Add av1 scalability structures for 2 spatial and 2 temporal layers.

Bug: webrtc:11404
Change-Id: I0a2257b88666a0158a85162974430c2c08d61b40
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176501
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Commit-Queue: Danil Chapovalov <danilchap@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31460}

modules/video_coding/codecs/av1/BUILD.gn

@@ -59,6 +59,12 @@ rtc_source_set("scalability_structures") {
     "scalability_structure_l2t1.h",
     "scalability_structure_l2t1_key.cc",
     "scalability_structure_l2t1_key.h",
+    "scalability_structure_l2t2.cc",
+    "scalability_structure_l2t2.h",
+    "scalability_structure_l2t2_key.cc",
+    "scalability_structure_l2t2_key.h",
+    "scalability_structure_l2t2_key_shift.cc",
+    "scalability_structure_l2t2_key_shift.h",
     "scalability_structure_s2t1.cc",
     "scalability_structure_s2t1.h",
   ]

modules/video_coding/codecs/av1/libaom_av1_unittest.cc

@@ -24,6 +24,9 @@
 #include "modules/video_coding/codecs/av1/scalability_structure_l1t2.h"
 #include "modules/video_coding/codecs/av1/scalability_structure_l2t1.h"
 #include "modules/video_coding/codecs/av1/scalability_structure_l2t1_key.h"
+#include "modules/video_coding/codecs/av1/scalability_structure_l2t2.h"
+#include "modules/video_coding/codecs/av1/scalability_structure_l2t2_key.h"
+#include "modules/video_coding/codecs/av1/scalability_structure_l2t2_key_shift.h"
 #include "modules/video_coding/codecs/av1/scalability_structure_s2t1.h"
 #include "modules/video_coding/codecs/av1/scalable_video_controller.h"
 #include "modules/video_coding/codecs/av1/scalable_video_controller_no_layering.h"
@@ -285,7 +288,13 @@ INSTANTIATE_TEST_SUITE_P(
            SvcTestParam{std::make_unique<ScalabilityStructureL2T1Key>,
                         /*num_frames_to_generate=*/3},
            SvcTestParam{std::make_unique<ScalabilityStructureS2T1>,
-                        /*num_frames_to_generate=*/3}));
+                        /*num_frames_to_generate=*/3},
+           SvcTestParam{std::make_unique<ScalabilityStructureL2T2>,
+                        /*num_frames_to_generate=*/4},
+           SvcTestParam{std::make_unique<ScalabilityStructureL2T2Key>,
+                        /*num_frames_to_generate=*/4},
+           SvcTestParam{std::make_unique<ScalabilityStructureL2T2KeyShift>,
+                        /*num_frames_to_generate=*/4}));
 
 }  // namespace
 }  // namespace webrtc

modules/video_coding/codecs/av1/scalability_structure_l2t2.cc

@@ -0,0 +1,175 @@
+/*
+ *  Copyright (c) 2020 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/codecs/av1/scalability_structure_l2t2.h"
+
+#include <utility>
+#include <vector>
+
+#include "absl/base/macros.h"
+#include "api/transport/rtp/dependency_descriptor.h"
+#include "rtc_base/checks.h"
+#include "rtc_base/logging.h"
+
+namespace webrtc {
+namespace {
+
+constexpr auto kNotPresent = DecodeTargetIndication::kNotPresent;
+constexpr auto kDiscardable = DecodeTargetIndication::kDiscardable;
+constexpr auto kSwitch = DecodeTargetIndication::kSwitch;
+constexpr auto kRequired = DecodeTargetIndication::kRequired;
+
+// decode targets: S0T0, S0T1, S1T0, S1T1
+constexpr DecodeTargetIndication kDtis[6][4] = {
+    {kSwitch, kSwitch, kSwitch, kSwitch},                   // kKey, S0
+    {kNotPresent, kNotPresent, kSwitch, kSwitch},           // kKey, S1
+    {kNotPresent, kDiscardable, kNotPresent, kRequired},    // kDeltaT1, S0
+    {kNotPresent, kNotPresent, kNotPresent, kDiscardable},  // kDeltaT1, S1
+    {kSwitch, kSwitch, kRequired, kRequired},               // kDeltaT0, S0
+    {kNotPresent, kNotPresent, kSwitch, kSwitch},           // kDeltaT0, S1
+};
+
+}  // namespace
+
+ScalabilityStructureL2T2::~ScalabilityStructureL2T2() = default;
+
+ScalableVideoController::StreamLayersConfig
+ScalabilityStructureL2T2::StreamConfig() const {
+  StreamLayersConfig result;
+  result.num_spatial_layers = 2;
+  result.num_temporal_layers = 2;
+  return result;
+}
+
+FrameDependencyStructure ScalabilityStructureL2T2::DependencyStructure() const {
+  using Builder = GenericFrameInfo::Builder;
+  FrameDependencyStructure structure;
+  structure.num_decode_targets = 4;
+  structure.num_chains = 2;
+  structure.decode_target_protected_by_chain = {0, 0, 1, 1};
+  structure.templates = {
+      Builder().S(0).T(0).Dtis("SSSS").ChainDiffs({0, 0}).Build(),
+      Builder().S(0).T(0).Dtis("SSRR").Fdiffs({4}).ChainDiffs({4, 3}).Build(),
+      Builder().S(0).T(1).Dtis("-D-R").Fdiffs({2}).ChainDiffs({2, 1}).Build(),
+      Builder().S(1).T(0).Dtis("--SS").Fdiffs({1}).ChainDiffs({1, 1}).Build(),
+      Builder()
+          .S(1)
+          .T(0)
+          .Dtis("--SS")
+          .Fdiffs({4, 1})
+          .ChainDiffs({1, 1})
+          .Build(),
+      Builder()
+          .S(1)
+          .T(1)
+          .Dtis("---D")
+          .Fdiffs({2, 1})
+          .ChainDiffs({3, 2})
+          .Build(),
+  };
+  return structure;
+}
+
+ScalableVideoController::LayerFrameConfig
+ScalabilityStructureL2T2::KeyFrameConfig() const {
+  LayerFrameConfig result;
+  result.id = 0;
+  result.is_keyframe = true;
+  result.spatial_id = 0;
+  result.temporal_id = 0;
+  result.buffers = {{/*id=*/0, /*referenced=*/false, /*updated=*/true}};
+  return result;
+}
+
+std::vector<ScalableVideoController::LayerFrameConfig>
+ScalabilityStructureL2T2::NextFrameConfig(bool restart) {
+  if (restart) {
+    next_pattern_ = kKey;
+  }
+  std::vector<LayerFrameConfig> result(2);
+
+  // Buffer0 keeps latest S0T0 frame,
+  // Buffer1 keeps latest S1T0 frame.
+  // Buffer2 keeps latest S0T1 frame.
+  switch (next_pattern_) {
+    case kKey:
+      result[0] = KeyFrameConfig();
+
+      result[1].id = 1;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 0;
+      result[1].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/false},
+                           {/*id=*/1, /*referenced=*/false, /*updated=*/true}};
+
+      next_pattern_ = kDeltaT1;
+      break;
+    case kDeltaT1:
+      result[0].id = 2;
+      result[0].is_keyframe = false;
+      result[0].spatial_id = 0;
+      result[0].temporal_id = 1;
+      result[0].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/false},
+                           {/*id=*/2, /*referenced=*/false, /*updated=*/true}};
+
+      result[1].id = 3;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 1;
+      result[1].buffers = {{/*id=*/2, /*referenced=*/true, /*updated=*/false},
+                           {/*id=*/1, /*referenced=*/true, /*updated=*/false}};
+
+      next_pattern_ = kDeltaT0;
+      break;
+    case kDeltaT0:
+      result[0].id = 4;
+      result[0].is_keyframe = false;
+      result[0].spatial_id = 0;
+      result[0].temporal_id = 0;
+      result[0].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/true}};
+
+      result[1].id = 5;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 0;
+      result[1].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/false},
+                           {/*id=*/1, /*referenced=*/true, /*updated=*/true}};
+
+      next_pattern_ = kDeltaT1;
+      break;
+  }
+  return result;
+}
+
+absl::optional<GenericFrameInfo> ScalabilityStructureL2T2::OnEncodeDone(
+    LayerFrameConfig config) {
+  if (config.is_keyframe) {
+    config = KeyFrameConfig();
+  }
+
+  absl::optional<GenericFrameInfo> frame_info;
+  if (config.id < 0 || config.id >= int{ABSL_ARRAYSIZE(kDtis)}) {
+    RTC_LOG(LS_ERROR) << "Unexpected config id " << config.id;
+    return frame_info;
+  }
+  frame_info.emplace();
+  frame_info->spatial_id = config.spatial_id;
+  frame_info->temporal_id = config.temporal_id;
+  frame_info->encoder_buffers = std::move(config.buffers);
+  frame_info->decode_target_indications.assign(std::begin(kDtis[config.id]),
+                                               std::end(kDtis[config.id]));
+  if (config.temporal_id == 0) {
+    frame_info->part_of_chain = {config.spatial_id == 0, true};
+  } else {
+    frame_info->part_of_chain = {false, false};
+  }
+  return frame_info;
+}
+
+}  // namespace webrtc

modules/video_coding/codecs/av1/scalability_structure_l2t2.h

@@ -0,0 +1,53 @@
+/*
+ *  Copyright (c) 2020 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_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_H_
+#define MODULES_VIDEO_CODING_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_H_
+
+#include <vector>
+
+#include "api/transport/rtp/dependency_descriptor.h"
+#include "common_video/generic_frame_descriptor/generic_frame_info.h"
+#include "modules/video_coding/codecs/av1/scalable_video_controller.h"
+
+namespace webrtc {
+
+// S1T1     0   0
+//         /|  /|  /
+// S1T0   0-+-0-+-0
+//        | | | | | ...
+// S0T1   | 0 | 0 |
+//        |/  |/  |/
+// S0T0   0---0---0--
+// Time-> 0 1 2 3 4
+class ScalabilityStructureL2T2 : public ScalableVideoController {
+ public:
+  ~ScalabilityStructureL2T2() override;
+
+  StreamLayersConfig StreamConfig() const override;
+  FrameDependencyStructure DependencyStructure() const override;
+
+  std::vector<LayerFrameConfig> NextFrameConfig(bool restart) override;
+  absl::optional<GenericFrameInfo> OnEncodeDone(
+      LayerFrameConfig config) override;
+
+ private:
+  enum FramePattern {
+    kKey,
+    kDeltaT1,
+    kDeltaT0,
+  };
+  LayerFrameConfig KeyFrameConfig() const;
+
+  FramePattern next_pattern_ = kKey;
+};
+
+}  // namespace webrtc
+
+#endif  // MODULES_VIDEO_CODING_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_H_

modules/video_coding/codecs/av1/scalability_structure_l2t2_key.cc

@@ -0,0 +1,162 @@
+/*
+ *  Copyright (c) 2020 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/codecs/av1/scalability_structure_l2t2_key.h"
+
+#include <utility>
+#include <vector>
+
+#include "absl/base/macros.h"
+#include "api/transport/rtp/dependency_descriptor.h"
+#include "rtc_base/checks.h"
+#include "rtc_base/logging.h"
+
+namespace webrtc {
+namespace {
+
+constexpr auto kNotPresent = DecodeTargetIndication::kNotPresent;
+constexpr auto kDiscardable = DecodeTargetIndication::kDiscardable;
+constexpr auto kSwitch = DecodeTargetIndication::kSwitch;
+
+// decode targets: S0T0, S0T1, S1T0, S1T1
+constexpr DecodeTargetIndication kDtis[6][4] = {
+    {kSwitch, kSwitch, kSwitch, kSwitch},                   //    kKey, S0
+    {kNotPresent, kNotPresent, kSwitch, kSwitch},           //    kKey, S1
+    {kNotPresent, kDiscardable, kNotPresent, kNotPresent},  //    kDeltaT1, S0
+    {kNotPresent, kNotPresent, kNotPresent, kDiscardable},  //    kDeltaT1, S1
+    {kSwitch, kSwitch, kNotPresent, kNotPresent},           //    kDeltaT0, S0
+    {kNotPresent, kNotPresent, kSwitch, kSwitch},           //    kDeltaT0, S1
+};
+
+}  // namespace
+
+ScalabilityStructureL2T2Key::~ScalabilityStructureL2T2Key() = default;
+
+ScalableVideoController::StreamLayersConfig
+ScalabilityStructureL2T2Key::StreamConfig() const {
+  StreamLayersConfig result;
+  result.num_spatial_layers = 2;
+  result.num_temporal_layers = 2;
+  return result;
+}
+
+FrameDependencyStructure ScalabilityStructureL2T2Key::DependencyStructure()
+    const {
+  using Builder = GenericFrameInfo::Builder;
+  FrameDependencyStructure structure;
+  structure.num_decode_targets = 4;
+  structure.num_chains = 2;
+  structure.decode_target_protected_by_chain = {0, 0, 1, 1};
+  structure.templates = {
+      Builder().S(0).T(0).Dtis("SSSS").ChainDiffs({0, 0}).Build(),
+      Builder().S(0).T(0).Dtis("SS--").Fdiffs({4}).ChainDiffs({4, 3}).Build(),
+      Builder().S(0).T(1).Dtis("-D--").Fdiffs({2}).ChainDiffs({2, 1}).Build(),
+      Builder().S(1).T(0).Dtis("--SS").Fdiffs({1}).ChainDiffs({1, 1}).Build(),
+      Builder().S(1).T(0).Dtis("--SS").Fdiffs({4}).ChainDiffs({1, 4}).Build(),
+      Builder().S(1).T(1).Dtis("---D").Fdiffs({2}).ChainDiffs({3, 2}).Build(),
+  };
+  return structure;
+}
+
+ScalableVideoController::LayerFrameConfig
+ScalabilityStructureL2T2Key::KeyFrameConfig() const {
+  LayerFrameConfig result;
+  result.id = 0;
+  result.is_keyframe = true;
+  result.spatial_id = 0;
+  result.temporal_id = 0;
+  result.buffers = {{/*id=*/0, /*referenced=*/false, /*updated=*/true}};
+  return result;
+}
+
+std::vector<ScalableVideoController::LayerFrameConfig>
+ScalabilityStructureL2T2Key::NextFrameConfig(bool restart) {
+  if (restart) {
+    next_pattern_ = kKey;
+  }
+  std::vector<LayerFrameConfig> result(2);
+
+  // Buffer0 keeps latest S0T0 frame,
+  // Buffer1 keeps latest S1T0 frame.
+  switch (next_pattern_) {
+    case kKey:
+      result[0] = KeyFrameConfig();
+
+      result[1].id = 1;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 0;
+      result[1].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/false},
+                           {/*id=*/1, /*referenced=*/false, /*updated=*/true}};
+
+      next_pattern_ = kDeltaT1;
+      break;
+    case kDeltaT1:
+      result[0].id = 2;
+      result[0].is_keyframe = false;
+      result[0].spatial_id = 0;
+      result[0].temporal_id = 1;
+      result[0].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/false}};
+
+      result[1].id = 3;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 1;
+      result[1].buffers = {{/*id=*/1, /*referenced=*/true, /*updated=*/false}};
+
+      next_pattern_ = kDeltaT0;
+      break;
+    case kDeltaT0:
+      result[0].id = 4;
+      result[0].is_keyframe = false;
+      result[0].spatial_id = 0;
+      result[0].temporal_id = 0;
+      result[0].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/true}};
+
+      result[1].id = 5;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 0;
+      result[1].buffers = {{/*id=*/1, /*referenced=*/true, /*updated=*/true}};
+
+      next_pattern_ = kDeltaT1;
+      break;
+  }
+  return result;
+}
+
+absl::optional<GenericFrameInfo> ScalabilityStructureL2T2Key::OnEncodeDone(
+    LayerFrameConfig config) {
+  if (config.is_keyframe) {
+    config = KeyFrameConfig();
+  }
+
+  absl::optional<GenericFrameInfo> frame_info;
+  if (config.id < 0 || config.id >= int{ABSL_ARRAYSIZE(kDtis)}) {
+    RTC_LOG(LS_ERROR) << "Unexpected config id " << config.id;
+    return frame_info;
+  }
+  frame_info.emplace();
+  frame_info->spatial_id = config.spatial_id;
+  frame_info->temporal_id = config.temporal_id;
+  frame_info->encoder_buffers = std::move(config.buffers);
+  frame_info->decode_target_indications.assign(std::begin(kDtis[config.id]),
+                                               std::end(kDtis[config.id]));
+  if (config.is_keyframe) {
+    frame_info->part_of_chain = {true, true};
+  } else if (config.temporal_id == 0) {
+    frame_info->part_of_chain = {config.spatial_id == 0,
+                                 config.spatial_id == 1};
+  } else {
+    frame_info->part_of_chain = {false, false};
+  }
+  return frame_info;
+}
+
+}  // namespace webrtc

modules/video_coding/codecs/av1/scalability_structure_l2t2_key.h

@@ -0,0 +1,53 @@
+/*
+ *  Copyright (c) 2020 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_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_KEY_H_
+#define MODULES_VIDEO_CODING_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_KEY_H_
+
+#include <vector>
+
+#include "api/transport/rtp/dependency_descriptor.h"
+#include "common_video/generic_frame_descriptor/generic_frame_info.h"
+#include "modules/video_coding/codecs/av1/scalable_video_controller.h"
+
+namespace webrtc {
+
+// S1T1     0   0
+//         /   /   /
+// S1T0   0---0---0
+//        |         ...
+// S0T1   | 0   0
+//        |/   /   /
+// S0T0   0---0---0
+// Time-> 0 1 2 3 4
+class ScalabilityStructureL2T2Key : public ScalableVideoController {
+ public:
+  ~ScalabilityStructureL2T2Key() override;
+
+  StreamLayersConfig StreamConfig() const override;
+  FrameDependencyStructure DependencyStructure() const override;
+
+  std::vector<LayerFrameConfig> NextFrameConfig(bool restart) override;
+  absl::optional<GenericFrameInfo> OnEncodeDone(
+      LayerFrameConfig config) override;
+
+ private:
+  enum FramePattern {
+    kKey,
+    kDeltaT1,
+    kDeltaT0,
+  };
+  LayerFrameConfig KeyFrameConfig() const;
+
+  FramePattern next_pattern_ = kKey;
+};
+
+}  // namespace webrtc
+
+#endif  // MODULES_VIDEO_CODING_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_KEY_H_

modules/video_coding/codecs/av1/scalability_structure_l2t2_key_shift.cc

@@ -0,0 +1,163 @@
+/*
+ *  Copyright (c) 2020 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/codecs/av1/scalability_structure_l2t2_key_shift.h"
+
+#include <utility>
+#include <vector>
+
+#include "absl/base/macros.h"
+#include "api/transport/rtp/dependency_descriptor.h"
+#include "rtc_base/checks.h"
+#include "rtc_base/logging.h"
+
+namespace webrtc {
+namespace {
+
+constexpr auto kNotPresent = DecodeTargetIndication::kNotPresent;
+constexpr auto kDiscardable = DecodeTargetIndication::kDiscardable;
+constexpr auto kSwitch = DecodeTargetIndication::kSwitch;
+
+// decode targets: S0T0, S0T1, S1T0, S1T1
+constexpr DecodeTargetIndication kDtis[6][4] = {
+    {kSwitch, kSwitch, kSwitch, kSwitch},                   //    kKey, S0T0
+    {kNotPresent, kNotPresent, kSwitch, kSwitch},           //    kKey, S1T0
+    {kSwitch, kSwitch, kNotPresent, kNotPresent},           //    kDelta0, S0T0
+    {kNotPresent, kNotPresent, kNotPresent, kDiscardable},  //    kDelta0, S1T1
+    {kNotPresent, kDiscardable, kNotPresent, kNotPresent},  //    kDelta1, S0T1
+    {kNotPresent, kNotPresent, kSwitch, kSwitch},           //    kDelta1, S1T0
+};
+
+}  // namespace
+
+ScalabilityStructureL2T2KeyShift::~ScalabilityStructureL2T2KeyShift() = default;
+
+ScalableVideoController::StreamLayersConfig
+ScalabilityStructureL2T2KeyShift::StreamConfig() const {
+  StreamLayersConfig result;
+  result.num_spatial_layers = 2;
+  result.num_temporal_layers = 2;
+  return result;
+}
+
+FrameDependencyStructure ScalabilityStructureL2T2KeyShift::DependencyStructure()
+    const {
+  using Builder = GenericFrameInfo::Builder;
+  FrameDependencyStructure structure;
+  structure.num_decode_targets = 4;
+  structure.num_chains = 2;
+  structure.decode_target_protected_by_chain = {0, 0, 1, 1};
+  structure.templates = {
+      Builder().S(0).T(0).Dtis("SSSS").ChainDiffs({0, 0}).Build(),
+      Builder().S(0).T(0).Dtis("SS--").Fdiffs({2}).ChainDiffs({2, 1}).Build(),
+      Builder().S(0).T(0).Dtis("SS--").Fdiffs({4}).ChainDiffs({4, 1}).Build(),
+      Builder().S(0).T(1).Dtis("-D--").Fdiffs({2}).ChainDiffs({2, 3}).Build(),
+      Builder().S(1).T(0).Dtis("--SS").Fdiffs({1}).ChainDiffs({1, 1}).Build(),
+      Builder().S(1).T(0).Dtis("--SS").Fdiffs({4}).ChainDiffs({3, 4}).Build(),
+      Builder().S(1).T(1).Dtis("---D").Fdiffs({2}).ChainDiffs({1, 2}).Build(),
+  };
+  return structure;
+}
+
+ScalableVideoController::LayerFrameConfig
+ScalabilityStructureL2T2KeyShift::KeyFrameConfig() const {
+  LayerFrameConfig result;
+  result.id = 0;
+  result.is_keyframe = true;
+  result.spatial_id = 0;
+  result.temporal_id = 0;
+  result.buffers = {{/*id=*/0, /*referenced=*/false, /*updated=*/true}};
+  return result;
+}
+
+std::vector<ScalableVideoController::LayerFrameConfig>
+ScalabilityStructureL2T2KeyShift::NextFrameConfig(bool restart) {
+  if (restart) {
+    next_pattern_ = kKey;
+  }
+  std::vector<LayerFrameConfig> result(2);
+
+  // Buffer0 keeps latest S0T0 frame,
+  // Buffer1 keeps latest S1T0 frame.
+  switch (next_pattern_) {
+    case kKey:
+      result[0] = KeyFrameConfig();
+
+      result[1].id = 1;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 0;
+      result[1].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/false},
+                           {/*id=*/1, /*referenced=*/false, /*updated=*/true}};
+
+      next_pattern_ = kDelta0;
+      break;
+    case kDelta0:
+      result[0].id = 2;
+      result[0].is_keyframe = false;
+      result[0].spatial_id = 0;
+      result[0].temporal_id = 0;
+      result[0].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/true}};
+
+      result[1].id = 3;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 1;
+      result[1].buffers = {{/*id=*/1, /*referenced=*/true, /*updated=*/false}};
+
+      next_pattern_ = kDelta1;
+      break;
+    case kDelta1:
+      result[0].id = 4;
+      result[0].is_keyframe = false;
+      result[0].spatial_id = 0;
+      result[0].temporal_id = 1;
+      result[0].buffers = {{/*id=*/0, /*referenced=*/true, /*updated=*/false}};
+
+      result[1].id = 5;
+      result[1].is_keyframe = false;
+      result[1].spatial_id = 1;
+      result[1].temporal_id = 0;
+      result[1].buffers = {{/*id=*/1, /*referenced=*/true, /*updated=*/true}};
+
+      next_pattern_ = kDelta0;
+      break;
+  }
+  return result;
+}
+
+absl::optional<GenericFrameInfo> ScalabilityStructureL2T2KeyShift::OnEncodeDone(
+    LayerFrameConfig config) {
+  if (config.is_keyframe) {
+    config = KeyFrameConfig();
+  }
+
+  absl::optional<GenericFrameInfo> frame_info;
+  if (config.id < 0 || config.id >= int{ABSL_ARRAYSIZE(kDtis)}) {
+    RTC_LOG(LS_ERROR) << "Unexpected config id " << config.id;
+    return frame_info;
+  }
+  frame_info.emplace();
+  frame_info->spatial_id = config.spatial_id;
+  frame_info->temporal_id = config.temporal_id;
+  frame_info->encoder_buffers = std::move(config.buffers);
+  frame_info->decode_target_indications.assign(std::begin(kDtis[config.id]),
+                                               std::end(kDtis[config.id]));
+  if (config.is_keyframe) {
+    frame_info->part_of_chain = {true, true};
+  } else if (config.temporal_id == 0) {
+    frame_info->part_of_chain = {config.spatial_id == 0,
+                                 config.spatial_id == 1};
+  } else {
+    frame_info->part_of_chain = {false, false};
+  }
+  return frame_info;
+}
+
+}  // namespace webrtc

modules/video_coding/codecs/av1/scalability_structure_l2t2_key_shift.h

@@ -0,0 +1,53 @@
+/*
+ *  Copyright (c) 2020 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_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_KEY_SHIFT_H_
+#define MODULES_VIDEO_CODING_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_KEY_SHIFT_H_
+
+#include <vector>
+
+#include "api/transport/rtp/dependency_descriptor.h"
+#include "common_video/generic_frame_descriptor/generic_frame_info.h"
+#include "modules/video_coding/codecs/av1/scalable_video_controller.h"
+
+namespace webrtc {
+
+// S1T1     0   0
+//         /   /   /
+// S1T0   0---0---0
+//        |        ...
+// S0T1   |   0   0
+//        |  /   /
+// S0T0   0-0---0--
+// Time-> 0 1 2 3 4
+class ScalabilityStructureL2T2KeyShift : public ScalableVideoController {
+ public:
+  ~ScalabilityStructureL2T2KeyShift() override;
+
+  StreamLayersConfig StreamConfig() const override;
+  FrameDependencyStructure DependencyStructure() const override;
+
+  std::vector<LayerFrameConfig> NextFrameConfig(bool restart) override;
+  absl::optional<GenericFrameInfo> OnEncodeDone(
+      LayerFrameConfig config) override;
+
+ private:
+  enum FramePattern {
+    kKey,
+    kDelta0,
+    kDelta1,
+  };
+  LayerFrameConfig KeyFrameConfig() const;
+
+  FramePattern next_pattern_ = kKey;
+};
+
+}  // namespace webrtc
+
+#endif  // MODULES_VIDEO_CODING_CODECS_AV1_SCALABILITY_STRUCTURE_L2T2_KEY_SHIFT_H_

modules/video_coding/codecs/av1/scalability_structure_unittest.cc

@@ -22,6 +22,9 @@
 #include "modules/video_coding/codecs/av1/scalability_structure_l1t2.h"
 #include "modules/video_coding/codecs/av1/scalability_structure_l2t1.h"
 #include "modules/video_coding/codecs/av1/scalability_structure_l2t1_key.h"
+#include "modules/video_coding/codecs/av1/scalability_structure_l2t2.h"
+#include "modules/video_coding/codecs/av1/scalability_structure_l2t2_key.h"
+#include "modules/video_coding/codecs/av1/scalability_structure_l2t2_key_shift.h"
 #include "modules/video_coding/codecs/av1/scalability_structure_s2t1.h"
 #include "modules/video_coding/codecs/av1/scalable_video_controller.h"
 #include "modules/video_coding/frame_dependencies_calculator.h"
@@ -287,7 +290,15 @@ INSTANTIATE_TEST_SUITE_P(
                         std::make_unique<ScalabilityStructureL2T1Key>,
                         /*num_temporal_units=*/3},
            SvcTestParam{"S2T1", std::make_unique<ScalabilityStructureS2T1>,
-                        /*num_temporal_units=*/3}),
+                        /*num_temporal_units=*/3},
+           SvcTestParam{"L2T2", std::make_unique<ScalabilityStructureL2T2>,
+                        /*num_temporal_units=*/4},
+           SvcTestParam{"L2T2Key",
+                        std::make_unique<ScalabilityStructureL2T2Key>,
+                        /*num_temporal_units=*/4},
+           SvcTestParam{"L2T2KeyShift",
+                        std::make_unique<ScalabilityStructureL2T2KeyShift>,
+                        /*num_temporal_units=*/4}),
     [](const testing::TestParamInfo<SvcTestParam>& info) {
       return info.param.name;
     });