Added basic framework for AEC3 in the audio processing module

webrtc/modules/audio_processing/aec3/echo_canceller3.cc

Index: webrtc/modules/audio_processing/aec3/echo_canceller3.cc
diff --git a/webrtc/modules/audio_processing/aec3/echo_canceller3.cc b/webrtc/modules/audio_processing/aec3/echo_canceller3.cc
new file mode 100644
index 0000000000000000000000000000000000000000..c8ba2eb72c12225a03cc32a39fba7fc95b395b1c
--- /dev/null
+++ b/webrtc/modules/audio_processing/aec3/echo_canceller3.cc
@@ -0,0 +1,56 @@
+/*
+ *  Copyright (c) 2016 The WebRTC project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+#include "webrtc/modules/audio_processing/aec3/echo_canceller3.h"
+
+#include "webrtc/system_wrappers/include/logging.h"
+
+namespace webrtc {
+
+int EchoCanceller3::instance_count_ = 0;

[hlundin-webrtc, 2016/12/12 21:35:21]

Is this thread-safe? See
https://sites.google.com/a/chromium.org/dev/developers/coding-style/cpp-dos-a....

[peah-webrtc, 2016/12/13 11:23:02]

I added an atomic operation for this. That should do it.
PTAL

Done.

+
+EchoCanceller3::EchoCanceller3(int sample_rate_hz, bool use_anti_hum_filter) {
+  int band_sample_rate_hz = (sample_rate_hz == 8000 ? sample_rate_hz : 16000);
+  frame_length_ = rtc::CheckedDivExact(band_sample_rate_hz, 100);
+
+  LOG(LS_INFO) << "AEC3 created : "
+               << "{ instance_count: " << instance_count_ << "}";
+  ++instance_count_;
+}
+
+EchoCanceller3::~EchoCanceller3() = default;
+
+bool EchoCanceller3::AnalyzeRender(AudioBuffer* render) const {

[hlundin-webrtc, 2016/12/12 21:35:21]

It's hard to tell from this dummy implementation, but does the parameter have to
be non-const?

[peah-webrtc, 2016/12/13 11:23:02]

In a perfect world, it shouldn't. But the AudioBuffer class works in such a way
that it does on-demand conversions between floating and fixed point samples.
This means, that even though AnalyzeRender only reads the AudioBuffer, that
reading may mutate the AudioBuffer in the sense that it may do an internal
on-demand conversion to floating point which then is cached internally inside
AudioBuffer. 

[hlundin-webrtc, 2016/12/14 08:15:56]

Acknowledged.

+  RTC_DCHECK_EQ(1u, render->num_channels());
+  RTC_DCHECK_EQ(frame_length_, render->num_frames_per_band());
+  return true;
+}
+
+void EchoCanceller3::AnalyzeCapture(AudioBuffer* capture) {}

[hlundin-webrtc, 2016/12/12 21:35:21]

Same here: Can you make the input const? The name of the method suggests that it
will just "look" at the data.

[peah-webrtc, 2016/12/13 11:23:02]

That would make sense. But with the current AudioBuffer implementation, it is
not possible. (see above explanation).

+
+void EchoCanceller3::ProcessCapture(AudioBuffer* capture,
+                                    bool known_echo_path_change) {
+  RTC_DCHECK_EQ(1u, capture->num_channels());
+  RTC_DCHECK_EQ(frame_length_, capture->num_frames_per_band());
+}
+
+std::string EchoCanceller3::ToString(
+    const AudioProcessing::Config::EchoCanceller3& config) {
+  std::stringstream ss;
+  ss << "{"
+     << "enabled: " << (config.enabled ? "true" : "false") << "}";
+  return ss.str();
+}
+
+bool EchoCanceller3::Validate(
+    const AudioProcessing::Config::EchoCanceller3& config) {
+  return true;
+}
+
+}  // namespace webrtc