UML顺序图:
AudioTrack.svg
https://download.csdn.net/download/u012906122/19589063
Audio里主要有三个功能:
AudioManager:这个主要是用来管理Audio系统的
AudioTrack:这个主要是用来播放声音的
AudioRecord:这个主要是用来录音的
其中AudioManager的理解需要考虑整个系统上声音的策略问题,例如来电话铃声,短信铃声等,主要是策略上的问题。
一个简单的API使用流程:
//根据采样率,采样精度,单双声道来得到frame的大小。
int bufsize = AudioTrack.getMinBufferSize(8000, //每秒8K个点
AudioFormat.CHANNEL_CONFIGURATION_STEREO, //双声道
AudioFormat.ENCODING_PCM_16BIT); //一个采样点16比特-2个字节
//创建AudioTrack
AudioTrack trackplayer = new AudioTrack(AudioManager.STREAM_MUSIC, 8000,
AudioFormat.CHANNEL_CONFIGURATION_STEREO,
AudioFormat.ENCODING_PCM_16BIT,
bufsize,
AudioTrack.MODE_STREAM);
trackplayer.play() ; //开始
trackplayer.write(bytes_pkg, 0, bytes_pkg.length); //往track中写数据
….
trackplayer.stop(); //停止播放
trackplayer.release(); //释放底层资源
这里需要解释下两个东西:
1 AudioTrack.MODE_STREAM:
AudioTrack中有MODE_STATIC和MODE_STREAM两种分类。STREAM的意思是由用户在应用层通过write方式把数据一次一次得写到audiotrack中。这种方式的坏处就是总是在JAVA层和Native层交互,效率损失较大。
而STATIC的意思是一开始创建的时候,就把音频数据放到一个固定的buffer,然后直接传给audiotrack,后续就不用一次次得write了。AudioTrack会自己播放这个buffer中的数据。这种方法对于铃声等内存占用较小,延时要求较高的声音来说很适用。
2 StreamType
这个在构造AudioTrack的第一个参数中使用。这个参数和Android中的AudioManager有关系,涉及到手机上的音频管理策略。
Android将系统的声音分为以下几类常见的(未写全):
STREAM_ALARM:警告声
STREAM_MUSCI:音乐声
STREAM_RING:铃声
STREAM_SYSTEM:系统声音
STREAM_VOCIE_CALL:电话声音
static public int getMinBufferSize(int sampleRateInHz, int channelConfig, int audioFormat) {
int channelCount = 0;
switch(channelConfig) {
case AudioFormat.CHANNEL_OUT_MONO:
case AudioFormat.CHANNEL_CONFIGURATION_MONO:
channelCount = 1;
break;
case AudioFormat.CHANNEL_OUT_STEREO:
case AudioFormat.CHANNEL_CONFIGURATION_STEREO:
channelCount = 2;
break;
default:
loge("getMinBufferSize(): Invalid channel configuration.");
return AudioTrack.ERROR_BAD_VALUE;
}
//目前只支持PCM8和PCM16精度的音频
if ((audioFormat != AudioFormat.ENCODING_PCM_16BIT)
&& (audioFormat != AudioFormat.ENCODING_PCM_8BIT)) {
loge("getMinBufferSize(): Invalid audio format.");
return AudioTrack.ERROR_BAD_VALUE;
}
//ft,对采样频率也有要求,太低或太高都不行,人耳分辨率在20HZ到40KHZ之间
if ( (sampleRateInHz < 4000) || (sampleRateInHz > 48000) ) {
loge("getMinBufferSize(): " + sampleRateInHz +"Hz is not a supported sample rate.");
return AudioTrack.ERROR_BAD_VALUE;
}
//调用native函数,JNI层
int size = native_get_min_buff_size(sampleRateInHz, channelCount, audioFormat);
if ((size == -1) || (size == 0)) {
loge("getMinBufferSize(): error querying hardware");
return AudioTrack.ERROR;
}
else {
return size;
}
}
2 native_get_min_buff_size
native_get_min_buff_size最终对应到函数android_media_AudioTrack_get_min_buff_size
3 android_media_AudioTrack_get_min_buff_size
在framework/base/core/jni/android_media_track.cpp中实现
static jint android_media_AudioTrack_get_min_buff_size(JNIEnv *env, jobject thiz, jint sampleRateInHertz, jint nbChannels, jint audioFormat) { int afSamplingRate;
int afFrameCount;
uint32_t afLatency;
if (AudioSystem::getOutputSamplingRate(&afSamplingRate) != NO_ERROR) {
return -1;
}
if (AudioSystem::getOutputFrameCount(&afFrameCount) != NO_ERROR) {
return -1;
}
if (AudioSystem::getOutputLatency(&afLatency) != NO_ERROR) {
return -1;
}
//音频中最常见的是frame这个单位,什么意思?经过多方查找,最后在ALSA的wiki中找到解释。一个frame就是1个采样点的字节数*声道。为啥搞个frame出来?因为对于多声道的话,用1个采样点的字节数表示不全,因为播放的时候肯定是多个声道的数据都要播出来才行。所以为了方便,就说1秒钟有多少个frame。
//Ensure that buffer depth covers at least audio hardware latency
uint32_t minBufCount = afLatency / ((1000 * afFrameCount)/afSamplingRate);
if (minBufCount < 2) minBufCount = 2;
uint32_t minFrameCount = (afFrameCount*sampleRateInHertz*minBufCount)/afSamplingRate;
//下面根据最小的framecount计算最小的buffersize
int minBuffSize = minFrameCount
* (audioFormat == javaAudioTrackFields.PCM16 ? 2 : 1)
* nbChannels;
return minBuffSize;
}
getMinBufSize函数完了后,我们得到一个满足最小要求的缓冲区大小。下面就需要创建AudioTrack对象了。
4 new AudioTrack
public AudioTrack(int streamType, int sampleRateInHz, int channelConfig, int audioFormat, int bufferSizeInBytes, int mode) throws IllegalArgumentException {
mState = STATE_UNINITIALIZED;
//获得主线程的Looper,这个在MediaScanner分析中已经讲过了
if ((mInitializationLooper = Looper.myLooper()) == null) {
mInitializationLooper = Looper.getMainLooper();
}
//检查参数是否合法之类的,可以不管它
audioParamCheck(streamType, sampleRateInHz, channelConfig, audioFormat, mode);
audioBuffSizeCheck(bufferSizeInBytes);
//调用native层的native_setup,把自己的WeakReference传进去了
//不了解JAVA WeakReference的可以上网自己查一下
int initResult = native_setup(new WeakReference<AudioTrack>(this),
mStreamType, 这个值是AudioManager.STREAM_MUSIC
mSampleRate, 这个值是8000
mChannels, 这个值是2
mAudioFormat, 这个值是AudioFormat.ENCODING_PCM_16BIT
mNativeBufferSizeInBytes, //这个是刚才getMinBufSize得到的
mDataLoadMode); //DataLoadMode是MODE_STREAM
....
}
5 native_setup
6 android_media_AudioTrack_native_setup
static int android_media_AudioTrack_native_setup(JNIEnv *env, jobject thiz, jobject weak_this,
jint streamType, jint sampleRateInHertz, jint channels,
jint audioFormat, jint buffSizeInBytes, jint memoryMode)
{
int afSampleRate;
int afFrameCount;
AudioSystem::getOutputFrameCount(&afFrameCount, streamType);
AudioSystem::getOutputSamplingRate(&afSampleRate, streamType);
AudioSystem::isOutputChannel(channels);
//popCount是统计一个整数中有多少位为1的算法
int nbChannels = AudioSystem::popCount(channels);
if (streamType == javaAudioTrackFields.STREAM_MUSIC) {
atStreamType = AudioSystem::MUSIC;
}
int bytesPerSample = audioFormat == javaAudioTrackFields.PCM16 ? 2 : 1;
int format = audioFormat == javaAudioTrackFields.PCM16 ?
AudioSystem::PCM_16_BIT : AudioSystem::PCM_8_BIT;
int frameCount = buffSizeInBytes / (nbChannels * bytesPerSample);
//上面是根据Buffer大小和一个Frame大小来计算帧数的。
//AudioTrackJniStorage,就是一个保存一些数据的地方
AudioTrackJniStorage* lpJniStorage = new AudioTrackJniStorage();
jclass clazz = env->GetObjectClass(thiz);
lpJniStorage->mCallbackData.audioTrack_class = (jclass)env->NewGlobalRef(clazz);
lpJniStorage->mCallbackData.audioTrack_ref = env->NewGlobalRef(weak_this);
lpJniStorage->mStreamType = atStreamType;
//创建真正的AudioTrack对象
AudioTrack* lpTrack = new AudioTrack();
if (memoryMode == javaAudioTrackFields.MODE_STREAM) {
//如果是STREAM流方式的话,把刚才那些参数设进去
lpTrack->set(
atStreamType,// stream type
sampleRateInHertz,
format,// word length, PCM
channels,
frameCount,
0,// flags
audioCallback,
&(lpJniStorage->mCallbackData),//callback, callback data (user)
0,// notificationFrames == 0 since not using EVENT_MORE_DATA to feed the AudioTrack
0,// 共享内存,STREAM模式需要用户一次次写,所以就不用共享内存了
true);// thread can call Java
} else if (memoryMode == javaAudioTrackFields.MODE_STATIC) {
//如果是static模式,需要用户一次性把数据写进去,然后再由audioTrack自己去把数据读出来,所以需要一个共享内存。这里的共享内存是指C++AudioTrack和AudioFlinger之间共享的内容,因为真正播放的工作是由AudioFlinger来完成的。
lpJniStorage->allocSharedMem(buffSizeInBytes);
lpTrack->set(
atStreamType,// stream type
sampleRateInHertz,
format,// word length, PCM
channels,
frameCount,
0,// flags
audioCallback,
&(lpJniStorage->mCallbackData),//callback, callback data (user));
0,// notificationFrames == 0 since not using EVENT_MORE_DATA to feed the AudioTrack
lpJniStorage->mMemBase,// shared mem
true);// thread can call Java
}
if (lpTrack->initCheck() != NO_ERROR) {
LOGE("Error initializing AudioTrack");
goto native_init_failure;
}
//把C++AudioTrack对象指针保存到JAVA对象的一个变量中。这样,Native层的AudioTrack对象就和JAVA层的AudioTrack对象关联起来了。
env->SetIntField(thiz, javaAudioTrackFields.nativeTrackInJavaObj, (int)lpTrack);
env->SetIntField(thiz, javaAudioTrackFields.jniData, (int)lpJniStorage);
}
7 AudioTrack* lpTrack = new AudioTrack
那么,我们就看看真正干活的的C++AudioTrack吧。
AudioTrack.cpp位于framework/base/libmedia/AudioTrack.cpp
AudioTrack::AudioTrack()
: mStatus(NO_INIT) //把状态初始化成NO_INIT。Android大量使用了设计模式中的state。
{
}
8 lpTrack->set
status_t AudioTrack::set(
int streamType,
uint32_t sampleRate,
int format,
int channels,
int frameCount,
uint32_t flags,
callback_t cbf,
void* user,
int notificationFrames,
const sp<IMemory>& sharedBuffer,
bool threadCanCallJava)
{
//...前面一堆的判断,等以后讲AudioSystem再说
audio_io_handle_t output =
AudioSystem::getOutput((AudioSystem::stream_type)streamType,
sampleRate, format, channels, (AudioSystem::output_flags)flags);
status_t status = createTrack(streamType, sampleRate, format, channelCount,
frameCount, flags, sharedBuffer, output);
//cbf是JNI传入的回调函数audioCallback
if (cbf != 0) { //看来,怎么着也要创建这个线程了!
mAudioTrackThread = new AudioTrackThread(*this, threadCanCallJava);
}
return NO_ERROR;
}
9 createTrack
status_t AudioTrack::createTrack(
int streamType,
uint32_t sampleRate,
int format,
int channelCount,
int frameCount,
uint32_t flags,
const sp<IMemory>& sharedBuffer,
audio_io_handle_t output)
{
status_t status;
//和audioFlinger挂上关系了
const sp<IAudioFlinger>& audioFlinger = AudioSystem::get_audio_flinger();
//下面这个调用最终会在AudioFlinger中出现。暂时不管它。
sp<IAudioTrack> track = audioFlinger->createTrack(getpid(),
streamType,
sampleRate,
format,
channelCount,
frameCount,
((uint16_t)flags) << 16,
sharedBuffer,
output,
&status);
//从track也就是AudioFlinger那边得到一个IMemory接口
//这个看来就是最终write写入的地方
sp<IMemory> cblk = track->getCblk();
mAudioTrack.clear();
mAudioTrack = track;
mCblkMemory.clear();
mCblkMemory = cblk;
mCblk = static_cast<audio_track_cblk_t*>(cblk->pointer());
mCblk->out = 1;
mFrameCount = mCblk->frameCount;
if (sharedBuffer == 0) {
//终于看到buffer相关的了。注意我们这里的情况STREAM模式没有传入共享buffer,但是数据确实又需要buffer承载。反正AudioTrack是没有创建buffer,那只能是刚才从AudioFlinger中得到的buffer了。
mCblk->buffers = (char*)mCblk + sizeof(audio_track_cblk_t);
}
return NO_ERROR;
}
10 audioFlinger->createTrack
AudioTrack得到AudioFlinger中的一个IAudioTrack对象。
AudioTrack创建了一个线程,叫AudioTrackThread。
AudioTrack调用write函数,肯定是把数据写到那块共享缓冲了,然后IAudioTrack在另外一个进程AudioFlinger中(其实AudioFlinger是一个服务,在mediaservice中运行)接收数据,并最终写到音频设备中。
11 new AudioTrackThread
创建了一个线程
AudioTrack::AudioTrackThread::AudioTrackThread(AudioTrack& receiver, bool bCanCallJava)
: Thread(bCanCallJava), mReceiver(receiver)
{ //mReceiver就是AudioTrack对象
// bCanCallJava为TRUE
}
12 play
13 android_media_AudioTrack_start
play函数对应的JNI函数android_media_AudioTrack_start
static void android_media_AudioTrack_start(JNIEnv *env, jobject thiz)
{
//从JAVA那个AudioTrack对象获取保存的C++层的AudioTrack对象指针
AudioTrack *lpTrack = (AudioTrack *)env->GetIntField(
thiz, javaAudioTrackFields.nativeTrackInJavaObj);
lpTrack->start();
}
14 start
start函数调用AudioTrackThread启动一个新的线程,执行mAudioTrackThread的threadLoop
void AudioTrack::start()
{
//start函数调用AudioTrackThread函数触发产生一个新的线程,执行mAudioTrackThread的threadLoop
sp<AudioTrackThread> t = mAudioTrackThread;
t->run("AudioTrackThread", THREAD_PRIORITY_AUDIO_CLIENT);
//让AudioFlinger中的track也start
status_t status = mAudioTrack->start();
}
15 threadLoop()
bool AudioTrack::AudioTrackThread::threadLoop()
{
//调用AudioTrack的processAudioBuffer函数
return mReceiver.processAudioBuffer(this);
}
16 processAudioBuffer
bool AudioTrack::processAudioBuffer(const sp<AudioTrackThread>& thread)
{
Buffer audioBuffer;
uint32_t frames;
size_t writtenSize;
//...回调1
mCbf(EVENT_UNDERRUN, mUserData, 0);
//...回调2 都是传递一些信息到JNI里边
mCbf(EVENT_BUFFER_END, mUserData, 0);
// Manage loop end callback
while (mLoopCount > mCblk->loopCount) {
mCbf(EVENT_LOOP_END, mUserData, (void *)&loopCount);
}
//下面好像有写数据的东西
do {
audioBuffer.frameCount = frames;
//获得buffer,
status_t err = obtainBuffer(&audioBuffer, 1);
size_t reqSize = audioBuffer.size;
//把buffer回调到JNI那去,这是单独一个线程,而我们还有上层用户在那不停地write呢,怎么会这样?
mCbf(EVENT_MORE_DATA, mUserData, &audioBuffer);
audioBuffer.size = writtenSize;
frames -= audioBuffer.frameCount;
releaseBuffer(&audioBuffer);
}
while (frames);
return true;
}
17 mCbf(EVENT_MORE_DATA, mUserData, &audioBuffer)
static void audioCallback(int event, void* user, void *info) {
if (event == AudioTrack::EVENT_MORE_DATA) {
//这个函数没往里边写数据
AudioTrack::Buffer* pBuff = (AudioTrack::Buffer*)info;
pBuff->size = 0;
}
}
从代码上看,本来google考虑是异步的回调方式来写数据,实际上没write,这个AudioTrackThread除了通知一下,也没什么实际有意义的操作了。
18 write
19 android_media_AudioTrack_native_write_short
20 writeToTrack
最终会调到重要函数writeToTrack
jint writeToTrack(AudioTrack* pTrack, jint audioFormat, jbyte* data, jint offsetInBytes, jint sizeInBytes) {
ssize_t written = 0;
//regular write() or copy the data to the AudioTrack's shared memory?
if (pTrack->sharedBuffer() == 0) {
//创建的是流的方式,所以没有共享内存在track中。还记得我们在native_setup中调用的set吗?流模式下AudioTrackJniStorage可没创建共享内存
written = pTrack->write(data + offsetInBytes, sizeInBytes);
} else {
if (audioFormat == javaAudioTrackFields.PCM16) {
//writing to shared memory, check for capacity
if ((size_t)sizeInBytes > pTrack->sharedBuffer()->size()) {
sizeInBytes = pTrack->sharedBuffer()->size();
}
//STATIC模式的,就直接把数据拷贝到共享内存里
//当然,这个共享内存是pTrack的,是我们在set时候把AudioTrackJniStorage的共享设进去的
memcpy(pTrack->sharedBuffer()->pointer(), data + offsetInBytes, sizeInBytes);
written = sizeInBytes;
} else if (audioFormat == javaAudioTrackFields.PCM8) {
//PCM8格式的要先转换成PCM16
...
}
return written;
}
21 pTrack->write
就是obtainBuffer,memcpy数据,然后releaseBuffer
ssize_t AudioTrack::write(const void* buffer, size_t userSize)
{
//就是obtainBuffer,memcpy数据,然后releaseBuffer
do {
audioBuffer.frameCount = userSize/frameSize();
status_t err = obtainBuffer(&audioBuffer, -1);
size_t toWrite;
toWrite = audioBuffer.size;
memcpy(audioBuffer.i8, src, toWrite);
src += toWrite;
}
userSize -= toWrite;
written += toWrite;
releaseBuffer(&audioBuffer);
} while (userSize);
return written;
}
AudioTrack总结:
1 通过AudioTrack的介绍,我们给后续深入分析AudioFlinger提供了一个切入点。
2 AudioTrack被new出来,然后set了一堆信息,同时会通过Binder机制调用另外一端的AudioFlinger,得到IAudioTrack对象,通过它和AudioFlinger交互。
3 调用start函数后,会启动一个线程专门做回调处理,代码里边也会有那种数据拷贝的回调,但是JNI层的回调函数实际并没有往里边write数据。真正的写数据如图18:write。
4 用户一次次地write,AudioTrack把数据memcpy到共享buffer中
5 可想而知,AudioFlinger那一定有一个线程在memcpy数据到音频设备中去。我们拭目以待。