Add frame_helpers

A number of utility functions to be shared between frame buffer 2
and the new frame scheduling implementation based on frame buffer 3.

Change-Id: Icc932c6c76fddeeedc8aa64ec27c7e0c955abfd0
Bug: webrtc:13343
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/241604
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Commit-Queue: Evan Shrubsole <eshr@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35743}
This commit is contained in:
Evan Shrubsole
2021-12-15 15:05:15 +01:00
committed by WebRTC LUCI CQ
parent 8e8b966d54
commit f83d4265b5
5 changed files with 142 additions and 65 deletions

View File

@ -148,6 +148,18 @@ rtc_library("h264_packet_buffer") {
absl_deps = [ "//third_party/abseil-cpp/absl/types:optional" ]
}
rtc_library("frame_helpers") {
sources = [
"frame_helpers.cc",
"frame_helpers.h",
]
deps = [
"../../api/video:encoded_frame",
"../../rtc_base:logging",
]
absl_deps = [ "//third_party/abseil-cpp/absl/container:inlined_vector" ]
}
rtc_library("frame_buffer") {
sources = [
"frame_buffer3.cc",
@ -227,6 +239,7 @@ rtc_library("video_coding") {
":codec_globals_headers",
":encoded_frame",
":frame_buffer",
":frame_helpers",
":packet_buffer",
":video_codec_interface",
":video_coding_utility",

View File

@ -18,8 +18,10 @@
#include <utility>
#include <vector>
#include "absl/container/inlined_vector.h"
#include "api/video/encoded_image.h"
#include "api/video/video_timing.h"
#include "modules/video_coding/frame_helpers.h"
#include "modules/video_coding/include/video_coding_defines.h"
#include "modules/video_coding/jitter_estimator.h"
#include "modules/video_coding/timing.h"
@ -250,7 +252,8 @@ std::unique_ptr<EncodedFrame> FrameBuffer::GetNextFrame() {
int64_t render_time_ms = first_frame.RenderTime();
int64_t receive_time_ms = first_frame.ReceivedTime();
// Gracefully handle bad RTP timestamps and render time issues.
if (HasBadRenderTiming(first_frame, now_ms)) {
if (FrameHasBadRenderTiming(first_frame.RenderTimeMs(), now_ms,
timing_->TargetVideoDelay())) {
jitter_estimator_.Reset();
timing_->Reset();
render_time_ms = timing_->RenderTimeMs(first_frame.Timestamp(), now_ms);
@ -318,35 +321,6 @@ std::unique_ptr<EncodedFrame> FrameBuffer::GetNextFrame() {
}
}
bool FrameBuffer::HasBadRenderTiming(const EncodedFrame& frame,
int64_t now_ms) {
// Assume that render timing errors are due to changes in the video stream.
int64_t render_time_ms = frame.RenderTimeMs();
// Zero render time means render immediately.
if (render_time_ms == 0) {
return false;
}
if (render_time_ms < 0) {
return true;
}
const int64_t kMaxVideoDelayMs = 10000;
if (std::abs(render_time_ms - now_ms) > kMaxVideoDelayMs) {
int frame_delay = static_cast<int>(std::abs(render_time_ms - now_ms));
RTC_LOG(LS_WARNING)
<< "A frame about to be decoded is out of the configured "
"delay bounds ("
<< frame_delay << " > " << kMaxVideoDelayMs
<< "). Resetting the video jitter buffer.";
return true;
}
if (static_cast<int>(timing_->TargetVideoDelay()) > kMaxVideoDelayMs) {
RTC_LOG(LS_WARNING) << "The video target delay has grown larger than "
<< kMaxVideoDelayMs << " ms.";
return true;
}
return false;
}
void FrameBuffer::SetProtectionMode(VCMVideoProtection mode) {
TRACE_EVENT0("webrtc", "FrameBuffer::SetProtectionMode");
MutexLock lock(&mutex_);
@ -660,39 +634,11 @@ void FrameBuffer::ClearFramesAndHistory() {
std::unique_ptr<EncodedFrame> FrameBuffer::CombineAndDeleteFrames(
std::vector<std::unique_ptr<EncodedFrame>> frames) const {
RTC_DCHECK(!frames.empty());
size_t total_length = 0;
for (const auto& frame : frames) {
total_length += frame->size();
absl::InlinedVector<std::unique_ptr<EncodedFrame>, 4> inlined;
for (auto& frame : frames) {
inlined.push_back(std::move(frame));
}
const EncodedFrame& last_frame = *frames.back();
std::unique_ptr<EncodedFrame> first_frame = std::move(frames[0]);
auto encoded_image_buffer = EncodedImageBuffer::Create(total_length);
uint8_t* buffer = encoded_image_buffer->data();
first_frame->SetSpatialLayerFrameSize(first_frame->SpatialIndex().value_or(0),
first_frame->size());
memcpy(buffer, first_frame->data(), first_frame->size());
buffer += first_frame->size();
// Spatial index of combined frame is set equal to spatial index of its top
// spatial layer.
first_frame->SetSpatialIndex(last_frame.SpatialIndex().value_or(0));
first_frame->video_timing_mutable()->network2_timestamp_ms =
last_frame.video_timing().network2_timestamp_ms;
first_frame->video_timing_mutable()->receive_finish_ms =
last_frame.video_timing().receive_finish_ms;
// Append all remaining frames to the first one.
for (size_t i = 1; i < frames.size(); ++i) {
// Let |next_frame| fall out of scope so it is deleted after copying.
std::unique_ptr<EncodedFrame> next_frame = std::move(frames[i]);
first_frame->SetSpatialLayerFrameSize(
next_frame->SpatialIndex().value_or(0), next_frame->size());
memcpy(buffer, next_frame->data(), next_frame->size());
buffer += next_frame->size();
}
first_frame->SetEncodedData(encoded_image_buffer);
return first_frame;
return webrtc::CombineAndDeleteFrames(std::move(inlined));
}
FrameBuffer::FrameInfo::FrameInfo() = default;

View File

@ -147,9 +147,6 @@ class FrameBuffer {
void ClearFramesAndHistory() RTC_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
bool HasBadRenderTiming(const EncodedFrame& frame, int64_t now_ms)
RTC_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
// The cleaner solution would be to have the NextFrame function return a
// vector of frames, but until the decoding pipeline can support decoding
// multiple frames at the same time we combine all frames to one frame and

View File

@ -0,0 +1,90 @@
/*
* Copyright (c) 2021 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/video_coding/frame_helpers.h"
#include <utility>
#include "rtc_base/logging.h"
namespace webrtc {
bool FrameHasBadRenderTiming(int64_t render_time_ms,
int64_t now_ms,
int target_video_delay) {
// Zero render time means render immediately.
if (render_time_ms == 0) {
return false;
}
if (render_time_ms < 0) {
return true;
}
const int64_t kMaxVideoDelayMs = 10000;
if (std::abs(render_time_ms - now_ms) > kMaxVideoDelayMs) {
int frame_delay = static_cast<int>(std::abs(render_time_ms - now_ms));
RTC_LOG(LS_WARNING)
<< "A frame about to be decoded is out of the configured "
"delay bounds ("
<< frame_delay << " > " << kMaxVideoDelayMs
<< "). Resetting the video jitter buffer.";
return true;
}
if (target_video_delay > kMaxVideoDelayMs) {
RTC_LOG(LS_WARNING) << "The video target delay has grown larger than "
<< kMaxVideoDelayMs << " ms.";
return true;
}
return false;
}
std::unique_ptr<EncodedFrame> CombineAndDeleteFrames(
absl::InlinedVector<std::unique_ptr<EncodedFrame>, 4> frames) {
RTC_DCHECK(!frames.empty());
if (frames.size() == 1) {
return std::move(frames[0]);
}
size_t total_length = 0;
for (const auto& frame : frames) {
total_length += frame->size();
}
const EncodedFrame& last_frame = *frames.back();
std::unique_ptr<EncodedFrame> first_frame = std::move(frames[0]);
auto encoded_image_buffer = EncodedImageBuffer::Create(total_length);
uint8_t* buffer = encoded_image_buffer->data();
first_frame->SetSpatialLayerFrameSize(first_frame->SpatialIndex().value_or(0),
first_frame->size());
memcpy(buffer, first_frame->data(), first_frame->size());
buffer += first_frame->size();
// Spatial index of combined frame is set equal to spatial index of its top
// spatial layer.
first_frame->SetSpatialIndex(last_frame.SpatialIndex().value_or(0));
first_frame->video_timing_mutable()->network2_timestamp_ms =
last_frame.video_timing().network2_timestamp_ms;
first_frame->video_timing_mutable()->receive_finish_ms =
last_frame.video_timing().receive_finish_ms;
// Append all remaining frames to the first one.
for (size_t i = 1; i < frames.size(); ++i) {
// Let |next_frame| fall out of scope so it is deleted after copying.
std::unique_ptr<EncodedFrame> next_frame = std::move(frames[i]);
first_frame->SetSpatialLayerFrameSize(
next_frame->SpatialIndex().value_or(0), next_frame->size());
memcpy(buffer, next_frame->data(), next_frame->size());
buffer += next_frame->size();
}
first_frame->SetEncodedData(encoded_image_buffer);
return first_frame;
}
} // namespace webrtc

View File

@ -0,0 +1,31 @@
/*
* Copyright (c) 2021 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.
*/
#ifndef MODULES_VIDEO_CODING_FRAME_HELPERS_H_
#define MODULES_VIDEO_CODING_FRAME_HELPERS_H_
#include <memory>
#include "absl/container/inlined_vector.h"
#include "api/video/encoded_frame.h"
namespace webrtc {
// TODO(https://bugs.webrtc.org/13589): Switch to using Timestamp and TimeDelta.
bool FrameHasBadRenderTiming(int64_t render_time_ms,
int64_t now_ms,
int target_video_delay);
std::unique_ptr<EncodedFrame> CombineAndDeleteFrames(
absl::InlinedVector<std::unique_ptr<EncodedFrame>, 4> frames);
} // namespace webrtc
#endif // MODULES_VIDEO_CODING_FRAME_HELPERS_H_