WIP: Adding a centralized NetEq Clock

webrtc/modules/audio_coding/neteq/neteq_impl.cc

Index: webrtc/modules/audio_coding/neteq/neteq_impl.cc
diff --git a/webrtc/modules/audio_coding/neteq/neteq_impl.cc b/webrtc/modules/audio_coding/neteq/neteq_impl.cc
index db37e716d66dde966a4cac42f10dd47166c4f42c..1b6e0c0c3e13856ad9c334d0efbe3daba8f65716 100644
--- a/webrtc/modules/audio_coding/neteq/neteq_impl.cc
+++ b/webrtc/modules/audio_coding/neteq/neteq_impl.cc
@@ -42,6 +42,7 @@
 #include "webrtc/modules/audio_coding/neteq/post_decode_vad.h"
 #include "webrtc/modules/audio_coding/neteq/preemptive_expand.h"
 #include "webrtc/modules/audio_coding/neteq/sync_buffer.h"
+#include "webrtc/modules/audio_coding/neteq/tick_timer.h"
 #include "webrtc/modules/audio_coding/neteq/timestamp_scaler.h"
 #include "webrtc/modules/include/module_common_types.h"
 
@@ -53,6 +54,7 @@
 namespace webrtc {
 
 NetEqImpl::NetEqImpl(const NetEq::Config& config,
+                     std::unique_ptr<TickTimer> tick_timer,
                      BufferLevelFilter* buffer_level_filter,
                      DecoderDatabase* decoder_database,
                      DelayManager* delay_manager,
@@ -66,7 +68,8 @@ NetEqImpl::NetEqImpl(const NetEq::Config& config,
                      ExpandFactory* expand_factory,
                      PreemptiveExpandFactory* preemptive_expand_factory,
                      bool create_components)
-    : buffer_level_filter_(buffer_level_filter),
+    : tick_timer_(std::move(tick_timer)),
+      buffer_level_filter_(buffer_level_filter),
       decoder_database_(decoder_database),
       delay_manager_(delay_manager),
       delay_peak_detector_(delay_peak_detector),
@@ -532,7 +535,8 @@ int NetEqImpl::InsertPacketInternal(const WebRtcRTPHeader& rtp_header,
     packet->header.numCSRCs = 0;
     packet->payload_length = payload.size();
     packet->primary = true;
-    packet->waiting_time = 0;
+    // Waiting time will be set upon inserting the packet in the buffer.
+    RTC_DCHECK(!packet->waiting_time);
     packet->payload = new uint8_t[packet->payload_length];
     packet->sync_packet = is_sync_packet;
     if (!packet->payload) {
@@ -788,6 +792,8 @@ int NetEqImpl::GetAudioInternal(AudioFrame* audio_frame) {
   DtmfEvent dtmf_event;
   Operations operation;
   bool play_dtmf;
+  tick_timer_->Increment();
+  stats_.IncreaseCounter(output_size_samples_, fs_hz_);
   int return_value = GetDecision(&operation, &packet_list, &dtmf_event,
                                  &play_dtmf);
   if (return_value != 0) {
@@ -806,6 +812,11 @@ int NetEqImpl::GetAudioInternal(AudioFrame* audio_frame) {
   vad_->Update(decoded_buffer_.get(), static_cast<size_t>(length), speech_type,
                sid_frame_available, fs_hz_);
 
+  if (sid_frame_available || speech_type == AudioDecoder::kComfortNoise) {
+    // Start a new stopwatch since we are decoding a new CNG packet.
+    generated_noise_stopwatch_ = tick_timer_->GetNewStopwatch();
+  }
+
   algorithm_buffer_->Clear();
   switch (operation) {
     case kNormal: {
@@ -978,6 +989,12 @@ int NetEqImpl::GetAudioInternal(AudioFrame* audio_frame) {
           : timestamp_scaler_->ToExternal(playout_timestamp_) -
                 static_cast<uint32_t>(audio_frame->samples_per_channel_);
 
+  if (!(last_mode_ == kModeRfc3389Cng ||
+      last_mode_ == kModeCodecInternalCng ||
+      last_mode_ == kModeExpand)) {
+    generated_noise_stopwatch_.reset();
+  }
+
   if (decode_return_value) return decode_return_value;
   return return_value;
 }
@@ -990,10 +1007,6 @@ int NetEqImpl::GetDecision(Operations* operation,
   *play_dtmf = false;
   *operation = kUndefined;
 
-  // Increment time counters.
-  packet_buffer_->IncrementWaitingTimes();
-  stats_.IncreaseCounter(output_size_samples_, fs_hz_);
-
   assert(sync_buffer_.get());
   uint32_t end_timestamp = sync_buffer_->end_timestamp();
   if (!new_codec_) {
@@ -1002,14 +1015,22 @@ int NetEqImpl::GetDecision(Operations* operation,
   }
   const RTPHeader* header = packet_buffer_->NextRtpHeader();
 
+  RTC_DCHECK(!generated_noise_stopwatch_ ||
+             generated_noise_stopwatch_->ElapsedTicks() >= 1);
+  uint64_t generated_noise_samples =
+      generated_noise_stopwatch_
+          ? (generated_noise_stopwatch_->ElapsedTicks() - 1) *
+                    output_size_samples_ +
+                decision_logic_->noise_fast_forward()
+          : 0;
+
   if (decision_logic_->CngRfc3389On() || last_mode_ == kModeRfc3389Cng) {
     // Because of timestamp peculiarities, we have to "manually" disallow using
     // a CNG packet with the same timestamp as the one that was last played.
     // This can happen when using redundancy and will cause the timing to shift.
     while (header && decoder_database_->IsComfortNoise(header->payloadType) &&
            (end_timestamp >= header->timestamp ||
-            end_timestamp + decision_logic_->generated_noise_samples() >
-                header->timestamp)) {
+            end_timestamp + generated_noise_samples > header->timestamp)) {
       // Don't use this packet, discard it.
       if (packet_buffer_->DiscardNextPacket() != PacketBuffer::kOK) {
         assert(false);  // Must be ok by design.
@@ -1037,7 +1058,7 @@ int NetEqImpl::GetDecision(Operations* operation,
   // Check if it is time to play a DTMF event.
   if (dtmf_buffer_->GetEvent(
       static_cast<uint32_t>(
-          end_timestamp + decision_logic_->generated_noise_samples()),
+          end_timestamp + generated_noise_samples),
       dtmf_event)) {
     *play_dtmf = true;
   }
@@ -1045,13 +1066,14 @@ int NetEqImpl::GetDecision(Operations* operation,
   // Get instruction.
   assert(sync_buffer_.get());
   assert(expand_.get());
-  *operation = decision_logic_->GetDecision(*sync_buffer_,
-                                            *expand_,
-                                            decoder_frame_length_,
-                                            header,
-                                            last_mode_,
-                                            *play_dtmf,
-                                            &reset_decoder_);
+  generated_noise_samples =
+      generated_noise_stopwatch_
+          ? generated_noise_stopwatch_->ElapsedTicks() * output_size_samples_ +
+                decision_logic_->noise_fast_forward()
+          : 0;
+  *operation = decision_logic_->GetDecision(
+      *sync_buffer_, *expand_, decoder_frame_length_, header, last_mode_,
+      *play_dtmf, generated_noise_samples, &reset_decoder_);
 
   // Check if we already have enough samples in the |sync_buffer_|. If so,
   // change decision to normal, unless the decision was merge, accelerate, or
@@ -1124,15 +1146,19 @@ int NetEqImpl::GetDecision(Operations* operation,
       // TODO(hlundin): Write test for this.
       // Update timestamp.
       timestamp_ = end_timestamp;
-      if (decision_logic_->generated_noise_samples() > 0 &&
-          last_mode_ != kModeDtmf) {
+      const uint64_t generated_noise_samples =
+          generated_noise_stopwatch_
+              ? generated_noise_stopwatch_->ElapsedTicks() *
+                        output_size_samples_ +
+                    decision_logic_->noise_fast_forward()
+              : 0;
+      if (generated_noise_samples > 0 && last_mode_ != kModeDtmf) {
         // Make a jump in timestamp due to the recently played comfort noise.
         uint32_t timestamp_jump =
-            static_cast<uint32_t>(decision_logic_->generated_noise_samples());
+            static_cast<uint32_t>(generated_noise_samples);
         sync_buffer_->IncreaseEndTimestamp(timestamp_jump);
         timestamp_ += timestamp_jump;
       }
-      decision_logic_->set_generated_noise_samples(0);
       return 0;
     }
     case kAccelerate:
@@ -1215,9 +1241,6 @@ int NetEqImpl::GetDecision(Operations* operation,
       // We are about to decode and use a non-CNG packet.
       decision_logic_->SetCngOff();
     }
-    // Reset CNG timestamp as a new packet will be delivered.
-    // (Also if this is a CNG packet, since playedOutTS is updated.)
-    decision_logic_->set_generated_noise_samples(0);
 
     extracted_samples = ExtractPackets(required_samples, packet_list);
     if (extracted_samples < 0) {
@@ -1550,6 +1573,12 @@ int NetEqImpl::DoExpand(bool play_dtmf) {
   if (!play_dtmf) {
     dtmf_tone_generator_->Reset();
   }
+
+  if (!generated_noise_stopwatch_) {
+    // Start a new stopwatch since we may be covering for a lost CNG packet.
+    generated_noise_stopwatch_ = tick_timer_->GetNewStopwatch();
+  }
+
   return 0;
 }
 
@@ -1920,8 +1949,7 @@ int NetEqImpl::ExtractPackets(size_t required_samples,
       return -1;
     }
     stats_.PacketsDiscarded(discard_count);
-    // Store waiting time in ms; packets->waiting_time is in "output blocks".
-    stats_.StoreWaitingTime(packet->waiting_time * kOutputSizeMs);
+    stats_.StoreWaitingTime(packet->waiting_time->ElapsedMs());
     assert(packet->payload_length > 0);
     packet_list->push_back(packet);  // Store packet in list.
 
@@ -2094,11 +2122,9 @@ NetEqImpl::OutputType NetEqImpl::LastOutputType() {
 }
 
 void NetEqImpl::CreateDecisionLogic() {
-  decision_logic_.reset(DecisionLogic::Create(fs_hz_, output_size_samples_,
-                                              playout_mode_,
-                                              decoder_database_.get(),
-                                              *packet_buffer_.get(),
-                                              delay_manager_.get(),
-                                              buffer_level_filter_.get()));
+  decision_logic_.reset(DecisionLogic::Create(
+      fs_hz_, output_size_samples_, playout_mode_, decoder_database_.get(),
+      *packet_buffer_.get(), delay_manager_.get(), buffer_level_filter_.get(),
+      *tick_timer_));
 }
 }  // namespace webrtc