1 // Copyright (c) 2012 The Chromium Authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style license that can be 3 // found in the LICENSE file. 4 5 #include "media/audio/win/audio_low_latency_input_win.h" 6 7 #include "base/logging.h" 8 #include "base/memory/scoped_ptr.h" 9 #include "base/strings/utf_string_conversions.h" 10 #include "media/audio/win/audio_manager_win.h" 11 #include "media/audio/win/avrt_wrapper_win.h" 12 #include "media/base/audio_bus.h" 13 14 using base::win::ScopedComPtr; 15 using base::win::ScopedCOMInitializer; 16 17 namespace media { 18 namespace { 19 20 // Returns true if |device| represents the default communication capture device. 21 bool IsDefaultCommunicationDevice(IMMDeviceEnumerator* enumerator, 22 IMMDevice* device) { 23 ScopedComPtr<IMMDevice> communications; 24 if (FAILED(enumerator->GetDefaultAudioEndpoint(eCapture, eCommunications, 25 communications.Receive()))) { 26 return false; 27 } 28 29 base::win::ScopedCoMem<WCHAR> communications_id, device_id; 30 device->GetId(&device_id); 31 communications->GetId(&communications_id); 32 return lstrcmpW(communications_id, device_id) == 0; 33 } 34 35 } // namespace 36 37 WASAPIAudioInputStream::WASAPIAudioInputStream(AudioManagerWin* manager, 38 const AudioParameters& params, 39 const std::string& device_id) 40 : manager_(manager), 41 capture_thread_(NULL), 42 opened_(false), 43 started_(false), 44 frame_size_(0), 45 packet_size_frames_(0), 46 packet_size_bytes_(0), 47 endpoint_buffer_size_frames_(0), 48 effects_(params.effects()), 49 device_id_(device_id), 50 perf_count_to_100ns_units_(0.0), 51 ms_to_frame_count_(0.0), 52 sink_(NULL), 53 audio_bus_(media::AudioBus::Create(params)) { 54 DCHECK(manager_); 55 56 // Load the Avrt DLL if not already loaded. Required to support MMCSS. 57 bool avrt_init = avrt::Initialize(); 58 DCHECK(avrt_init) << "Failed to load the Avrt.dll"; 59 60 // Set up the desired capture format specified by the client. 61 format_.nSamplesPerSec = params.sample_rate(); 62 format_.wFormatTag = WAVE_FORMAT_PCM; 63 format_.wBitsPerSample = params.bits_per_sample(); 64 format_.nChannels = params.channels(); 65 format_.nBlockAlign = (format_.wBitsPerSample / 8) * format_.nChannels; 66 format_.nAvgBytesPerSec = format_.nSamplesPerSec * format_.nBlockAlign; 67 format_.cbSize = 0; 68 69 // Size in bytes of each audio frame. 70 frame_size_ = format_.nBlockAlign; 71 // Store size of audio packets which we expect to get from the audio 72 // endpoint device in each capture event. 73 packet_size_frames_ = params.GetBytesPerBuffer() / format_.nBlockAlign; 74 packet_size_bytes_ = params.GetBytesPerBuffer(); 75 DVLOG(1) << "Number of bytes per audio frame : " << frame_size_; 76 DVLOG(1) << "Number of audio frames per packet: " << packet_size_frames_; 77 78 // All events are auto-reset events and non-signaled initially. 79 80 // Create the event which the audio engine will signal each time 81 // a buffer becomes ready to be processed by the client. 82 audio_samples_ready_event_.Set(CreateEvent(NULL, FALSE, FALSE, NULL)); 83 DCHECK(audio_samples_ready_event_.IsValid()); 84 85 // Create the event which will be set in Stop() when capturing shall stop. 86 stop_capture_event_.Set(CreateEvent(NULL, FALSE, FALSE, NULL)); 87 DCHECK(stop_capture_event_.IsValid()); 88 89 ms_to_frame_count_ = static_cast<double>(params.sample_rate()) / 1000.0; 90 91 LARGE_INTEGER performance_frequency; 92 if (QueryPerformanceFrequency(&performance_frequency)) { 93 perf_count_to_100ns_units_ = 94 (10000000.0 / static_cast<double>(performance_frequency.QuadPart)); 95 } else { 96 DLOG(ERROR) << "High-resolution performance counters are not supported."; 97 } 98 } 99 100 WASAPIAudioInputStream::~WASAPIAudioInputStream() {} 101 102 bool WASAPIAudioInputStream::Open() { 103 DCHECK(CalledOnValidThread()); 104 // Verify that we are not already opened. 105 if (opened_) 106 return false; 107 108 // Obtain a reference to the IMMDevice interface of the capturing 109 // device with the specified unique identifier or role which was 110 // set at construction. 111 HRESULT hr = SetCaptureDevice(); 112 if (FAILED(hr)) 113 return false; 114 115 // Obtain an IAudioClient interface which enables us to create and initialize 116 // an audio stream between an audio application and the audio engine. 117 hr = ActivateCaptureDevice(); 118 if (FAILED(hr)) 119 return false; 120 121 // Retrieve the stream format which the audio engine uses for its internal 122 // processing/mixing of shared-mode streams. This function call is for 123 // diagnostic purposes only and only in debug mode. 124 #ifndef NDEBUG 125 hr = GetAudioEngineStreamFormat(); 126 #endif 127 128 // Verify that the selected audio endpoint supports the specified format 129 // set during construction. 130 if (!DesiredFormatIsSupported()) 131 return false; 132 133 // Initialize the audio stream between the client and the device using 134 // shared mode and a lowest possible glitch-free latency. 135 hr = InitializeAudioEngine(); 136 137 opened_ = SUCCEEDED(hr); 138 return opened_; 139 } 140 141 void WASAPIAudioInputStream::Start(AudioInputCallback* callback) { 142 DCHECK(CalledOnValidThread()); 143 DCHECK(callback); 144 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 145 if (!opened_) 146 return; 147 148 if (started_) 149 return; 150 151 DCHECK(!sink_); 152 sink_ = callback; 153 154 // Starts periodic AGC microphone measurements if the AGC has been enabled 155 // using SetAutomaticGainControl(). 156 StartAgc(); 157 158 // Create and start the thread that will drive the capturing by waiting for 159 // capture events. 160 capture_thread_ = 161 new base::DelegateSimpleThread(this, "wasapi_capture_thread"); 162 capture_thread_->Start(); 163 164 // Start streaming data between the endpoint buffer and the audio engine. 165 HRESULT hr = audio_client_->Start(); 166 DLOG_IF(ERROR, FAILED(hr)) << "Failed to start input streaming."; 167 168 if (SUCCEEDED(hr) && audio_render_client_for_loopback_) 169 hr = audio_render_client_for_loopback_->Start(); 170 171 started_ = SUCCEEDED(hr); 172 } 173 174 void WASAPIAudioInputStream::Stop() { 175 DCHECK(CalledOnValidThread()); 176 DVLOG(1) << "WASAPIAudioInputStream::Stop()"; 177 if (!started_) 178 return; 179 180 // Stops periodic AGC microphone measurements. 181 StopAgc(); 182 183 // Shut down the capture thread. 184 if (stop_capture_event_.IsValid()) { 185 SetEvent(stop_capture_event_.Get()); 186 } 187 188 // Stop the input audio streaming. 189 HRESULT hr = audio_client_->Stop(); 190 if (FAILED(hr)) { 191 LOG(ERROR) << "Failed to stop input streaming."; 192 } 193 194 // Wait until the thread completes and perform cleanup. 195 if (capture_thread_) { 196 SetEvent(stop_capture_event_.Get()); 197 capture_thread_->Join(); 198 capture_thread_ = NULL; 199 } 200 201 started_ = false; 202 sink_ = NULL; 203 } 204 205 void WASAPIAudioInputStream::Close() { 206 DVLOG(1) << "WASAPIAudioInputStream::Close()"; 207 // It is valid to call Close() before calling open or Start(). 208 // It is also valid to call Close() after Start() has been called. 209 Stop(); 210 211 // Inform the audio manager that we have been closed. This will cause our 212 // destruction. 213 manager_->ReleaseInputStream(this); 214 } 215 216 double WASAPIAudioInputStream::GetMaxVolume() { 217 // Verify that Open() has been called succesfully, to ensure that an audio 218 // session exists and that an ISimpleAudioVolume interface has been created. 219 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 220 if (!opened_) 221 return 0.0; 222 223 // The effective volume value is always in the range 0.0 to 1.0, hence 224 // we can return a fixed value (=1.0) here. 225 return 1.0; 226 } 227 228 void WASAPIAudioInputStream::SetVolume(double volume) { 229 DVLOG(1) << "SetVolume(volume=" << volume << ")"; 230 DCHECK(CalledOnValidThread()); 231 DCHECK_GE(volume, 0.0); 232 DCHECK_LE(volume, 1.0); 233 234 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 235 if (!opened_) 236 return; 237 238 // Set a new master volume level. Valid volume levels are in the range 239 // 0.0 to 1.0. Ignore volume-change events. 240 HRESULT hr = simple_audio_volume_->SetMasterVolume(static_cast<float>(volume), 241 NULL); 242 DLOG_IF(WARNING, FAILED(hr)) << "Failed to set new input master volume."; 243 244 // Update the AGC volume level based on the last setting above. Note that, 245 // the volume-level resolution is not infinite and it is therefore not 246 // possible to assume that the volume provided as input parameter can be 247 // used directly. Instead, a new query to the audio hardware is required. 248 // This method does nothing if AGC is disabled. 249 UpdateAgcVolume(); 250 } 251 252 double WASAPIAudioInputStream::GetVolume() { 253 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 254 if (!opened_) 255 return 0.0; 256 257 // Retrieve the current volume level. The value is in the range 0.0 to 1.0. 258 float level = 0.0f; 259 HRESULT hr = simple_audio_volume_->GetMasterVolume(&level); 260 DLOG_IF(WARNING, FAILED(hr)) << "Failed to get input master volume."; 261 262 return static_cast<double>(level); 263 } 264 265 // static 266 AudioParameters WASAPIAudioInputStream::GetInputStreamParameters( 267 const std::string& device_id) { 268 int sample_rate = 48000; 269 ChannelLayout channel_layout = CHANNEL_LAYOUT_STEREO; 270 271 base::win::ScopedCoMem<WAVEFORMATEX> audio_engine_mix_format; 272 int effects = AudioParameters::NO_EFFECTS; 273 if (SUCCEEDED(GetMixFormat(device_id, &audio_engine_mix_format, &effects))) { 274 sample_rate = static_cast<int>(audio_engine_mix_format->nSamplesPerSec); 275 channel_layout = audio_engine_mix_format->nChannels == 1 ? 276 CHANNEL_LAYOUT_MONO : CHANNEL_LAYOUT_STEREO; 277 } 278 279 // Use 10ms frame size as default. 280 int frames_per_buffer = sample_rate / 100; 281 return AudioParameters( 282 AudioParameters::AUDIO_PCM_LOW_LATENCY, channel_layout, 0, sample_rate, 283 16, frames_per_buffer, effects); 284 } 285 286 // static 287 HRESULT WASAPIAudioInputStream::GetMixFormat(const std::string& device_id, 288 WAVEFORMATEX** device_format, 289 int* effects) { 290 DCHECK(effects); 291 292 // It is assumed that this static method is called from a COM thread, i.e., 293 // CoInitializeEx() is not called here to avoid STA/MTA conflicts. 294 ScopedComPtr<IMMDeviceEnumerator> enumerator; 295 HRESULT hr = enumerator.CreateInstance(__uuidof(MMDeviceEnumerator), NULL, 296 CLSCTX_INPROC_SERVER); 297 if (FAILED(hr)) 298 return hr; 299 300 ScopedComPtr<IMMDevice> endpoint_device; 301 if (device_id == AudioManagerBase::kDefaultDeviceId) { 302 // Retrieve the default capture audio endpoint. 303 hr = enumerator->GetDefaultAudioEndpoint(eCapture, eConsole, 304 endpoint_device.Receive()); 305 } else if (device_id == AudioManagerBase::kLoopbackInputDeviceId) { 306 // Get the mix format of the default playback stream. 307 hr = enumerator->GetDefaultAudioEndpoint(eRender, eConsole, 308 endpoint_device.Receive()); 309 } else { 310 // Retrieve a capture endpoint device that is specified by an endpoint 311 // device-identification string. 312 hr = enumerator->GetDevice(base::UTF8ToUTF16(device_id).c_str(), 313 endpoint_device.Receive()); 314 } 315 316 if (FAILED(hr)) 317 return hr; 318 319 *effects = IsDefaultCommunicationDevice(enumerator, endpoint_device) ? 320 AudioParameters::DUCKING : AudioParameters::NO_EFFECTS; 321 322 ScopedComPtr<IAudioClient> audio_client; 323 hr = endpoint_device->Activate(__uuidof(IAudioClient), 324 CLSCTX_INPROC_SERVER, 325 NULL, 326 audio_client.ReceiveVoid()); 327 return SUCCEEDED(hr) ? audio_client->GetMixFormat(device_format) : hr; 328 } 329 330 void WASAPIAudioInputStream::Run() { 331 ScopedCOMInitializer com_init(ScopedCOMInitializer::kMTA); 332 333 // Increase the thread priority. 334 capture_thread_->SetThreadPriority(base::kThreadPriority_RealtimeAudio); 335 336 // Enable MMCSS to ensure that this thread receives prioritized access to 337 // CPU resources. 338 DWORD task_index = 0; 339 HANDLE mm_task = avrt::AvSetMmThreadCharacteristics(L"Pro Audio", 340 &task_index); 341 bool mmcss_is_ok = 342 (mm_task && avrt::AvSetMmThreadPriority(mm_task, AVRT_PRIORITY_CRITICAL)); 343 if (!mmcss_is_ok) { 344 // Failed to enable MMCSS on this thread. It is not fatal but can lead 345 // to reduced QoS at high load. 346 DWORD err = GetLastError(); 347 LOG(WARNING) << "Failed to enable MMCSS (error code=" << err << ")."; 348 } 349 350 // Allocate a buffer with a size that enables us to take care of cases like: 351 // 1) The recorded buffer size is smaller, or does not match exactly with, 352 // the selected packet size used in each callback. 353 // 2) The selected buffer size is larger than the recorded buffer size in 354 // each event. 355 size_t buffer_frame_index = 0; 356 size_t capture_buffer_size = std::max( 357 2 * endpoint_buffer_size_frames_ * frame_size_, 358 2 * packet_size_frames_ * frame_size_); 359 scoped_ptr<uint8[]> capture_buffer(new uint8[capture_buffer_size]); 360 361 LARGE_INTEGER now_count; 362 bool recording = true; 363 bool error = false; 364 double volume = GetVolume(); 365 HANDLE wait_array[2] = {stop_capture_event_, audio_samples_ready_event_}; 366 367 while (recording && !error) { 368 HRESULT hr = S_FALSE; 369 370 // Wait for a close-down event or a new capture event. 371 DWORD wait_result = WaitForMultipleObjects(2, wait_array, FALSE, INFINITE); 372 switch (wait_result) { 373 case WAIT_FAILED: 374 error = true; 375 break; 376 case WAIT_OBJECT_0 + 0: 377 // |stop_capture_event_| has been set. 378 recording = false; 379 break; 380 case WAIT_OBJECT_0 + 1: 381 { 382 // |audio_samples_ready_event_| has been set. 383 BYTE* data_ptr = NULL; 384 UINT32 num_frames_to_read = 0; 385 DWORD flags = 0; 386 UINT64 device_position = 0; 387 UINT64 first_audio_frame_timestamp = 0; 388 389 // Retrieve the amount of data in the capture endpoint buffer, 390 // replace it with silence if required, create callbacks for each 391 // packet and store non-delivered data for the next event. 392 hr = audio_capture_client_->GetBuffer(&data_ptr, 393 &num_frames_to_read, 394 &flags, 395 &device_position, 396 &first_audio_frame_timestamp); 397 if (FAILED(hr)) { 398 DLOG(ERROR) << "Failed to get data from the capture buffer"; 399 continue; 400 } 401 402 if (num_frames_to_read != 0) { 403 size_t pos = buffer_frame_index * frame_size_; 404 size_t num_bytes = num_frames_to_read * frame_size_; 405 DCHECK_GE(capture_buffer_size, pos + num_bytes); 406 407 if (flags & AUDCLNT_BUFFERFLAGS_SILENT) { 408 // Clear out the local buffer since silence is reported. 409 memset(&capture_buffer[pos], 0, num_bytes); 410 } else { 411 // Copy captured data from audio engine buffer to local buffer. 412 memcpy(&capture_buffer[pos], data_ptr, num_bytes); 413 } 414 415 buffer_frame_index += num_frames_to_read; 416 } 417 418 hr = audio_capture_client_->ReleaseBuffer(num_frames_to_read); 419 DLOG_IF(ERROR, FAILED(hr)) << "Failed to release capture buffer"; 420 421 // Derive a delay estimate for the captured audio packet. 422 // The value contains two parts (A+B), where A is the delay of the 423 // first audio frame in the packet and B is the extra delay 424 // contained in any stored data. Unit is in audio frames. 425 QueryPerformanceCounter(&now_count); 426 double audio_delay_frames = 427 ((perf_count_to_100ns_units_ * now_count.QuadPart - 428 first_audio_frame_timestamp) / 10000.0) * ms_to_frame_count_ + 429 buffer_frame_index - num_frames_to_read; 430 431 // Get a cached AGC volume level which is updated once every second 432 // on the audio manager thread. Note that, |volume| is also updated 433 // each time SetVolume() is called through IPC by the render-side AGC. 434 GetAgcVolume(&volume); 435 436 // Deliver captured data to the registered consumer using a packet 437 // size which was specified at construction. 438 uint32 delay_frames = static_cast<uint32>(audio_delay_frames + 0.5); 439 while (buffer_frame_index >= packet_size_frames_) { 440 // Copy data to audio bus to match the OnData interface. 441 uint8* audio_data = reinterpret_cast<uint8*>(capture_buffer.get()); 442 audio_bus_->FromInterleaved( 443 audio_data, audio_bus_->frames(), format_.wBitsPerSample / 8); 444 445 // Deliver data packet, delay estimation and volume level to 446 // the user. 447 sink_->OnData( 448 this, audio_bus_.get(), delay_frames * frame_size_, volume); 449 450 // Store parts of the recorded data which can't be delivered 451 // using the current packet size. The stored section will be used 452 // either in the next while-loop iteration or in the next 453 // capture event. 454 memmove(&capture_buffer[0], 455 &capture_buffer[packet_size_bytes_], 456 (buffer_frame_index - packet_size_frames_) * frame_size_); 457 458 buffer_frame_index -= packet_size_frames_; 459 delay_frames -= packet_size_frames_; 460 } 461 } 462 break; 463 default: 464 error = true; 465 break; 466 } 467 } 468 469 if (recording && error) { 470 // TODO(henrika): perhaps it worth improving the cleanup here by e.g. 471 // stopping the audio client, joining the thread etc.? 472 NOTREACHED() << "WASAPI capturing failed with error code " 473 << GetLastError(); 474 } 475 476 // Disable MMCSS. 477 if (mm_task && !avrt::AvRevertMmThreadCharacteristics(mm_task)) { 478 PLOG(WARNING) << "Failed to disable MMCSS"; 479 } 480 } 481 482 void WASAPIAudioInputStream::HandleError(HRESULT err) { 483 NOTREACHED() << "Error code: " << err; 484 if (sink_) 485 sink_->OnError(this); 486 } 487 488 HRESULT WASAPIAudioInputStream::SetCaptureDevice() { 489 DCHECK(!endpoint_device_); 490 491 ScopedComPtr<IMMDeviceEnumerator> enumerator; 492 HRESULT hr = enumerator.CreateInstance(__uuidof(MMDeviceEnumerator), 493 NULL, CLSCTX_INPROC_SERVER); 494 if (FAILED(hr)) 495 return hr; 496 497 // Retrieve the IMMDevice by using the specified role or the specified 498 // unique endpoint device-identification string. 499 500 if (effects_ & AudioParameters::DUCKING) { 501 // Ducking has been requested and it is only supported for the default 502 // communication device. So, let's open up the communication device and 503 // see if the ID of that device matches the requested ID. 504 // We consider a kDefaultDeviceId as well as an explicit device id match, 505 // to be valid matches. 506 hr = enumerator->GetDefaultAudioEndpoint(eCapture, eCommunications, 507 endpoint_device_.Receive()); 508 if (endpoint_device_ && device_id_ != AudioManagerBase::kDefaultDeviceId) { 509 base::win::ScopedCoMem<WCHAR> communications_id; 510 endpoint_device_->GetId(&communications_id); 511 if (device_id_ != 512 base::WideToUTF8(static_cast<WCHAR*>(communications_id))) { 513 DLOG(WARNING) << "Ducking has been requested for a non-default device." 514 "Not supported."; 515 endpoint_device_.Release(); // Fall back on code below. 516 } 517 } 518 } 519 520 if (!endpoint_device_) { 521 if (device_id_ == AudioManagerBase::kDefaultDeviceId) { 522 // Retrieve the default capture audio endpoint for the specified role. 523 // Note that, in Windows Vista, the MMDevice API supports device roles 524 // but the system-supplied user interface programs do not. 525 hr = enumerator->GetDefaultAudioEndpoint(eCapture, eConsole, 526 endpoint_device_.Receive()); 527 } else if (device_id_ == AudioManagerBase::kLoopbackInputDeviceId) { 528 // Capture the default playback stream. 529 hr = enumerator->GetDefaultAudioEndpoint(eRender, eConsole, 530 endpoint_device_.Receive()); 531 } else { 532 hr = enumerator->GetDevice(base::UTF8ToUTF16(device_id_).c_str(), 533 endpoint_device_.Receive()); 534 } 535 } 536 537 if (FAILED(hr)) 538 return hr; 539 540 // Verify that the audio endpoint device is active, i.e., the audio 541 // adapter that connects to the endpoint device is present and enabled. 542 DWORD state = DEVICE_STATE_DISABLED; 543 hr = endpoint_device_->GetState(&state); 544 if (FAILED(hr)) 545 return hr; 546 547 if (!(state & DEVICE_STATE_ACTIVE)) { 548 DLOG(ERROR) << "Selected capture device is not active."; 549 hr = E_ACCESSDENIED; 550 } 551 552 return hr; 553 } 554 555 HRESULT WASAPIAudioInputStream::ActivateCaptureDevice() { 556 // Creates and activates an IAudioClient COM object given the selected 557 // capture endpoint device. 558 HRESULT hr = endpoint_device_->Activate(__uuidof(IAudioClient), 559 CLSCTX_INPROC_SERVER, 560 NULL, 561 audio_client_.ReceiveVoid()); 562 return hr; 563 } 564 565 HRESULT WASAPIAudioInputStream::GetAudioEngineStreamFormat() { 566 HRESULT hr = S_OK; 567 #ifndef NDEBUG 568 // The GetMixFormat() method retrieves the stream format that the 569 // audio engine uses for its internal processing of shared-mode streams. 570 // The method always uses a WAVEFORMATEXTENSIBLE structure, instead 571 // of a stand-alone WAVEFORMATEX structure, to specify the format. 572 // An WAVEFORMATEXTENSIBLE structure can specify both the mapping of 573 // channels to speakers and the number of bits of precision in each sample. 574 base::win::ScopedCoMem<WAVEFORMATEXTENSIBLE> format_ex; 575 hr = audio_client_->GetMixFormat( 576 reinterpret_cast<WAVEFORMATEX**>(&format_ex)); 577 578 // See http://msdn.microsoft.com/en-us/windows/hardware/gg463006#EFH 579 // for details on the WAVE file format. 580 WAVEFORMATEX format = format_ex->Format; 581 DVLOG(2) << "WAVEFORMATEX:"; 582 DVLOG(2) << " wFormatTags : 0x" << std::hex << format.wFormatTag; 583 DVLOG(2) << " nChannels : " << format.nChannels; 584 DVLOG(2) << " nSamplesPerSec : " << format.nSamplesPerSec; 585 DVLOG(2) << " nAvgBytesPerSec: " << format.nAvgBytesPerSec; 586 DVLOG(2) << " nBlockAlign : " << format.nBlockAlign; 587 DVLOG(2) << " wBitsPerSample : " << format.wBitsPerSample; 588 DVLOG(2) << " cbSize : " << format.cbSize; 589 590 DVLOG(2) << "WAVEFORMATEXTENSIBLE:"; 591 DVLOG(2) << " wValidBitsPerSample: " << 592 format_ex->Samples.wValidBitsPerSample; 593 DVLOG(2) << " dwChannelMask : 0x" << std::hex << 594 format_ex->dwChannelMask; 595 if (format_ex->SubFormat == KSDATAFORMAT_SUBTYPE_PCM) 596 DVLOG(2) << " SubFormat : KSDATAFORMAT_SUBTYPE_PCM"; 597 else if (format_ex->SubFormat == KSDATAFORMAT_SUBTYPE_IEEE_FLOAT) 598 DVLOG(2) << " SubFormat : KSDATAFORMAT_SUBTYPE_IEEE_FLOAT"; 599 else if (format_ex->SubFormat == KSDATAFORMAT_SUBTYPE_WAVEFORMATEX) 600 DVLOG(2) << " SubFormat : KSDATAFORMAT_SUBTYPE_WAVEFORMATEX"; 601 #endif 602 return hr; 603 } 604 605 bool WASAPIAudioInputStream::DesiredFormatIsSupported() { 606 // An application that uses WASAPI to manage shared-mode streams can rely 607 // on the audio engine to perform only limited format conversions. The audio 608 // engine can convert between a standard PCM sample size used by the 609 // application and the floating-point samples that the engine uses for its 610 // internal processing. However, the format for an application stream 611 // typically must have the same number of channels and the same sample 612 // rate as the stream format used by the device. 613 // Many audio devices support both PCM and non-PCM stream formats. However, 614 // the audio engine can mix only PCM streams. 615 base::win::ScopedCoMem<WAVEFORMATEX> closest_match; 616 HRESULT hr = audio_client_->IsFormatSupported(AUDCLNT_SHAREMODE_SHARED, 617 &format_, 618 &closest_match); 619 DLOG_IF(ERROR, hr == S_FALSE) << "Format is not supported " 620 << "but a closest match exists."; 621 return (hr == S_OK); 622 } 623 624 HRESULT WASAPIAudioInputStream::InitializeAudioEngine() { 625 DWORD flags; 626 // Use event-driven mode only fo regular input devices. For loopback the 627 // EVENTCALLBACK flag is specified when intializing 628 // |audio_render_client_for_loopback_|. 629 if (device_id_ == AudioManagerBase::kLoopbackInputDeviceId) { 630 flags = AUDCLNT_STREAMFLAGS_LOOPBACK | AUDCLNT_STREAMFLAGS_NOPERSIST; 631 } else { 632 flags = 633 AUDCLNT_STREAMFLAGS_EVENTCALLBACK | AUDCLNT_STREAMFLAGS_NOPERSIST; 634 } 635 636 // Initialize the audio stream between the client and the device. 637 // We connect indirectly through the audio engine by using shared mode. 638 // Note that, |hnsBufferDuration| is set of 0, which ensures that the 639 // buffer is never smaller than the minimum buffer size needed to ensure 640 // that glitches do not occur between the periodic processing passes. 641 // This setting should lead to lowest possible latency. 642 HRESULT hr = audio_client_->Initialize(AUDCLNT_SHAREMODE_SHARED, 643 flags, 644 0, // hnsBufferDuration 645 0, 646 &format_, 647 NULL); 648 if (FAILED(hr)) 649 return hr; 650 651 // Retrieve the length of the endpoint buffer shared between the client 652 // and the audio engine. The buffer length determines the maximum amount 653 // of capture data that the audio engine can read from the endpoint buffer 654 // during a single processing pass. 655 // A typical value is 960 audio frames <=> 20ms @ 48kHz sample rate. 656 hr = audio_client_->GetBufferSize(&endpoint_buffer_size_frames_); 657 if (FAILED(hr)) 658 return hr; 659 660 DVLOG(1) << "endpoint buffer size: " << endpoint_buffer_size_frames_ 661 << " [frames]"; 662 663 #ifndef NDEBUG 664 // The period between processing passes by the audio engine is fixed for a 665 // particular audio endpoint device and represents the smallest processing 666 // quantum for the audio engine. This period plus the stream latency between 667 // the buffer and endpoint device represents the minimum possible latency 668 // that an audio application can achieve. 669 // TODO(henrika): possibly remove this section when all parts are ready. 670 REFERENCE_TIME device_period_shared_mode = 0; 671 REFERENCE_TIME device_period_exclusive_mode = 0; 672 HRESULT hr_dbg = audio_client_->GetDevicePeriod( 673 &device_period_shared_mode, &device_period_exclusive_mode); 674 if (SUCCEEDED(hr_dbg)) { 675 DVLOG(1) << "device period: " 676 << static_cast<double>(device_period_shared_mode / 10000.0) 677 << " [ms]"; 678 } 679 680 REFERENCE_TIME latency = 0; 681 hr_dbg = audio_client_->GetStreamLatency(&latency); 682 if (SUCCEEDED(hr_dbg)) { 683 DVLOG(1) << "stream latency: " << static_cast<double>(latency / 10000.0) 684 << " [ms]"; 685 } 686 #endif 687 688 // Set the event handle that the audio engine will signal each time a buffer 689 // becomes ready to be processed by the client. 690 // 691 // In loopback case the capture device doesn't receive any events, so we 692 // need to create a separate playback client to get notifications. According 693 // to MSDN: 694 // 695 // A pull-mode capture client does not receive any events when a stream is 696 // initialized with event-driven buffering and is loopback-enabled. To 697 // work around this, initialize a render stream in event-driven mode. Each 698 // time the client receives an event for the render stream, it must signal 699 // the capture client to run the capture thread that reads the next set of 700 // samples from the capture endpoint buffer. 701 // 702 // http://msdn.microsoft.com/en-us/library/windows/desktop/dd316551(v=vs.85).aspx 703 if (device_id_ == AudioManagerBase::kLoopbackInputDeviceId) { 704 hr = endpoint_device_->Activate( 705 __uuidof(IAudioClient), CLSCTX_INPROC_SERVER, NULL, 706 audio_render_client_for_loopback_.ReceiveVoid()); 707 if (FAILED(hr)) 708 return hr; 709 710 hr = audio_render_client_for_loopback_->Initialize( 711 AUDCLNT_SHAREMODE_SHARED, 712 AUDCLNT_STREAMFLAGS_EVENTCALLBACK | AUDCLNT_STREAMFLAGS_NOPERSIST, 713 0, 0, &format_, NULL); 714 if (FAILED(hr)) 715 return hr; 716 717 hr = audio_render_client_for_loopback_->SetEventHandle( 718 audio_samples_ready_event_.Get()); 719 } else { 720 hr = audio_client_->SetEventHandle(audio_samples_ready_event_.Get()); 721 } 722 723 if (FAILED(hr)) 724 return hr; 725 726 // Get access to the IAudioCaptureClient interface. This interface 727 // enables us to read input data from the capture endpoint buffer. 728 hr = audio_client_->GetService(__uuidof(IAudioCaptureClient), 729 audio_capture_client_.ReceiveVoid()); 730 if (FAILED(hr)) 731 return hr; 732 733 // Obtain a reference to the ISimpleAudioVolume interface which enables 734 // us to control the master volume level of an audio session. 735 hr = audio_client_->GetService(__uuidof(ISimpleAudioVolume), 736 simple_audio_volume_.ReceiveVoid()); 737 return hr; 738 } 739 740 } // namespace media 741