godot/servers/audio/audio_server_sw.h

287 lines
8.8 KiB
C
Raw Normal View History

2014-02-10 01:10:30 +00:00
/*************************************************************************/
/* audio_server_sw.h */
/*************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* http://www.godotengine.org */
/*************************************************************************/
2015-04-18 17:38:54 +00:00
/* Copyright (c) 2007-2015 Juan Linietsky, Ariel Manzur. */
2014-02-10 01:10:30 +00:00
/* */
/* 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. */
/*************************************************************************/
#ifndef AUDIO_SERVER_SW_H
#define AUDIO_SERVER_SW_H
#include "servers/audio_server.h"
#include "servers/audio/audio_mixer_sw.h"
#include "servers/audio/voice_rb_sw.h"
#include "self_list.h"
#include "os/thread_safe.h"
#include "os/thread.h"
class AudioServerSW : public AudioServer {
OBJ_TYPE( AudioServerSW, AudioServer );
_THREAD_SAFE_CLASS_
enum {
INTERNAL_BUFFER_SIZE=4096,
STREAM_SCALE_BITS=12
};
SampleManagerSW *sample_manager;
AudioMixerSW *mixer;
virtual AudioMixer *get_mixer();
virtual void audio_mixer_chunk_callback(int p_frames);
struct Voice {
float volume;
volatile bool active;
SelfList<Voice> active_item;
AudioMixer::ChannelID channel;
Voice () : active_item(this) { channel=AudioMixer::INVALID_CHANNEL; active=false;}
};
mutable RID_Owner<Voice> voice_owner;
SelfList<Voice>::List active_list;
struct Stream {
bool active;
List<Stream*>::Element *E;
AudioStream *audio_stream;
EventStream *event_stream;
float volume_scale;
};
List<Stream*> active_audio_streams;
//List<Stream*> event_streams;
int32_t * internal_buffer;
int internal_buffer_channels;
int32_t * stream_buffer;
mutable RID_Owner<Stream> stream_owner;
float stream_volume;
float stream_volume_scale;
float fx_volume_scale;
float event_voice_volume_scale;
float peak_left,peak_right;
uint32_t max_peak;
double _output_delay;
2014-02-10 01:10:30 +00:00
VoiceRBSW voice_rb;
bool exit_update_thread;
Thread *thread;
static void _thread_func(void *self);
void _update_streams(bool p_thread);
void driver_process_chunk(int p_frames,int32_t *p_buffer);
AudioMixerSW::InterpolationType mixer_interp;
bool mixer_use_fx;
uint64_t mixer_step_usecs;
static void _mixer_callback(void *p_udata);
friend class AudioDriverSW;
void driver_process(int p_frames,int32_t *p_buffer);
public:
/* SAMPLE API */
virtual RID sample_create(SampleFormat p_format, bool p_stereo, int p_length);
virtual void sample_set_description(RID p_sample, const String& p_description);
virtual String sample_get_description(RID p_sample, const String& p_description) const;
virtual SampleFormat sample_get_format(RID p_sample) const;
virtual bool sample_is_stereo(RID p_sample) const;
virtual int sample_get_length(RID p_sample) const;
const void* sample_get_data_ptr(RID p_sample) const;
virtual void sample_set_data(RID p_sample, const DVector<uint8_t>& p_buffer);
virtual const DVector<uint8_t> sample_get_data(RID p_sample) const;
virtual void sample_set_mix_rate(RID p_sample,int p_rate);
virtual int sample_get_mix_rate(RID p_sample) const;
virtual void sample_set_loop_format(RID p_sample,SampleLoopFormat p_format);
virtual SampleLoopFormat sample_get_loop_format(RID p_sample) const;
virtual void sample_set_loop_begin(RID p_sample,int p_pos);
virtual int sample_get_loop_begin(RID p_sample) const;
virtual void sample_set_loop_end(RID p_sample,int p_pos);
virtual int sample_get_loop_end(RID p_sample) const;
/* VOICE API */
virtual RID voice_create();
virtual void voice_play(RID p_voice, RID p_sample);
virtual void voice_set_volume(RID p_voice, float p_volume);
2014-02-10 01:10:30 +00:00
virtual void voice_set_pan(RID p_voice, float p_pan, float p_depth=0,float height=0); //pan and depth go from -1 to 1
virtual void voice_set_filter(RID p_voice, FilterType p_type, float p_cutoff, float p_resonance,float p_gain=0);
virtual void voice_set_chorus(RID p_voice, float p_chorus );
virtual void voice_set_reverb(RID p_voice, ReverbRoomType p_room_type, float p_reverb);
virtual void voice_set_mix_rate(RID p_voice, int p_mix_rate);
virtual void voice_set_positional(RID p_voice, bool p_positional);
virtual float voice_get_volume(RID p_voice) const;
virtual float voice_get_pan(RID p_voice) const; //pan and depth go from -1 to 1
virtual float voice_get_pan_depth(RID p_voice) const; //pan and depth go from -1 to 1
virtual float voice_get_pan_height(RID p_voice) const; //pan and depth go from -1 to 1
virtual FilterType voice_get_filter_type(RID p_voice) const;
virtual float voice_get_filter_cutoff(RID p_voice) const;
virtual float voice_get_filter_resonance(RID p_voice) const;
virtual float voice_get_chorus(RID p_voice) const;
virtual ReverbRoomType voice_get_reverb_type(RID p_voice) const;
virtual float voice_get_reverb(RID p_voice) const;
virtual int voice_get_mix_rate(RID p_voice) const;
virtual bool voice_is_positional(RID p_voice) const;
virtual void voice_stop(RID p_voice);
virtual bool voice_is_active(RID p_voice) const;
/* STREAM API */
virtual RID audio_stream_create(AudioStream *p_stream);
virtual RID event_stream_create(EventStream *p_stream);
virtual void stream_set_active(RID p_stream, bool p_active);
virtual bool stream_is_active(RID p_stream) const;
virtual void stream_set_volume_scale(RID p_stream, float p_scale);
virtual float stream_set_volume_scale(RID p_stream) const;
virtual void free(RID p_id);
virtual void init();
virtual void finish();
virtual void update();
virtual void lock();
virtual void unlock();
virtual int get_default_channel_count() const;
virtual int get_default_mix_rate() const;
void set_mixer_params(AudioMixerSW::InterpolationType p_interp, bool p_use_fx);
virtual void set_stream_global_volume_scale(float p_volume);
virtual void set_fx_global_volume_scale(float p_volume);
virtual void set_event_voice_global_volume_scale(float p_volume);
virtual float get_stream_global_volume_scale() const;
virtual float get_fx_global_volume_scale() const;
virtual float get_event_voice_global_volume_scale() const;
virtual uint32_t read_output_peak() const;
virtual double get_mix_time() const; //useful for video -> audio sync
virtual double get_output_delay() const;
2014-02-10 01:10:30 +00:00
AudioServerSW(SampleManagerSW *p_sample_manager);
~AudioServerSW();
};
class AudioDriverSW {
static AudioDriverSW *singleton;
uint64_t _last_mix_time;
uint64_t _mix_amount;
protected:
void audio_server_process(int p_frames,int32_t *p_buffer,bool p_update_mix_time=true);
void update_mix_time(int p_frames);
public:
double get_mix_time() const; //useful for video -> audio sync
enum OutputFormat {
OUTPUT_MONO,
OUTPUT_STEREO,
OUTPUT_QUAD,
OUTPUT_5_1
};
static AudioDriverSW *get_singleton();
void set_singleton();
virtual const char* get_name() const=0;
virtual Error init()=0;
virtual void start()=0;
virtual int get_mix_rate() const =0;
virtual OutputFormat get_output_format() const=0;
virtual void lock()=0;
virtual void unlock()=0;
virtual void finish()=0;
virtual float get_latency() { return 0; }
2014-02-10 01:10:30 +00:00
AudioDriverSW();
virtual ~AudioDriverSW() {};
};
class AudioDriverManagerSW {
enum {
MAX_DRIVERS=10
};
static AudioDriverSW *drivers[MAX_DRIVERS];
static int driver_count;
public:
static void add_driver(AudioDriverSW *p_driver);
static int get_driver_count();
static AudioDriverSW *get_driver(int p_driver);
};
#endif // AUDIO_SERVER_SW_H