2nd reland of https://webrtc-review.googlesource.com/c/src/+/114883
The difference to the original is new bitexactness strings. The
reason for reland is breaking downstream projects.
Original CL description:
Tests for multi-stream Opus.
This CL (mainly) adds bit-exactness tests for multi-stream Opus. The
tests are in audio_coding_unittest.cc. Some refactoring of
AcmSendTestOldApi, AcmSenderBitExactnessOldApi is done to make it
possible. A few checks for "channels \in {1, 2}" are replaced with
"channels \in {1, 2, 4, 6, 8}" in the WebRTC Opus codec wrapper. A few
other changes are made to be able to write and read multi-channel WAV
files.
The SDP changes are NOT included; as of this CL there is no way to set
up a multi-stream opus en/de-coder from SDP strings.
TBR=ossu@webrtc.org
Bug: webrtc:8649
Change-Id: I6261b18c69fd666d43ab34ed8f1bc9d5cc82b21f
Reviewed-on: https://webrtc-review.googlesource.com/c/123882
Reviewed-by: Alex Loiko <aleloi@webrtc.org>
Commit-Queue: Alex Loiko <aleloi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#26809}diff --git a/api/audio_codecs/opus/audio_encoder_opus_config.cc b/api/audio_codecs/opus/audio_encoder_opus_config.cc
index 301c256..2847cea 100644
--- a/api/audio_codecs/opus/audio_encoder_opus_config.cc
+++ b/api/audio_codecs/opus/audio_encoder_opus_config.cc
@@ -55,7 +55,8 @@
bool AudioEncoderOpusConfig::IsOk() const {
if (frame_size_ms <= 0 || frame_size_ms % 10 != 0)
return false;
- if (num_channels != 1 && num_channels != 2)
+ if (num_channels != 1 && num_channels != 2 && num_channels != 4 &&
+ num_channels != 6 && num_channels != 8)
return false;
if (!bitrate_bps)
return false;
diff --git a/modules/BUILD.gn b/modules/BUILD.gn
index 40ea128..501ba2a 100644
--- a/modules/BUILD.gn
+++ b/modules/BUILD.gn
@@ -126,6 +126,7 @@
"../resources/audio_coding/speech_mono_32_48kHz.pcm",
"../resources/audio_coding/speech_4_channels_48k_one_second.wav",
"../resources/audio_coding/testfile32kHz.pcm",
+ "../resources/audio_coding/testfile_fake_stereo_32kHz.pcm",
"../resources/audio_coding/teststereo32kHz.pcm",
"../resources/audio_device/audio_short16.pcm",
"../resources/audio_device/audio_short44.pcm",
diff --git a/modules/audio_coding/acm2/acm_receive_test.h b/modules/audio_coding/acm2/acm_receive_test.h
index fce3d6b..9d004c6 100644
--- a/modules/audio_coding/acm2/acm_receive_test.h
+++ b/modules/audio_coding/acm2/acm_receive_test.h
@@ -33,7 +33,8 @@
enum NumOutputChannels : size_t {
kArbitraryChannels = 0,
kMonoOutput = 1,
- kStereoOutput = 2
+ kStereoOutput = 2,
+ kQuadOutput = 4
};
AcmReceiveTestOldApi(PacketSource* packet_source,
diff --git a/modules/audio_coding/acm2/acm_send_test.cc b/modules/audio_coding/acm2/acm_send_test.cc
index 98d673f..b6110b6 100644
--- a/modules/audio_coding/acm2/acm_send_test.cc
+++ b/modules/audio_coding/acm2/acm_send_test.cc
@@ -106,13 +106,9 @@
// Insert audio and process until one packet is produced.
while (clock_.TimeInMilliseconds() < test_duration_ms_) {
clock_.AdvanceTimeMilliseconds(kBlockSizeMs);
- RTC_CHECK(audio_source_->Read(input_block_size_samples_,
- input_frame_.mutable_data()));
- if (input_frame_.num_channels_ > 1) {
- InputAudioFile::DuplicateInterleaved(
- input_frame_.data(), input_block_size_samples_,
- input_frame_.num_channels_, input_frame_.mutable_data());
- }
+ RTC_CHECK(audio_source_->Read(
+ input_block_size_samples_ * input_frame_.num_channels_,
+ input_frame_.mutable_data()));
data_to_send_ = false;
RTC_CHECK_GE(acm_->Add10MsData(input_frame_), 0);
input_frame_.timestamp_ += static_cast<uint32_t>(input_block_size_samples_);
diff --git a/modules/audio_coding/acm2/audio_coding_module.cc b/modules/audio_coding/acm2/audio_coding_module.cc
index 67ef556..1547b37 100644
--- a/modules/audio_coding/acm2/audio_coding_module.cc
+++ b/modules/audio_coding/acm2/audio_coding_module.cc
@@ -477,7 +477,9 @@
return -1;
}
- if (audio_frame.num_channels_ != 1 && audio_frame.num_channels_ != 2) {
+ if (audio_frame.num_channels_ != 1 && audio_frame.num_channels_ != 2 &&
+ audio_frame.num_channels_ != 4 && audio_frame.num_channels_ != 6 &&
+ audio_frame.num_channels_ != 8) {
RTC_LOG(LS_ERROR) << "Cannot Add 10 ms audio, invalid number of channels.";
return -1;
}
diff --git a/modules/audio_coding/acm2/audio_coding_module_unittest.cc b/modules/audio_coding/acm2/audio_coding_module_unittest.cc
index c9a03a1..a609f98 100644
--- a/modules/audio_coding/acm2/audio_coding_module_unittest.cc
+++ b/modules/audio_coding/acm2/audio_coding_module_unittest.cc
@@ -17,6 +17,7 @@
#include "api/audio_codecs/audio_encoder.h"
#include "api/audio_codecs/builtin_audio_decoder_factory.h"
#include "api/audio_codecs/builtin_audio_encoder_factory.h"
+#include "api/audio_codecs/opus/audio_decoder_opus.h"
#include "api/audio_codecs/opus/audio_encoder_opus.h"
#include "modules/audio_coding/acm2/acm_receive_test.h"
#include "modules/audio_coding/acm2/acm_send_test.h"
@@ -24,6 +25,8 @@
#include "modules/audio_coding/codecs/g711/audio_decoder_pcm.h"
#include "modules/audio_coding/codecs/g711/audio_encoder_pcm.h"
#include "modules/audio_coding/codecs/isac/main/include/audio_encoder_isac.h"
+#include "modules/audio_coding/codecs/opus/audio_decoder_opus.h"
+#include "modules/audio_coding/codecs/opus/audio_encoder_opus.h"
#include "modules/audio_coding/include/audio_coding_module.h"
#include "modules/audio_coding/include/audio_coding_module_typedefs.h"
#include "modules/audio_coding/neteq/tools/audio_checksum.h"
@@ -44,6 +47,7 @@
#include "rtc_base/thread_annotations.h"
#include "system_wrappers/include/clock.h"
#include "system_wrappers/include/sleep.h"
+#include "test/audio_decoder_proxy_factory.h"
#include "test/gtest.h"
#include "test/mock_audio_decoder.h"
#include "test/mock_audio_encoder.h"
@@ -935,7 +939,7 @@
->test_case_name() +
"_" + ::testing::UnitTest::GetInstance()->current_test_info()->name() +
"_output.wav";
- test::OutputWavFile output_file(output_file_name, output_freq_hz);
+ test::OutputWavFile output_file(output_file_name, output_freq_hz, 1);
test::AudioSinkFork output(&checksum, &output_file);
test::AcmReceiveTestOldApi test(
@@ -1116,15 +1120,12 @@
// Sets up the test::AcmSendTest object. Returns true on success, otherwise
// false.
- bool SetUpSender() {
- const std::string input_file_name =
- webrtc::test::ResourcePath("audio_coding/testfile32kHz", "pcm");
+ bool SetUpSender(std::string input_file_name, int source_rate) {
// Note that |audio_source_| will loop forever. The test duration is set
// explicitly by |kTestDurationMs|.
audio_source_.reset(new test::InputAudioFile(input_file_name));
- static const int kSourceRateHz = 32000;
- send_test_.reset(new test::AcmSendTestOldApi(
- audio_source_.get(), kSourceRateHz, kTestDurationMs));
+ send_test_.reset(new test::AcmSendTestOldApi(audio_source_.get(),
+ source_rate, kTestDurationMs));
return send_test_.get() != NULL;
}
@@ -1157,7 +1158,11 @@
void Run(const std::string& audio_checksum_ref,
const std::string& payload_checksum_ref,
int expected_packets,
- test::AcmReceiveTestOldApi::NumOutputChannels expected_channels) {
+ test::AcmReceiveTestOldApi::NumOutputChannels expected_channels,
+ rtc::scoped_refptr<AudioDecoderFactory> decoder_factory = nullptr) {
+ if (!decoder_factory) {
+ decoder_factory = CreateBuiltinAudioDecoderFactory();
+ }
// Set up the receiver used to decode the packets and verify the decoded
// output.
test::AudioChecksum audio_checksum;
@@ -1169,12 +1174,12 @@
"_" + ::testing::UnitTest::GetInstance()->current_test_info()->name() +
"_output.wav";
const int kOutputFreqHz = 8000;
- test::OutputWavFile output_file(output_file_name, kOutputFreqHz);
+ test::OutputWavFile output_file(output_file_name, kOutputFreqHz,
+ expected_channels);
// Have the output audio sent both to file and to the checksum calculator.
test::AudioSinkFork output(&audio_checksum, &output_file);
test::AcmReceiveTestOldApi receive_test(this, &output, kOutputFreqHz,
- expected_channels,
- CreateBuiltinAudioDecoderFactory());
+ expected_channels, decoder_factory);
ASSERT_NO_FATAL_FAILURE(receive_test.RegisterDefaultCodecs());
// This is where the actual test is executed.
@@ -1250,7 +1255,8 @@
int payload_type,
int codec_frame_size_samples,
int codec_frame_size_rtp_timestamps) {
- ASSERT_TRUE(SetUpSender());
+ ASSERT_TRUE(SetUpSender(
+ channels == 1 ? kTestFileMono32kHz : kTestFileFakeStereo32kHz, 32000));
ASSERT_TRUE(RegisterSendCodec(codec_name, codec_sample_rate_hz, channels,
payload_type, codec_frame_size_samples,
codec_frame_size_rtp_timestamps));
@@ -1259,7 +1265,7 @@
void SetUpTestExternalEncoder(
std::unique_ptr<AudioEncoder> external_speech_encoder,
int payload_type) {
- ASSERT_TRUE(SetUpSender());
+ ASSERT_TRUE(send_test_);
RegisterExternalSendCodec(std::move(external_speech_encoder), payload_type);
}
@@ -1271,6 +1277,14 @@
uint16_t last_sequence_number_;
uint32_t last_timestamp_;
std::unique_ptr<rtc::MessageDigest> payload_checksum_;
+ const std::string kTestFileMono32kHz =
+ webrtc::test::ResourcePath("audio_coding/testfile32kHz", "pcm");
+ const std::string kTestFileFakeStereo32kHz =
+ webrtc::test::ResourcePath("audio_coding/testfile_fake_stereo_32kHz",
+ "pcm");
+ const std::string kTestFileQuad48kHz = webrtc::test::ResourcePath(
+ "audio_coding/speech_4_channels_48k_one_second",
+ "wav");
};
class AcmSenderBitExactnessNewApi : public AcmSenderBitExactnessOldApi {};
@@ -1481,17 +1495,59 @@
TEST_F(AcmSenderBitExactnessNewApi, MAYBE_OpusFromFormat_stereo_20ms) {
const auto config = AudioEncoderOpus::SdpToConfig(
SdpAudioFormat("opus", 48000, 2, {{"stereo", "1"}}));
+ ASSERT_TRUE(SetUpSender(kTestFileFakeStereo32kHz, 32000));
ASSERT_NO_FATAL_FAILURE(SetUpTestExternalEncoder(
AudioEncoderOpus::MakeAudioEncoder(*config, 120), 120));
Run(audio_checksum, payload_checksum, 50,
test::AcmReceiveTestOldApi::kStereoOutput);
}
+TEST_F(AcmSenderBitExactnessNewApi, OpusManyChannels) {
+ constexpr int kNumChannels = 4;
+ constexpr int kOpusPayloadType = 120;
+ constexpr int kBitrateBps = 128000;
+
+ // Read a 4 channel file at 48kHz.
+ ASSERT_TRUE(SetUpSender(kTestFileQuad48kHz, 48000));
+
+ // TODO(webrtc:8649): change to higher level
+ // AudioEncoderOpus::MakeAudioEncoder once a multistream encoder can be set up
+ // from SDP.
+ AudioEncoderOpusConfig config = *AudioEncoderOpus::SdpToConfig(
+ SdpAudioFormat("opus", 48000, 2, {{"stereo", "1"}}));
+ config.num_channels = kNumChannels;
+ config.bitrate_bps = kBitrateBps;
+
+ ASSERT_NO_FATAL_FAILURE(SetUpTestExternalEncoder(
+ absl::make_unique<AudioEncoderOpusImpl>(config, kOpusPayloadType),
+ kOpusPayloadType));
+
+ AudioDecoderOpusImpl opus_decoder(kNumChannels);
+
+ rtc::scoped_refptr<AudioDecoderFactory> decoder_factory =
+ new rtc::RefCountedObject<test::AudioDecoderProxyFactory>(&opus_decoder);
+
+ // Set up an EXTERNAL DECODER to parse 4 channels.
+ Run(AcmReceiverBitExactnessOldApi::PlatformChecksum( // audio checksum
+ "b70470884d9a8613eff019b0d1c8876e|d0a73d377e0ca1be6b06e989e0ad2c35",
+ "d0a73d377e0ca1be6b06e989e0ad2c35",
+ "b45d2ce5fc4723e9eb41350af9c68f56", "android arm64 audio checksum",
+ "1c9a3c9dacdd4b8fc9ff608227e531f2"),
+ // payload_checksum,
+ AcmReceiverBitExactnessOldApi::PlatformChecksum( // payload checksum
+ "c2e7d40f8269ef754bd86d6be9623fa7|76de0f4992e3937ca60d35bbb0d308d6",
+ "76de0f4992e3937ca60d35bbb0d308d6",
+ "2a310aca965c16c2dfd61a9f9fc0c877", "android arm64 payload checksum",
+ "2294f4b61fb8f174f5196776a0a49be7"),
+ 50, test::AcmReceiveTestOldApi::kQuadOutput, decoder_factory);
+}
+
TEST_F(AcmSenderBitExactnessNewApi, OpusFromFormat_stereo_20ms_voip) {
auto config = AudioEncoderOpus::SdpToConfig(
SdpAudioFormat("opus", 48000, 2, {{"stereo", "1"}}));
// If not set, default will be kAudio in case of stereo.
config->application = AudioEncoderOpusConfig::ApplicationMode::kVoip;
+ ASSERT_TRUE(SetUpSender(kTestFileFakeStereo32kHz, 32000));
ASSERT_NO_FATAL_FAILURE(SetUpTestExternalEncoder(
AudioEncoderOpus::MakeAudioEncoder(*config, 120), 120));
// Checksum depends on libopus being compiled with or without SSE.
@@ -1775,6 +1831,7 @@
&encoder, static_cast<AudioEncoder::EncodedInfo (AudioEncoder::*)(
uint32_t, rtc::ArrayView<const int16_t>, rtc::Buffer*)>(
&AudioEncoderPcmU::Encode)));
+ ASSERT_TRUE(SetUpSender(kTestFileMono32kHz, 32000));
ASSERT_NO_FATAL_FAILURE(
SetUpTestExternalEncoder(std::move(mock_encoder), config.payload_type));
Run("81a9d4c0bb72e9becc43aef124c981e9", "8f9b8750bd80fe26b6cbf6659b89f0f9",
diff --git a/modules/audio_coding/codecs/opus/audio_decoder_opus.cc b/modules/audio_coding/codecs/opus/audio_decoder_opus.cc
index 1accfe4..b6eada9 100644
--- a/modules/audio_coding/codecs/opus/audio_decoder_opus.cc
+++ b/modules/audio_coding/codecs/opus/audio_decoder_opus.cc
@@ -71,7 +71,8 @@
AudioDecoderOpusImpl::AudioDecoderOpusImpl(size_t num_channels)
: channels_(num_channels) {
- RTC_DCHECK(num_channels == 1 || num_channels == 2);
+ RTC_DCHECK(num_channels == 1 || num_channels == 2 || num_channels == 4 ||
+ num_channels == 6 || num_channels == 8);
const int error = WebRtcOpus_DecoderCreate(&dec_state_, channels_);
RTC_DCHECK(error == 0);
WebRtcOpus_DecoderInit(dec_state_);
diff --git a/modules/audio_coding/neteq/tools/input_audio_file.cc b/modules/audio_coding/neteq/tools/input_audio_file.cc
index 6d11064..d5e2862 100644
--- a/modules/audio_coding/neteq/tools/input_audio_file.cc
+++ b/modules/audio_coding/neteq/tools/input_audio_file.cc
@@ -18,9 +18,11 @@
InputAudioFile::InputAudioFile(const std::string file_name, bool loop_at_end)
: loop_at_end_(loop_at_end) {
fp_ = fopen(file_name.c_str(), "rb");
+ RTC_DCHECK(fp_) << file_name << " could not be opened.";
}
InputAudioFile::~InputAudioFile() {
+ RTC_DCHECK(fp_);
fclose(fp_);
}
diff --git a/modules/audio_coding/neteq/tools/output_wav_file.h b/modules/audio_coding/neteq/tools/output_wav_file.h
index 3ffcfc6..6982a76 100644
--- a/modules/audio_coding/neteq/tools/output_wav_file.h
+++ b/modules/audio_coding/neteq/tools/output_wav_file.h
@@ -24,8 +24,10 @@
public:
// Creates an OutputWavFile, opening a file named |file_name| for writing.
// The output file is a PCM encoded wav file.
- OutputWavFile(const std::string& file_name, int sample_rate_hz)
- : wav_writer_(file_name, sample_rate_hz, 1) {}
+ OutputWavFile(const std::string& file_name,
+ int sample_rate_hz,
+ int num_channels = 1)
+ : wav_writer_(file_name, sample_rate_hz, num_channels) {}
bool WriteArray(const int16_t* audio, size_t num_samples) override {
wav_writer_.WriteSamples(audio, num_samples);
diff --git a/resources/audio_coding/testfile_fake_stereo_32kHz.pcm.sha1 b/resources/audio_coding/testfile_fake_stereo_32kHz.pcm.sha1
new file mode 100644
index 0000000..004f8bb
--- /dev/null
+++ b/resources/audio_coding/testfile_fake_stereo_32kHz.pcm.sha1
@@ -0,0 +1 @@
+4f382602b5605dbbbf78451810ce644788681262
\ No newline at end of file
