audio_low_latency_input_win.cc revision a1401311d1ab56c4ed0a474bd38c108f75cb0cd9
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 13using base::win::ScopedComPtr; 14using base::win::ScopedCOMInitializer; 15 16namespace media { 17namespace { 18 19// Returns true if |device| represents the default communication capture device. 20bool IsDefaultCommunicationDevice(IMMDeviceEnumerator* enumerator, 21 IMMDevice* device) { 22 ScopedComPtr<IMMDevice> communications; 23 if (FAILED(enumerator->GetDefaultAudioEndpoint(eCapture, eCommunications, 24 communications.Receive()))) { 25 return false; 26 } 27 28 base::win::ScopedCoMem<WCHAR> communications_id, device_id; 29 device->GetId(&device_id); 30 communications->GetId(&communications_id); 31 return lstrcmpW(communications_id, device_id) == 0; 32} 33 34} // namespace 35 36WASAPIAudioInputStream::WASAPIAudioInputStream( 37 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 DCHECK(manager_); 54 55 // Load the Avrt DLL if not already loaded. Required to support MMCSS. 56 bool avrt_init = avrt::Initialize(); 57 DCHECK(avrt_init) << "Failed to load the Avrt.dll"; 58 59 // Set up the desired capture format specified by the client. 60 format_.nSamplesPerSec = params.sample_rate(); 61 format_.wFormatTag = WAVE_FORMAT_PCM; 62 format_.wBitsPerSample = params.bits_per_sample(); 63 format_.nChannels = params.channels(); 64 format_.nBlockAlign = (format_.wBitsPerSample / 8) * format_.nChannels; 65 format_.nAvgBytesPerSec = format_.nSamplesPerSec * format_.nBlockAlign; 66 format_.cbSize = 0; 67 68 // Size in bytes of each audio frame. 69 frame_size_ = format_.nBlockAlign; 70 // Store size of audio packets which we expect to get from the audio 71 // endpoint device in each capture event. 72 packet_size_frames_ = params.GetBytesPerBuffer() / format_.nBlockAlign; 73 packet_size_bytes_ = params.GetBytesPerBuffer(); 74 DVLOG(1) << "Number of bytes per audio frame : " << frame_size_; 75 DVLOG(1) << "Number of audio frames per packet: " << packet_size_frames_; 76 77 // All events are auto-reset events and non-signaled initially. 78 79 // Create the event which the audio engine will signal each time 80 // a buffer becomes ready to be processed by the client. 81 audio_samples_ready_event_.Set(CreateEvent(NULL, FALSE, FALSE, NULL)); 82 DCHECK(audio_samples_ready_event_.IsValid()); 83 84 // Create the event which will be set in Stop() when capturing shall stop. 85 stop_capture_event_.Set(CreateEvent(NULL, FALSE, FALSE, NULL)); 86 DCHECK(stop_capture_event_.IsValid()); 87 88 ms_to_frame_count_ = static_cast<double>(params.sample_rate()) / 1000.0; 89 90 LARGE_INTEGER performance_frequency; 91 if (QueryPerformanceFrequency(&performance_frequency)) { 92 perf_count_to_100ns_units_ = 93 (10000000.0 / static_cast<double>(performance_frequency.QuadPart)); 94 } else { 95 DLOG(ERROR) << "High-resolution performance counters are not supported."; 96 } 97} 98 99WASAPIAudioInputStream::~WASAPIAudioInputStream() {} 100 101bool WASAPIAudioInputStream::Open() { 102 DCHECK(CalledOnValidThread()); 103 // Verify that we are not already opened. 104 if (opened_) 105 return false; 106 107 // Obtain a reference to the IMMDevice interface of the capturing 108 // device with the specified unique identifier or role which was 109 // set at construction. 110 HRESULT hr = SetCaptureDevice(); 111 if (FAILED(hr)) 112 return false; 113 114 // Obtain an IAudioClient interface which enables us to create and initialize 115 // an audio stream between an audio application and the audio engine. 116 hr = ActivateCaptureDevice(); 117 if (FAILED(hr)) 118 return false; 119 120 // Retrieve the stream format which the audio engine uses for its internal 121 // processing/mixing of shared-mode streams. This function call is for 122 // diagnostic purposes only and only in debug mode. 123#ifndef NDEBUG 124 hr = GetAudioEngineStreamFormat(); 125#endif 126 127 // Verify that the selected audio endpoint supports the specified format 128 // set during construction. 129 if (!DesiredFormatIsSupported()) 130 return false; 131 132 // Initialize the audio stream between the client and the device using 133 // shared mode and a lowest possible glitch-free latency. 134 hr = InitializeAudioEngine(); 135 136 opened_ = SUCCEEDED(hr); 137 return opened_; 138} 139 140void WASAPIAudioInputStream::Start(AudioInputCallback* callback) { 141 DCHECK(CalledOnValidThread()); 142 DCHECK(callback); 143 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 144 if (!opened_) 145 return; 146 147 if (started_) 148 return; 149 150 DCHECK(!sink_); 151 sink_ = callback; 152 153 // Starts periodic AGC microphone measurements if the AGC has been enabled 154 // using SetAutomaticGainControl(). 155 StartAgc(); 156 157 // Create and start the thread that will drive the capturing by waiting for 158 // capture events. 159 capture_thread_ = 160 new base::DelegateSimpleThread(this, "wasapi_capture_thread"); 161 capture_thread_->Start(); 162 163 // Start streaming data between the endpoint buffer and the audio engine. 164 HRESULT hr = audio_client_->Start(); 165 DLOG_IF(ERROR, FAILED(hr)) << "Failed to start input streaming."; 166 167 if (SUCCEEDED(hr) && audio_render_client_for_loopback_) 168 hr = audio_render_client_for_loopback_->Start(); 169 170 started_ = SUCCEEDED(hr); 171} 172 173void WASAPIAudioInputStream::Stop() { 174 DCHECK(CalledOnValidThread()); 175 DVLOG(1) << "WASAPIAudioInputStream::Stop()"; 176 if (!started_) 177 return; 178 179 // Stops periodic AGC microphone measurements. 180 StopAgc(); 181 182 // Shut down the capture thread. 183 if (stop_capture_event_.IsValid()) { 184 SetEvent(stop_capture_event_.Get()); 185 } 186 187 // Stop the input audio streaming. 188 HRESULT hr = audio_client_->Stop(); 189 if (FAILED(hr)) { 190 LOG(ERROR) << "Failed to stop input streaming."; 191 } 192 193 // Wait until the thread completes and perform cleanup. 194 if (capture_thread_) { 195 SetEvent(stop_capture_event_.Get()); 196 capture_thread_->Join(); 197 capture_thread_ = NULL; 198 } 199 200 started_ = false; 201 sink_ = NULL; 202} 203 204void WASAPIAudioInputStream::Close() { 205 DVLOG(1) << "WASAPIAudioInputStream::Close()"; 206 // It is valid to call Close() before calling open or Start(). 207 // It is also valid to call Close() after Start() has been called. 208 Stop(); 209 210 // Inform the audio manager that we have been closed. This will cause our 211 // destruction. 212 manager_->ReleaseInputStream(this); 213} 214 215double WASAPIAudioInputStream::GetMaxVolume() { 216 // Verify that Open() has been called succesfully, to ensure that an audio 217 // session exists and that an ISimpleAudioVolume interface has been created. 218 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 219 if (!opened_) 220 return 0.0; 221 222 // The effective volume value is always in the range 0.0 to 1.0, hence 223 // we can return a fixed value (=1.0) here. 224 return 1.0; 225} 226 227void WASAPIAudioInputStream::SetVolume(double volume) { 228 DVLOG(1) << "SetVolume(volume=" << volume << ")"; 229 DCHECK(CalledOnValidThread()); 230 DCHECK_GE(volume, 0.0); 231 DCHECK_LE(volume, 1.0); 232 233 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 234 if (!opened_) 235 return; 236 237 // Set a new master volume level. Valid volume levels are in the range 238 // 0.0 to 1.0. Ignore volume-change events. 239 HRESULT hr = simple_audio_volume_->SetMasterVolume(static_cast<float>(volume), 240 NULL); 241 DLOG_IF(WARNING, FAILED(hr)) << "Failed to set new input master volume."; 242 243 // Update the AGC volume level based on the last setting above. Note that, 244 // the volume-level resolution is not infinite and it is therefore not 245 // possible to assume that the volume provided as input parameter can be 246 // used directly. Instead, a new query to the audio hardware is required. 247 // This method does nothing if AGC is disabled. 248 UpdateAgcVolume(); 249} 250 251double WASAPIAudioInputStream::GetVolume() { 252 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 253 if (!opened_) 254 return 0.0; 255 256 // Retrieve the current volume level. The value is in the range 0.0 to 1.0. 257 float level = 0.0f; 258 HRESULT hr = simple_audio_volume_->GetMasterVolume(&level); 259 DLOG_IF(WARNING, FAILED(hr)) << "Failed to get input master volume."; 260 261 return static_cast<double>(level); 262} 263 264// static 265AudioParameters WASAPIAudioInputStream::GetInputStreamParameters( 266 const std::string& device_id) { 267 int sample_rate = 48000; 268 ChannelLayout channel_layout = CHANNEL_LAYOUT_STEREO; 269 270 base::win::ScopedCoMem<WAVEFORMATEX> audio_engine_mix_format; 271 int effects = AudioParameters::NO_EFFECTS; 272 if (SUCCEEDED(GetMixFormat(device_id, &audio_engine_mix_format, &effects))) { 273 sample_rate = static_cast<int>(audio_engine_mix_format->nSamplesPerSec); 274 channel_layout = audio_engine_mix_format->nChannels == 1 ? 275 CHANNEL_LAYOUT_MONO : CHANNEL_LAYOUT_STEREO; 276 } 277 278 // Use 10ms frame size as default. 279 int frames_per_buffer = sample_rate / 100; 280 return AudioParameters( 281 AudioParameters::AUDIO_PCM_LOW_LATENCY, channel_layout, 0, sample_rate, 282 16, frames_per_buffer, effects); 283} 284 285// static 286HRESULT WASAPIAudioInputStream::GetMixFormat(const std::string& device_id, 287 WAVEFORMATEX** device_format, 288 int* effects) { 289 DCHECK(effects); 290 291 // It is assumed that this static method is called from a COM thread, i.e., 292 // CoInitializeEx() is not called here to avoid STA/MTA conflicts. 293 ScopedComPtr<IMMDeviceEnumerator> enumerator; 294 HRESULT hr = enumerator.CreateInstance(__uuidof(MMDeviceEnumerator), NULL, 295 CLSCTX_INPROC_SERVER); 296 if (FAILED(hr)) 297 return hr; 298 299 ScopedComPtr<IMMDevice> endpoint_device; 300 if (device_id == AudioManagerBase::kDefaultDeviceId) { 301 // Retrieve the default capture audio endpoint. 302 hr = enumerator->GetDefaultAudioEndpoint(eCapture, eConsole, 303 endpoint_device.Receive()); 304 } else if (device_id == AudioManagerBase::kLoopbackInputDeviceId) { 305 // Get the mix format of the default playback stream. 306 hr = enumerator->GetDefaultAudioEndpoint(eRender, eConsole, 307 endpoint_device.Receive()); 308 } else { 309 // Retrieve a capture endpoint device that is specified by an endpoint 310 // device-identification string. 311 hr = enumerator->GetDevice(base::UTF8ToUTF16(device_id).c_str(), 312 endpoint_device.Receive()); 313 } 314 315 if (FAILED(hr)) 316 return hr; 317 318 *effects = IsDefaultCommunicationDevice(enumerator, endpoint_device) ? 319 AudioParameters::DUCKING : AudioParameters::NO_EFFECTS; 320 321 ScopedComPtr<IAudioClient> audio_client; 322 hr = endpoint_device->Activate(__uuidof(IAudioClient), 323 CLSCTX_INPROC_SERVER, 324 NULL, 325 audio_client.ReceiveVoid()); 326 return SUCCEEDED(hr) ? audio_client->GetMixFormat(device_format) : hr; 327} 328 329void WASAPIAudioInputStream::Run() { 330 ScopedCOMInitializer com_init(ScopedCOMInitializer::kMTA); 331 332 // Increase the thread priority. 333 capture_thread_->SetThreadPriority(base::kThreadPriority_RealtimeAudio); 334 335 // Enable MMCSS to ensure that this thread receives prioritized access to 336 // CPU resources. 337 DWORD task_index = 0; 338 HANDLE mm_task = avrt::AvSetMmThreadCharacteristics(L"Pro Audio", 339 &task_index); 340 bool mmcss_is_ok = 341 (mm_task && avrt::AvSetMmThreadPriority(mm_task, AVRT_PRIORITY_CRITICAL)); 342 if (!mmcss_is_ok) { 343 // Failed to enable MMCSS on this thread. It is not fatal but can lead 344 // to reduced QoS at high load. 345 DWORD err = GetLastError(); 346 LOG(WARNING) << "Failed to enable MMCSS (error code=" << err << ")."; 347 } 348 349 // Allocate a buffer with a size that enables us to take care of cases like: 350 // 1) The recorded buffer size is smaller, or does not match exactly with, 351 // the selected packet size used in each callback. 352 // 2) The selected buffer size is larger than the recorded buffer size in 353 // each event. 354 size_t buffer_frame_index = 0; 355 size_t capture_buffer_size = std::max( 356 2 * endpoint_buffer_size_frames_ * frame_size_, 357 2 * packet_size_frames_ * frame_size_); 358 scoped_ptr<uint8[]> capture_buffer(new uint8[capture_buffer_size]); 359 360 LARGE_INTEGER now_count; 361 bool recording = true; 362 bool error = false; 363 double volume = GetVolume(); 364 HANDLE wait_array[2] = {stop_capture_event_, audio_samples_ready_event_}; 365 366 while (recording && !error) { 367 HRESULT hr = S_FALSE; 368 369 // Wait for a close-down event or a new capture event. 370 DWORD wait_result = WaitForMultipleObjects(2, wait_array, FALSE, INFINITE); 371 switch (wait_result) { 372 case WAIT_FAILED: 373 error = true; 374 break; 375 case WAIT_OBJECT_0 + 0: 376 // |stop_capture_event_| has been set. 377 recording = false; 378 break; 379 case WAIT_OBJECT_0 + 1: 380 { 381 // |audio_samples_ready_event_| has been set. 382 BYTE* data_ptr = NULL; 383 UINT32 num_frames_to_read = 0; 384 DWORD flags = 0; 385 UINT64 device_position = 0; 386 UINT64 first_audio_frame_timestamp = 0; 387 388 // Retrieve the amount of data in the capture endpoint buffer, 389 // replace it with silence if required, create callbacks for each 390 // packet and store non-delivered data for the next event. 391 hr = audio_capture_client_->GetBuffer(&data_ptr, 392 &num_frames_to_read, 393 &flags, 394 &device_position, 395 &first_audio_frame_timestamp); 396 if (FAILED(hr)) { 397 DLOG(ERROR) << "Failed to get data from the capture buffer"; 398 continue; 399 } 400 401 if (num_frames_to_read != 0) { 402 size_t pos = buffer_frame_index * frame_size_; 403 size_t num_bytes = num_frames_to_read * frame_size_; 404 DCHECK_GE(capture_buffer_size, pos + num_bytes); 405 406 if (flags & AUDCLNT_BUFFERFLAGS_SILENT) { 407 // Clear out the local buffer since silence is reported. 408 memset(&capture_buffer[pos], 0, num_bytes); 409 } else { 410 // Copy captured data from audio engine buffer to local buffer. 411 memcpy(&capture_buffer[pos], data_ptr, num_bytes); 412 } 413 414 buffer_frame_index += num_frames_to_read; 415 } 416 417 hr = audio_capture_client_->ReleaseBuffer(num_frames_to_read); 418 DLOG_IF(ERROR, FAILED(hr)) << "Failed to release capture buffer"; 419 420 // Derive a delay estimate for the captured audio packet. 421 // The value contains two parts (A+B), where A is the delay of the 422 // first audio frame in the packet and B is the extra delay 423 // contained in any stored data. Unit is in audio frames. 424 QueryPerformanceCounter(&now_count); 425 double audio_delay_frames = 426 ((perf_count_to_100ns_units_ * now_count.QuadPart - 427 first_audio_frame_timestamp) / 10000.0) * ms_to_frame_count_ + 428 buffer_frame_index - num_frames_to_read; 429 430 // Get a cached AGC volume level which is updated once every second 431 // on the audio manager thread. Note that, |volume| is also updated 432 // each time SetVolume() is called through IPC by the render-side AGC. 433 GetAgcVolume(&volume); 434 435 // Deliver captured data to the registered consumer using a packet 436 // size which was specified at construction. 437 uint32 delay_frames = static_cast<uint32>(audio_delay_frames + 0.5); 438 while (buffer_frame_index >= packet_size_frames_) { 439 uint8* audio_data = 440 reinterpret_cast<uint8*>(capture_buffer.get()); 441 442 // Deliver data packet, delay estimation and volume level to 443 // the user. 444 sink_->OnData(this, 445 audio_data, 446 packet_size_bytes_, 447 delay_frames * frame_size_, 448 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 482void WASAPIAudioInputStream::HandleError(HRESULT err) { 483 NOTREACHED() << "Error code: " << err; 484 if (sink_) 485 sink_->OnError(this); 486} 487 488HRESULT 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 555HRESULT 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 565HRESULT 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 605bool 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 624HRESULT 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