audio_low_latency_input_win.cc revision 1320f92c476a1ad9d19dba2a48c72b75566198e9
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/audio/win/core_audio_util_win.h" 13#include "media/base/audio_bus.h" 14 15using base::win::ScopedComPtr; 16using base::win::ScopedCOMInitializer; 17 18namespace media { 19namespace { 20 21// Returns true if |device| represents the default communication capture device. 22bool IsDefaultCommunicationDevice(IMMDeviceEnumerator* enumerator, 23 IMMDevice* device) { 24 ScopedComPtr<IMMDevice> communications; 25 if (FAILED(enumerator->GetDefaultAudioEndpoint(eCapture, eCommunications, 26 communications.Receive()))) { 27 return false; 28 } 29 30 base::win::ScopedCoMem<WCHAR> communications_id, device_id; 31 device->GetId(&device_id); 32 communications->GetId(&communications_id); 33 return lstrcmpW(communications_id, device_id) == 0; 34} 35 36} // namespace 37 38WASAPIAudioInputStream::WASAPIAudioInputStream(AudioManagerWin* manager, 39 const AudioParameters& params, 40 const std::string& device_id) 41 : manager_(manager), 42 capture_thread_(NULL), 43 opened_(false), 44 started_(false), 45 frame_size_(0), 46 packet_size_frames_(0), 47 packet_size_bytes_(0), 48 endpoint_buffer_size_frames_(0), 49 effects_(params.effects()), 50 device_id_(device_id), 51 perf_count_to_100ns_units_(0.0), 52 ms_to_frame_count_(0.0), 53 sink_(NULL), 54 audio_bus_(media::AudioBus::Create(params)) { 55 DCHECK(manager_); 56 57 // Load the Avrt DLL if not already loaded. Required to support MMCSS. 58 bool avrt_init = avrt::Initialize(); 59 DCHECK(avrt_init) << "Failed to load the Avrt.dll"; 60 61 // Set up the desired capture format specified by the client. 62 format_.nSamplesPerSec = params.sample_rate(); 63 format_.wFormatTag = WAVE_FORMAT_PCM; 64 format_.wBitsPerSample = params.bits_per_sample(); 65 format_.nChannels = params.channels(); 66 format_.nBlockAlign = (format_.wBitsPerSample / 8) * format_.nChannels; 67 format_.nAvgBytesPerSec = format_.nSamplesPerSec * format_.nBlockAlign; 68 format_.cbSize = 0; 69 70 // Size in bytes of each audio frame. 71 frame_size_ = format_.nBlockAlign; 72 // Store size of audio packets which we expect to get from the audio 73 // endpoint device in each capture event. 74 packet_size_frames_ = params.GetBytesPerBuffer() / format_.nBlockAlign; 75 packet_size_bytes_ = params.GetBytesPerBuffer(); 76 DVLOG(1) << "Number of bytes per audio frame : " << frame_size_; 77 DVLOG(1) << "Number of audio frames per packet: " << packet_size_frames_; 78 79 // All events are auto-reset events and non-signaled initially. 80 81 // Create the event which the audio engine will signal each time 82 // a buffer becomes ready to be processed by the client. 83 audio_samples_ready_event_.Set(CreateEvent(NULL, FALSE, FALSE, NULL)); 84 DCHECK(audio_samples_ready_event_.IsValid()); 85 86 // Create the event which will be set in Stop() when capturing shall stop. 87 stop_capture_event_.Set(CreateEvent(NULL, FALSE, FALSE, NULL)); 88 DCHECK(stop_capture_event_.IsValid()); 89 90 ms_to_frame_count_ = static_cast<double>(params.sample_rate()) / 1000.0; 91 92 LARGE_INTEGER performance_frequency; 93 if (QueryPerformanceFrequency(&performance_frequency)) { 94 perf_count_to_100ns_units_ = 95 (10000000.0 / static_cast<double>(performance_frequency.QuadPart)); 96 } else { 97 DLOG(ERROR) << "High-resolution performance counters are not supported."; 98 } 99} 100 101WASAPIAudioInputStream::~WASAPIAudioInputStream() { 102 DCHECK(CalledOnValidThread()); 103} 104 105bool WASAPIAudioInputStream::Open() { 106 DCHECK(CalledOnValidThread()); 107 // Verify that we are not already opened. 108 if (opened_) 109 return false; 110 111 // Obtain a reference to the IMMDevice interface of the capturing 112 // device with the specified unique identifier or role which was 113 // set at construction. 114 HRESULT hr = SetCaptureDevice(); 115 if (FAILED(hr)) 116 return false; 117 118 // Obtain an IAudioClient interface which enables us to create and initialize 119 // an audio stream between an audio application and the audio engine. 120 hr = ActivateCaptureDevice(); 121 if (FAILED(hr)) 122 return false; 123 124 // Retrieve the stream format which the audio engine uses for its internal 125 // processing/mixing of shared-mode streams. This function call is for 126 // diagnostic purposes only and only in debug mode. 127#ifndef NDEBUG 128 hr = GetAudioEngineStreamFormat(); 129#endif 130 131 // Verify that the selected audio endpoint supports the specified format 132 // set during construction. 133 if (!DesiredFormatIsSupported()) 134 return false; 135 136 // Initialize the audio stream between the client and the device using 137 // shared mode and a lowest possible glitch-free latency. 138 hr = InitializeAudioEngine(); 139 140 opened_ = SUCCEEDED(hr); 141 return opened_; 142} 143 144void WASAPIAudioInputStream::Start(AudioInputCallback* callback) { 145 DCHECK(CalledOnValidThread()); 146 DCHECK(callback); 147 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 148 if (!opened_) 149 return; 150 151 if (started_) 152 return; 153 154 DCHECK(!sink_); 155 sink_ = callback; 156 157 // Starts periodic AGC microphone measurements if the AGC has been enabled 158 // using SetAutomaticGainControl(). 159 StartAgc(); 160 161 // Create and start the thread that will drive the capturing by waiting for 162 // capture events. 163 capture_thread_ = 164 new base::DelegateSimpleThread(this, "wasapi_capture_thread"); 165 capture_thread_->Start(); 166 167 // Start streaming data between the endpoint buffer and the audio engine. 168 HRESULT hr = audio_client_->Start(); 169 DLOG_IF(ERROR, FAILED(hr)) << "Failed to start input streaming."; 170 171 if (SUCCEEDED(hr) && audio_render_client_for_loopback_) 172 hr = audio_render_client_for_loopback_->Start(); 173 174 started_ = SUCCEEDED(hr); 175} 176 177void WASAPIAudioInputStream::Stop() { 178 DCHECK(CalledOnValidThread()); 179 DVLOG(1) << "WASAPIAudioInputStream::Stop()"; 180 if (!started_) 181 return; 182 183 // Stops periodic AGC microphone measurements. 184 StopAgc(); 185 186 // Shut down the capture thread. 187 if (stop_capture_event_.IsValid()) { 188 SetEvent(stop_capture_event_.Get()); 189 } 190 191 // Stop the input audio streaming. 192 HRESULT hr = audio_client_->Stop(); 193 if (FAILED(hr)) { 194 LOG(ERROR) << "Failed to stop input streaming."; 195 } 196 197 // Wait until the thread completes and perform cleanup. 198 if (capture_thread_) { 199 SetEvent(stop_capture_event_.Get()); 200 capture_thread_->Join(); 201 capture_thread_ = NULL; 202 } 203 204 started_ = false; 205 sink_ = NULL; 206} 207 208void WASAPIAudioInputStream::Close() { 209 DVLOG(1) << "WASAPIAudioInputStream::Close()"; 210 // It is valid to call Close() before calling open or Start(). 211 // It is also valid to call Close() after Start() has been called. 212 Stop(); 213 214 // Inform the audio manager that we have been closed. This will cause our 215 // destruction. 216 manager_->ReleaseInputStream(this); 217} 218 219double WASAPIAudioInputStream::GetMaxVolume() { 220 // Verify that Open() has been called succesfully, to ensure that an audio 221 // session exists and that an ISimpleAudioVolume interface has been created. 222 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 223 if (!opened_) 224 return 0.0; 225 226 // The effective volume value is always in the range 0.0 to 1.0, hence 227 // we can return a fixed value (=1.0) here. 228 return 1.0; 229} 230 231void WASAPIAudioInputStream::SetVolume(double volume) { 232 DVLOG(1) << "SetVolume(volume=" << volume << ")"; 233 DCHECK(CalledOnValidThread()); 234 DCHECK_GE(volume, 0.0); 235 DCHECK_LE(volume, 1.0); 236 237 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 238 if (!opened_) 239 return; 240 241 // Set a new master volume level. Valid volume levels are in the range 242 // 0.0 to 1.0. Ignore volume-change events. 243 HRESULT hr = simple_audio_volume_->SetMasterVolume(static_cast<float>(volume), 244 NULL); 245 DLOG_IF(WARNING, FAILED(hr)) << "Failed to set new input master volume."; 246 247 // Update the AGC volume level based on the last setting above. Note that, 248 // the volume-level resolution is not infinite and it is therefore not 249 // possible to assume that the volume provided as input parameter can be 250 // used directly. Instead, a new query to the audio hardware is required. 251 // This method does nothing if AGC is disabled. 252 UpdateAgcVolume(); 253} 254 255double WASAPIAudioInputStream::GetVolume() { 256 DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully"; 257 if (!opened_) 258 return 0.0; 259 260 // Retrieve the current volume level. The value is in the range 0.0 to 1.0. 261 float level = 0.0f; 262 HRESULT hr = simple_audio_volume_->GetMasterVolume(&level); 263 DLOG_IF(WARNING, FAILED(hr)) << "Failed to get input master volume."; 264 265 return static_cast<double>(level); 266} 267 268// static 269AudioParameters WASAPIAudioInputStream::GetInputStreamParameters( 270 const std::string& device_id) { 271 int sample_rate = 48000; 272 ChannelLayout channel_layout = CHANNEL_LAYOUT_STEREO; 273 274 base::win::ScopedCoMem<WAVEFORMATEX> audio_engine_mix_format; 275 int effects = AudioParameters::NO_EFFECTS; 276 if (SUCCEEDED(GetMixFormat(device_id, &audio_engine_mix_format, &effects))) { 277 sample_rate = static_cast<int>(audio_engine_mix_format->nSamplesPerSec); 278 channel_layout = audio_engine_mix_format->nChannels == 1 ? 279 CHANNEL_LAYOUT_MONO : CHANNEL_LAYOUT_STEREO; 280 } 281 282 // Use 10ms frame size as default. 283 int frames_per_buffer = sample_rate / 100; 284 return AudioParameters( 285 AudioParameters::AUDIO_PCM_LOW_LATENCY, channel_layout, sample_rate, 286 16, frames_per_buffer, effects); 287} 288 289// static 290HRESULT WASAPIAudioInputStream::GetMixFormat(const std::string& device_id, 291 WAVEFORMATEX** device_format, 292 int* effects) { 293 DCHECK(effects); 294 295 // It is assumed that this static method is called from a COM thread, i.e., 296 // CoInitializeEx() is not called here to avoid STA/MTA conflicts. 297 ScopedComPtr<IMMDeviceEnumerator> enumerator; 298 HRESULT hr = enumerator.CreateInstance(__uuidof(MMDeviceEnumerator), NULL, 299 CLSCTX_INPROC_SERVER); 300 if (FAILED(hr)) 301 return hr; 302 303 ScopedComPtr<IMMDevice> endpoint_device; 304 if (device_id == AudioManagerBase::kDefaultDeviceId) { 305 // Retrieve the default capture audio endpoint. 306 hr = enumerator->GetDefaultAudioEndpoint(eCapture, eConsole, 307 endpoint_device.Receive()); 308 } else if (device_id == AudioManagerBase::kLoopbackInputDeviceId) { 309 // Get the mix format of the default playback stream. 310 hr = enumerator->GetDefaultAudioEndpoint(eRender, eConsole, 311 endpoint_device.Receive()); 312 } else { 313 // Retrieve a capture endpoint device that is specified by an endpoint 314 // device-identification string. 315 hr = enumerator->GetDevice(base::UTF8ToUTF16(device_id).c_str(), 316 endpoint_device.Receive()); 317 } 318 319 if (FAILED(hr)) 320 return hr; 321 322 *effects = IsDefaultCommunicationDevice(enumerator, endpoint_device) ? 323 AudioParameters::DUCKING : AudioParameters::NO_EFFECTS; 324 325 ScopedComPtr<IAudioClient> audio_client; 326 hr = endpoint_device->Activate(__uuidof(IAudioClient), 327 CLSCTX_INPROC_SERVER, 328 NULL, 329 audio_client.ReceiveVoid()); 330 return SUCCEEDED(hr) ? audio_client->GetMixFormat(device_format) : hr; 331} 332 333void WASAPIAudioInputStream::Run() { 334 ScopedCOMInitializer com_init(ScopedCOMInitializer::kMTA); 335 336 // Increase the thread priority. 337 capture_thread_->SetThreadPriority(base::kThreadPriority_RealtimeAudio); 338 339 // Enable MMCSS to ensure that this thread receives prioritized access to 340 // CPU resources. 341 DWORD task_index = 0; 342 HANDLE mm_task = avrt::AvSetMmThreadCharacteristics(L"Pro Audio", 343 &task_index); 344 bool mmcss_is_ok = 345 (mm_task && avrt::AvSetMmThreadPriority(mm_task, AVRT_PRIORITY_CRITICAL)); 346 if (!mmcss_is_ok) { 347 // Failed to enable MMCSS on this thread. It is not fatal but can lead 348 // to reduced QoS at high load. 349 DWORD err = GetLastError(); 350 LOG(WARNING) << "Failed to enable MMCSS (error code=" << err << ")."; 351 } 352 353 // Allocate a buffer with a size that enables us to take care of cases like: 354 // 1) The recorded buffer size is smaller, or does not match exactly with, 355 // the selected packet size used in each callback. 356 // 2) The selected buffer size is larger than the recorded buffer size in 357 // each event. 358 size_t buffer_frame_index = 0; 359 size_t capture_buffer_size = std::max( 360 2 * endpoint_buffer_size_frames_ * frame_size_, 361 2 * packet_size_frames_ * frame_size_); 362 scoped_ptr<uint8[]> capture_buffer(new uint8[capture_buffer_size]); 363 364 LARGE_INTEGER now_count; 365 bool recording = true; 366 bool error = false; 367 double volume = GetVolume(); 368 HANDLE wait_array[2] = 369 { stop_capture_event_.Get(), audio_samples_ready_event_.Get() }; 370 371 while (recording && !error) { 372 HRESULT hr = S_FALSE; 373 374 // Wait for a close-down event or a new capture event. 375 DWORD wait_result = WaitForMultipleObjects(2, wait_array, FALSE, INFINITE); 376 switch (wait_result) { 377 case WAIT_FAILED: 378 error = true; 379 break; 380 case WAIT_OBJECT_0 + 0: 381 // |stop_capture_event_| has been set. 382 recording = false; 383 break; 384 case WAIT_OBJECT_0 + 1: 385 { 386 // |audio_samples_ready_event_| has been set. 387 BYTE* data_ptr = NULL; 388 UINT32 num_frames_to_read = 0; 389 DWORD flags = 0; 390 UINT64 device_position = 0; 391 UINT64 first_audio_frame_timestamp = 0; 392 393 // Retrieve the amount of data in the capture endpoint buffer, 394 // replace it with silence if required, create callbacks for each 395 // packet and store non-delivered data for the next event. 396 hr = audio_capture_client_->GetBuffer(&data_ptr, 397 &num_frames_to_read, 398 &flags, 399 &device_position, 400 &first_audio_frame_timestamp); 401 if (FAILED(hr)) { 402 DLOG(ERROR) << "Failed to get data from the capture buffer"; 403 continue; 404 } 405 406 if (num_frames_to_read != 0) { 407 size_t pos = buffer_frame_index * frame_size_; 408 size_t num_bytes = num_frames_to_read * frame_size_; 409 DCHECK_GE(capture_buffer_size, pos + num_bytes); 410 411 if (flags & AUDCLNT_BUFFERFLAGS_SILENT) { 412 // Clear out the local buffer since silence is reported. 413 memset(&capture_buffer[pos], 0, num_bytes); 414 } else { 415 // Copy captured data from audio engine buffer to local buffer. 416 memcpy(&capture_buffer[pos], data_ptr, num_bytes); 417 } 418 419 buffer_frame_index += num_frames_to_read; 420 } 421 422 hr = audio_capture_client_->ReleaseBuffer(num_frames_to_read); 423 DLOG_IF(ERROR, FAILED(hr)) << "Failed to release capture buffer"; 424 425 // Derive a delay estimate for the captured audio packet. 426 // The value contains two parts (A+B), where A is the delay of the 427 // first audio frame in the packet and B is the extra delay 428 // contained in any stored data. Unit is in audio frames. 429 QueryPerformanceCounter(&now_count); 430 double audio_delay_frames = 431 ((perf_count_to_100ns_units_ * now_count.QuadPart - 432 first_audio_frame_timestamp) / 10000.0) * ms_to_frame_count_ + 433 buffer_frame_index - num_frames_to_read; 434 435 // Get a cached AGC volume level which is updated once every second 436 // on the audio manager thread. Note that, |volume| is also updated 437 // each time SetVolume() is called through IPC by the render-side AGC. 438 GetAgcVolume(&volume); 439 440 // Deliver captured data to the registered consumer using a packet 441 // size which was specified at construction. 442 uint32 delay_frames = static_cast<uint32>(audio_delay_frames + 0.5); 443 while (buffer_frame_index >= packet_size_frames_) { 444 // Copy data to audio bus to match the OnData interface. 445 uint8* audio_data = reinterpret_cast<uint8*>(capture_buffer.get()); 446 audio_bus_->FromInterleaved( 447 audio_data, audio_bus_->frames(), format_.wBitsPerSample / 8); 448 449 // Deliver data packet, delay estimation and volume level to 450 // the user. 451 sink_->OnData( 452 this, audio_bus_.get(), delay_frames * frame_size_, volume); 453 454 // Store parts of the recorded data which can't be delivered 455 // using the current packet size. The stored section will be used 456 // either in the next while-loop iteration or in the next 457 // capture event. 458 memmove(&capture_buffer[0], 459 &capture_buffer[packet_size_bytes_], 460 (buffer_frame_index - packet_size_frames_) * frame_size_); 461 462 buffer_frame_index -= packet_size_frames_; 463 delay_frames -= packet_size_frames_; 464 } 465 } 466 break; 467 default: 468 error = true; 469 break; 470 } 471 } 472 473 if (recording && error) { 474 // TODO(henrika): perhaps it worth improving the cleanup here by e.g. 475 // stopping the audio client, joining the thread etc.? 476 NOTREACHED() << "WASAPI capturing failed with error code " 477 << GetLastError(); 478 } 479 480 // Disable MMCSS. 481 if (mm_task && !avrt::AvRevertMmThreadCharacteristics(mm_task)) { 482 PLOG(WARNING) << "Failed to disable MMCSS"; 483 } 484} 485 486void WASAPIAudioInputStream::HandleError(HRESULT err) { 487 NOTREACHED() << "Error code: " << err; 488 if (sink_) 489 sink_->OnError(this); 490} 491 492HRESULT WASAPIAudioInputStream::SetCaptureDevice() { 493 DCHECK(!endpoint_device_); 494 495 ScopedComPtr<IMMDeviceEnumerator> enumerator; 496 HRESULT hr = enumerator.CreateInstance(__uuidof(MMDeviceEnumerator), 497 NULL, CLSCTX_INPROC_SERVER); 498 if (FAILED(hr)) 499 return hr; 500 501 // Retrieve the IMMDevice by using the specified role or the specified 502 // unique endpoint device-identification string. 503 504 if (effects_ & AudioParameters::DUCKING) { 505 // Ducking has been requested and it is only supported for the default 506 // communication device. So, let's open up the communication device and 507 // see if the ID of that device matches the requested ID. 508 // We consider a kDefaultDeviceId as well as an explicit device id match, 509 // to be valid matches. 510 hr = enumerator->GetDefaultAudioEndpoint(eCapture, eCommunications, 511 endpoint_device_.Receive()); 512 if (endpoint_device_ && device_id_ != AudioManagerBase::kDefaultDeviceId) { 513 base::win::ScopedCoMem<WCHAR> communications_id; 514 endpoint_device_->GetId(&communications_id); 515 if (device_id_ != 516 base::WideToUTF8(static_cast<WCHAR*>(communications_id))) { 517 DLOG(WARNING) << "Ducking has been requested for a non-default device." 518 "Not supported."; 519 // We can't honor the requested effect flag, so turn it off and 520 // continue. We'll check this flag later to see if we've actually 521 // opened up the communications device, so it's important that it 522 // reflects the active state. 523 effects_ &= ~AudioParameters::DUCKING; 524 endpoint_device_.Release(); // Fall back on code below. 525 } 526 } 527 } 528 529 if (!endpoint_device_) { 530 if (device_id_ == AudioManagerBase::kDefaultDeviceId) { 531 // Retrieve the default capture audio endpoint for the specified role. 532 // Note that, in Windows Vista, the MMDevice API supports device roles 533 // but the system-supplied user interface programs do not. 534 hr = enumerator->GetDefaultAudioEndpoint(eCapture, eConsole, 535 endpoint_device_.Receive()); 536 } else if (device_id_ == AudioManagerBase::kLoopbackInputDeviceId) { 537 // Capture the default playback stream. 538 hr = enumerator->GetDefaultAudioEndpoint(eRender, eConsole, 539 endpoint_device_.Receive()); 540 } else { 541 hr = enumerator->GetDevice(base::UTF8ToUTF16(device_id_).c_str(), 542 endpoint_device_.Receive()); 543 } 544 } 545 546 if (FAILED(hr)) 547 return hr; 548 549 // Verify that the audio endpoint device is active, i.e., the audio 550 // adapter that connects to the endpoint device is present and enabled. 551 DWORD state = DEVICE_STATE_DISABLED; 552 hr = endpoint_device_->GetState(&state); 553 if (FAILED(hr)) 554 return hr; 555 556 if (!(state & DEVICE_STATE_ACTIVE)) { 557 DLOG(ERROR) << "Selected capture device is not active."; 558 hr = E_ACCESSDENIED; 559 } 560 561 return hr; 562} 563 564HRESULT WASAPIAudioInputStream::ActivateCaptureDevice() { 565 // Creates and activates an IAudioClient COM object given the selected 566 // capture endpoint device. 567 HRESULT hr = endpoint_device_->Activate(__uuidof(IAudioClient), 568 CLSCTX_INPROC_SERVER, 569 NULL, 570 audio_client_.ReceiveVoid()); 571 return hr; 572} 573 574HRESULT WASAPIAudioInputStream::GetAudioEngineStreamFormat() { 575 HRESULT hr = S_OK; 576#ifndef NDEBUG 577 // The GetMixFormat() method retrieves the stream format that the 578 // audio engine uses for its internal processing of shared-mode streams. 579 // The method always uses a WAVEFORMATEXTENSIBLE structure, instead 580 // of a stand-alone WAVEFORMATEX structure, to specify the format. 581 // An WAVEFORMATEXTENSIBLE structure can specify both the mapping of 582 // channels to speakers and the number of bits of precision in each sample. 583 base::win::ScopedCoMem<WAVEFORMATEXTENSIBLE> format_ex; 584 hr = audio_client_->GetMixFormat( 585 reinterpret_cast<WAVEFORMATEX**>(&format_ex)); 586 587 // See http://msdn.microsoft.com/en-us/windows/hardware/gg463006#EFH 588 // for details on the WAVE file format. 589 WAVEFORMATEX format = format_ex->Format; 590 DVLOG(2) << "WAVEFORMATEX:"; 591 DVLOG(2) << " wFormatTags : 0x" << std::hex << format.wFormatTag; 592 DVLOG(2) << " nChannels : " << format.nChannels; 593 DVLOG(2) << " nSamplesPerSec : " << format.nSamplesPerSec; 594 DVLOG(2) << " nAvgBytesPerSec: " << format.nAvgBytesPerSec; 595 DVLOG(2) << " nBlockAlign : " << format.nBlockAlign; 596 DVLOG(2) << " wBitsPerSample : " << format.wBitsPerSample; 597 DVLOG(2) << " cbSize : " << format.cbSize; 598 599 DVLOG(2) << "WAVEFORMATEXTENSIBLE:"; 600 DVLOG(2) << " wValidBitsPerSample: " << 601 format_ex->Samples.wValidBitsPerSample; 602 DVLOG(2) << " dwChannelMask : 0x" << std::hex << 603 format_ex->dwChannelMask; 604 if (format_ex->SubFormat == KSDATAFORMAT_SUBTYPE_PCM) 605 DVLOG(2) << " SubFormat : KSDATAFORMAT_SUBTYPE_PCM"; 606 else if (format_ex->SubFormat == KSDATAFORMAT_SUBTYPE_IEEE_FLOAT) 607 DVLOG(2) << " SubFormat : KSDATAFORMAT_SUBTYPE_IEEE_FLOAT"; 608 else if (format_ex->SubFormat == KSDATAFORMAT_SUBTYPE_WAVEFORMATEX) 609 DVLOG(2) << " SubFormat : KSDATAFORMAT_SUBTYPE_WAVEFORMATEX"; 610#endif 611 return hr; 612} 613 614bool WASAPIAudioInputStream::DesiredFormatIsSupported() { 615 // An application that uses WASAPI to manage shared-mode streams can rely 616 // on the audio engine to perform only limited format conversions. The audio 617 // engine can convert between a standard PCM sample size used by the 618 // application and the floating-point samples that the engine uses for its 619 // internal processing. However, the format for an application stream 620 // typically must have the same number of channels and the same sample 621 // rate as the stream format used by the device. 622 // Many audio devices support both PCM and non-PCM stream formats. However, 623 // the audio engine can mix only PCM streams. 624 base::win::ScopedCoMem<WAVEFORMATEX> closest_match; 625 HRESULT hr = audio_client_->IsFormatSupported(AUDCLNT_SHAREMODE_SHARED, 626 &format_, 627 &closest_match); 628 DLOG_IF(ERROR, hr == S_FALSE) << "Format is not supported " 629 << "but a closest match exists."; 630 return (hr == S_OK); 631} 632 633HRESULT WASAPIAudioInputStream::InitializeAudioEngine() { 634 DWORD flags; 635 // Use event-driven mode only fo regular input devices. For loopback the 636 // EVENTCALLBACK flag is specified when intializing 637 // |audio_render_client_for_loopback_|. 638 if (device_id_ == AudioManagerBase::kLoopbackInputDeviceId) { 639 flags = AUDCLNT_STREAMFLAGS_LOOPBACK | AUDCLNT_STREAMFLAGS_NOPERSIST; 640 } else { 641 flags = 642 AUDCLNT_STREAMFLAGS_EVENTCALLBACK | AUDCLNT_STREAMFLAGS_NOPERSIST; 643 } 644 645 // Initialize the audio stream between the client and the device. 646 // We connect indirectly through the audio engine by using shared mode. 647 // Note that, |hnsBufferDuration| is set of 0, which ensures that the 648 // buffer is never smaller than the minimum buffer size needed to ensure 649 // that glitches do not occur between the periodic processing passes. 650 // This setting should lead to lowest possible latency. 651 HRESULT hr = audio_client_->Initialize( 652 AUDCLNT_SHAREMODE_SHARED, 653 flags, 654 0, // hnsBufferDuration 655 0, 656 &format_, 657 (effects_ & AudioParameters::DUCKING) ? &kCommunicationsSessionId : NULL); 658 659 if (FAILED(hr)) 660 return hr; 661 662 // Retrieve the length of the endpoint buffer shared between the client 663 // and the audio engine. The buffer length determines the maximum amount 664 // of capture data that the audio engine can read from the endpoint buffer 665 // during a single processing pass. 666 // A typical value is 960 audio frames <=> 20ms @ 48kHz sample rate. 667 hr = audio_client_->GetBufferSize(&endpoint_buffer_size_frames_); 668 if (FAILED(hr)) 669 return hr; 670 671 DVLOG(1) << "endpoint buffer size: " << endpoint_buffer_size_frames_ 672 << " [frames]"; 673 674#ifndef NDEBUG 675 // The period between processing passes by the audio engine is fixed for a 676 // particular audio endpoint device and represents the smallest processing 677 // quantum for the audio engine. This period plus the stream latency between 678 // the buffer and endpoint device represents the minimum possible latency 679 // that an audio application can achieve. 680 // TODO(henrika): possibly remove this section when all parts are ready. 681 REFERENCE_TIME device_period_shared_mode = 0; 682 REFERENCE_TIME device_period_exclusive_mode = 0; 683 HRESULT hr_dbg = audio_client_->GetDevicePeriod( 684 &device_period_shared_mode, &device_period_exclusive_mode); 685 if (SUCCEEDED(hr_dbg)) { 686 DVLOG(1) << "device period: " 687 << static_cast<double>(device_period_shared_mode / 10000.0) 688 << " [ms]"; 689 } 690 691 REFERENCE_TIME latency = 0; 692 hr_dbg = audio_client_->GetStreamLatency(&latency); 693 if (SUCCEEDED(hr_dbg)) { 694 DVLOG(1) << "stream latency: " << static_cast<double>(latency / 10000.0) 695 << " [ms]"; 696 } 697#endif 698 699 // Set the event handle that the audio engine will signal each time a buffer 700 // becomes ready to be processed by the client. 701 // 702 // In loopback case the capture device doesn't receive any events, so we 703 // need to create a separate playback client to get notifications. According 704 // to MSDN: 705 // 706 // A pull-mode capture client does not receive any events when a stream is 707 // initialized with event-driven buffering and is loopback-enabled. To 708 // work around this, initialize a render stream in event-driven mode. Each 709 // time the client receives an event for the render stream, it must signal 710 // the capture client to run the capture thread that reads the next set of 711 // samples from the capture endpoint buffer. 712 // 713 // http://msdn.microsoft.com/en-us/library/windows/desktop/dd316551(v=vs.85).aspx 714 if (device_id_ == AudioManagerBase::kLoopbackInputDeviceId) { 715 hr = endpoint_device_->Activate( 716 __uuidof(IAudioClient), CLSCTX_INPROC_SERVER, NULL, 717 audio_render_client_for_loopback_.ReceiveVoid()); 718 if (FAILED(hr)) 719 return hr; 720 721 hr = audio_render_client_for_loopback_->Initialize( 722 AUDCLNT_SHAREMODE_SHARED, 723 AUDCLNT_STREAMFLAGS_EVENTCALLBACK | AUDCLNT_STREAMFLAGS_NOPERSIST, 724 0, 0, &format_, NULL); 725 if (FAILED(hr)) 726 return hr; 727 728 hr = audio_render_client_for_loopback_->SetEventHandle( 729 audio_samples_ready_event_.Get()); 730 } else { 731 hr = audio_client_->SetEventHandle(audio_samples_ready_event_.Get()); 732 } 733 734 if (FAILED(hr)) 735 return hr; 736 737 // Get access to the IAudioCaptureClient interface. This interface 738 // enables us to read input data from the capture endpoint buffer. 739 hr = audio_client_->GetService(__uuidof(IAudioCaptureClient), 740 audio_capture_client_.ReceiveVoid()); 741 if (FAILED(hr)) 742 return hr; 743 744 // Obtain a reference to the ISimpleAudioVolume interface which enables 745 // us to control the master volume level of an audio session. 746 hr = audio_client_->GetService(__uuidof(ISimpleAudioVolume), 747 simple_audio_volume_.ReceiveVoid()); 748 return hr; 749} 750 751} // namespace media 752