mirror of
https://github.com/mollyim/webrtc.git
synced 2025-05-13 05:40:42 +01:00

Bug: None Change-Id: I93760b501ff712ca2f7a9dfa3cba6ed5245e4f4b Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/208080 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Commit-Queue: Danil Chapovalov <danilchap@webrtc.org> Cr-Commit-Position: refs/heads/master@{#33297}
89 lines
3.1 KiB
C++
89 lines
3.1 KiB
C++
/*
|
|
* 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_SVC_SCALABILITY_STRUCTURE_SIMULCAST_H_
|
|
#define MODULES_VIDEO_CODING_SVC_SCALABILITY_STRUCTURE_SIMULCAST_H_
|
|
|
|
#include <vector>
|
|
|
|
#include "api/transport/rtp/dependency_descriptor.h"
|
|
#include "api/video/video_bitrate_allocation.h"
|
|
#include "common_video/generic_frame_descriptor/generic_frame_info.h"
|
|
#include "modules/video_coding/svc/scalable_video_controller.h"
|
|
|
|
namespace webrtc {
|
|
|
|
// Scalability structure with multiple independent spatial layers each with the
|
|
// same temporal layering.
|
|
class ScalabilityStructureSimulcast : public ScalableVideoController {
|
|
public:
|
|
ScalabilityStructureSimulcast(int num_spatial_layers,
|
|
int num_temporal_layers);
|
|
~ScalabilityStructureSimulcast() override;
|
|
|
|
StreamLayersConfig StreamConfig() const override;
|
|
std::vector<LayerFrameConfig> NextFrameConfig(bool restart) override;
|
|
GenericFrameInfo OnEncodeDone(const LayerFrameConfig& config) override;
|
|
void OnRatesUpdated(const VideoBitrateAllocation& bitrates) override;
|
|
|
|
private:
|
|
enum FramePattern {
|
|
kNone,
|
|
kDeltaT2A,
|
|
kDeltaT1,
|
|
kDeltaT2B,
|
|
kDeltaT0,
|
|
};
|
|
static constexpr int kMaxNumSpatialLayers = 3;
|
|
static constexpr int kMaxNumTemporalLayers = 3;
|
|
|
|
// Index of the buffer to store last frame for layer (`sid`, `tid`)
|
|
int BufferIndex(int sid, int tid) const {
|
|
return tid * num_spatial_layers_ + sid;
|
|
}
|
|
bool DecodeTargetIsActive(int sid, int tid) const {
|
|
return active_decode_targets_[sid * num_temporal_layers_ + tid];
|
|
}
|
|
void SetDecodeTargetIsActive(int sid, int tid, bool value) {
|
|
active_decode_targets_.set(sid * num_temporal_layers_ + tid, value);
|
|
}
|
|
FramePattern NextPattern() const;
|
|
bool TemporalLayerIsActive(int tid) const;
|
|
|
|
const int num_spatial_layers_;
|
|
const int num_temporal_layers_;
|
|
|
|
FramePattern last_pattern_ = kNone;
|
|
std::bitset<kMaxNumSpatialLayers> can_reference_t0_frame_for_spatial_id_ = 0;
|
|
std::bitset<kMaxNumSpatialLayers> can_reference_t1_frame_for_spatial_id_ = 0;
|
|
std::bitset<32> active_decode_targets_;
|
|
};
|
|
|
|
// S1 0--0--0-
|
|
// ...
|
|
// S0 0--0--0-
|
|
class ScalabilityStructureS2T1 : public ScalabilityStructureSimulcast {
|
|
public:
|
|
ScalabilityStructureS2T1() : ScalabilityStructureSimulcast(2, 1) {}
|
|
~ScalabilityStructureS2T1() override = default;
|
|
|
|
FrameDependencyStructure DependencyStructure() const override;
|
|
};
|
|
|
|
class ScalabilityStructureS3T3 : public ScalabilityStructureSimulcast {
|
|
public:
|
|
ScalabilityStructureS3T3() : ScalabilityStructureSimulcast(3, 3) {}
|
|
~ScalabilityStructureS3T3() override = default;
|
|
|
|
FrameDependencyStructure DependencyStructure() const override;
|
|
};
|
|
|
|
} // namespace webrtc
|
|
|
|
#endif // MODULES_VIDEO_CODING_SVC_SCALABILITY_STRUCTURE_SIMULCAST_H_
|