red: implement RED with distance 2

Extends the RED implementation to support a distance of two, i.e. two
packets redundancy.

BUG=webrtc:11640

Change-Id: I5113a97a4e3d45d836d7952a0c19c5381069c158
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/178565
Reviewed-by: Henrik Lundin <henrik.lundin@webrtc.org>
Commit-Queue: Henrik Lundin <henrik.lundin@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31625}

3 files changed, 104 insertions, 33 deletions

diff --git a/modules/audio_coding/codecs/red/audio_encoder_copy_red.cc b/modules/audio_coding/codecs/red/audio_encoder_copy_red.cc
index 8d028c9b9a..2bfd2c44df 100644
--- a/modules/audio_coding/codecs/red/audio_encoder_copy_red.cc
+++ b/modules/audio_coding/codecs/red/audio_encoder_copy_red.cc
@@ -19,6 +19,7 @@
 #include "rtc_base/checks.h"
 
 namespace webrtc {
+static const int kRedMaxPacketSize = 1 << 10;
 
 AudioEncoderCopyRed::Config::Config() = default;
 AudioEncoderCopyRed::Config::Config(Config&&) = default;
@@ -56,70 +57,101 @@ int AudioEncoderCopyRed::GetTargetBitrate() const {
   return speech_encoder_->GetTargetBitrate();
 }
 
+size_t AudioEncoderCopyRed::CalculateHeaderLength() const {
+  size_t header_size = 1;
+  if (secondary_info_.encoded_bytes > 0) {
+    header_size += 4;
+  }
+  if (tertiary_info_.encoded_bytes > 0) {
+    header_size += 4;
+  }
+  return header_size > 1 ? header_size : 0;
+}
+
 AudioEncoder::EncodedInfo AudioEncoderCopyRed::EncodeImpl(
     uint32_t rtp_timestamp,
     rtc::ArrayView<const int16_t> audio,
     rtc::Buffer* encoded) {
+  rtc::Buffer primary_encoded;
+  EncodedInfo info =
+      speech_encoder_->Encode(rtp_timestamp, audio, &primary_encoded);
+  RTC_CHECK(info.redundant.empty()) << "Cannot use nested redundant encoders.";
+  RTC_DCHECK_EQ(primary_encoded.size(), info.encoded_bytes);
+
+  if (info.encoded_bytes == 0) {
+    return info;
+  }
+
   // Allocate room for RFC 2198 header if there is redundant data.
   // Otherwise this will send the primary payload type without
   // wrapping in RED.
-  const size_t header_length_bytes = secondary_info_.encoded_bytes > 0 ? 5 : 0;
-  size_t secondary_length_bytes = 0;
+  const size_t header_length_bytes = CalculateHeaderLength();
+  encoded->SetSize(header_length_bytes);
 
-  if (secondary_info_.encoded_bytes > 0) {
-    encoded->SetSize(header_length_bytes);
-    encoded->AppendData(secondary_encoded_);
-    secondary_length_bytes = secondary_info_.encoded_bytes;
-  }
-  EncodedInfo info = speech_encoder_->Encode(rtp_timestamp, audio, encoded);
+  size_t header_offset = 0;
+  if (tertiary_info_.encoded_bytes > 0 &&
+      tertiary_info_.encoded_bytes < kRedMaxPacketSize) {
+    encoded->AppendData(tertiary_encoded_);
 
-  if (info.encoded_bytes == 0) {
-    encoded->Clear();
-    return info;
+    const uint32_t timestamp_delta =
+        info.encoded_timestamp - tertiary_info_.encoded_timestamp;
+
+    encoded->data()[header_offset] = tertiary_info_.payload_type | 0x80;
+    rtc::SetBE16(static_cast<uint8_t*>(encoded->data()) + header_offset + 1,
+                 (timestamp_delta << 2) | (tertiary_info_.encoded_bytes >> 8));
+    encoded->data()[header_offset + 3] = tertiary_info_.encoded_bytes & 0xff;
+    header_offset += 4;
   }
 
-  // Actually construct the RFC 2198 header.
-  if (secondary_info_.encoded_bytes > 0) {
+  if (secondary_info_.encoded_bytes > 0 &&
+      secondary_info_.encoded_bytes < kRedMaxPacketSize) {
+    encoded->AppendData(secondary_encoded_);
+
     const uint32_t timestamp_delta =
         info.encoded_timestamp - secondary_info_.encoded_timestamp;
 
-    encoded->data()[0] = secondary_info_.payload_type | 0x80;
-    RTC_DCHECK_LT(secondary_info_.encoded_bytes, 1 << 10);
-    rtc::SetBE16(static_cast<uint8_t*>(encoded->data()) + 1,
+    encoded->data()[header_offset] = secondary_info_.payload_type | 0x80;
+    rtc::SetBE16(static_cast<uint8_t*>(encoded->data()) + header_offset + 1,
                  (timestamp_delta << 2) | (secondary_info_.encoded_bytes >> 8));
-    encoded->data()[3] = secondary_info_.encoded_bytes & 0xff;
-    encoded->data()[4] = info.payload_type;
+    encoded->data()[header_offset + 3] = secondary_info_.encoded_bytes & 0xff;
+    header_offset += 4;
   }
 
-  RTC_CHECK(info.redundant.empty()) << "Cannot use nested redundant encoders.";
-  RTC_DCHECK_EQ(encoded->size() - header_length_bytes - secondary_length_bytes,
-                info.encoded_bytes);
+  encoded->AppendData(primary_encoded);
+  if (header_length_bytes > 0) {
+    RTC_DCHECK_EQ(header_offset, header_length_bytes - 1);
+    encoded->data()[header_offset] = info.payload_type;
+  }
 
   // |info| will be implicitly cast to an EncodedInfoLeaf struct, effectively
   // discarding the (empty) vector of redundant information. This is
   // intentional.
   info.redundant.push_back(info);
   RTC_DCHECK_EQ(info.redundant.size(), 1);
+  RTC_DCHECK_EQ(info.speech, info.redundant[0].speech);
   if (secondary_info_.encoded_bytes > 0) {
     info.redundant.push_back(secondary_info_);
     RTC_DCHECK_EQ(info.redundant.size(), 2);
   }
+  if (tertiary_info_.encoded_bytes > 0) {
+    info.redundant.push_back(tertiary_info_);
+    RTC_DCHECK_EQ(info.redundant.size(),
+                  2 + (secondary_info_.encoded_bytes > 0 ? 1 : 0));
+  }
+
+  // Save secondary to tertiary.
+  tertiary_encoded_.SetData(secondary_encoded_);
+  tertiary_info_ = secondary_info_;
+
   // Save primary to secondary.
-  secondary_encoded_.SetData(
-      &encoded->data()[header_length_bytes + secondary_info_.encoded_bytes],
-      info.encoded_bytes);
+  secondary_encoded_.SetData(primary_encoded);
   secondary_info_ = info;
-  RTC_DCHECK_EQ(info.speech, info.redundant[0].speech);
 
   // Update main EncodedInfo.
   if (header_length_bytes > 0) {
     info.payload_type = red_payload_type_;
   }
-  info.encoded_bytes = header_length_bytes;
-  for (std::vector<EncodedInfoLeaf>::const_iterator it = info.redundant.begin();
-       it != info.redundant.end(); ++it) {
-    info.encoded_bytes += it->encoded_bytes;
-  }
+  info.encoded_bytes = encoded->size();
   return info;
 }
 
diff --git a/modules/audio_coding/codecs/red/audio_encoder_copy_red.h b/modules/audio_coding/codecs/red/audio_encoder_copy_red.h
index c6e829eeb6..4d7fc404f1 100644
--- a/modules/audio_coding/codecs/red/audio_encoder_copy_red.h
+++ b/modules/audio_coding/codecs/red/audio_encoder_copy_red.h
@@ -71,10 +71,13 @@ class AudioEncoderCopyRed final : public AudioEncoder {
                          rtc::Buffer* encoded) override;
 
  private:
+  size_t CalculateHeaderLength() const;
   std::unique_ptr<AudioEncoder> speech_encoder_;
   int red_payload_type_;
   rtc::Buffer secondary_encoded_;
   EncodedInfoLeaf secondary_info_;
+  rtc::Buffer tertiary_encoded_;
+  EncodedInfoLeaf tertiary_info_;
   RTC_DISALLOW_COPY_AND_ASSIGN(AudioEncoderCopyRed);
 };
 
diff --git a/modules/audio_coding/codecs/red/audio_encoder_copy_red_unittest.cc b/modules/audio_coding/codecs/red/audio_encoder_copy_red_unittest.cc
index 720acb4f87..fbc0b8aa38 100644
--- a/modules/audio_coding/codecs/red/audio_encoder_copy_red_unittest.cc
+++ b/modules/audio_coding/codecs/red/audio_encoder_copy_red_unittest.cc
@@ -183,12 +183,19 @@ TEST_F(AudioEncoderCopyRedTest, CheckPayloadSizes) {
   EXPECT_EQ(1u, encoded_info_.redundant.size());
   EXPECT_EQ(1u, encoded_info_.encoded_bytes);
 
-  for (size_t i = 2; i <= kNumPackets; ++i) {
+  // Second call is also special since it does not include a ternary
+  // payload.
+  Encode();
+  EXPECT_EQ(2u, encoded_info_.redundant.size());
+  EXPECT_EQ(8u, encoded_info_.encoded_bytes);
+
+  for (size_t i = 3; i <= kNumPackets; ++i) {
     Encode();
-    ASSERT_EQ(2u, encoded_info_.redundant.size());
+    ASSERT_EQ(3u, encoded_info_.redundant.size());
     EXPECT_EQ(i, encoded_info_.redundant[0].encoded_bytes);
     EXPECT_EQ(i - 1, encoded_info_.redundant[1].encoded_bytes);
-    EXPECT_EQ(5 + i + i - 1, encoded_info_.encoded_bytes);
+    EXPECT_EQ(i - 2, encoded_info_.redundant[2].encoded_bytes);
+    EXPECT_EQ(9 + i + (i - 1) + (i - 2), encoded_info_.encoded_bytes);
   }
 }
 
@@ -317,6 +324,35 @@ TEST_F(AudioEncoderCopyRedTest, CheckRFC2198Header) {
   EXPECT_EQ(encoded_[2] & 0x3u, encoded_info_.redundant[1].encoded_bytes >> 8);
   EXPECT_EQ(encoded_[3], encoded_info_.redundant[1].encoded_bytes & 0xff);
   EXPECT_EQ(encoded_[4], primary_payload_type);
+
+  EXPECT_CALL(*mock_encoder_, EncodeImpl(_, _, _))
+      .WillOnce(Invoke(MockAudioEncoder::FakeEncoding(info)));
+  Encode();  // Third call will produce a redundant encoding with double
+             // redundancy.
+
+  EXPECT_EQ(encoded_.size(),
+            9u + 3 * 10u);  // header size + two encoded payloads.
+  EXPECT_EQ(encoded_[0], primary_payload_type | 0x80);
+
+  timestamp_delta = encoded_info_.encoded_timestamp -
+                    encoded_info_.redundant[2].encoded_timestamp;
+  // Timestamp delta is encoded as a 14 bit value.
+  EXPECT_EQ(encoded_[1], timestamp_delta >> 6);
+  EXPECT_EQ(static_cast<uint8_t>(encoded_[2] >> 2), timestamp_delta & 0x3f);
+  // Redundant length is encoded as 10 bit value.
+  EXPECT_EQ(encoded_[2] & 0x3u, encoded_info_.redundant[2].encoded_bytes >> 8);
+  EXPECT_EQ(encoded_[3], encoded_info_.redundant[2].encoded_bytes & 0xff);
+
+  EXPECT_EQ(encoded_[4], primary_payload_type | 0x80);
+  timestamp_delta = encoded_info_.encoded_timestamp -
+                    encoded_info_.redundant[1].encoded_timestamp;
+  // Timestamp delta is encoded as a 14 bit value.
+  EXPECT_EQ(encoded_[5], timestamp_delta >> 6);
+  EXPECT_EQ(static_cast<uint8_t>(encoded_[6] >> 2), timestamp_delta & 0x3f);
+  // Redundant length is encoded as 10 bit value.
+  EXPECT_EQ(encoded_[6] & 0x3u, encoded_info_.redundant[2].encoded_bytes >> 8);
+  EXPECT_EQ(encoded_[7], encoded_info_.redundant[2].encoded_bytes & 0xff);
+  EXPECT_EQ(encoded_[8], primary_payload_type);
 }
 
 #if GTEST_HAS_DEATH_TEST && !defined(WEBRTC_ANDROID)