton()->get_time_to_next_mix();
|
|
}
|
|
|
|
double AudioServer::get_time_since_last_mix() const {
|
|
return AudioDriver::get_singleton()->get_time_since_last_mix();
|
|
}
|
|
|
|
AudioServer *AudioServer::singleton = nullptr;
|
|
|
|
void AudioServer::add_update_callback(AudioCallback p_callback, void *p_userdata) {
|
|
CallbackItem *ci = new CallbackItem();
|
|
ci->callback = p_callback;
|
|
ci->userdata = p_userdata;
|
|
update_callback_list.insert(ci);
|
|
}
|
|
|
|
void AudioServer::remove_update_callback(AudioCallback p_callback, void *p_userdata) {
|
|
for (CallbackItem *ci : update_callback_list) {
|
|
if (ci->callback == p_callback && ci->userdata == p_userdata) {
|
|
update_callback_list.erase(ci, [](CallbackItem *c) { delete c; });
|
|
}
|
|
}
|
|
}
|
|
|
|
void AudioServer::add_mix_callback(AudioCallback p_callback, void *p_userdata) {
|
|
CallbackItem *ci = new CallbackItem();
|
|
ci->callback = p_callback;
|
|
ci->userdata = p_userdata;
|
|
mix_callback_list.insert(ci);
|
|
}
|
|
|
|
void AudioServer::remove_mix_callback(AudioCallback p_callback, void *p_userdata) {
|
|
for (CallbackItem *ci : mix_callback_list) {
|
|
if (ci->callback == p_callback && ci->userdata == p_userdata) {
|
|
mix_callback_list.erase(ci, [](CallbackItem *c) { delete c; });
|
|
}
|
|
}
|
|
}
|
|
|
|
void AudioServer::add_listener_changed_callback(AudioCallback p_callback, void *p_userdata) {
|
|
CallbackItem *ci = new CallbackItem();
|
|
ci->callback = p_callback;
|
|
ci->userdata = p_userdata;
|
|
listener_changed_callback_list.insert(ci);
|
|
}
|
|
|
|
void AudioServer::remove_listener_changed_callback(AudioCallback p_callback, void *p_userdata) {
|
|
for (CallbackItem *ci : listener_changed_callback_list) {
|
|
if (ci->callback == p_callback && ci->userdata == p_userdata) {
|
|
listener_changed_callback_list.erase(ci, [](CallbackItem *c) { delete c; });
|
|
}
|
|
}
|
|
}
|
|
|
|
void AudioServer::set_bus_layout(const Ref<AudioBusLayout> &p_bus_layout) {
|
|
ERR_FAIL_COND(p_bus_layout.is_null() || p_bus_layout->buses.size() == 0);
|
|
|
|
lock();
|
|
for (int i = 0; i < buses.size(); i++) {
|
|
memdelete(buses[i]);
|
|
}
|
|
buses.resize(p_bus_layout->buses.size());
|
|
bus_map.clear();
|
|
for (int i = 0; i < p_bus_layout->buses.size(); i++) {
|
|
Bus *bus = memnew(Bus);
|
|
if (i == 0) {
|
|
bus->name = "Master";
|
|
} else {
|
|
bus->name = p_bus_layout->buses[i].name;
|
|
bus->send = p_bus_layout->buses[i].send;
|
|
}
|
|
|
|
bus->solo = p_bus_layout->buses[i].solo;
|
|
bus->mute = p_bus_layout->buses[i].mute;
|
|
bus->bypass = p_bus_layout->buses[i].bypass;
|
|
bus->volume_db = p_bus_layout->buses[i].volume_db;
|
|
|
|
for (int j = 0; j < p_bus_layout->buses[i].effects.size(); j++) {
|
|
Ref<AudioEffect> fx = p_bus_layout->buses[i].effects[j].effect;
|
|
|
|
if (fx.is_valid()) {
|
|
Bus::Effect bfx;
|
|
bfx.effect = fx;
|
|
bfx.enabled = p_bus_layout->buses[i].effects[j].enabled;
|
|
#ifdef DEBUG_ENABLED
|
|
bfx.prof_time = 0;
|
|
#endif
|
|
bus->effects.push_back(bfx);
|
|
}
|
|
}
|
|
|
|
bus_map[bus->name] = bus;
|
|
buses.write[i] = bus;
|
|
|
|
buses[i]->channels.resize(channel_count);
|
|
for (int j = 0; j < channel_count; j++) {
|
|
buses.write[i]->channels.write[j].buffer.resize(buffer_size);
|
|
}
|
|
_update_bus_effects(i);
|
|
}
|
|
#ifdef TOOLS_ENABLED
|
|
set_edited(false);
|
|
#endif
|
|
unlock();
|
|
}
|
|
|
|
Ref<AudioBusLayout> AudioServer::generate_bus_layout() const {
|
|
Ref<AudioBusLayout> state;
|
|
state.instantiate();
|
|
|
|
state->buses.resize(buses.size());
|
|
|
|
for (int i = 0; i < buses.size(); i++) {
|
|
state->buses.write[i].name = buses[i]->name;
|
|
state->buses.write[i].send = buses[i]->send;
|
|
state->buses.write[i].mute = buses[i]->mute;
|
|
state->buses.write[i].solo = buses[i]->solo;
|
|
state->buses.write[i].bypass = buses[i]->bypass;
|
|
state->buses.write[i].volume_db = buses[i]->volume_db;
|
|
for (int j = 0; j < buses[i]->effects.size(); j++) {
|
|
AudioBusLayout::Bus::Effect fx;
|
|
fx.effect = buses[i]->effects[j].effect;
|
|
fx.enabled = buses[i]->effects[j].enabled;
|
|
state->buses.write[i].effects.push_back(fx);
|
|
}
|
|
}
|
|
|
|
return state;
|
|
}
|
|
|
|
PackedStringArray AudioServer::get_device_list() {
|
|
return AudioDriver::get_singleton()->get_device_list();
|
|
}
|
|
|
|
String AudioServer::get_device() {
|
|
return AudioDriver::get_singleton()->get_device();
|
|
}
|
|
|
|
void AudioServer::set_device(String device) {
|
|
AudioDriver::get_singleton()->set_device(device);
|
|
}
|
|
|
|
PackedStringArray AudioServer::capture_get_device_list() {
|
|
return AudioDriver::get_singleton()->capture_get_device_list();
|
|
}
|
|
|
|
String AudioServer::capture_get_device() {
|
|
return AudioDriver::get_singleton()->capture_get_device();
|
|
}
|
|
|
|
void AudioServer::capture_set_device(const String &p_name) {
|
|
AudioDriver::get_singleton()->capture_set_device(p_name);
|
|
}
|
|
|
|
void AudioServer::set_enable_tagging_used_audio_streams(bool p_enable) {
|
|
tag_used_audio_streams = p_enable;
|
|
}
|
|
|
|
void AudioServer::_bind_methods() {
|
|
ClassDB::bind_method(D_METHOD("set_bus_count", "amount"), &AudioServer::set_bus_count);
|
|
ClassDB::bind_method(D_METHOD("get_bus_count"), &AudioServer::get_bus_count);
|
|
|
|
ClassDB::bind_method(D_METHOD("remove_bus", "index"), &AudioServer::remove_bus);
|
|
ClassDB::bind_method(D_METHOD("add_bus", "at_position"), &AudioServer::add_bus, DEFVAL(-1));
|
|
ClassDB::bind_method(D_METHOD("move_bus", "index", "to_index"), &AudioServer::move_bus);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_bus_name", "bus_idx", "name"), &AudioServer::set_bus_name);
|
|
ClassDB::bind_method(D_METHOD("get_bus_name", "bus_idx"), &AudioServer::get_bus_name);
|
|
ClassDB::bind_method(D_METHOD("get_bus_index", "bus_name"), &AudioServer::get_bus_index);
|
|
|
|
ClassDB::bind_method(D_METHOD("get_bus_channels", "bus_idx"), &AudioServer::get_bus_channels);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_bus_volume_db", "bus_idx", "volume_db"), &AudioServer::set_bus_volume_db);
|
|
ClassDB::bind_method(D_METHOD("get_bus_volume_db", "bus_idx"), &AudioServer::get_bus_volume_db);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_bus_send", "bus_idx", "send"), &AudioServer::set_bus_send);
|
|
ClassDB::bind_method(D_METHOD("get_bus_send", "bus_idx"), &AudioServer::get_bus_send);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_bus_solo", "bus_idx", "enable"), &AudioServer::set_bus_solo);
|
|
ClassDB::bind_method(D_METHOD("is_bus_solo", "bus_idx"), &AudioServer::is_bus_solo);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_bus_mute", "bus_idx", "enable"), &AudioServer::set_bus_mute);
|
|
ClassDB::bind_method(D_METHOD("is_bus_mute", "bus_idx"), &AudioServer::is_bus_mute);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_bus_bypass_effects", "bus_idx", "enable"), &AudioServer::set_bus_bypass_effects);
|
|
ClassDB::bind_method(D_METHOD("is_bus_bypassing_effects", "bus_idx"), &AudioServer::is_bus_bypassing_effects);
|
|
|
|
ClassDB::bind_method(D_METHOD("add_bus_effect", "bus_idx", "effect", "at_position"), &AudioServer::add_bus_effect, DEFVAL(-1));
|
|
ClassDB::bind_method(D_METHOD("remove_bus_effect", "bus_idx", "effect_idx"), &AudioServer::remove_bus_effect);
|
|
|
|
ClassDB::bind_method(D_METHOD("get_bus_effect_count", "bus_idx"), &AudioServer::get_bus_effect_count);
|
|
ClassDB::bind_method(D_METHOD("get_bus_effect", "bus_idx", "effect_idx"), &AudioServer::get_bus_effect);
|
|
ClassDB::bind_method(D_METHOD("get_bus_effect_instance", "bus_idx", "effect_idx", "channel"), &AudioServer::get_bus_effect_instance, DEFVAL(0));
|
|
ClassDB::bind_method(D_METHOD("swap_bus_effects", "bus_idx", "effect_idx", "by_effect_idx"), &AudioServer::swap_bus_effects);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_bus_effect_enabled", "bus_idx", "effect_idx", "enabled"), &AudioServer::set_bus_effect_enabled);
|
|
ClassDB::bind_method(D_METHOD("is_bus_effect_enabled", "bus_idx", "effect_idx"), &AudioServer::is_bus_effect_enabled);
|
|
|
|
ClassDB::bind_method(D_METHOD("get_bus_peak_volume_left_db", "bus_idx", "channel"), &AudioServer::get_bus_peak_volume_left_db);
|
|
ClassDB::bind_method(D_METHOD("get_bus_peak_volume_right_db", "bus_idx", "channel"), &AudioServer::get_bus_peak_volume_right_db);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_playback_speed_scale", "scale"), &AudioServer::set_playback_speed_scale);
|
|
ClassDB::bind_method(D_METHOD("get_playback_speed_scale"), &AudioServer::get_playback_speed_scale);
|
|
|
|
ClassDB::bind_method(D_METHOD("lock"), &AudioServer::lock);
|
|
ClassDB::bind_method(D_METHOD("unlock"), &AudioServer::unlock);
|
|
|
|
ClassDB::bind_method(D_METHOD("get_speaker_mode"), &AudioServer::get_speaker_mode);
|
|
ClassDB::bind_method(D_METHOD("get_mix_rate"), &AudioServer::get_mix_rate);
|
|
ClassDB::bind_method(D_METHOD("get_device_list"), &AudioServer::get_device_list);
|
|
ClassDB::bind_method(D_METHOD("get_device"), &AudioServer::get_device);
|
|
ClassDB::bind_method(D_METHOD("set_device", "device"), &AudioServer::set_device);
|
|
|
|
ClassDB::bind_method(D_METHOD("get_time_to_next_mix"), &AudioServer::get_time_to_next_mix);
|
|
ClassDB::bind_method(D_METHOD("get_time_since_last_mix"), &AudioServer::get_time_since_last_mix);
|
|
ClassDB::bind_method(D_METHOD("get_output_latency"), &AudioServer::get_output_latency);
|
|
|
|
ClassDB::bind_method(D_METHOD("capture_get_device_list"), &AudioServer::capture_get_device_list);
|
|
ClassDB::bind_method(D_METHOD("capture_get_device"), &AudioServer::capture_get_device);
|
|
ClassDB::bind_method(D_METHOD("capture_set_device", "name"), &AudioServer::capture_set_device);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_bus_layout", "bus_layout"), &AudioServer::set_bus_layout);
|
|
ClassDB::bind_method(D_METHOD("generate_bus_layout"), &AudioServer::generate_bus_layout);
|
|
|
|
ClassDB::bind_method(D_METHOD("set_enable_tagging_used_audio_streams", "enable"), &AudioServer::set_enable_tagging_used_audio_streams);
|
|
|
|
ADD_PROPERTY(PropertyInfo(Variant::INT, "bus_count"), "set_bus_count", "get_bus_count");
|
|
ADD_PROPERTY(PropertyInfo(Variant::STRING, "device"), "set_device", "get_device");
|
|
ADD_PROPERTY(PropertyInfo(Variant::STRING, "capture_device"), "capture_set_device", "capture_get_device");
|
|
// The default value may be set to an empty string by the platform-specific audio driver.
|
|
// Override for class reference generation purposes.
|
|
ADD_PROPERTY_DEFAULT("capture_device", "Default");
|
|
ADD_PROPERTY(PropertyInfo(Variant::FLOAT, "playback_speed_scale"), "set_playback_speed_scale", "get_playback_speed_scale");
|
|
|
|
ADD_SIGNAL(MethodInfo("bus_layout_changed"));
|
|
|
|
BIND_ENUM_CONSTANT(SPEAKER_MODE_STEREO);
|
|
BIND_ENUM_CONSTANT(SPEAKER_SURROUND_31);
|
|
BIND_ENUM_CONSTANT(SPEAKER_SURROUND_51);
|
|
BIND_ENUM_CONSTANT(SPEAKER_SURROUND_71);
|
|
}
|
|
|
|
AudioServer::AudioServer() {
|
|
singleton = this;
|
|
}
|
|
|
|
AudioServer::~AudioServer() {
|
|
singleton = nullptr;
|
|
}
|
|
|
|
/////////////////////////////////
|
|
|
|
bool AudioBusLayout::_set(const StringName &p_name, const Variant &p_value) {
|
|
String s = p_name;
|
|
if (s.begins_with("bus/")) {
|
|
int index = s.get_slice("/", 1).to_int();
|
|
if (buses.size() <= index) {
|
|
buses.resize(index + 1);
|
|
}
|
|
|
|
Bus &bus = buses.write[index];
|
|
|
|
String what = s.get_slice("/", 2);
|
|
|
|
if (what == "name") {
|
|
bus.name = p_value;
|
|
} else if (what == "solo") {
|
|
bus.solo = p_value;
|
|
} else if (what == "mute") {
|
|
bus.mute = p_value;
|
|
} else if (what == "bypass_fx") {
|
|
bus.bypass = p_value;
|
|
} else if (what == "volume_db") {
|
|
bus.volume_db = p_value;
|
|
} else if (what == "send") {
|
|
bus.send = p_value;
|
|
} else if (what == "effect") {
|
|
int which = s.get_slice("/", 3).to_int();
|
|
if (bus.effects.size() <= which) {
|
|
bus.effects.resize(which + 1);
|
|
}
|
|
|
|
Bus::Effect &fx = bus.effects.write[which];
|
|
|
|
String fxwhat = s.get_slice("/", 4);
|
|
if (fxwhat == "effect") {
|
|
fx.effect = p_value;
|
|
} else if (fxwhat == "enabled") {
|
|
fx.enabled = p_value;
|
|
} else {
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
} else {
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
bool AudioBusLayout::_get(const StringName &p_name, Variant &r_ret) const {
|
|
String s = p_name;
|
|
if (s.begins_with("bus/")) {
|
|
int index = s.get_slice("/", 1).to_int();
|
|
if (index < 0 || index >= buses.size()) {
|
|
return false;
|
|
}
|
|
|
|
const Bus &bus = buses[index];
|
|
|
|
String what = s.get_slice("/", 2);
|
|
|
|
if (what == "name") {
|
|
r_ret = bus.name;
|
|
} else if (what == "solo") {
|
|
r_ret = bus.solo;
|
|
} else if (what == "mute") {
|
|
r_ret = bus.mute;
|
|
} else if (what == "bypass_fx") {
|
|
r_ret = bus.bypass;
|
|
} else if (what == "volume_db") {
|
|
r_ret = bus.volume_db;
|
|
} else if (what == "send") {
|
|
r_ret = bus.send;
|
|
} else if (what == "effect") {
|
|
int which = s.get_slice("/", 3).to_int();
|
|
if (which < 0 || which >= bus.effects.size()) {
|
|
return false;
|
|
}
|
|
|
|
const Bus::Effect &fx = bus.effects[which];
|
|
|
|
String fxwhat = s.get_slice("/", 4);
|
|
if (fxwhat == "effect") {
|
|
r_ret = fx.effect;
|
|
} else if (fxwhat == "enabled") {
|
|
r_ret = fx.enabled;
|
|
} else {
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
} else {
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
void AudioBusLayout::_get_property_list(List<PropertyInfo> *p_list) const {
|
|
for (int i = 0; i < buses.size(); i++) {
|
|
p_list->push_back(PropertyInfo(Variant::STRING, "bus/" + itos(i) + "/name", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR | PROPERTY_USAGE_INTERNAL));
|
|
p_list->push_back(PropertyInfo(Variant::BOOL, "bus/" + itos(i) + "/solo", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR | PROPERTY_USAGE_INTERNAL));
|
|
p_list->push_back(PropertyInfo(Variant::BOOL, "bus/" + itos(i) + "/mute", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR | PROPERTY_USAGE_INTERNAL));
|
|
p_list->push_back(PropertyInfo(Variant::BOOL, "bus/" + itos(i) + "/bypass_fx", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR | PROPERTY_USAGE_INTERNAL));
|
|
p_list->push_back(PropertyInfo(Variant::FLOAT, "bus/" + itos(i) + "/volume_db", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR | PROPERTY_USAGE_INTERNAL));
|
|
p_list->push_back(PropertyInfo(Variant::FLOAT, "bus/" + itos(i) + "/send", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR | PROPERTY_USAGE_INTERNAL));
|
|
|
|
for (int j = 0; j < buses[i].effects.size(); j++) {
|
|
p_list->push_back(PropertyInfo(Variant::OBJECT, "bus/" + itos(i) + "/effect/" + itos(j) + "/effect", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR | PROPERTY_USAGE_INTERNAL));
|
|
p_list->push_back(PropertyInfo(Variant::BOOL, "bus/" + itos(i) + "/effect/" + itos(j) + "/enabled", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NO_EDITOR | PROPERTY_USAGE_INTERNAL));
|
|
}
|
|
}
|
|
}
|
|
|
|
AudioBusLayout::AudioBusLayout() {
|
|
buses.resize(1);
|
|
buses.write[0].name = "Master";
|
|
}
|