ACM: Corrected temporary buffer size

This CL corrects the temporary buffers size in the
pre-processing of the capture audio before encoding.

As part of this it removes the ACM-specific hardcoding
of the size and instead ensures that the size of the
temporary buffer matches that of the AudioFrame.

Bug: webrtc:11242
Change-Id: I56dd6cadfd4e140e8e159966c33d1027383ea9fa
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/170340
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Henrik Lundin <henrik.lundin@webrtc.org>
Reviewed-by: Henrik Andreassson <henrika@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#30775}

modules/audio_coding/acm2/audio_coding_module.cc

@@ -37,6 +37,8 @@ namespace {
 // 48 kHz data.
 constexpr size_t kInitialInputDataBufferSize = 6 * 480;
 
+constexpr int32_t kMaxInputSampleRateHz = 192000;
+
 class AudioCodingModuleImpl final : public AudioCodingModule {
  public:
   explicit AudioCodingModuleImpl(const AudioCodingModule::Config& config);
@@ -346,7 +348,7 @@ int AudioCodingModuleImpl::Add10MsDataInternal(const AudioFrame& audio_frame,
     return -1;
   }
 
-  if (audio_frame.sample_rate_hz_ > 192000) {
+  if (audio_frame.sample_rate_hz_ > kMaxInputSampleRateHz) {
     assert(false);
     RTC_LOG(LS_ERROR) << "Cannot Add 10 ms audio, input frequency not valid";
     return -1;
@@ -463,20 +465,25 @@ int AudioCodingModuleImpl::PreprocessToAddData(const AudioFrame& in_frame,
   *ptr_out = &preprocess_frame_;
   preprocess_frame_.num_channels_ = in_frame.num_channels_;
   preprocess_frame_.samples_per_channel_ = in_frame.samples_per_channel_;
-  std::array<int16_t, WEBRTC_10MS_PCM_AUDIO> audio;
+  std::array<int16_t, AudioFrame::kMaxDataSizeSamples> audio;
-  const int16_t* src_ptr_audio = in_frame.data();
+  const int16_t* src_ptr_audio;
   if (down_mix) {
-    // If a resampling is required the output of a down-mix is written into a
+    // If a resampling is required, the output of a down-mix is written into a
     // local buffer, otherwise, it will be written to the output frame.
     int16_t* dest_ptr_audio =
         resample ? audio.data() : preprocess_frame_.mutable_data();
+    RTC_DCHECK_GE(audio.size(), preprocess_frame_.samples_per_channel_);
     RTC_DCHECK_GE(audio.size(), in_frame.samples_per_channel_);
     DownMixFrame(in_frame,
                  rtc::ArrayView<int16_t>(
                      dest_ptr_audio, preprocess_frame_.samples_per_channel_));
     preprocess_frame_.num_channels_ = 1;
-    // Set the input of the resampler is the down-mixed signal.
+
+    // Set the input of the resampler to the down-mixed signal.
     src_ptr_audio = audio.data();
+  } else {
+    // Set the input of the resampler to the original data.
+    src_ptr_audio = in_frame.data();
   }
 
   preprocess_frame_.timestamp_ = expected_codec_ts_;

modules/audio_coding/include/audio_coding_module.h

@@ -33,8 +33,6 @@ class AudioEncoder;
 class AudioFrame;
 struct RTPHeader;
 
-#define WEBRTC_10MS_PCM_AUDIO 960  // 16 bits super wideband 48 kHz
-
 // Callback class used for sending data ready to be packetized
 class AudioPacketizationCallback {
  public:

modules/audio_coding/test/EncodeDecodeTest.cc

@@ -24,6 +24,12 @@
 
 namespace webrtc {
 
+namespace {
+// Buffer size for stereo 48 kHz audio.
+constexpr size_t kWebRtc10MsPcmAudio = 960;
+
+}  // namespace
+
 TestPacketization::TestPacketization(RTPStream* rtpStream, uint16_t frequency)
     : _rtpStream(rtpStream), _frequency(frequency), _seqNo(0) {}
 
@@ -92,7 +98,7 @@ void Sender::Run() {
 }
 
 Receiver::Receiver()
-    : _playoutLengthSmpls(WEBRTC_10MS_PCM_AUDIO),
+    : _playoutLengthSmpls(kWebRtc10MsPcmAudio),
       _payloadSizeBytes(MAX_INCOMING_PAYLOAD) {}
 
 void Receiver::Setup(AudioCodingModule* acm,
@@ -139,7 +145,7 @@ void Receiver::Setup(AudioCodingModule* acm,
   _pcmFile.Open(file_name, 32000, "wb+");
 
   _realPayloadSizeBytes = 0;
-  _playoutBuffer = new int16_t[WEBRTC_10MS_PCM_AUDIO];
+  _playoutBuffer = new int16_t[kWebRtc10MsPcmAudio];
   _frequency = playSampFreq;
   _acm = acm;
   _firstTime = true;