Add basic resample support to WASAPIAudioInputStream.

media/audio/win/audio_low_latency_input_win.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "media/audio/win/audio_low_latency_input_win.h"
 
 #include <memory>
 
 #include "base/logging.h"
 #include "base/metrics/histogram_macros.h"
 #include "base/strings/utf_string_conversions.h"
 #include "base/trace_event/trace_event.h"
 #include "media/audio/audio_device_description.h"
 #include "media/audio/win/audio_manager_win.h"
 #include "media/audio/win/avrt_wrapper_win.h"
 #include "media/audio/win/core_audio_util_win.h"
 #include "media/base/audio_bus.h"
+#include "media/base/multi_channel_resampler.h"
 
 using base::win::ScopedComPtr;
 using base::win::ScopedCOMInitializer;
 
 namespace media {
 
 WASAPIAudioInputStream::WASAPIAudioInputStream(AudioManagerWin* manager,
                                                const AudioParameters& params,
                                                const std::string& device_id)
     : manager_(manager),
@@ skipping 88 matching lines @@
   // set during construction.
   if (!DesiredFormatIsSupported()) {
     open_result_ = OPEN_RESULT_FORMAT_NOT_SUPPORTED;
     ReportOpenResult();
     return false;
   }
 
   // Initialize the audio stream between the client and the device using
   // shared mode and a lowest possible glitch-free latency.
   hr = InitializeAudioEngine();
+  if (SUCCEEDED(hr) && resampler_.get() != nullptr)

[DaleCurtis, 2017/02/13 19:40:42]

just "&& resampler_" it has safe-bool conversion built in these days.

[tommi (sloooow) - chröme, 2017/02/16 17:08:46]

Done.

+    open_result_ = OPEN_RESULT_OK_WITH_RESAMPLING;
   ReportOpenResult();  // Report before we assign a value to |opened_|.
   opened_ = SUCCEEDED(hr);
-  DCHECK(open_result_ == OPEN_RESULT_OK || !opened_);
 
   return opened_;
 }
 
 void WASAPIAudioInputStream::Start(AudioInputCallback* callback) {
   DCHECK(CalledOnValidThread());
   DCHECK(callback);
   DLOG_IF(ERROR, !opened_) << "Open() has not been called successfully";
   if (!opened_)
     return;
@@ skipping 169 matching lines @@
       avrt::AvSetMmThreadCharacteristics(L"Pro Audio", &task_index);
   bool mmcss_is_ok =
       (mm_task && avrt::AvSetMmThreadPriority(mm_task, AVRT_PRIORITY_CRITICAL));
   if (!mmcss_is_ok) {
     // Failed to enable MMCSS on this thread. It is not fatal but can lead
     // to reduced QoS at high load.
     DWORD err = GetLastError();
     LOG(WARNING) << "Failed to enable MMCSS (error code=" << err << ").";
   }
 
   // Allocate a buffer with a size that enables us to take care of cases like:

[DaleCurtis, 2017/02/13 19:40:42]

You might consider that instead of uint8_t array + resampler you use a
AudioBlockFIFO + Resampler here; this way the resampler can pull from subsequent
buffers if it needs to.

[tommi (sloooow) - chröme, 2017/02/16 17:08:47]

Acknowledged.

   // 1) The recorded buffer size is smaller, or does not match exactly with,
   //    the selected packet size used in each callback.
   // 2) The selected buffer size is larger than the recorded buffer size in
   //    each event.
   size_t buffer_frame_index = 0;
   size_t capture_buffer_size =
       std::max(2 * endpoint_buffer_size_frames_ * frame_size_,
                2 * packet_size_frames_ * frame_size_);
   std::unique_ptr<uint8_t[]> capture_buffer(new uint8_t[capture_buffer_size]);
 
@@ skipping 88 matching lines @@
         // each time SetVolume() is called through IPC by the render-side AGC.
         GetAgcVolume(&volume);
 
         // Deliver captured data to the registered consumer using a packet
         // size which was specified at construction.
         uint32_t delay_frames = static_cast<uint32_t>(audio_delay_frames + 0.5);
         while (buffer_frame_index >= packet_size_frames_) {
           // Copy data to audio bus to match the OnData interface.
           uint8_t* audio_data =
               reinterpret_cast<uint8_t*>(capture_buffer.get());
-          audio_bus_->FromInterleaved(audio_data, audio_bus_->frames(),
-                                      format_.wBitsPerSample / 8);
+
+          if (resampler_.get()) {

[DaleCurtis, 2017/02/13 19:40:42]

Ditto.

[tommi (sloooow) - chröme, 2017/02/16 17:08:46]

Done.

+            resample_bus_->FromInterleaved(audio_data, resample_bus_->frames(),
+                                           format_.wBitsPerSample / 8);
+            resampler_->Resample(audio_bus_->frames(), audio_bus_.get());
+          } else {
+            audio_bus_->FromInterleaved(audio_data, audio_bus_->frames(),
+                                        format_.wBitsPerSample / 8);
+          }
 
           // Deliver data packet, delay estimation and volume level to
           // the user.
           sink_->OnData(this, audio_bus_.get(), delay_frames * frame_size_,
                         volume);
 
           // Store parts of the recorded data which can't be delivered
           // using the current packet size. The stored section will be used
           // either in the next while-loop iteration or in the next
           // capture event.
@@ skipping 149 matching lines @@
   // internal processing. However, the format for an application stream
   // typically must have the same number of channels and the same sample
   // rate as the stream format used by the device.
   // Many audio devices support both PCM and non-PCM stream formats. However,
   // the audio engine can mix only PCM streams.
   base::win::ScopedCoMem<WAVEFORMATEX> closest_match;
   HRESULT hr = audio_client_->IsFormatSupported(AUDCLNT_SHAREMODE_SHARED,
                                                 &format_, &closest_match);
   DLOG_IF(ERROR, hr == S_FALSE) << "Format is not supported "
                                 << "but a closest match exists.";
+  if (hr == S_FALSE && closest_match->nChannels == format_.nChannels) {

[DaleCurtis, 2017/02/13 19:40:42]

If you use a AudioConverter it'll handle mismatched channel counts for you.

[DaleCurtis, 2017/02/13 23:59:00]

You should also sanity check the closest match sample rate here. I.e. that it
isn't 0 or something outside the bound of our limits::kMin/MaxSampleRate.

[tommi (sloooow) - chröme, 2017/02/16 17:08:46]

Done. Also switched to AudioConverter.

+    // We want a 1:1 ratio between the buffers we get and the buffers we
+    // give to OnData so that each buffer we receive from the OS can be directly
+    // resampled to a buffer that matches with what the client asked for.
+    const double buffer_ratio =
+        format_.nSamplesPerSec / static_cast<double>(audio_bus_->frames());
+    const size_t new_frames_per_buffer =
+        static_cast<size_t>(closest_match->nSamplesPerSec / buffer_ratio);
+
+    const double sample_rate_ratio =
+        static_cast<double>(closest_match->nSamplesPerSec) /
+        static_cast<double>(format_.nSamplesPerSec);
+    resampler_.reset(new MultiChannelResampler(

[DaleCurtis, 2017/02/13 19:40:42]

I think this works, though not how we've done it in the past see
AudioBufferConverter. Will take a closer look later, have to run now.

[DaleCurtis, 2017/02/13 23:59:00]

Over a long enough time frame I think this might lose a sample every now and
then. I.e. Resample() will consume all but the last frame and then the next
FromInterleaved() call will overwrite it. To make this correct I think you need
to use the AudioBlockFifo as suggested above.

E.g., if the desired is 44.1kHz/1024 frames and the closest match is 16kHz, I
think you'll lose the fractional part on occasion.

[tommi (sloooow) - chröme, 2017/02/16 17:08:47]

I tried several ways of reproducing that, including recording at 16kHz and
converting to 44.1kHz/1024 for 10 minutes, but was unable to reproduce the
problem (ProvideInput was called for all buffers).  Any further hints on how to
test that?

[tommi (sloooow) - chröme, 2017/02/16 17:09:49]

Oh please ignore this, I did more testing and figured out what I was missing.

+        format_.nChannels, sample_rate_ratio, new_frames_per_buffer,
+        base::Bind(&WASAPIAudioInputStream::ProvideInput,
+                   base::Unretained(this))));
+    resampler_->PrimeWithSilence();
+    resample_bus_ = AudioBus::Create(format_.nChannels, new_frames_per_buffer);
+
+    // Now change the format we're going to ask for to better match with what
+    // the OS can provide.  If we succeed in opening the stream with these
+    // params, we can take care of the required resampling.
+    format_.nSamplesPerSec = closest_match->nSamplesPerSec;
+    format_.nAvgBytesPerSec = format_.nSamplesPerSec * format_.nBlockAlign;
+
+    // Update our packet size assumptions based on the new format.
+    const auto new_bytes_per_buffer = resample_bus_->frames() *
+                                      closest_match->nChannels *
+                                      (format_.wBitsPerSample / 8);
+    packet_size_frames_ = new_bytes_per_buffer / format_.nBlockAlign;
+    packet_size_bytes_ = new_bytes_per_buffer;
+    ms_to_frame_count_ =
+        static_cast<double>(closest_match->nSamplesPerSec) / 1000.0;
+
+    // Indicate that we're good to go with a close match.
+    hr = S_OK;
+  }
+
   return (hr == S_OK);
 }
 
 HRESULT WASAPIAudioInputStream::InitializeAudioEngine() {
   DCHECK_EQ(OPEN_RESULT_OK, open_result_);
   DWORD flags;
   // Use event-driven mode only fo regular input devices. For loopback the
   // EVENTCALLBACK flag is specified when intializing
   // |audio_render_client_for_loopback_|.
   if (device_id_ == AudioDeviceDescription::kLoopbackInputDeviceId ||
@@ skipping 123 matching lines @@
 
   return hr;
 }
 
 void WASAPIAudioInputStream::ReportOpenResult() const {
   DCHECK(!opened_);  // This method must be called before we set this flag.
   UMA_HISTOGRAM_ENUMERATION("Media.Audio.Capture.Win.Open", open_result_,
                             OPEN_RESULT_MAX + 1);
 }
 
+void WASAPIAudioInputStream::ProvideInput(int resampler_frame_delay,
+                                          AudioBus* audio_bus) {
+  resample_bus_->CopyTo(audio_bus);

[DaleCurtis, 2017/02/13 19:40:42]

Probably want a DCHECK this isn't called twice without resample_bus_ being
refilled.

[tommi (sloooow) - chröme, 2017/02/16 17:08:47]

Done.

+}
+
 }  // namespace media