mirror of
https://github.com/mollyim/webrtc.git
synced 2025-05-13 05:40:42 +01:00
Start using ArrayView in AudioFrame, update PushResampler
Start introducing ArrayView to AudioFrame and code that flows down from there. In this first step: * Add `data_view()` that returns a read-only ArrayView for the audio buffer. When AudioFrame is not initialized however, data_view() will return a nullptr whereas the current data() method never returns nullptr. * Add `mutable_data()` that requires two arguments for properly setting the samples per channel and number of channels that's required for accurately reserving the returned mutable ArrayView. A notable behavior change is that if the requested number of channels is larger than supported or the calculated buffer size is too large, the function will trigger a check. * Add TODOs for following work. Bug: chromium:335805780 Change-Id: I2937de800422589ebe6a3840b3caadf3d9ff8b00 Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/347982 Reviewed-by: Per Åhgren <peah@webrtc.org> Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org> Cr-Commit-Position: refs/heads/main@{#42202}
This commit is contained in:
parent
652bd288b3
commit
1f3679884c
14 changed files with 229 additions and 60 deletions
|
@ -34,6 +34,7 @@ rtc_library("audio_frame_api") {
|
||||||
]
|
]
|
||||||
|
|
||||||
deps = [
|
deps = [
|
||||||
|
"..:array_view",
|
||||||
"..:rtp_packet_info",
|
"..:rtp_packet_info",
|
||||||
"../../rtc_base:checks",
|
"../../rtc_base:checks",
|
||||||
"../../rtc_base:logging",
|
"../../rtc_base:logging",
|
||||||
|
|
|
@ -22,6 +22,20 @@ AudioFrame::AudioFrame() {
|
||||||
static_assert(sizeof(data_) == kMaxDataSizeBytes, "kMaxDataSizeBytes");
|
static_assert(sizeof(data_) == kMaxDataSizeBytes, "kMaxDataSizeBytes");
|
||||||
}
|
}
|
||||||
|
|
||||||
|
AudioFrame::AudioFrame(int sample_rate_hz,
|
||||||
|
size_t num_channels,
|
||||||
|
ChannelLayout layout /*= CHANNEL_LAYOUT_UNSUPPORTED*/)
|
||||||
|
: samples_per_channel_(SampleRateToDefaultChannelSize(sample_rate_hz)),
|
||||||
|
sample_rate_hz_(sample_rate_hz),
|
||||||
|
num_channels_(num_channels),
|
||||||
|
channel_layout_(layout == CHANNEL_LAYOUT_UNSUPPORTED
|
||||||
|
? GuessChannelLayout(num_channels)
|
||||||
|
: layout) {
|
||||||
|
RTC_DCHECK_LE(num_channels_, kMaxConcurrentChannels);
|
||||||
|
RTC_DCHECK_GT(sample_rate_hz_, 0);
|
||||||
|
RTC_DCHECK_GT(samples_per_channel_, 0u);
|
||||||
|
}
|
||||||
|
|
||||||
void AudioFrame::Reset() {
|
void AudioFrame::Reset() {
|
||||||
ResetWithoutMuting();
|
ResetWithoutMuting();
|
||||||
muted_ = true;
|
muted_ = true;
|
||||||
|
@ -51,6 +65,7 @@ void AudioFrame::UpdateFrame(uint32_t timestamp,
|
||||||
SpeechType speech_type,
|
SpeechType speech_type,
|
||||||
VADActivity vad_activity,
|
VADActivity vad_activity,
|
||||||
size_t num_channels) {
|
size_t num_channels) {
|
||||||
|
RTC_CHECK_LE(num_channels, kMaxConcurrentChannels);
|
||||||
timestamp_ = timestamp;
|
timestamp_ = timestamp;
|
||||||
samples_per_channel_ = samples_per_channel;
|
samples_per_channel_ = samples_per_channel;
|
||||||
sample_rate_hz_ = sample_rate_hz;
|
sample_rate_hz_ = sample_rate_hz;
|
||||||
|
@ -110,12 +125,26 @@ int64_t AudioFrame::ElapsedProfileTimeMs() const {
|
||||||
}
|
}
|
||||||
|
|
||||||
const int16_t* AudioFrame::data() const {
|
const int16_t* AudioFrame::data() const {
|
||||||
return muted_ ? empty_data() : data_;
|
return muted_ ? zeroed_data().begin() : data_;
|
||||||
|
}
|
||||||
|
|
||||||
|
rtc::ArrayView<const int16_t> AudioFrame::data_view() const {
|
||||||
|
const auto samples = samples_per_channel_ * num_channels_;
|
||||||
|
// If you get a nullptr from `data_view()`, it's likely because the
|
||||||
|
// samples_per_channel_ and/or num_channels_ haven't been properly set.
|
||||||
|
// Since `data_view()` returns an rtc::ArrayView<>, we inherit the behavior
|
||||||
|
// in ArrayView when the view size is 0 that ArrayView<>::data() will always
|
||||||
|
// return nullptr. So, even when an AudioFrame is muted and we want to
|
||||||
|
// return `zeroed_data()`, if samples_per_channel_ or num_channels_ is 0,
|
||||||
|
// the view will point to nullptr.
|
||||||
|
return muted_ ? zeroed_data().subview(0, samples)
|
||||||
|
: rtc::ArrayView<const int16_t>(&data_[0], samples);
|
||||||
}
|
}
|
||||||
|
|
||||||
// TODO(henrik.lundin) Can we skip zeroing the buffer?
|
|
||||||
// See https://bugs.chromium.org/p/webrtc/issues/detail?id=5647.
|
|
||||||
int16_t* AudioFrame::mutable_data() {
|
int16_t* AudioFrame::mutable_data() {
|
||||||
|
// TODO: bugs.webrtc.org/5647 - Can we skip zeroing the buffer?
|
||||||
|
// Consider instead if we should rather zero the buffer when `muted_` is set
|
||||||
|
// to `true`.
|
||||||
if (muted_) {
|
if (muted_) {
|
||||||
memset(data_, 0, kMaxDataSizeBytes);
|
memset(data_, 0, kMaxDataSizeBytes);
|
||||||
muted_ = false;
|
muted_ = false;
|
||||||
|
@ -123,6 +152,29 @@ int16_t* AudioFrame::mutable_data() {
|
||||||
return data_;
|
return data_;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
rtc::ArrayView<int16_t> AudioFrame::mutable_data(size_t samples_per_channel,
|
||||||
|
size_t num_channels) {
|
||||||
|
const size_t total_samples = samples_per_channel * num_channels;
|
||||||
|
RTC_CHECK_LE(total_samples, kMaxDataSizeSamples);
|
||||||
|
RTC_CHECK_LE(num_channels, kMaxConcurrentChannels);
|
||||||
|
// Sanity check for valid argument values during development.
|
||||||
|
// If `samples_per_channel` is <= kMaxConcurrentChannels but larger than 0,
|
||||||
|
// then chances are the order of arguments is incorrect.
|
||||||
|
RTC_DCHECK((samples_per_channel == 0 && num_channels == 0) ||
|
||||||
|
samples_per_channel > kMaxConcurrentChannels);
|
||||||
|
|
||||||
|
// TODO: bugs.webrtc.org/5647 - Can we skip zeroing the buffer?
|
||||||
|
// Consider instead if we should rather zero the whole buffer when `muted_` is
|
||||||
|
// set to `true`.
|
||||||
|
if (muted_) {
|
||||||
|
memset(data_, 0, total_samples * sizeof(int16_t));
|
||||||
|
muted_ = false;
|
||||||
|
}
|
||||||
|
samples_per_channel_ = samples_per_channel;
|
||||||
|
num_channels_ = num_channels;
|
||||||
|
return rtc::ArrayView<int16_t>(&data_[0], total_samples);
|
||||||
|
}
|
||||||
|
|
||||||
void AudioFrame::Mute() {
|
void AudioFrame::Mute() {
|
||||||
muted_ = true;
|
muted_ = true;
|
||||||
}
|
}
|
||||||
|
@ -146,10 +198,20 @@ void AudioFrame::SetLayoutAndNumChannels(ChannelLayout layout,
|
||||||
RTC_CHECK_LE(samples_per_channel_ * num_channels_, kMaxDataSizeSamples);
|
RTC_CHECK_LE(samples_per_channel_ * num_channels_, kMaxDataSizeSamples);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
void AudioFrame::SetSampleRateAndChannelSize(int sample_rate) {
|
||||||
|
sample_rate_hz_ = sample_rate;
|
||||||
|
// We could call `AudioProcessing::GetFrameSize()` here, but that requires
|
||||||
|
// adding a dependency on the ":audio_processing" build target, which can
|
||||||
|
// complicate the dependency tree. Some refactoring is probably in order to
|
||||||
|
// get some consistency around this since there are many places across the
|
||||||
|
// code that assume this default buffer size.
|
||||||
|
samples_per_channel_ = SampleRateToDefaultChannelSize(sample_rate_hz_);
|
||||||
|
}
|
||||||
|
|
||||||
// static
|
// static
|
||||||
const int16_t* AudioFrame::empty_data() {
|
rtc::ArrayView<const int16_t> AudioFrame::zeroed_data() {
|
||||||
static int16_t* null_data = new int16_t[kMaxDataSizeSamples]();
|
static int16_t* null_data = new int16_t[kMaxDataSizeSamples]();
|
||||||
return &null_data[0];
|
return rtc::ArrayView<const int16_t>(null_data, kMaxDataSizeSamples);
|
||||||
}
|
}
|
||||||
|
|
||||||
} // namespace webrtc
|
} // namespace webrtc
|
||||||
|
|
|
@ -14,11 +14,30 @@
|
||||||
#include <stddef.h>
|
#include <stddef.h>
|
||||||
#include <stdint.h>
|
#include <stdint.h>
|
||||||
|
|
||||||
|
#include "api/array_view.h"
|
||||||
#include "api/audio/channel_layout.h"
|
#include "api/audio/channel_layout.h"
|
||||||
#include "api/rtp_packet_infos.h"
|
#include "api/rtp_packet_infos.h"
|
||||||
|
#include "rtc_base/checks.h"
|
||||||
|
|
||||||
namespace webrtc {
|
namespace webrtc {
|
||||||
|
|
||||||
|
// Default webrtc buffer size in milliseconds.
|
||||||
|
constexpr size_t kDefaultAudioBufferLengthMs = 10u;
|
||||||
|
|
||||||
|
// Default total number of audio buffers per second based on the default length.
|
||||||
|
constexpr size_t kDefaultAudioBuffersPerSec =
|
||||||
|
1000u / kDefaultAudioBufferLengthMs;
|
||||||
|
|
||||||
|
// Returns the number of samples a buffer needs to hold for ~10ms of a single
|
||||||
|
// audio channel at a given sample rate.
|
||||||
|
// See also `AudioProcessing::GetFrameSize()`.
|
||||||
|
inline size_t SampleRateToDefaultChannelSize(size_t sample_rate) {
|
||||||
|
// Basic sanity check. 192kHz is the highest supported input sample rate.
|
||||||
|
RTC_DCHECK_LE(sample_rate, 192000);
|
||||||
|
return sample_rate / kDefaultAudioBuffersPerSec;
|
||||||
|
}
|
||||||
|
/////////////////////////////////////////////////////////////////////
|
||||||
|
|
||||||
/* This class holds up to 120 ms of super-wideband (32 kHz) stereo audio. It
|
/* This class holds up to 120 ms of super-wideband (32 kHz) stereo audio. It
|
||||||
* allows for adding and subtracting frames while keeping track of the resulting
|
* allows for adding and subtracting frames while keeping track of the resulting
|
||||||
* states.
|
* states.
|
||||||
|
@ -57,6 +76,15 @@ class AudioFrame {
|
||||||
|
|
||||||
AudioFrame();
|
AudioFrame();
|
||||||
|
|
||||||
|
// Construct an audio frame with frame length properties and channel
|
||||||
|
// information. `samples_per_channel()` will be initialized to a 10ms buffer
|
||||||
|
// size and if `layout` is not specified (default value of
|
||||||
|
// CHANNEL_LAYOUT_UNSUPPORTED is set), then the channel layout is derived
|
||||||
|
// (guessed) from `num_channels`.
|
||||||
|
AudioFrame(int sample_rate_hz,
|
||||||
|
size_t num_channels,
|
||||||
|
ChannelLayout layout = CHANNEL_LAYOUT_UNSUPPORTED);
|
||||||
|
|
||||||
AudioFrame(const AudioFrame&) = delete;
|
AudioFrame(const AudioFrame&) = delete;
|
||||||
AudioFrame& operator=(const AudioFrame&) = delete;
|
AudioFrame& operator=(const AudioFrame&) = delete;
|
||||||
|
|
||||||
|
@ -68,6 +96,7 @@ class AudioFrame {
|
||||||
// ResetWithoutMuting() to skip this wasteful zeroing.
|
// ResetWithoutMuting() to skip this wasteful zeroing.
|
||||||
void ResetWithoutMuting();
|
void ResetWithoutMuting();
|
||||||
|
|
||||||
|
// TODO: b/335805780 - Accept ArrayView.
|
||||||
void UpdateFrame(uint32_t timestamp,
|
void UpdateFrame(uint32_t timestamp,
|
||||||
const int16_t* data,
|
const int16_t* data,
|
||||||
size_t samples_per_channel,
|
size_t samples_per_channel,
|
||||||
|
@ -90,11 +119,29 @@ class AudioFrame {
|
||||||
int64_t ElapsedProfileTimeMs() const;
|
int64_t ElapsedProfileTimeMs() const;
|
||||||
|
|
||||||
// data() returns a zeroed static buffer if the frame is muted.
|
// data() returns a zeroed static buffer if the frame is muted.
|
||||||
// mutable_frame() always returns a non-static buffer; the first call to
|
// TODO: b/335805780 - Return ArrayView.
|
||||||
// mutable_frame() zeros the non-static buffer and marks the frame unmuted.
|
|
||||||
const int16_t* data() const;
|
const int16_t* data() const;
|
||||||
|
|
||||||
|
// Returns a read-only view of all the valid samples held by the AudioFrame.
|
||||||
|
// Note that for a muted AudioFrame, the size of the returned view will be
|
||||||
|
// 0u and the contained data will be nullptr.
|
||||||
|
rtc::ArrayView<const int16_t> data_view() const;
|
||||||
|
|
||||||
|
// mutable_frame() always returns a non-static buffer; the first call to
|
||||||
|
// mutable_frame() zeros the buffer and marks the frame as unmuted.
|
||||||
|
// TODO: b/335805780 - Return ArrayView based on the current values for
|
||||||
|
// samples per channel and num channels.
|
||||||
int16_t* mutable_data();
|
int16_t* mutable_data();
|
||||||
|
|
||||||
|
// Grants write access to the audio buffer. The size of the returned writable
|
||||||
|
// view is determined by the `samples_per_channel` and `num_channels`
|
||||||
|
// dimensions which the function checks for correctness and stores in the
|
||||||
|
// internal member variables; `samples_per_channel()` and `num_channels()`
|
||||||
|
// respectively.
|
||||||
|
// If the state is currently muted, the returned view will be zeroed out.
|
||||||
|
rtc::ArrayView<int16_t> mutable_data(size_t samples_per_channel,
|
||||||
|
size_t num_channels);
|
||||||
|
|
||||||
// Prefer to mute frames using AudioFrameOperations::Mute.
|
// Prefer to mute frames using AudioFrameOperations::Mute.
|
||||||
void Mute();
|
void Mute();
|
||||||
// Frame is muted by default.
|
// Frame is muted by default.
|
||||||
|
@ -119,6 +166,10 @@ class AudioFrame {
|
||||||
return absolute_capture_timestamp_ms_;
|
return absolute_capture_timestamp_ms_;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
// Sets the sample_rate_hz and samples_per_channel properties based on a
|
||||||
|
// given sample rate and calculates a default 10ms samples_per_channel value.
|
||||||
|
void SetSampleRateAndChannelSize(int sample_rate);
|
||||||
|
|
||||||
// RTP timestamp of the first sample in the AudioFrame.
|
// RTP timestamp of the first sample in the AudioFrame.
|
||||||
uint32_t timestamp_ = 0;
|
uint32_t timestamp_ = 0;
|
||||||
// Time since the first frame in milliseconds.
|
// Time since the first frame in milliseconds.
|
||||||
|
@ -157,9 +208,9 @@ class AudioFrame {
|
||||||
|
|
||||||
private:
|
private:
|
||||||
// A permanently zeroed out buffer to represent muted frames. This is a
|
// A permanently zeroed out buffer to represent muted frames. This is a
|
||||||
// header-only class, so the only way to avoid creating a separate empty
|
// header-only class, so the only way to avoid creating a separate zeroed
|
||||||
// buffer per translation unit is to wrap a static in an inline function.
|
// buffer per translation unit is to wrap a static in an inline function.
|
||||||
static const int16_t* empty_data();
|
static rtc::ArrayView<const int16_t> zeroed_data();
|
||||||
|
|
||||||
int16_t data_[kMaxDataSizeSamples];
|
int16_t data_[kMaxDataSizeSamples];
|
||||||
bool muted_ = true;
|
bool muted_ = true;
|
||||||
|
|
|
@ -19,10 +19,27 @@ namespace webrtc {
|
||||||
|
|
||||||
namespace {
|
namespace {
|
||||||
|
|
||||||
|
bool AllSamplesAre(int16_t sample, rtc::ArrayView<const int16_t> samples) {
|
||||||
|
for (const auto s : samples) {
|
||||||
|
if (s != sample) {
|
||||||
|
return false;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
return true;
|
||||||
|
}
|
||||||
|
|
||||||
bool AllSamplesAre(int16_t sample, const AudioFrame& frame) {
|
bool AllSamplesAre(int16_t sample, const AudioFrame& frame) {
|
||||||
const int16_t* frame_data = frame.data();
|
return AllSamplesAre(sample, frame.data_view());
|
||||||
for (size_t i = 0; i < frame.max_16bit_samples(); i++) {
|
}
|
||||||
if (frame_data[i] != sample) {
|
|
||||||
|
// Checks the values of samples in the AudioFrame buffer, regardless of whether
|
||||||
|
// they're valid or not, and disregard the `muted()` state of the frame.
|
||||||
|
// I.e. use `max_16bit_samples()` instead of the audio properties
|
||||||
|
// `num_samples * samples_per_channel`.
|
||||||
|
bool AllBufferSamplesAre(int16_t sample, const AudioFrame& frame) {
|
||||||
|
const auto* data = frame.data_view().data();
|
||||||
|
for (size_t i = 0; i < frame.max_16bit_samples(); ++i) {
|
||||||
|
if (data[i] != sample) {
|
||||||
return false;
|
return false;
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
@ -38,29 +55,46 @@ constexpr size_t kSamplesPerChannel = kSampleRateHz / 100;
|
||||||
|
|
||||||
} // namespace
|
} // namespace
|
||||||
|
|
||||||
TEST(AudioFrameTest, FrameStartsMuted) {
|
TEST(AudioFrameTest, FrameStartsZeroedAndMuted) {
|
||||||
AudioFrame frame;
|
AudioFrame frame;
|
||||||
EXPECT_TRUE(frame.muted());
|
EXPECT_TRUE(frame.muted());
|
||||||
|
EXPECT_TRUE(frame.data_view().empty());
|
||||||
EXPECT_TRUE(AllSamplesAre(0, frame));
|
EXPECT_TRUE(AllSamplesAre(0, frame));
|
||||||
}
|
}
|
||||||
|
|
||||||
|
// TODO: b/335805780 - Delete test when `mutable_data()` returns ArrayView.
|
||||||
|
TEST(AudioFrameTest, UnmutedFrameIsInitiallyZeroedLegacy) {
|
||||||
|
AudioFrame frame(kSampleRateHz, kNumChannelsMono, CHANNEL_LAYOUT_NONE);
|
||||||
|
frame.mutable_data();
|
||||||
|
EXPECT_FALSE(frame.muted());
|
||||||
|
EXPECT_TRUE(AllSamplesAre(0, frame));
|
||||||
|
EXPECT_TRUE(AllBufferSamplesAre(0, frame));
|
||||||
|
}
|
||||||
|
|
||||||
TEST(AudioFrameTest, UnmutedFrameIsInitiallyZeroed) {
|
TEST(AudioFrameTest, UnmutedFrameIsInitiallyZeroed) {
|
||||||
AudioFrame frame;
|
AudioFrame frame;
|
||||||
frame.mutable_data();
|
auto data = frame.mutable_data(kSamplesPerChannel, kNumChannelsMono);
|
||||||
EXPECT_FALSE(frame.muted());
|
EXPECT_FALSE(frame.muted());
|
||||||
|
EXPECT_EQ(frame.data_view().size(), kSamplesPerChannel);
|
||||||
|
EXPECT_EQ(data.size(), kSamplesPerChannel);
|
||||||
EXPECT_TRUE(AllSamplesAre(0, frame));
|
EXPECT_TRUE(AllSamplesAre(0, frame));
|
||||||
}
|
}
|
||||||
|
|
||||||
TEST(AudioFrameTest, MutedFrameBufferIsZeroed) {
|
TEST(AudioFrameTest, MutedFrameBufferIsZeroed) {
|
||||||
AudioFrame frame;
|
AudioFrame frame;
|
||||||
int16_t* frame_data = frame.mutable_data();
|
int16_t* frame_data =
|
||||||
|
frame.mutable_data(kSamplesPerChannel, kNumChannelsMono).begin();
|
||||||
|
EXPECT_FALSE(frame.muted());
|
||||||
|
// Fill the reserved buffer with non-zero data.
|
||||||
for (size_t i = 0; i < frame.max_16bit_samples(); i++) {
|
for (size_t i = 0; i < frame.max_16bit_samples(); i++) {
|
||||||
frame_data[i] = 17;
|
frame_data[i] = 17;
|
||||||
}
|
}
|
||||||
ASSERT_TRUE(AllSamplesAre(17, frame));
|
ASSERT_TRUE(AllSamplesAre(17, frame));
|
||||||
|
ASSERT_TRUE(AllBufferSamplesAre(17, frame));
|
||||||
frame.Mute();
|
frame.Mute();
|
||||||
EXPECT_TRUE(frame.muted());
|
EXPECT_TRUE(frame.muted());
|
||||||
EXPECT_TRUE(AllSamplesAre(0, frame));
|
EXPECT_TRUE(AllSamplesAre(0, frame));
|
||||||
|
ASSERT_TRUE(AllBufferSamplesAre(0, frame));
|
||||||
}
|
}
|
||||||
|
|
||||||
TEST(AudioFrameTest, UpdateFrameMono) {
|
TEST(AudioFrameTest, UpdateFrameMono) {
|
||||||
|
@ -95,11 +129,17 @@ TEST(AudioFrameTest, UpdateFrameMultiChannel) {
|
||||||
EXPECT_EQ(kSamplesPerChannel, frame.samples_per_channel());
|
EXPECT_EQ(kSamplesPerChannel, frame.samples_per_channel());
|
||||||
EXPECT_EQ(kNumChannelsStereo, frame.num_channels());
|
EXPECT_EQ(kNumChannelsStereo, frame.num_channels());
|
||||||
EXPECT_EQ(CHANNEL_LAYOUT_STEREO, frame.channel_layout());
|
EXPECT_EQ(CHANNEL_LAYOUT_STEREO, frame.channel_layout());
|
||||||
|
EXPECT_TRUE(frame.muted());
|
||||||
|
|
||||||
frame.UpdateFrame(kTimestamp, nullptr /* data */, kSamplesPerChannel,
|
// Initialize the frame with valid `kNumChannels5_1` data to make sure we
|
||||||
|
// get an unmuted frame with valid samples.
|
||||||
|
int16_t samples[kSamplesPerChannel * kNumChannels5_1] = {17};
|
||||||
|
frame.UpdateFrame(kTimestamp, samples /* data */, kSamplesPerChannel,
|
||||||
kSampleRateHz, AudioFrame::kPLC, AudioFrame::kVadActive,
|
kSampleRateHz, AudioFrame::kPLC, AudioFrame::kVadActive,
|
||||||
kNumChannels5_1);
|
kNumChannels5_1);
|
||||||
|
EXPECT_FALSE(frame.muted());
|
||||||
EXPECT_EQ(kSamplesPerChannel, frame.samples_per_channel());
|
EXPECT_EQ(kSamplesPerChannel, frame.samples_per_channel());
|
||||||
|
EXPECT_EQ(kSamplesPerChannel * kNumChannels5_1, frame.data_view().size());
|
||||||
EXPECT_EQ(kNumChannels5_1, frame.num_channels());
|
EXPECT_EQ(kNumChannels5_1, frame.num_channels());
|
||||||
EXPECT_EQ(CHANNEL_LAYOUT_5_1, frame.channel_layout());
|
EXPECT_EQ(CHANNEL_LAYOUT_5_1, frame.channel_layout());
|
||||||
}
|
}
|
||||||
|
@ -121,6 +161,7 @@ TEST(AudioFrameTest, CopyFrom) {
|
||||||
EXPECT_EQ(frame2.vad_activity_, frame1.vad_activity_);
|
EXPECT_EQ(frame2.vad_activity_, frame1.vad_activity_);
|
||||||
EXPECT_EQ(frame2.num_channels_, frame1.num_channels_);
|
EXPECT_EQ(frame2.num_channels_, frame1.num_channels_);
|
||||||
|
|
||||||
|
EXPECT_EQ(frame2.data_view().size(), frame1.data_view().size());
|
||||||
EXPECT_EQ(frame2.muted(), frame1.muted());
|
EXPECT_EQ(frame2.muted(), frame1.muted());
|
||||||
EXPECT_EQ(0, memcmp(frame2.data(), frame1.data(), sizeof(samples)));
|
EXPECT_EQ(0, memcmp(frame2.data(), frame1.data(), sizeof(samples)));
|
||||||
|
|
||||||
|
|
|
@ -70,20 +70,21 @@ void ProcessCaptureFrame(uint32_t delay_ms,
|
||||||
int Resample(const AudioFrame& frame,
|
int Resample(const AudioFrame& frame,
|
||||||
const int destination_sample_rate,
|
const int destination_sample_rate,
|
||||||
PushResampler<int16_t>* resampler,
|
PushResampler<int16_t>* resampler,
|
||||||
int16_t* destination) {
|
rtc::ArrayView<int16_t> destination) {
|
||||||
TRACE_EVENT2("webrtc", "Resample", "frame sample rate", frame.sample_rate_hz_,
|
TRACE_EVENT2("webrtc", "Resample", "frame sample rate", frame.sample_rate_hz_,
|
||||||
"destination_sample_rate", destination_sample_rate);
|
"destination_sample_rate", destination_sample_rate);
|
||||||
const int number_of_channels = static_cast<int>(frame.num_channels_);
|
const int number_of_channels = static_cast<int>(frame.num_channels_);
|
||||||
const int target_number_of_samples_per_channel =
|
const int target_number_of_samples_per_channel =
|
||||||
destination_sample_rate / 100;
|
destination_sample_rate / 100;
|
||||||
|
RTC_CHECK_EQ(destination.size(),
|
||||||
|
frame.num_channels_ * target_number_of_samples_per_channel);
|
||||||
|
|
||||||
resampler->InitializeIfNeeded(frame.sample_rate_hz_, destination_sample_rate,
|
resampler->InitializeIfNeeded(frame.sample_rate_hz_, destination_sample_rate,
|
||||||
number_of_channels);
|
number_of_channels);
|
||||||
|
|
||||||
// TODO(yujo): make resampler take an AudioFrame, and add special case
|
// TODO(yujo): make resampler take an AudioFrame, and add special case
|
||||||
// handling of muted frames.
|
// handling of muted frames.
|
||||||
return resampler->Resample(
|
return resampler->Resample(frame.data_view(), destination);
|
||||||
frame.data(), frame.samples_per_channel_ * number_of_channels,
|
|
||||||
destination, number_of_channels * target_number_of_samples_per_channel);
|
|
||||||
}
|
}
|
||||||
} // namespace
|
} // namespace
|
||||||
|
|
||||||
|
@ -232,8 +233,10 @@ int32_t AudioTransportImpl::NeedMorePlayData(const size_t nSamples,
|
||||||
RTC_DCHECK_EQ(error, AudioProcessing::kNoError);
|
RTC_DCHECK_EQ(error, AudioProcessing::kNoError);
|
||||||
}
|
}
|
||||||
|
|
||||||
nSamplesOut = Resample(mixed_frame_, samplesPerSec, &render_resampler_,
|
nSamplesOut =
|
||||||
static_cast<int16_t*>(audioSamples));
|
Resample(mixed_frame_, samplesPerSec, &render_resampler_,
|
||||||
|
rtc::ArrayView<int16_t>(static_cast<int16_t*>(audioSamples),
|
||||||
|
nSamples * nChannels));
|
||||||
RTC_DCHECK_EQ(nSamplesOut, nChannels * nSamples);
|
RTC_DCHECK_EQ(nSamplesOut, nChannels * nSamples);
|
||||||
return 0;
|
return 0;
|
||||||
}
|
}
|
||||||
|
@ -263,8 +266,10 @@ void AudioTransportImpl::PullRenderData(int bits_per_sample,
|
||||||
*elapsed_time_ms = mixed_frame_.elapsed_time_ms_;
|
*elapsed_time_ms = mixed_frame_.elapsed_time_ms_;
|
||||||
*ntp_time_ms = mixed_frame_.ntp_time_ms_;
|
*ntp_time_ms = mixed_frame_.ntp_time_ms_;
|
||||||
|
|
||||||
auto output_samples = Resample(mixed_frame_, sample_rate, &render_resampler_,
|
int output_samples =
|
||||||
static_cast<int16_t*>(audio_data));
|
Resample(mixed_frame_, sample_rate, &render_resampler_,
|
||||||
|
rtc::ArrayView<int16_t>(static_cast<int16_t*>(audio_data),
|
||||||
|
number_of_channels * number_of_frames));
|
||||||
RTC_DCHECK_EQ(output_samples, number_of_channels * number_of_frames);
|
RTC_DCHECK_EQ(output_samples, number_of_channels * number_of_frames);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
|
@ -14,6 +14,7 @@
|
||||||
#include "audio/utility/audio_frame_operations.h"
|
#include "audio/utility/audio_frame_operations.h"
|
||||||
#include "common_audio/resampler/include/push_resampler.h"
|
#include "common_audio/resampler/include/push_resampler.h"
|
||||||
#include "rtc_base/checks.h"
|
#include "rtc_base/checks.h"
|
||||||
|
#include "rtc_base/logging.h"
|
||||||
|
|
||||||
namespace webrtc {
|
namespace webrtc {
|
||||||
namespace voe {
|
namespace voe {
|
||||||
|
@ -67,15 +68,22 @@ void RemixAndResample(const int16_t* src_data,
|
||||||
// how much to zero here; or 2) make resampler accept a hint that the input is
|
// how much to zero here; or 2) make resampler accept a hint that the input is
|
||||||
// zeroed.
|
// zeroed.
|
||||||
const size_t src_length = samples_per_channel * audio_ptr_num_channels;
|
const size_t src_length = samples_per_channel * audio_ptr_num_channels;
|
||||||
int out_length =
|
// Ensure the `samples_per_channel_` member is set correctly based on the
|
||||||
resampler->Resample(audio_ptr, src_length, dst_frame->mutable_data(),
|
// destination sample rate, number of channels and assumed 10ms buffer size.
|
||||||
AudioFrame::kMaxDataSizeSamples);
|
// TODO(tommi): Could we rather assume that this has been done by the caller?
|
||||||
|
dst_frame->SetSampleRateAndChannelSize(dst_frame->sample_rate_hz_);
|
||||||
|
|
||||||
|
int out_length = resampler->Resample(
|
||||||
|
rtc::ArrayView<const int16_t>(audio_ptr, src_length),
|
||||||
|
dst_frame->mutable_data(dst_frame->samples_per_channel_,
|
||||||
|
dst_frame->num_channels_));
|
||||||
if (out_length == -1) {
|
if (out_length == -1) {
|
||||||
RTC_FATAL() << "Resample failed: audio_ptr = " << audio_ptr
|
RTC_FATAL() << "Resample failed: audio_ptr = " << audio_ptr
|
||||||
<< ", src_length = " << src_length
|
<< ", src_length = " << src_length
|
||||||
<< ", dst_frame->mutable_data() = "
|
<< ", dst_frame->mutable_data() = "
|
||||||
<< dst_frame->mutable_data();
|
<< dst_frame->mutable_data();
|
||||||
}
|
}
|
||||||
|
|
||||||
dst_frame->samples_per_channel_ = out_length / audio_ptr_num_channels;
|
dst_frame->samples_per_channel_ = out_length / audio_ptr_num_channels;
|
||||||
|
|
||||||
// Upmix after resampling.
|
// Upmix after resampling.
|
||||||
|
|
|
@ -17,6 +17,8 @@
|
||||||
namespace webrtc {
|
namespace webrtc {
|
||||||
namespace voe {
|
namespace voe {
|
||||||
|
|
||||||
|
// Note: The RemixAndResample methods assume 10ms buffer sizes.
|
||||||
|
|
||||||
// Upmix or downmix and resample the audio to `dst_frame`. Expects `dst_frame`
|
// Upmix or downmix and resample the audio to `dst_frame`. Expects `dst_frame`
|
||||||
// to have its sample rate and channels members set to the desired values.
|
// to have its sample rate and channels members set to the desired values.
|
||||||
// Updates the `samples_per_channel_` member accordingly.
|
// Updates the `samples_per_channel_` member accordingly.
|
||||||
|
|
|
@ -14,11 +14,14 @@
|
||||||
#include <memory>
|
#include <memory>
|
||||||
#include <vector>
|
#include <vector>
|
||||||
|
|
||||||
|
#include "api/array_view.h"
|
||||||
|
|
||||||
namespace webrtc {
|
namespace webrtc {
|
||||||
|
|
||||||
class PushSincResampler;
|
class PushSincResampler;
|
||||||
|
|
||||||
// Wraps PushSincResampler to provide stereo support.
|
// Wraps PushSincResampler to provide stereo support.
|
||||||
|
// Note: This implementation assumes 10ms buffer sizes throughout.
|
||||||
// TODO(ajm): add support for an arbitrary number of channels.
|
// TODO(ajm): add support for an arbitrary number of channels.
|
||||||
template <typename T>
|
template <typename T>
|
||||||
class PushResampler {
|
class PushResampler {
|
||||||
|
@ -34,7 +37,7 @@ class PushResampler {
|
||||||
|
|
||||||
// Returns the total number of samples provided in destination (e.g. 32 kHz,
|
// Returns the total number of samples provided in destination (e.g. 32 kHz,
|
||||||
// 2 channel audio gives 640 samples).
|
// 2 channel audio gives 640 samples).
|
||||||
int Resample(const T* src, size_t src_length, T* dst, size_t dst_capacity);
|
int Resample(rtc::ArrayView<const T> src, rtc::ArrayView<T> dst);
|
||||||
|
|
||||||
private:
|
private:
|
||||||
int src_sample_rate_hz_;
|
int src_sample_rate_hz_;
|
||||||
|
|
|
@ -73,32 +73,31 @@ int PushResampler<T>::InitializeIfNeeded(int src_sample_rate_hz,
|
||||||
}
|
}
|
||||||
|
|
||||||
template <typename T>
|
template <typename T>
|
||||||
int PushResampler<T>::Resample(const T* src,
|
int PushResampler<T>::Resample(rtc::ArrayView<const T> src,
|
||||||
size_t src_length,
|
rtc::ArrayView<T> dst) {
|
||||||
T* dst,
|
|
||||||
size_t dst_capacity) {
|
|
||||||
// These checks used to be factored out of this template function due to
|
// These checks used to be factored out of this template function due to
|
||||||
// Windows debug build issues with clang. http://crbug.com/615050
|
// Windows debug build issues with clang. http://crbug.com/615050
|
||||||
const size_t src_size_10ms = (src_sample_rate_hz_ / 100) * num_channels_;
|
const size_t src_size_10ms = (src_sample_rate_hz_ / 100) * num_channels_;
|
||||||
const size_t dst_size_10ms = (dst_sample_rate_hz_ / 100) * num_channels_;
|
const size_t dst_size_10ms = (dst_sample_rate_hz_ / 100) * num_channels_;
|
||||||
RTC_DCHECK_EQ(src_length, src_size_10ms);
|
RTC_DCHECK_EQ(src.size(), src_size_10ms);
|
||||||
RTC_DCHECK_GE(dst_capacity, dst_size_10ms);
|
RTC_DCHECK_GE(dst.size(), dst_size_10ms);
|
||||||
|
|
||||||
if (src_sample_rate_hz_ == dst_sample_rate_hz_) {
|
if (src_sample_rate_hz_ == dst_sample_rate_hz_) {
|
||||||
// The old resampler provides this memcpy facility in the case of matching
|
// The old resampler provides this memcpy facility in the case of matching
|
||||||
// sample rates, so reproduce it here for the sinc resampler.
|
// sample rates, so reproduce it here for the sinc resampler.
|
||||||
memcpy(dst, src, src_length * sizeof(T));
|
memcpy(dst.data(), src.data(), src.size() * sizeof(T));
|
||||||
return static_cast<int>(src_length);
|
return static_cast<int>(src.size());
|
||||||
}
|
}
|
||||||
|
|
||||||
const size_t src_length_mono = src_length / num_channels_;
|
const size_t src_length_mono = src.size() / num_channels_;
|
||||||
const size_t dst_capacity_mono = dst_capacity / num_channels_;
|
const size_t dst_capacity_mono = dst.size() / num_channels_;
|
||||||
|
|
||||||
for (size_t ch = 0; ch < num_channels_; ++ch) {
|
for (size_t ch = 0; ch < num_channels_; ++ch) {
|
||||||
channel_data_array_[ch] = channel_resamplers_[ch].source.data();
|
channel_data_array_[ch] = channel_resamplers_[ch].source.data();
|
||||||
}
|
}
|
||||||
|
|
||||||
Deinterleave(src, src_length_mono, num_channels_, channel_data_array_.data());
|
Deinterleave(src.data(), src_length_mono, num_channels_,
|
||||||
|
channel_data_array_.data());
|
||||||
|
|
||||||
size_t dst_length_mono = 0;
|
size_t dst_length_mono = 0;
|
||||||
|
|
||||||
|
@ -112,7 +111,8 @@ int PushResampler<T>::Resample(const T* src,
|
||||||
channel_data_array_[ch] = channel_resamplers_[ch].destination.data();
|
channel_data_array_[ch] = channel_resamplers_[ch].destination.data();
|
||||||
}
|
}
|
||||||
|
|
||||||
Interleave(channel_data_array_.data(), dst_length_mono, num_channels_, dst);
|
Interleave(channel_data_array_.data(), dst_length_mono, num_channels_,
|
||||||
|
dst.data());
|
||||||
return static_cast<int>(dst_length_mono * num_channels_);
|
return static_cast<int>(dst_length_mono * num_channels_);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
|
@ -45,8 +45,9 @@ int ACMResampler::Resample10Msec(const int16_t* in_audio,
|
||||||
return -1;
|
return -1;
|
||||||
}
|
}
|
||||||
|
|
||||||
int out_length =
|
int out_length = resampler_.Resample(
|
||||||
resampler_.Resample(in_audio, in_length, out_audio, out_capacity_samples);
|
rtc::ArrayView<const int16_t>(in_audio, in_length),
|
||||||
|
rtc::ArrayView<int16_t>(out_audio, out_capacity_samples));
|
||||||
if (out_length == -1) {
|
if (out_length == -1) {
|
||||||
RTC_LOG(LS_ERROR) << "Resample(" << in_audio << ", " << in_length << ", "
|
RTC_LOG(LS_ERROR) << "Resample(" << in_audio << ", " << in_length << ", "
|
||||||
<< out_audio << ", " << out_capacity_samples
|
<< out_audio << ", " << out_capacity_samples
|
||||||
|
|
|
@ -517,13 +517,8 @@ TEST(AudioMixerDeathTest, MultipleChannelsAndHighRate) {
|
||||||
other_frame->samples_per_channel_ = kSamplesPerChannel;
|
other_frame->samples_per_channel_ = kSamplesPerChannel;
|
||||||
mixer->AddSource(&other_source);
|
mixer->AddSource(&other_source);
|
||||||
|
|
||||||
#if RTC_DCHECK_IS_ON && GTEST_HAS_DEATH_TEST && !defined(WEBRTC_ANDROID)
|
#if GTEST_HAS_DEATH_TEST && !defined(WEBRTC_ANDROID)
|
||||||
EXPECT_DEATH(mixer->Mix(kNumberOfChannels, &frame_for_mixing), "");
|
EXPECT_DEATH(mixer->Mix(kNumberOfChannels, &frame_for_mixing), "");
|
||||||
#elif !RTC_DCHECK_IS_ON
|
|
||||||
mixer->Mix(kNumberOfChannels, &frame_for_mixing);
|
|
||||||
EXPECT_EQ(frame_for_mixing.num_channels_, kNumberOfChannels);
|
|
||||||
EXPECT_EQ(frame_for_mixing.sample_rate_hz_,
|
|
||||||
HighOutputRateCalculator::kDefaultFrequency);
|
|
||||||
#endif
|
#endif
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
|
@ -139,8 +139,9 @@ TEST(FrameCombiner, ContainsAllRtpPacketInfos) {
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
// There are DCHECKs in place to check for invalid parameters.
|
#if GTEST_HAS_DEATH_TEST && !defined(WEBRTC_ANDROID)
|
||||||
TEST(FrameCombinerDeathTest, DebugBuildCrashesWithManyChannels) {
|
// There are CHECKs in place to check for invalid parameters.
|
||||||
|
TEST(FrameCombinerDeathTest, BuildCrashesWithManyChannels) {
|
||||||
FrameCombiner combiner(true);
|
FrameCombiner combiner(true);
|
||||||
for (const int rate : {8000, 18000, 34000, 48000}) {
|
for (const int rate : {8000, 18000, 34000, 48000}) {
|
||||||
for (const int number_of_channels : {10, 20, 21}) {
|
for (const int number_of_channels : {10, 20, 21}) {
|
||||||
|
@ -149,7 +150,9 @@ TEST(FrameCombinerDeathTest, DebugBuildCrashesWithManyChannels) {
|
||||||
continue;
|
continue;
|
||||||
}
|
}
|
||||||
const std::vector<AudioFrame*> all_frames = {&frame1, &frame2};
|
const std::vector<AudioFrame*> all_frames = {&frame1, &frame2};
|
||||||
SetUpFrames(rate, number_of_channels);
|
// With an unsupported channel count, this will crash in
|
||||||
|
// `AudioFrame::UpdateFrame`.
|
||||||
|
EXPECT_DEATH(SetUpFrames(rate, number_of_channels), "");
|
||||||
|
|
||||||
const int number_of_frames = 2;
|
const int number_of_frames = 2;
|
||||||
SCOPED_TRACE(
|
SCOPED_TRACE(
|
||||||
|
@ -157,18 +160,14 @@ TEST(FrameCombinerDeathTest, DebugBuildCrashesWithManyChannels) {
|
||||||
const std::vector<AudioFrame*> frames_to_combine(
|
const std::vector<AudioFrame*> frames_to_combine(
|
||||||
all_frames.begin(), all_frames.begin() + number_of_frames);
|
all_frames.begin(), all_frames.begin() + number_of_frames);
|
||||||
AudioFrame audio_frame_for_mixing;
|
AudioFrame audio_frame_for_mixing;
|
||||||
#if RTC_DCHECK_IS_ON && GTEST_HAS_DEATH_TEST && !defined(WEBRTC_ANDROID)
|
|
||||||
EXPECT_DEATH(
|
EXPECT_DEATH(
|
||||||
combiner.Combine(frames_to_combine, number_of_channels, rate,
|
combiner.Combine(frames_to_combine, number_of_channels, rate,
|
||||||
frames_to_combine.size(), &audio_frame_for_mixing),
|
frames_to_combine.size(), &audio_frame_for_mixing),
|
||||||
"");
|
"");
|
||||||
#elif !RTC_DCHECK_IS_ON
|
|
||||||
combiner.Combine(frames_to_combine, number_of_channels, rate,
|
|
||||||
frames_to_combine.size(), &audio_frame_for_mixing);
|
|
||||||
#endif
|
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
#endif // GTEST_HAS_DEATH_TEST && !defined(WEBRTC_ANDROID)
|
||||||
|
|
||||||
TEST(FrameCombinerDeathTest, DebugBuildCrashesWithHighRate) {
|
TEST(FrameCombinerDeathTest, DebugBuildCrashesWithHighRate) {
|
||||||
FrameCombiner combiner(true);
|
FrameCombiner combiner(true);
|
||||||
|
@ -249,7 +248,8 @@ TEST(FrameCombiner, CombiningZeroFramesShouldProduceSilence) {
|
||||||
TEST(FrameCombiner, CombiningOneFrameShouldNotChangeFrame) {
|
TEST(FrameCombiner, CombiningOneFrameShouldNotChangeFrame) {
|
||||||
FrameCombiner combiner(false);
|
FrameCombiner combiner(false);
|
||||||
for (const int rate : {8000, 10000, 11000, 32000, 44100}) {
|
for (const int rate : {8000, 10000, 11000, 32000, 44100}) {
|
||||||
for (const int number_of_channels : {1, 2, 4, 8, 10}) {
|
// kMaxConcurrentChannels is 8.
|
||||||
|
for (const int number_of_channels : {1, 2, 4, kMaxConcurrentChannels}) {
|
||||||
SCOPED_TRACE(ProduceDebugText(rate, number_of_channels, 1));
|
SCOPED_TRACE(ProduceDebugText(rate, number_of_channels, 1));
|
||||||
|
|
||||||
AudioFrame audio_frame_for_mixing;
|
AudioFrame audio_frame_for_mixing;
|
||||||
|
|
|
@ -104,8 +104,7 @@ float VoiceActivityDetectorWrapper::Analyze(AudioFrameView<const float> frame) {
|
||||||
}
|
}
|
||||||
// Resample the first channel of `frame`.
|
// Resample the first channel of `frame`.
|
||||||
RTC_DCHECK_EQ(frame.samples_per_channel(), frame_size_);
|
RTC_DCHECK_EQ(frame.samples_per_channel(), frame_size_);
|
||||||
resampler_.Resample(frame.channel(0).data(), frame_size_,
|
resampler_.Resample(frame.channel(0), resampled_buffer_);
|
||||||
resampled_buffer_.data(), resampled_buffer_.size());
|
|
||||||
|
|
||||||
return vad_->Analyze(resampled_buffer_);
|
return vad_->Analyze(resampled_buffer_);
|
||||||
}
|
}
|
||||||
|
|
|
@ -2198,7 +2198,8 @@ TEST_P(AudioProcessingTest, Formats) {
|
||||||
// necessary.
|
// necessary.
|
||||||
ASSERT_EQ(ref_length,
|
ASSERT_EQ(ref_length,
|
||||||
static_cast<size_t>(resampler.Resample(
|
static_cast<size_t>(resampler.Resample(
|
||||||
out_ptr, out_length, cmp_data.get(), ref_length)));
|
rtc::ArrayView<const float>(out_ptr, out_length),
|
||||||
|
rtc::ArrayView<float>(cmp_data.get(), ref_length))));
|
||||||
out_ptr = cmp_data.get();
|
out_ptr = cmp_data.get();
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
Loading…
Reference in a new issue