(一)Audio子系统之AudioRecord.getMinBufferSize
在文章《基于Allwinner的Audio子系统分析(Android-5.1)》中已经介绍了Audio的系统架构以及应用层调用的流程,接下来,继续分析AudioRecorder方法中的getMinBufferSize的实现
函数原型:
public static int getMinBufferSize (int sampleRateInHz, int channelConfig, int audioFormat)
作用:
返回成功创建AudioRecord对象所需要的最小缓冲区大小
参数:
sampleRateInHz:默认采样率,单位Hz,这里设置为44100,44100Hz是当前唯一能保证在所有设备上工作的采样率;
channelConfig: 描述音频声道设置,这里设置为AudioFormat.CHANNEL_CONFIGURATION_MONO,CHANNEL_CONFIGURATION_MONO保证能在所有设备上工作;
audioFormat:音频数据的采样精度,这里设置为AudioFormat.ENCODING_16BIT;
返回值:
返回成功创建AudioRecord对象所需要的最小缓冲区大小。 注意:这个大小并不保证在负荷下的流畅录制,应根据预期的频率来选择更高的值,AudioRecord实例在推送新数据时使用此值
如果硬件不支持录制参数,或输入了一个无效的参数,则返回ERROR_BAD_VALUE(-2),如果硬件查询到输出属性没有实现,或最小缓冲区用byte表示,则返回ERROR(-1)
接下来进入系统分析具体实现
frameworks/base/media/java/android/media/AudioRecord.java
static public int getMinBufferSize(int sampleRateInHz, int channelConfig, int audioFormat) { int channelCount = 0; switch (channelConfig) { case AudioFormat.CHANNEL_IN_DEFAULT: // AudioFormat.CHANNEL_CONFIGURATION_DEFAULT //1 case AudioFormat.CHANNEL_IN_MONO: //16 case AudioFormat.CHANNEL_CONFIGURATION_MONO://2 channelCount = 1; break; case AudioFormat.CHANNEL_IN_STEREO: //12 case AudioFormat.CHANNEL_CONFIGURATION_STEREO://3 case (AudioFormat.CHANNEL_IN_FRONT | AudioFormat.CHANNEL_IN_BACK): // 16||32 channelCount = 2; break; case AudioFormat.CHANNEL_INVALID://0 default: loge("getMinBufferSize(): Invalid channel configuration."); return ERROR_BAD_VALUE; } // PCM_8BIT is not supported at the moment if (audioFormat != AudioFormat.ENCODING_PCM_16BIT) { loge("getMinBufferSize(): Invalid audio format."); return ERROR_BAD_VALUE; } int size = native_get_min_buff_size(sampleRateInHz, channelCount, audioFormat); if (size == 0) { return ERROR_BAD_VALUE; } else if (size == -1) { return ERROR; } else { return size; } }
对音频通道与音频采样精度进行判断,单声道(MONO)时channelCount为1,立体声(STEREO)时channelCount为2,且A64上仅支持PCM_16BIT采样,其值为2,然后继续调用native函数
frameworks/base/core/jni/android_media_AudioRecord.cpp
static jint android_media_AudioRecord_get_min_buff_size(JNIEnv *env, jobject thiz, jint sampleRateInHertz, jint channelCount, jint audioFormat) { ALOGV(">> android_media_AudioRecord_get_min_buff_size(%d, %d, %d)", sampleRateInHertz, channelCount, audioFormat); size_t frameCount = 0; //从java转成jni的format类型 audio_format_t format = audioFormatToNative(audioFormat);//AUDIO_FORMAT_PCM_16_BIT=0x1 //获取frameCount,并判断硬件是否支持 status_t result = AudioRecord::getMinFrameCount(&frameCount, sampleRateInHertz, format, audio_channel_in_mask_from_count(channelCount)); if (result == BAD_VALUE) { return 0; } if (result != NO_ERROR) { return -1; } return frameCount * channelCount * audio_bytes_per_sample(format); }
调用服务端的函数,获取frameCount大小,最后返回了frameCount*声道数*采样精度,其中frameCount表示最小采样帧数,继续分析frameCount的计算方法
frameworks/av/media/libmedia/AudioRecord.cpp
status_t AudioRecord::getMinFrameCount( size_t* frameCount, uint32_t sampleRate, audio_format_t format, audio_channel_mask_t channelMask) { if (frameCount == NULL) { return BAD_VALUE; } size_t size; status_t status = AudioSystem::getInputBufferSize(sampleRate, format, channelMask, &size); if (status != NO_ERROR) { ALOGE("AudioSystem could not query the input buffer size for sampleRate %u, format %#x, " "channelMask %#x; status %d", sampleRate, format, channelMask, status); return status; } //计算出最小的frame // We double the size of input buffer for ping pong use of record buffer. // Assumes audio_is_linear_pcm(format) if ((*frameCount = (size * 2) / (audio_channel_count_from_in_mask(channelMask) * audio_bytes_per_sample(format))) == 0) { ALOGE("Unsupported configuration: sampleRate %u, format %#x, channelMask %#x", sampleRate, format, channelMask); return BAD_VALUE; } return NO_ERROR; }
此时frameCount= size*2/(声道数*采样精度),注意这里需要double一下,而size是由hal层得到的,AudioSystem::getInputBufferSize()函数最终会调用到HAL层
frameworks/av/media/libmedia/AudioSystem.cpp
status_t AudioSystem::getInputBufferSize(uint32_t sampleRate, audio_format_t format, audio_channel_mask_t channelMask, size_t* buffSize) { const sp<IAudioFlinger>& af = AudioSystem::get_audio_flinger(); if (af == 0) { return PERMISSION_DENIED; } Mutex::Autolock _l(gLockCache); // Do we have a stale gInBufferSize or are we requesting the input buffer size for new values size_t inBuffSize = gInBuffSize; if ((inBuffSize == 0) || (sampleRate != gPrevInSamplingRate) || (format != gPrevInFormat) || (channelMask != gPrevInChannelMask)) { gLockCache.unlock(); inBuffSize = af->getInputBufferSize(sampleRate, format, channelMask); gLockCache.lock(); if (inBuffSize == 0) { ALOGE("AudioSystem::getInputBufferSize failed sampleRate %d format %#x channelMask %x", sampleRate, format, channelMask); return BAD_VALUE; } // A benign race is possible here: we could overwrite a fresher cache entry // save the request params gPrevInSamplingRate = sampleRate; gPrevInFormat = format; gPrevInChannelMask = channelMask; gInBuffSize = inBuffSize; } *buffSize = inBuffSize; return NO_ERROR; }
这里通过get_audio_flinger获取到了一个AudioFlinger对象
const sp<IAudioFlinger> AudioSystem::get_audio_flinger() { sp<IAudioFlinger> af; sp<AudioFlingerClient> afc; { Mutex::Autolock _l(gLock); if (gAudioFlinger == 0) { sp<IServiceManager> sm = defaultServiceManager(); sp<IBinder> binder; do { binder = sm->getService(String16("media.audio_flinger")); if (binder != 0) break; ALOGW("AudioFlinger not published, waiting..."); usleep(500000); // 0.5 s } while (true); if (gAudioFlingerClient == NULL) { gAudioFlingerClient = new AudioFlingerClient(); } else { if (gAudioErrorCallback) { gAudioErrorCallback(NO_ERROR); } } binder->linkToDeath(gAudioFlingerClient); gAudioFlinger = interface_cast<IAudioFlinger>(binder); LOG_ALWAYS_FATAL_IF(gAudioFlinger == 0); afc = gAudioFlingerClient; } af = gAudioFlinger; } if (afc != 0) { af->registerClient(afc); } return af; }
然后判断是否参数是之前配置过的参数,这样做是为了防止重复多次调用getMinBufferSize导致占用硬件资源,所以当第一次调用或更新参数调用后,则调用AF中的getInputBufferSize方法获取BuffSize,而af是IAudioFlinger类型的智能指针,所以实际上会通过binder到达AudioFlinger中
frameworks\av\services\audioflinger\AudioFlinger.cpp
size_t AudioFlinger::getInputBufferSize(uint32_t sampleRate, audio_format_t format, audio_channel_mask_t channelMask) const { status_t ret = initCheck(); if (ret != NO_ERROR) { return 0; } AutoMutex lock(mHardwareLock); mHardwareStatus = AUDIO_HW_GET_INPUT_BUFFER_SIZE; audio_config_t config; memset(&config, 0, sizeof(config)); config.sample_rate = sampleRate; config.channel_mask = channelMask; config.format = format; audio_hw_device_t *dev = mPrimaryHardwareDev->hwDevice(); size_t size = dev->get_input_buffer_size(dev, &config); mHardwareStatus = AUDIO_HW_IDLE; return size; }
把参数传递给hal层,获取buffer大小
hardware\aw\audio\tulip\audio_hw.c
static size_t adev_get_input_buffer_size(const struct audio_hw_device *dev, const struct audio_config *config) { size_t size; int channel_count = popcount(config->channel_mask); if (check_input_parameters(config->sample_rate, config->format, channel_count) != 0) return 0; return get_input_buffer_size(config->sample_rate, config->format, channel_count); }
再次检查一次参数是否正确,为什么在很多函数里面都做一次检查参数呢?可能在其他的地方也调用到了这个函数,所以最好做一次检查,确保万无一失
static size_t get_input_buffer_size(uint32_t sample_rate, int format, int channel_count) { size_t size; size_t device_rate; if (check_input_parameters(sample_rate, format, channel_count) != 0) return 0; /* take resampling into account and return the closest majoring multiple of 16 frames, as audioflinger expects audio buffers to be a multiple of 16 frames */ size = (pcm_config_mm_in.period_size * sample_rate) / pcm_config_mm_in.rate; size = ((size + 15) / 16) * 16; return size * channel_count * sizeof(short); }
这里包含一个结构体struct pcm_config,定义了一个周期包含了多少采样帧,并根据结构体的rate数据进行重采样计算,这里的rate是以MM_SAMPLING_RATE为标准,即44100,一个采样周期有1024个采样帧,然后计算出重采样之后的size
同时audioflinger的音频buffer是16的整数倍,所以再次计算得出一个最接近16倍的整数,最后返回size*声道数*1帧数据所占字节数
struct pcm_config pcm_config_mm_in = { .channels = 2, .rate = MM_SAMPLING_RATE, .period_size = 1024, .period_count = CAPTURE_PERIOD_COUNT, .format = PCM_FORMAT_S16_LE, };
总结:
minBuffSize = ((((((((pcm_config_mm_in.period_size * sample_rate) / pcm_config_mm_in.rate) + 15) / 16) * 16) * channel_count * sizeof(short)) * 2) / (audio_channel_count_from_in_mask(channelMask) * audio_bytes_per_sample(format))) * channelCount * audio_bytes_per_sample(format);
=(((((((pcm_config_mm_in.period_size * sample_rate) / pcm_config_mm_in.rate) + 15) / 16) * 16) * channel_count * sizeof(short)) * 2)
其中:pcm_config_mm_in.period_size=1024;pcm_config_mm_in.rate=44100;这里我们可以看到他除掉(channelCount*format),后面又乘回来了,这个是因为在AudioRecord.cpp对frameCount进行了一次校验,判断是否支持该参数的设置。
以getMinBufferSize(44100, MONO, 16BIT);为例,即sample_rate=44100,channel_count=1, format=2,那么
BufferSize = (((1024*sample_rate/44100)+15)/16)*16*channel_count*sizeof(short)*2 = 4096
即最小缓冲区大小为:周期大小 * 重采样 * 采样声道数 * 2 * 采样精度所占字节数;这里的2的解释为We double the size of input buffer for ping pong use of record buffer,采样精度:PCM_8_BIT为unsigned char,PCM_16_BIT为short,PCM_32_BIT为int。
由于作者内功有限,若文章中存在错误或不足的地方,还请给位大佬指出,不胜感激!
作者:pngcui
博客园:http://www.cnblogs.com/pngcui/
github:https://github.com/pngcui
本文版权归作者和博客园共有,欢迎转载,但未经作者同意必须保留此段声明。