blob: 911bd50f67b0d92a8dd81479b192b97ad0810b28 [file] [log] [blame]
/*
* Copyright (c) 2014 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 "modules/audio_coding/include/audio_coding_module.h"
#include <stdio.h>
#include <string.h>
#include <atomic>
#include <memory>
#include <vector>
#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_multi_channel_opus.h"
#include "api/audio_codecs/opus/audio_decoder_opus.h"
#include "api/audio_codecs/opus/audio_encoder_multi_channel_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"
#include "modules/audio_coding/codecs/cng/audio_encoder_cng.h"
#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/include/audio_coding_module_typedefs.h"
#include "modules/audio_coding/neteq/tools/audio_checksum.h"
#include "modules/audio_coding/neteq/tools/audio_loop.h"
#include "modules/audio_coding/neteq/tools/constant_pcm_packet_source.h"
#include "modules/audio_coding/neteq/tools/input_audio_file.h"
#include "modules/audio_coding/neteq/tools/output_audio_file.h"
#include "modules/audio_coding/neteq/tools/output_wav_file.h"
#include "modules/audio_coding/neteq/tools/packet.h"
#include "modules/audio_coding/neteq/tools/rtp_file_source.h"
#include "rtc_base/event.h"
#include "rtc_base/message_digest.h"
#include "rtc_base/numerics/safe_conversions.h"
#include "rtc_base/platform_thread.h"
#include "rtc_base/synchronization/mutex.h"
#include "rtc_base/system/arch.h"
#include "rtc_base/thread_annotations.h"
#include "system_wrappers/include/clock.h"
#include "system_wrappers/include/cpu_features_wrapper.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"
#include "test/testsupport/file_utils.h"
#include "test/testsupport/rtc_expect_death.h"
using ::testing::_;
using ::testing::AtLeast;
using ::testing::Invoke;
namespace webrtc {
namespace {
const int kSampleRateHz = 16000;
const int kNumSamples10ms = kSampleRateHz / 100;
const int kFrameSizeMs = 10; // Multiple of 10.
const int kFrameSizeSamples = kFrameSizeMs / 10 * kNumSamples10ms;
const int kPayloadSizeBytes = kFrameSizeSamples * sizeof(int16_t);
const uint8_t kPayloadType = 111;
} // namespace
class RtpData {
public:
RtpData(int samples_per_packet, uint8_t payload_type)
: samples_per_packet_(samples_per_packet), payload_type_(payload_type) {}
virtual ~RtpData() {}
void Populate(RTPHeader* rtp_header) {
rtp_header->sequenceNumber = 0xABCD;
rtp_header->timestamp = 0xABCDEF01;
rtp_header->payloadType = payload_type_;
rtp_header->markerBit = false;
rtp_header->ssrc = 0x1234;
rtp_header->numCSRCs = 0;
rtp_header->payload_type_frequency = kSampleRateHz;
}
void Forward(RTPHeader* rtp_header) {
++rtp_header->sequenceNumber;
rtp_header->timestamp += samples_per_packet_;
}
private:
int samples_per_packet_;
uint8_t payload_type_;
};
class PacketizationCallbackStubOldApi : public AudioPacketizationCallback {
public:
PacketizationCallbackStubOldApi()
: num_calls_(0),
last_frame_type_(AudioFrameType::kEmptyFrame),
last_payload_type_(-1),
last_timestamp_(0) {}
int32_t SendData(AudioFrameType frame_type,
uint8_t payload_type,
uint32_t timestamp,
const uint8_t* payload_data,
size_t payload_len_bytes,
int64_t absolute_capture_timestamp_ms) override {
MutexLock lock(&mutex_);
++num_calls_;
last_frame_type_ = frame_type;
last_payload_type_ = payload_type;
last_timestamp_ = timestamp;
last_payload_vec_.assign(payload_data, payload_data + payload_len_bytes);
return 0;
}
int num_calls() const {
MutexLock lock(&mutex_);
return num_calls_;
}
int last_payload_len_bytes() const {
MutexLock lock(&mutex_);
return rtc::checked_cast<int>(last_payload_vec_.size());
}
AudioFrameType last_frame_type() const {
MutexLock lock(&mutex_);
return last_frame_type_;
}
int last_payload_type() const {
MutexLock lock(&mutex_);
return last_payload_type_;
}
uint32_t last_timestamp() const {
MutexLock lock(&mutex_);
return last_timestamp_;
}
void SwapBuffers(std::vector<uint8_t>* payload) {
MutexLock lock(&mutex_);
last_payload_vec_.swap(*payload);
}
private:
int num_calls_ RTC_GUARDED_BY(mutex_);
AudioFrameType last_frame_type_ RTC_GUARDED_BY(mutex_);
int last_payload_type_ RTC_GUARDED_BY(mutex_);
uint32_t last_timestamp_ RTC_GUARDED_BY(mutex_);
std::vector<uint8_t> last_payload_vec_ RTC_GUARDED_BY(mutex_);
mutable Mutex mutex_;
};
class AudioCodingModuleTestOldApi : public ::testing::Test {
protected:
AudioCodingModuleTestOldApi()
: rtp_utility_(new RtpData(kFrameSizeSamples, kPayloadType)),
clock_(Clock::GetRealTimeClock()) {}
~AudioCodingModuleTestOldApi() {}
void TearDown() {}
void SetUp() {
acm_.reset(AudioCodingModule::Create([this] {
AudioCodingModule::Config config;
config.clock = clock_;
config.decoder_factory = CreateBuiltinAudioDecoderFactory();
return config;
}()));
rtp_utility_->Populate(&rtp_header_);
input_frame_.sample_rate_hz_ = kSampleRateHz;
input_frame_.num_channels_ = 1;
input_frame_.samples_per_channel_ = kSampleRateHz * 10 / 1000; // 10 ms.
static_assert(kSampleRateHz * 10 / 1000 <= AudioFrame::kMaxDataSizeSamples,
"audio frame too small");
input_frame_.Mute();
ASSERT_EQ(0, acm_->RegisterTransportCallback(&packet_cb_));
SetUpL16Codec();
}
// Set up L16 codec.
virtual void SetUpL16Codec() {
audio_format_ = SdpAudioFormat("L16", kSampleRateHz, 1);
pac_size_ = 160;
}
virtual void RegisterCodec() {
acm_->SetReceiveCodecs({{kPayloadType, *audio_format_}});
acm_->SetEncoder(CreateBuiltinAudioEncoderFactory()->MakeAudioEncoder(
kPayloadType, *audio_format_, absl::nullopt));
}
virtual void InsertPacketAndPullAudio() {
InsertPacket();
PullAudio();
}
virtual void InsertPacket() {
const uint8_t kPayload[kPayloadSizeBytes] = {0};
ASSERT_EQ(0,
acm_->IncomingPacket(kPayload, kPayloadSizeBytes, rtp_header_));
rtp_utility_->Forward(&rtp_header_);
}
virtual void PullAudio() {
AudioFrame audio_frame;
bool muted;
ASSERT_EQ(0, acm_->PlayoutData10Ms(-1, &audio_frame, &muted));
ASSERT_FALSE(muted);
}
virtual void InsertAudio() {
ASSERT_GE(acm_->Add10MsData(input_frame_), 0);
input_frame_.timestamp_ += kNumSamples10ms;
}
virtual void VerifyEncoding() {
int last_length = packet_cb_.last_payload_len_bytes();
EXPECT_TRUE(last_length == 2 * pac_size_ || last_length == 0)
<< "Last encoded packet was " << last_length << " bytes.";
}
virtual void InsertAudioAndVerifyEncoding() {
InsertAudio();
VerifyEncoding();
}
std::unique_ptr<RtpData> rtp_utility_;
std::unique_ptr<AudioCodingModule> acm_;
PacketizationCallbackStubOldApi packet_cb_;
RTPHeader rtp_header_;
AudioFrame input_frame_;
absl::optional<SdpAudioFormat> audio_format_;
int pac_size_ = -1;
Clock* clock_;
};
class AudioCodingModuleTestOldApiDeathTest
: public AudioCodingModuleTestOldApi {};
TEST_F(AudioCodingModuleTestOldApi, VerifyOutputFrame) {
AudioFrame audio_frame;
const int kSampleRateHz = 32000;
bool muted;
EXPECT_EQ(0, acm_->PlayoutData10Ms(kSampleRateHz, &audio_frame, &muted));
ASSERT_FALSE(muted);
EXPECT_EQ(0u, audio_frame.timestamp_);
EXPECT_GT(audio_frame.num_channels_, 0u);
EXPECT_EQ(static_cast<size_t>(kSampleRateHz / 100),
audio_frame.samples_per_channel_);
EXPECT_EQ(kSampleRateHz, audio_frame.sample_rate_hz_);
}
// The below test is temporarily disabled on Windows due to problems
// with clang debug builds.
// TODO(tommi): Re-enable when we've figured out what the problem is.
// http://crbug.com/615050
#if !defined(WEBRTC_WIN) && defined(__clang__) && RTC_DCHECK_IS_ON && \
GTEST_HAS_DEATH_TEST && !defined(WEBRTC_ANDROID)
TEST_F(AudioCodingModuleTestOldApiDeathTest, FailOnZeroDesiredFrequency) {
AudioFrame audio_frame;
bool muted;
RTC_EXPECT_DEATH(acm_->PlayoutData10Ms(0, &audio_frame, &muted),
"dst_sample_rate_hz");
}
#endif
// Checks that the transport callback is invoked once for each speech packet.
// Also checks that the frame type is kAudioFrameSpeech.
TEST_F(AudioCodingModuleTestOldApi, TransportCallbackIsInvokedForEachPacket) {
const int k10MsBlocksPerPacket = 3;
pac_size_ = k10MsBlocksPerPacket * kSampleRateHz / 100;
audio_format_->parameters["ptime"] = "30";
RegisterCodec();
const int kLoops = 10;
for (int i = 0; i < kLoops; ++i) {
EXPECT_EQ(i / k10MsBlocksPerPacket, packet_cb_.num_calls());
if (packet_cb_.num_calls() > 0)
EXPECT_EQ(AudioFrameType::kAudioFrameSpeech,
packet_cb_.last_frame_type());
InsertAudioAndVerifyEncoding();
}
EXPECT_EQ(kLoops / k10MsBlocksPerPacket, packet_cb_.num_calls());
EXPECT_EQ(AudioFrameType::kAudioFrameSpeech, packet_cb_.last_frame_type());
}
#if defined(WEBRTC_CODEC_ISAC) || defined(WEBRTC_CODEC_ISACFX)
// Verifies that the RTP timestamp series is not reset when the codec is
// changed.
TEST_F(AudioCodingModuleTestOldApi, TimestampSeriesContinuesWhenCodecChanges) {
RegisterCodec(); // This registers the default codec.
uint32_t expected_ts = input_frame_.timestamp_;
int blocks_per_packet = pac_size_ / (kSampleRateHz / 100);
// Encode 5 packets of the first codec type.
const int kNumPackets1 = 5;
for (int j = 0; j < kNumPackets1; ++j) {
for (int i = 0; i < blocks_per_packet; ++i) {
EXPECT_EQ(j, packet_cb_.num_calls());
InsertAudio();
}
EXPECT_EQ(j + 1, packet_cb_.num_calls());
EXPECT_EQ(expected_ts, packet_cb_.last_timestamp());
expected_ts += pac_size_;
}
// Change codec.
audio_format_ = SdpAudioFormat("ISAC", kSampleRateHz, 1);
pac_size_ = 480;
RegisterCodec();
blocks_per_packet = pac_size_ / (kSampleRateHz / 100);
// Encode another 5 packets.
const int kNumPackets2 = 5;
for (int j = 0; j < kNumPackets2; ++j) {
for (int i = 0; i < blocks_per_packet; ++i) {
EXPECT_EQ(kNumPackets1 + j, packet_cb_.num_calls());
InsertAudio();
}
EXPECT_EQ(kNumPackets1 + j + 1, packet_cb_.num_calls());
EXPECT_EQ(expected_ts, packet_cb_.last_timestamp());
expected_ts += pac_size_;
}
}
#endif
// Introduce this class to set different expectations on the number of encoded
// bytes. This class expects all encoded packets to be 9 bytes (matching one
// CNG SID frame) or 0 bytes. This test depends on `input_frame_` containing
// (near-)zero values. It also introduces a way to register comfort noise with
// a custom payload type.
class AudioCodingModuleTestWithComfortNoiseOldApi
: public AudioCodingModuleTestOldApi {
protected:
void RegisterCngCodec(int rtp_payload_type) {
acm_->SetReceiveCodecs({{kPayloadType, *audio_format_},
{rtp_payload_type, {"cn", kSampleRateHz, 1}}});
acm_->ModifyEncoder([&](std::unique_ptr<AudioEncoder>* enc) {
AudioEncoderCngConfig config;
config.speech_encoder = std::move(*enc);
config.num_channels = 1;
config.payload_type = rtp_payload_type;
config.vad_mode = Vad::kVadNormal;
*enc = CreateComfortNoiseEncoder(std::move(config));
});
}
void VerifyEncoding() override {
int last_length = packet_cb_.last_payload_len_bytes();
EXPECT_TRUE(last_length == 9 || last_length == 0)
<< "Last encoded packet was " << last_length << " bytes.";
}
void DoTest(int blocks_per_packet, int cng_pt) {
const int kLoops = 40;
// This array defines the expected frame types, and when they should arrive.
// We expect a frame to arrive each time the speech encoder would have
// produced a packet, and once every 100 ms the frame should be non-empty,
// that is contain comfort noise.
const struct {
int ix;
AudioFrameType type;
} expectation[] = {{2, AudioFrameType::kAudioFrameCN},
{5, AudioFrameType::kEmptyFrame},
{8, AudioFrameType::kEmptyFrame},
{11, AudioFrameType::kAudioFrameCN},
{14, AudioFrameType::kEmptyFrame},
{17, AudioFrameType::kEmptyFrame},
{20, AudioFrameType::kAudioFrameCN},
{23, AudioFrameType::kEmptyFrame},
{26, AudioFrameType::kEmptyFrame},
{29, AudioFrameType::kEmptyFrame},
{32, AudioFrameType::kAudioFrameCN},
{35, AudioFrameType::kEmptyFrame},
{38, AudioFrameType::kEmptyFrame}};
for (int i = 0; i < kLoops; ++i) {
int num_calls_before = packet_cb_.num_calls();
EXPECT_EQ(i / blocks_per_packet, num_calls_before);
InsertAudioAndVerifyEncoding();
int num_calls = packet_cb_.num_calls();
if (num_calls == num_calls_before + 1) {
EXPECT_EQ(expectation[num_calls - 1].ix, i);
EXPECT_EQ(expectation[num_calls - 1].type, packet_cb_.last_frame_type())
<< "Wrong frame type for lap " << i;
EXPECT_EQ(cng_pt, packet_cb_.last_payload_type());
} else {
EXPECT_EQ(num_calls, num_calls_before);
}
}
}
};
// Checks that the transport callback is invoked once per frame period of the
// underlying speech encoder, even when comfort noise is produced.
// Also checks that the frame type is kAudioFrameCN or kEmptyFrame.
TEST_F(AudioCodingModuleTestWithComfortNoiseOldApi,
TransportCallbackTestForComfortNoiseRegisterCngLast) {
const int k10MsBlocksPerPacket = 3;
pac_size_ = k10MsBlocksPerPacket * kSampleRateHz / 100;
audio_format_->parameters["ptime"] = "30";
RegisterCodec();
const int kCngPayloadType = 105;
RegisterCngCodec(kCngPayloadType);
DoTest(k10MsBlocksPerPacket, kCngPayloadType);
}
// A multi-threaded test for ACM. This base class is using the PCM16b 16 kHz
// codec, while the derive class AcmIsacMtTest is using iSAC.
class AudioCodingModuleMtTestOldApi : public AudioCodingModuleTestOldApi {
protected:
static const int kNumPackets = 500;
static const int kNumPullCalls = 500;
AudioCodingModuleMtTestOldApi()
: AudioCodingModuleTestOldApi(),
send_count_(0),
insert_packet_count_(0),
pull_audio_count_(0),
next_insert_packet_time_ms_(0),
fake_clock_(new SimulatedClock(0)) {
clock_ = fake_clock_.get();
}
void SetUp() {
AudioCodingModuleTestOldApi::SetUp();
RegisterCodec(); // Must be called before the threads start below.
StartThreads();
}
void StartThreads() {
quit_.store(false);
const auto attributes =
rtc::ThreadAttributes().SetPriority(rtc::ThreadPriority::kRealtime);
send_thread_ = rtc::PlatformThread::SpawnJoinable(
[this] {
while (!quit_.load()) {
CbSendImpl();
}
},
"send", attributes);
insert_packet_thread_ = rtc::PlatformThread::SpawnJoinable(
[this] {
while (!quit_.load()) {
CbInsertPacketImpl();
}
},
"insert_packet", attributes);
pull_audio_thread_ = rtc::PlatformThread::SpawnJoinable(
[this] {
while (!quit_.load()) {
CbPullAudioImpl();
}
},
"pull_audio", attributes);
}
void TearDown() {
AudioCodingModuleTestOldApi::TearDown();
quit_.store(true);
pull_audio_thread_.Finalize();
send_thread_.Finalize();
insert_packet_thread_.Finalize();
}
bool RunTest() {
return test_complete_.Wait(10 * 60 * 1000); // 10 minutes' timeout.
}
virtual bool TestDone() {
if (packet_cb_.num_calls() > kNumPackets) {
MutexLock lock(&mutex_);
if (pull_audio_count_ > kNumPullCalls) {
// Both conditions for completion are met. End the test.
return true;
}
}
return false;
}
// The send thread doesn't have to care about the current simulated time,
// since only the AcmReceiver is using the clock.
void CbSendImpl() {
SleepMs(1);
if (HasFatalFailure()) {
// End the test early if a fatal failure (ASSERT_*) has occurred.
test_complete_.Set();
}
++send_count_;
InsertAudioAndVerifyEncoding();
if (TestDone()) {
test_complete_.Set();
}
}
void CbInsertPacketImpl() {
SleepMs(1);
{
MutexLock lock(&mutex_);
if (clock_->TimeInMilliseconds() < next_insert_packet_time_ms_) {
return;
}
next_insert_packet_time_ms_ += 10;
}
// Now we're not holding the crit sect when calling ACM.
++insert_packet_count_;
InsertPacket();
}
void CbPullAudioImpl() {
SleepMs(1);
{
MutexLock lock(&mutex_);
// Don't let the insert thread fall behind.
if (next_insert_packet_time_ms_ < clock_->TimeInMilliseconds()) {
return;
}
++pull_audio_count_;
}
// Now we're not holding the crit sect when calling ACM.
PullAudio();
fake_clock_->AdvanceTimeMilliseconds(10);
}
rtc::PlatformThread send_thread_;
rtc::PlatformThread insert_packet_thread_;
rtc::PlatformThread pull_audio_thread_;
// Used to force worker threads to stop looping.
std::atomic<bool> quit_;
rtc::Event test_complete_;
int send_count_;
int insert_packet_count_;
int pull_audio_count_ RTC_GUARDED_BY(mutex_);
Mutex mutex_;
int64_t next_insert_packet_time_ms_ RTC_GUARDED_BY(mutex_);
std::unique_ptr<SimulatedClock> fake_clock_;
};
#if defined(WEBRTC_IOS)
#define MAYBE_DoTest DISABLED_DoTest
#else
#define MAYBE_DoTest DoTest
#endif
TEST_F(AudioCodingModuleMtTestOldApi, MAYBE_DoTest) {
EXPECT_TRUE(RunTest());
}
// This is a multi-threaded ACM test using iSAC. The test encodes audio
// from a PCM file. The most recent encoded frame is used as input to the
// receiving part. Depending on timing, it may happen that the same RTP packet
// is inserted into the receiver multiple times, but this is a valid use-case,
// and simplifies the test code a lot.
class AcmIsacMtTestOldApi : public AudioCodingModuleMtTestOldApi {
protected:
static const int kNumPackets = 500;
static const int kNumPullCalls = 500;
AcmIsacMtTestOldApi()
: AudioCodingModuleMtTestOldApi(), last_packet_number_(0) {}
~AcmIsacMtTestOldApi() {}
void SetUp() override {
AudioCodingModuleTestOldApi::SetUp();
RegisterCodec(); // Must be called before the threads start below.
// Set up input audio source to read from specified file, loop after 5
// seconds, and deliver blocks of 10 ms.
const std::string input_file_name =
webrtc::test::ResourcePath("audio_coding/speech_mono_16kHz", "pcm");
audio_loop_.Init(input_file_name, 5 * kSampleRateHz, kNumSamples10ms);
// Generate one packet to have something to insert.
int loop_counter = 0;
while (packet_cb_.last_payload_len_bytes() == 0) {
InsertAudio();
ASSERT_LT(loop_counter++, 10);
}
// Set `last_packet_number_` to one less that `num_calls` so that the packet
// will be fetched in the next InsertPacket() call.
last_packet_number_ = packet_cb_.num_calls() - 1;
StartThreads();
}
void RegisterCodec() override {
static_assert(kSampleRateHz == 16000, "test designed for iSAC 16 kHz");
audio_format_ = SdpAudioFormat("isac", kSampleRateHz, 1);
pac_size_ = 480;
// Register iSAC codec in ACM, effectively unregistering the PCM16B codec
// registered in AudioCodingModuleTestOldApi::SetUp();
acm_->SetReceiveCodecs({{kPayloadType, *audio_format_}});
acm_->SetEncoder(CreateBuiltinAudioEncoderFactory()->MakeAudioEncoder(
kPayloadType, *audio_format_, absl::nullopt));
}
void InsertPacket() override {
int num_calls = packet_cb_.num_calls(); // Store locally for thread safety.
if (num_calls > last_packet_number_) {
// Get the new payload out from the callback handler.
// Note that since we swap buffers here instead of directly inserting
// a pointer to the data in `packet_cb_`, we avoid locking the callback
// for the duration of the IncomingPacket() call.
packet_cb_.SwapBuffers(&last_payload_vec_);
ASSERT_GT(last_payload_vec_.size(), 0u);
rtp_utility_->Forward(&rtp_header_);
last_packet_number_ = num_calls;
}
ASSERT_GT(last_payload_vec_.size(), 0u);
ASSERT_EQ(0, acm_->IncomingPacket(&last_payload_vec_[0],
last_payload_vec_.size(), rtp_header_));
}
void InsertAudio() override {
// TODO(kwiberg): Use std::copy here. Might be complications because AFAICS
// this call confuses the number of samples with the number of bytes, and
// ends up copying only half of what it should.
memcpy(input_frame_.mutable_data(), audio_loop_.GetNextBlock().data(),
kNumSamples10ms);
AudioCodingModuleTestOldApi::InsertAudio();
}
// Override the verification function with no-op, since iSAC produces variable
// payload sizes.
void VerifyEncoding() override {}
// This method is the same as AudioCodingModuleMtTestOldApi::TestDone(), but
// here it is using the constants defined in this class (i.e., shorter test
// run).
bool TestDone() override {
if (packet_cb_.num_calls() > kNumPackets) {
MutexLock lock(&mutex_);
if (pull_audio_count_ > kNumPullCalls) {
// Both conditions for completion are met. End the test.
return true;
}
}
return false;
}
int last_packet_number_;
std::vector<uint8_t> last_payload_vec_;
test::AudioLoop audio_loop_;
};
#if defined(WEBRTC_IOS)
#define MAYBE_DoTest DISABLED_DoTest
#else
#define MAYBE_DoTest DoTest
#endif
#if defined(WEBRTC_CODEC_ISAC) || defined(WEBRTC_CODEC_ISACFX)
TEST_F(AcmIsacMtTestOldApi, MAYBE_DoTest) {
EXPECT_TRUE(RunTest());
}
#endif
class AcmReRegisterIsacMtTestOldApi : public AudioCodingModuleTestOldApi {
protected:
static const int kRegisterAfterNumPackets = 5;
static const int kNumPackets = 10;
static const int kPacketSizeMs = 30;
static const int kPacketSizeSamples = kPacketSizeMs * 16;
AcmReRegisterIsacMtTestOldApi()
: AudioCodingModuleTestOldApi(),
codec_registered_(false),
receive_packet_count_(0),
next_insert_packet_time_ms_(0),
fake_clock_(new SimulatedClock(0)) {
AudioEncoderIsacFloatImpl::Config config;
config.payload_type = kPayloadType;
isac_encoder_.reset(new AudioEncoderIsacFloatImpl(config));
clock_ = fake_clock_.get();
}
void SetUp() override {
AudioCodingModuleTestOldApi::SetUp();
// Set up input audio source to read from specified file, loop after 5
// seconds, and deliver blocks of 10 ms.
const std::string input_file_name =
webrtc::test::ResourcePath("audio_coding/speech_mono_16kHz", "pcm");
audio_loop_.Init(input_file_name, 5 * kSampleRateHz, kNumSamples10ms);
RegisterCodec(); // Must be called before the threads start below.
StartThreads();
}
void RegisterCodec() override {
// Register iSAC codec in ACM, effectively unregistering the PCM16B codec
// registered in AudioCodingModuleTestOldApi::SetUp();
// Only register the decoder for now. The encoder is registered later.
static_assert(kSampleRateHz == 16000, "test designed for iSAC 16 kHz");
acm_->SetReceiveCodecs({{kPayloadType, {"ISAC", kSampleRateHz, 1}}});
}
void StartThreads() {
quit_.store(false);
const auto attributes =
rtc::ThreadAttributes().SetPriority(rtc::ThreadPriority::kRealtime);
receive_thread_ = rtc::PlatformThread::SpawnJoinable(
[this] {
while (!quit_.load() && CbReceiveImpl()) {
}
},
"receive", attributes);
codec_registration_thread_ = rtc::PlatformThread::SpawnJoinable(
[this] {
while (!quit_.load()) {
CbCodecRegistrationImpl();
}
},
"codec_registration", attributes);
}
void TearDown() override {
AudioCodingModuleTestOldApi::TearDown();
quit_.store(true);
receive_thread_.Finalize();
codec_registration_thread_.Finalize();
}
bool RunTest() {
return test_complete_.Wait(10 * 60 * 1000); // 10 minutes' timeout.
}
bool CbReceiveImpl() {
SleepMs(1);
rtc::Buffer encoded;
AudioEncoder::EncodedInfo info;
{
MutexLock lock(&mutex_);
if (clock_->TimeInMilliseconds() < next_insert_packet_time_ms_) {
return true;
}
next_insert_packet_time_ms_ += kPacketSizeMs;
++receive_packet_count_;
// Encode new frame.
uint32_t input_timestamp = rtp_header_.timestamp;
while (info.encoded_bytes == 0) {
info = isac_encoder_->Encode(input_timestamp,
audio_loop_.GetNextBlock(), &encoded);
input_timestamp += 160; // 10 ms at 16 kHz.
}
EXPECT_EQ(rtp_header_.timestamp + kPacketSizeSamples, input_timestamp);
EXPECT_EQ(rtp_header_.timestamp, info.encoded_timestamp);
EXPECT_EQ(rtp_header_.payloadType, info.payload_type);
}
// Now we're not holding the crit sect when calling ACM.
// Insert into ACM.
EXPECT_EQ(0, acm_->IncomingPacket(encoded.data(), info.encoded_bytes,
rtp_header_));
// Pull audio.
for (int i = 0; i < rtc::CheckedDivExact(kPacketSizeMs, 10); ++i) {
AudioFrame audio_frame;
bool muted;
EXPECT_EQ(0, acm_->PlayoutData10Ms(-1 /* default output frequency */,
&audio_frame, &muted));
if (muted) {
ADD_FAILURE();
return false;
}
fake_clock_->AdvanceTimeMilliseconds(10);
}
rtp_utility_->Forward(&rtp_header_);
return true;
}
void CbCodecRegistrationImpl() {
SleepMs(1);
if (HasFatalFailure()) {
// End the test early if a fatal failure (ASSERT_*) has occurred.
test_complete_.Set();
}
MutexLock lock(&mutex_);
if (!codec_registered_ &&
receive_packet_count_ > kRegisterAfterNumPackets) {
// Register the iSAC encoder.
acm_->SetEncoder(CreateBuiltinAudioEncoderFactory()->MakeAudioEncoder(
kPayloadType, *audio_format_, absl::nullopt));
codec_registered_ = true;
}
if (codec_registered_ && receive_packet_count_ > kNumPackets) {
test_complete_.Set();
}
}
rtc::PlatformThread receive_thread_;
rtc::PlatformThread codec_registration_thread_;
// Used to force worker threads to stop looping.
std::atomic<bool> quit_;
rtc::Event test_complete_;
Mutex mutex_;
bool codec_registered_ RTC_GUARDED_BY(mutex_);
int receive_packet_count_ RTC_GUARDED_BY(mutex_);
int64_t next_insert_packet_time_ms_ RTC_GUARDED_BY(mutex_);
std::unique_ptr<AudioEncoderIsacFloatImpl> isac_encoder_;
std::unique_ptr<SimulatedClock> fake_clock_;
test::AudioLoop audio_loop_;
};
#if defined(WEBRTC_IOS)
#define MAYBE_DoTest DISABLED_DoTest
#else
#define MAYBE_DoTest DoTest
#endif
#if defined(WEBRTC_CODEC_ISAC) || defined(WEBRTC_CODEC_ISACFX)
TEST_F(AcmReRegisterIsacMtTestOldApi, MAYBE_DoTest) {
EXPECT_TRUE(RunTest());
}
#endif
// Disabling all of these tests on iOS until file support has been added.
// See https://code.google.com/p/webrtc/issues/detail?id=4752 for details.
#if !defined(WEBRTC_IOS)
// This test verifies bit exactness for the send-side of ACM. The test setup is
// a chain of three different test classes:
//
// test::AcmSendTest -> AcmSenderBitExactness -> test::AcmReceiveTest
//
// The receiver side is driving the test by requesting new packets from
// AcmSenderBitExactness::NextPacket(). This method, in turn, asks for the
// packet from test::AcmSendTest::NextPacket, which inserts audio from the
// input file until one packet is produced. (The input file loops indefinitely.)
// Before passing the packet to the receiver, this test class verifies the
// packet header and updates a payload checksum with the new payload. The
// decoded output from the receiver is also verified with a (separate) checksum.
class AcmSenderBitExactnessOldApi : public ::testing::Test,
public test::PacketSource {
protected:
static const int kTestDurationMs = 1000;
AcmSenderBitExactnessOldApi()
: frame_size_rtp_timestamps_(0),
packet_count_(0),
payload_type_(0),
last_sequence_number_(0),
last_timestamp_(0),
payload_checksum_(rtc::MessageDigestFactory::Create(rtc::DIGEST_MD5)) {}
// Sets up the test::AcmSendTest object. Returns true on success, otherwise
// false.
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));
send_test_.reset(new test::AcmSendTestOldApi(audio_source_.get(),
source_rate, kTestDurationMs));
return send_test_.get() != NULL;
}
// Registers a send codec in the test::AcmSendTest object. Returns true on
// success, false on failure.
bool RegisterSendCodec(const char* payload_name,
int sampling_freq_hz,
int channels,
int payload_type,
int frame_size_samples,
int frame_size_rtp_timestamps) {
payload_type_ = payload_type;
frame_size_rtp_timestamps_ = frame_size_rtp_timestamps;
return send_test_->RegisterCodec(payload_name, sampling_freq_hz, channels,
payload_type, frame_size_samples);
}
void RegisterExternalSendCodec(
std::unique_ptr<AudioEncoder> external_speech_encoder,
int payload_type) {
payload_type_ = payload_type;
frame_size_rtp_timestamps_ = rtc::checked_cast<uint32_t>(
external_speech_encoder->Num10MsFramesInNextPacket() *
external_speech_encoder->RtpTimestampRateHz() / 100);
send_test_->RegisterExternalCodec(std::move(external_speech_encoder));
}
// Runs the test. SetUpSender() and RegisterSendCodec() must have been called
// before calling this method.
void Run(const std::string& audio_checksum_ref,
const std::string& payload_checksum_ref,
int expected_packets,
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;
const std::string output_file_name =
webrtc::test::OutputPath() +
::testing::UnitTest::GetInstance()
->current_test_info()
->test_case_name() +
"_" + ::testing::UnitTest::GetInstance()->current_test_info()->name() +
"_output.wav";
const int kOutputFreqHz = 8000;
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, decoder_factory);
ASSERT_NO_FATAL_FAILURE(receive_test.RegisterDefaultCodecs());
// This is where the actual test is executed.
receive_test.Run();
// Extract and verify the audio checksum.
std::string checksum_string = audio_checksum.Finish();
ExpectChecksumEq(audio_checksum_ref, checksum_string);
// Extract and verify the payload checksum.
rtc::Buffer checksum_result(payload_checksum_->Size());
payload_checksum_->Finish(checksum_result.data(), checksum_result.size());
checksum_string = rtc::hex_encode(checksum_result);
ExpectChecksumEq(payload_checksum_ref, checksum_string);
// Verify number of packets produced.
EXPECT_EQ(expected_packets, packet_count_);
// Delete the output file.
remove(output_file_name.c_str());
}
// Helper: result must be one the "|"-separated checksums.
void ExpectChecksumEq(std::string ref, std::string result) {
if (ref.size() == result.size()) {
// Only one checksum: clearer message.
EXPECT_EQ(ref, result);
} else {
EXPECT_NE(ref.find(result), std::string::npos)
<< result << " must be one of these:\n"
<< ref;
}
}
// Inherited from test::PacketSource.
std::unique_ptr<test::Packet> NextPacket() override {
auto packet = send_test_->NextPacket();
if (!packet)
return NULL;
VerifyPacket(packet.get());
// TODO(henrik.lundin) Save the packet to file as well.
// Pass it on to the caller. The caller becomes the owner of `packet`.
return packet;
}
// Verifies the packet.
void VerifyPacket(const test::Packet* packet) {
EXPECT_TRUE(packet->valid_header());
// (We can check the header fields even if valid_header() is false.)
EXPECT_EQ(payload_type_, packet->header().payloadType);
if (packet_count_ > 0) {
// This is not the first packet.
uint16_t sequence_number_diff =
packet->header().sequenceNumber - last_sequence_number_;
EXPECT_EQ(1, sequence_number_diff);
uint32_t timestamp_diff = packet->header().timestamp - last_timestamp_;
EXPECT_EQ(frame_size_rtp_timestamps_, timestamp_diff);
}
++packet_count_;
last_sequence_number_ = packet->header().sequenceNumber;
last_timestamp_ = packet->header().timestamp;
// Update the checksum.
payload_checksum_->Update(packet->payload(),
packet->payload_length_bytes());
}
void SetUpTest(const char* codec_name,
int codec_sample_rate_hz,
int channels,
int payload_type,
int codec_frame_size_samples,
int codec_frame_size_rtp_timestamps) {
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));
}
void SetUpTestExternalEncoder(
std::unique_ptr<AudioEncoder> external_speech_encoder,
int payload_type) {
ASSERT_TRUE(send_test_);
RegisterExternalSendCodec(std::move(external_speech_encoder), payload_type);
}
std::unique_ptr<test::AcmSendTestOldApi> send_test_;
std::unique_ptr<test::InputAudioFile> audio_source_;
uint32_t frame_size_rtp_timestamps_;
int packet_count_;
uint8_t payload_type_;
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 {};
// Run bit exactness tests only for release builds.
#if (defined(WEBRTC_CODEC_ISAC) || defined(WEBRTC_CODEC_ISACFX)) && \
defined(NDEBUG) && defined(WEBRTC_LINUX) && defined(WEBRTC_ARCH_X86_64)
TEST_F(AcmSenderBitExactnessOldApi, IsacWb30ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("ISAC", 16000, 1, 103, 480, 480));
Run(/*audio_checksum_ref=*/"a3077ac01b0137e8bbc237fb1f9816a5",
/*payload_checksum_ref=*/"3c79f16f34218271f3dca4e2b1dfe1bb",
/*expected_packets=*/33,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, IsacWb60ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("ISAC", 16000, 1, 103, 960, 960));
Run(/*audio_checksum_ref=*/"76da9b7514f986fc2bb32b1c3170e8d4",
/*payload_checksum_ref=*/"9e0a0ab743ad987b55b8e14802769c56",
/*expected_packets=*/16,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
#endif
// Run bit exactness test only for release build.
#if defined(WEBRTC_CODEC_ISAC) && defined(NDEBUG) && defined(WEBRTC_LINUX) && \
defined(WEBRTC_ARCH_X86_64)
TEST_F(AcmSenderBitExactnessOldApi, IsacSwb30ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("ISAC", 32000, 1, 104, 960, 960));
Run(/*audio_checksum_ref=*/"f4cf577f28a0dcbac33358b757518e0c",
/*payload_checksum_ref=*/"ce86106a93419aefb063097108ec94ab",
/*expected_packets=*/33,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
#endif
TEST_F(AcmSenderBitExactnessOldApi, Pcm16_8000khz_10ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("L16", 8000, 1, 107, 80, 80));
Run(/*audio_checksum_ref=*/"69118ed438ac76252d023e0463819471",
/*payload_checksum_ref=*/"c1edd36339ce0326cc4550041ad719a0",
/*expected_packets=*/100,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcm16_16000khz_10ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("L16", 16000, 1, 108, 160, 160));
Run(/*audio_checksum_ref=*/"bc6ab94d12a464921763d7544fdbd07e",
/*payload_checksum_ref=*/"ad786526383178b08d80d6eee06e9bad",
/*expected_packets=*/100,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcm16_32000khz_10ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("L16", 32000, 1, 109, 320, 320));
Run(/*audio_checksum_ref=*/"c50244419c5c3a2f04cc69a022c266a2",
/*payload_checksum_ref=*/"5ef82ea885e922263606c6fdbc49f651",
/*expected_packets=*/100,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcm16_stereo_8000khz_10ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("L16", 8000, 2, 111, 80, 80));
Run(/*audio_checksum_ref=*/"4fccf4cc96f1e8e8de4b9fadf62ded9e",
/*payload_checksum_ref=*/"62ce5adb0d4965d0a52ec98ae7f98974",
/*expected_packets=*/100,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcm16_stereo_16000khz_10ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("L16", 16000, 2, 112, 160, 160));
Run(/*audio_checksum_ref=*/"e15e388d9d4af8c02a59fe1552fedee3",
/*payload_checksum_ref=*/"41ca8edac4b8c71cd54fd9f25ec14870",
/*expected_packets=*/100,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcm16_stereo_32000khz_10ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("L16", 32000, 2, 113, 320, 320));
Run(/*audio_checksum_ref=*/"b240520c0d05003fde7a174ae5957286",
/*payload_checksum_ref=*/"50e58502fb04421bf5b857dda4c96879",
/*expected_packets=*/100,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcmu_20ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("PCMU", 8000, 1, 0, 160, 160));
Run(/*audio_checksum_ref=*/"c8d1fc677f33c2022ec5f83c7f302280",
/*payload_checksum_ref=*/"8f9b8750bd80fe26b6cbf6659b89f0f9",
/*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcma_20ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("PCMA", 8000, 1, 8, 160, 160));
Run(/*audio_checksum_ref=*/"47eb60e855eb12d1b0e6da9c975754a4",
/*payload_checksum_ref=*/"6ad745e55aa48981bfc790d0eeef2dd1",
/*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcmu_stereo_20ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("PCMU", 8000, 2, 110, 160, 160));
Run(/*audio_checksum_ref=*/"6ef2f57d4934714787fd0a834e3ea18e",
/*payload_checksum_ref=*/"60b6f25e8d1e74cb679cfe756dd9bca5",
/*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
TEST_F(AcmSenderBitExactnessOldApi, Pcma_stereo_20ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("PCMA", 8000, 2, 118, 160, 160));
Run(/*audio_checksum_ref=*/"a84d75e098d87ab6b260687eb4b612a2",
/*payload_checksum_ref=*/"92b282c83efd20e7eeef52ba40842cf7",
/*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
#if defined(WEBRTC_CODEC_ILBC) && defined(WEBRTC_LINUX) && \
defined(WEBRTC_ARCH_X86_64)
TEST_F(AcmSenderBitExactnessOldApi, Ilbc_30ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("ILBC", 8000, 1, 102, 240, 240));
Run(/*audio_checksum_ref=*/"b14dba0de36efa5ec88a32c0b320b70f",
/*payload_checksum_ref=*/"cfae2e9f6aba96e145f2bcdd5050ce78",
/*expected_packets=*/33,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
#endif
#if defined(WEBRTC_LINUX) && defined(WEBRTC_ARCH_X86_64)
TEST_F(AcmSenderBitExactnessOldApi, G722_20ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("G722", 16000, 1, 9, 320, 160));
Run(/*audio_checksum_ref=*/"a87a91ec0124510a64967f5d768554ff",
/*payload_checksum_ref=*/"fc68a87e1380614e658087cb35d5ca10",
/*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kMonoOutput);
}
#endif
#if defined(WEBRTC_LINUX) && defined(WEBRTC_ARCH_X86_64)
TEST_F(AcmSenderBitExactnessOldApi, G722_stereo_20ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("G722", 16000, 2, 119, 320, 160));
Run(/*audio_checksum_ref=*/"be0b8528ff9db3a2219f55ddd36faf7f",
/*payload_checksum_ref=*/"66516152eeaa1e650ad94ff85f668dac",
/*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
#endif
namespace {
// Checksum depends on libopus being compiled with or without SSE.
const std::string audio_checksum =
"6a76fe2ffba057c06eb63239b3c47abe"
"|0c4f9d33b4a7379a34ee0c0d5718afe6";
const std::string payload_checksum =
"b43bdf7638b2bc2a5a6f30bdc640b9ed"
"|c30d463e7ed10bdd1da9045f80561f27";
} // namespace
#if defined(WEBRTC_LINUX) && defined(WEBRTC_ARCH_X86_64)
TEST_F(AcmSenderBitExactnessOldApi, Opus_stereo_20ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("opus", 48000, 2, 120, 960, 960));
Run(audio_checksum, payload_checksum, /*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
#endif
#if defined(WEBRTC_LINUX) && defined(WEBRTC_ARCH_X86_64)
TEST_F(AcmSenderBitExactnessNewApi, 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, /*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
#endif
// TODO(webrtc:8649): Disabled until the Encoder counterpart of
// https://webrtc-review.googlesource.com/c/src/+/129768 lands.
#if defined(WEBRTC_LINUX) && defined(WEBRTC_ARCH_X86_64)
TEST_F(AcmSenderBitExactnessNewApi, DISABLED_OpusManyChannels) {
constexpr int kNumChannels = 4;
constexpr int kOpusPayloadType = 120;
// Read a 4 channel file at 48kHz.
ASSERT_TRUE(SetUpSender(kTestFileQuad48kHz, 48000));
const auto sdp_format = SdpAudioFormat("multiopus", 48000, kNumChannels,
{{"channel_mapping", "0,1,2,3"},
{"coupled_streams", "2"},
{"num_streams", "2"}});
const auto encoder_config =
AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format);
ASSERT_TRUE(encoder_config.has_value());
ASSERT_NO_FATAL_FAILURE(
SetUpTestExternalEncoder(AudioEncoderMultiChannelOpus::MakeAudioEncoder(
*encoder_config, kOpusPayloadType),
kOpusPayloadType));
const auto decoder_config =
AudioDecoderMultiChannelOpus::SdpToConfig(sdp_format);
const auto opus_decoder =
AudioDecoderMultiChannelOpus::MakeAudioDecoder(*decoder_config);
rtc::scoped_refptr<AudioDecoderFactory> decoder_factory =
rtc::make_ref_counted<test::AudioDecoderProxyFactory>(opus_decoder.get());
// Set up an EXTERNAL DECODER to parse 4 channels.
Run("audio checksum check downstream|8051617907766bec5f4e4a4f7c6d5291",
"payload checksum check downstream|b09c52e44b2bdd9a0809e3a5b1623a76",
/*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kQuadOutput,
decoder_factory);
}
#endif
#if defined(WEBRTC_LINUX) && defined(WEBRTC_ARCH_X86_64)
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));
const std::string audio_maybe_sse =
"058c03ca2c9bb5c0066d4c15ce50d772"
"|ca54661b220cc35239c6864ab858d29a";
const std::string payload_maybe_sse =
"f270ec7be7a5ed60c203c2317c4e1011"
"|eb0752ce1b6f2436fefc2e19bd084fb5";
Run(audio_maybe_sse, payload_maybe_sse, /*expected_packets=*/50,
/*expected_channels=*/test::AcmReceiveTestOldApi::kStereoOutput);
}
#endif
// This test is for verifying the SetBitRate function. The bitrate is changed at
// the beginning, and the number of generated bytes are checked.
class AcmSetBitRateTest : public ::testing::Test {
protected:
static const int kTestDurationMs = 1000;
// 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");
// 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));
return send_test_.get();
}
// Registers a send codec in the test::AcmSendTest object. Returns true on
// success, false on failure.
virtual bool RegisterSendCodec(const char* payload_name,
int sampling_freq_hz,
int channels,
int payload_type,
int frame_size_samples,
int frame_size_rtp_timestamps) {
return send_test_->RegisterCodec(payload_name, sampling_freq_hz, channels,
payload_type, frame_size_samples);
}
void RegisterExternalSendCodec(
std::unique_ptr<AudioEncoder> external_speech_encoder,
int payload_type) {
send_test_->RegisterExternalCodec(std::move(external_speech_encoder));
}
void RunInner(int min_expected_total_bits, int max_expected_total_bits) {
int nr_bytes = 0;
while (std::unique_ptr<test::Packet> next_packet =
send_test_->NextPacket()) {
nr_bytes += rtc::checked_cast<int>(next_packet->payload_length_bytes());
}
EXPECT_LE(min_expected_total_bits, nr_bytes * 8);
EXPECT_GE(max_expected_total_bits, nr_bytes * 8);
}
void SetUpTest(const char* codec_name,
int codec_sample_rate_hz,
int channels,
int payload_type,
int codec_frame_size_samples,
int codec_frame_size_rtp_timestamps) {
ASSERT_TRUE(SetUpSender());
ASSERT_TRUE(RegisterSendCodec(codec_name, codec_sample_rate_hz, channels,
payload_type, codec_frame_size_samples,
codec_frame_size_rtp_timestamps));
}
std::unique_ptr<test::AcmSendTestOldApi> send_test_;
std::unique_ptr<test::InputAudioFile> audio_source_;
};
class AcmSetBitRateNewApi : public AcmSetBitRateTest {
protected:
// Runs the test. SetUpSender() must have been called and a codec must be set
// up before calling this method.
void Run(int min_expected_total_bits, int max_expected_total_bits) {
RunInner(min_expected_total_bits, max_expected_total_bits);
}
};
TEST_F(AcmSetBitRateNewApi, OpusFromFormat_48khz_20ms_10kbps) {
const auto config = AudioEncoderOpus::SdpToConfig(
SdpAudioFormat("opus", 48000, 2, {{"maxaveragebitrate", "10000"}}));
ASSERT_TRUE(SetUpSender());
RegisterExternalSendCodec(AudioEncoderOpus::MakeAudioEncoder(*config, 107),
107);
RunInner(7000, 12000);
}
TEST_F(AcmSetBitRateNewApi, OpusFromFormat_48khz_20ms_50kbps) {
const auto config = AudioEncoderOpus::SdpToConfig(
SdpAudioFormat("opus", 48000, 2, {{"maxaveragebitrate", "50000"}}));
ASSERT_TRUE(SetUpSender());
RegisterExternalSendCodec(AudioEncoderOpus::MakeAudioEncoder(*config, 107),
107);
RunInner(40000, 60000);
}
// Verify that it works when the data to send is mono and the encoder is set to
// send surround audio.
TEST_F(AudioCodingModuleTestOldApi, SendingMultiChannelForMonoInput) {
constexpr int kSampleRateHz = 48000;
constexpr int kSamplesPerChannel = kSampleRateHz * 10 / 1000;
audio_format_ = SdpAudioFormat({"multiopus",
kSampleRateHz,
6,
{{"minptime", "10"},
{"useinbandfec", "1"},
{"channel_mapping", "0,4,1,2,3,5"},
{"num_streams", "4"},
{"coupled_streams", "2"}}});
RegisterCodec();
input_frame_.sample_rate_hz_ = kSampleRateHz;
input_frame_.num_channels_ = 1;
input_frame_.samples_per_channel_ = kSamplesPerChannel;
for (size_t k = 0; k < 10; ++k) {
ASSERT_GE(acm_->Add10MsData(input_frame_), 0);
input_frame_.timestamp_ += kSamplesPerChannel;
}
}
// Verify that it works when the data to send is stereo and the encoder is set
// to send surround audio.
TEST_F(AudioCodingModuleTestOldApi, SendingMultiChannelForStereoInput) {
constexpr int kSampleRateHz = 48000;
constexpr int kSamplesPerChannel = (kSampleRateHz * 10) / 1000;
audio_format_ = SdpAudioFormat({"multiopus",
kSampleRateHz,
6,
{{"minptime", "10"},
{"useinbandfec", "1"},
{"channel_mapping", "0,4,1,2,3,5"},
{"num_streams", "4"},
{"coupled_streams", "2"}}});
RegisterCodec();
input_frame_.sample_rate_hz_ = kSampleRateHz;
input_frame_.num_channels_ = 2;
input_frame_.samples_per_channel_ = kSamplesPerChannel;
for (size_t k = 0; k < 10; ++k) {
ASSERT_GE(acm_->Add10MsData(input_frame_), 0);
input_frame_.timestamp_ += kSamplesPerChannel;
}
}
// Verify that it works when the data to send is mono and the encoder is set to
// send stereo audio.
TEST_F(AudioCodingModuleTestOldApi, SendingStereoForMonoInput) {
constexpr int kSampleRateHz = 48000;
constexpr int kSamplesPerChannel = (kSampleRateHz * 10) / 1000;
audio_format_ = SdpAudioFormat("L16", kSampleRateHz, 2);
RegisterCodec();
input_frame_.sample_rate_hz_ = kSampleRateHz;
input_frame_.num_channels_ = 1;
input_frame_.samples_per_channel_ = kSamplesPerChannel;
for (size_t k = 0; k < 10; ++k) {
ASSERT_GE(acm_->Add10MsData(input_frame_), 0);
input_frame_.timestamp_ += kSamplesPerChannel;
}
}
// Verify that it works when the data to send is stereo and the encoder is set
// to send mono audio.
TEST_F(AudioCodingModuleTestOldApi, SendingMonoForStereoInput) {
constexpr int kSampleRateHz = 48000;
constexpr int kSamplesPerChannel = (kSampleRateHz * 10) / 1000;
audio_format_ = SdpAudioFormat("L16", kSampleRateHz, 1);
RegisterCodec();
input_frame_.sample_rate_hz_ = kSampleRateHz;
input_frame_.num_channels_ = 1;
input_frame_.samples_per_channel_ = kSamplesPerChannel;
for (size_t k = 0; k < 10; ++k) {
ASSERT_GE(acm_->Add10MsData(input_frame_), 0);
input_frame_.timestamp_ += kSamplesPerChannel;
}
}
// The result on the Android platforms is inconsistent for this test case.
// On android_rel the result is different from android and android arm64 rel.
#if defined(WEBRTC_ANDROID)
#define MAYBE_OpusFromFormat_48khz_20ms_100kbps \
DISABLED_OpusFromFormat_48khz_20ms_100kbps
#else
#define MAYBE_OpusFromFormat_48khz_20ms_100kbps \
OpusFromFormat_48khz_20ms_100kbps
#endif
TEST_F(AcmSetBitRateNewApi, MAYBE_OpusFromFormat_48khz_20ms_100kbps) {
const auto config = AudioEncoderOpus::SdpToConfig(
SdpAudioFormat("opus", 48000, 2, {{"maxaveragebitrate", "100000"}}));
ASSERT_TRUE(SetUpSender());
RegisterExternalSendCodec(AudioEncoderOpus::MakeAudioEncoder(*config, 107),
107);
RunInner(80000, 120000);
}
TEST_F(AcmSenderBitExactnessOldApi, External_Pcmu_20ms) {
AudioEncoderPcmU::Config config;
config.frame_size_ms = 20;
config.num_channels = 1;
config.payload_type = 0;
AudioEncoderPcmU encoder(config);
auto mock_encoder = std::make_unique<MockAudioEncoder>();
// Set expectations on the mock encoder and also delegate the calls to the
// real encoder.
EXPECT_CALL(*mock_encoder, SampleRateHz())
.Times(AtLeast(1))
.WillRepeatedly(Invoke(&encoder, &AudioEncoderPcmU::SampleRateHz));
EXPECT_CALL(*mock_encoder, NumChannels())
.Times(AtLeast(1))
.WillRepeatedly(Invoke(&encoder, &AudioEncoderPcmU::NumChannels));
EXPECT_CALL(*mock_encoder, RtpTimestampRateHz())
.Times(AtLeast(1))
.WillRepeatedly(Invoke(&encoder, &AudioEncoderPcmU::RtpTimestampRateHz));
EXPECT_CALL(*mock_encoder, Num10MsFramesInNextPacket())
.Times(AtLeast(1))
.WillRepeatedly(
Invoke(&encoder, &AudioEncoderPcmU::Num10MsFramesInNextPacket));
EXPECT_CALL(*mock_encoder, GetTargetBitrate())
.Times(AtLeast(1))
.WillRepeatedly(Invoke(&encoder, &AudioEncoderPcmU::GetTargetBitrate));
EXPECT_CALL(*mock_encoder, EncodeImpl(_, _, _))
.Times(AtLeast(1))
.WillRepeatedly(Invoke(
&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("c8d1fc677f33c2022ec5f83c7f302280", "8f9b8750bd80fe26b6cbf6659b89f0f9",
50, test::AcmReceiveTestOldApi::kMonoOutput);
}
// This test fixture is implemented to run ACM and change the desired output
// frequency during the call. The input packets are simply PCM16b-wb encoded
// payloads with a constant value of `kSampleValue`. The test fixture itself
// acts as PacketSource in between the receive test class and the constant-
// payload packet source class. The output is both written to file, and analyzed
// in this test fixture.
class AcmSwitchingOutputFrequencyOldApi : public ::testing::Test,
public test::PacketSource,
public test::AudioSink {
protected:
static const size_t kTestNumPackets = 50;
static const int kEncodedSampleRateHz = 16000;
static const size_t kPayloadLenSamples = 30 * kEncodedSampleRateHz / 1000;
static const int kPayloadType = 108; // Default payload type for PCM16b-wb.
AcmSwitchingOutputFrequencyOldApi()
: first_output_(true),
num_packets_(0),
packet_source_(kPayloadLenSamples,
kSampleValue,
kEncodedSampleRateHz,
kPayloadType),
output_freq_2_(0),
has_toggled_(false) {}
void Run(int output_freq_1, int output_freq_2, int toggle_period_ms) {
// Set up the receiver used to decode the packets and verify the decoded
// output.
const std::string output_file_name =
webrtc::test::OutputPath() +
::testing::UnitTest::GetInstance()
->current_test_info()
->test_case_name() +
"_" + ::testing::UnitTest::GetInstance()->current_test_info()->name() +
"_output.pcm";
test::OutputAudioFile output_file(output_file_name);
// Have the output audio sent both to file and to the WriteArray method in
// this class.
test::AudioSinkFork output(this, &output_file);
test::AcmReceiveTestToggleOutputFreqOldApi receive_test(
this, &output, output_freq_1, output_freq_2, toggle_period_ms,
test::AcmReceiveTestOldApi::kMonoOutput);
ASSERT_NO_FATAL_FAILURE(receive_test.RegisterDefaultCodecs());
output_freq_2_ = output_freq_2;
// This is where the actual test is executed.
receive_test.Run();
// Delete output file.
remove(output_file_name.c_str());
}
// Inherited from test::PacketSource.
std::unique_ptr<test::Packet> NextPacket() override {
// Check if it is time to terminate the test. The packet source is of type
// ConstantPcmPacketSource, which is infinite, so we must end the test
// "manually".
if (num_packets_++ > kTestNumPackets) {
EXPECT_TRUE(has_toggled_);
return NULL; // Test ended.
}
// Get the next packet from the source.
return packet_source_.NextPacket();
}
// Inherited from test::AudioSink.
bool WriteArray(const int16_t* audio, size_t num_samples) override {
// Skip checking the first output frame, since it has a number of zeros
// due to how NetEq is initialized.
if (first_output_) {
first_output_ = false;
return true;
}
for (size_t i = 0; i < num_samples; ++i) {
EXPECT_EQ(kSampleValue, audio[i]);
}
if (num_samples ==
static_cast<size_t>(output_freq_2_ / 100)) // Size of 10 ms frame.
has_toggled_ = true;
// The return value does not say if the values match the expectation, just
// that the method could process the samples.
return true;
}
const int16_t kSampleValue = 1000;
bool first_output_;
size_t num_packets_;
test::ConstantPcmPacketSource packet_source_;
int output_freq_2_;
bool has_toggled_;
};
TEST_F(AcmSwitchingOutputFrequencyOldApi, TestWithoutToggling) {
Run(16000, 16000, 1000);
}
TEST_F(AcmSwitchingOutputFrequencyOldApi, Toggle16KhzTo32Khz) {
Run(16000, 32000, 1000);
}
TEST_F(AcmSwitchingOutputFrequencyOldApi, Toggle32KhzTo16Khz) {
Run(32000, 16000, 1000);
}
TEST_F(AcmSwitchingOutputFrequencyOldApi, Toggle16KhzTo8Khz) {
Run(16000, 8000, 1000);
}
TEST_F(AcmSwitchingOutputFrequencyOldApi, Toggle8KhzTo16Khz) {
Run(8000, 16000, 1000);
}
#endif
} // namespace webrtc