[VEA]Remove redundant temporal_idx for AV1/H265

Since we have SVCGenericMetadata to provide generic svc layer
metadata for video encode accelerator, remove the redundant
definition for codec AV1 and H265.

Bug: 40763991
Change-Id: I342bfd9f83c1a52e7c2a123c5081f7201c1e4903
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/5950554
Commit-Queue: Yingying Ma <yingying.ma@intel.com>
Reviewed-by: Alex Gough <ajgo@chromium.org>
Reviewed-by: Eugene Zemtsov <eugene@chromium.org>
Cr-Commit-Position: refs/heads/main@{#1373183}

media/gpu/mac/vt_video_encode_accelerator_mac.mm

@@ -854,7 +854,7 @@ void VTVideoEncodeAccelerator::ReturnBitstreamBuffer(
       md.h264.emplace().temporal_idx = belongs_to_base_layer ? 0 : 1;
       break;
     case VideoCodec::kHEVC:
-      md.h265.emplace().temporal_idx = belongs_to_base_layer ? 0 : 1;
+      md.svc_generic.emplace().temporal_idx = belongs_to_base_layer ? 0 : 1;
       break;
     default:
       NOTREACHED_IN_MIGRATION();

media/gpu/test/video_encoder/decoder_buffer_validator.cc

@@ -994,7 +994,7 @@ bool AV1Validator::Validate(const DecoderBuffer* buffer,
       LOG(ERROR) << "Don't drop key frame";
       return false;
     }
-    if (metadata.av1.has_value()) {
+    if (metadata.svc_generic.has_value()) {
       LOG(ERROR) << "BitstreamBufferMetadata has Av1Metadata on dropped frame";
       return false;
     }

media/gpu/windows/media_foundation_video_encode_accelerator_win.cc

@@ -2682,7 +2682,7 @@ void MediaFoundationVideoEncodeAccelerator::ProcessOutput() {
     if (codec_ == VideoCodec::kH264) {
       md.h264.emplace().temporal_idx = temporal_id;
     } else if (codec_ == VideoCodec::kHEVC) {
-      md.h265.emplace().temporal_idx = temporal_id;
+      md.svc_generic.emplace().temporal_idx = temporal_id;
     } else if (codec_ == VideoCodec::kVP9) {
       Vp9Metadata& vp9 = md.vp9.emplace();
       if (keyframe) {

media/mojo/mojom/video_encode_accelerator.mojom

@@ -266,10 +266,9 @@ interface VideoEncodeAccelerator {
   Flush() => (bool result);
 };
 
-// DropFrameMetadata, H264Metadata, H265metadata, Vp8Metadata, Vp9Metadata and
-// Av1Metadata define mojo transport formats for media::DropFrameMetadata,
-// media::H264Metadata, media::H265Metadata, media::Vp8Metadata,
-// media::Vp9Metadata and media::Av1Metadata, respectively. See the structures
+// DropFrameMetadata, H264Metadata, Vp8Metadata and Vp9Metadata define mojo
+// transport formats for media::DropFrameMetadata, media::H264Metadata,
+// media::Vp8Metadata and media::Vp9Metadata, respectively. See the structures
 // defined video_encode_accelerator.h for the descriptions of the variables.
 // Either of them is filled in GPU process only in the case of temporal/spatial
 // SVC encoding. That is, none of them is filled in the case of non
@@ -288,10 +287,6 @@ struct H264Metadata {
   bool layer_sync;
 };
 
-struct H265Metadata {
-  uint8 temporal_idx;
-};
-
 struct Vp8Metadata {
   bool non_reference;
   uint8 temporal_idx;
@@ -312,10 +307,6 @@ struct Vp9Metadata {
   array<uint8> p_diffs;
 };
 
-struct Av1Metadata {
-  uint8 temporal_idx;
-};
-
 struct SVCGenericMetadata {
   bool follow_svc_spec;
   uint8 temporal_idx;
@@ -328,10 +319,8 @@ struct SVCGenericMetadata {
 union OptionalMetadata {
   DropFrameMetadata drop;
   H264Metadata h264;
-  H265Metadata h265;
   Vp8Metadata vp8;
   Vp9Metadata vp9;
-  Av1Metadata av1;
 };
 
 struct BitstreamBufferMetadata {

media/mojo/mojom/video_encode_accelerator_mojom_traits.cc

@@ -207,12 +207,6 @@ bool UnionTraits<media::mojom::OptionalMetadataDataView,
     case media::mojom::OptionalMetadataDataView::Tag::kVp9: {
       return data.ReadVp9(&out->vp9);
     }
-    case media::mojom::OptionalMetadataDataView::Tag::kAv1: {
-      return data.ReadAv1(&out->av1);
-    }
-    case media::mojom::OptionalMetadataDataView::Tag::kH265: {
-      return data.ReadH265(&out->h265);
-    }
   }
   NOTREACHED();
 }
@@ -260,14 +254,6 @@ bool StructTraits<media::mojom::H264MetadataDataView, media::H264Metadata>::
   return true;
 }
 
-// static
-bool StructTraits<media::mojom::H265MetadataDataView, media::H265Metadata>::
-    Read(media::mojom::H265MetadataDataView data,
-         media::H265Metadata* out_metadata) {
-  out_metadata->temporal_idx = data.temporal_idx();
-  return true;
-}
-
 // static
 bool StructTraits<media::mojom::Vp8MetadataDataView, media::Vp8Metadata>::Read(
     media::mojom::Vp8MetadataDataView data,
@@ -300,14 +286,6 @@ bool StructTraits<media::mojom::Vp9MetadataDataView, media::Vp9Metadata>::Read(
          data.ReadPDiffs(&out_metadata->p_diffs);
 }
 
-// static
-bool StructTraits<media::mojom::Av1MetadataDataView, media::Av1Metadata>::Read(
-    media::mojom::Av1MetadataDataView data,
-    media::Av1Metadata* out_metadata) {
-  out_metadata->temporal_idx = data.temporal_idx();
-  return true;
-}
-
 // static
 bool StructTraits<media::mojom::SVCGenericMetadataDataView,
                   media::SVCGenericMetadata>::

media/mojo/mojom/video_encode_accelerator_mojom_traits.h

@@ -162,17 +162,12 @@ struct UnionTraits<media::mojom::OptionalMetadataDataView,
       return media::mojom::OptionalMetadataDataView::Tag::kVp8;
     } else if (metadata.vp9) {
       return media::mojom::OptionalMetadataDataView::Tag::kVp9;
-    } else if (metadata.av1) {
-      return media::mojom::OptionalMetadataDataView::Tag::kAv1;
-    } else if (metadata.h265) {
-      return media::mojom::OptionalMetadataDataView::Tag::kH265;
     }
     NOTREACHED();
   }
 
   static bool IsNull(const media::BitstreamBufferMetadata& metadata) {
-    return !metadata.drop && !metadata.h264 && !metadata.vp8 && !metadata.vp9 &&
-           !metadata.av1 && !metadata.h265;
+    return !metadata.drop && !metadata.h264 && !metadata.vp8 && !metadata.vp9;
   }
 
   static void SetToNull(media::BitstreamBufferMetadata* metadata) {
@@ -180,8 +175,6 @@ struct UnionTraits<media::mojom::OptionalMetadataDataView,
     metadata->h264.reset();
     metadata->vp8.reset();
     metadata->vp9.reset();
-    metadata->av1.reset();
-    metadata->h265.reset();
   }
 
   static const media::DropFrameMetadata& drop(
@@ -203,16 +196,6 @@ struct UnionTraits<media::mojom::OptionalMetadataDataView,
     return *metadata.vp9;
   }
 
-  static const media::Av1Metadata& av1(
-      const media::BitstreamBufferMetadata& metadata) {
-    return *metadata.av1;
-  }
-
-  static const media::H265Metadata& h265(
-      const media::BitstreamBufferMetadata& metadata) {
-    return *metadata.h265;
-  }
-
   static bool Read(media::mojom::OptionalMetadataDataView data,
                    media::BitstreamBufferMetadata* metadata);
 };
@@ -282,17 +265,6 @@ class StructTraits<media::mojom::H264MetadataDataView, media::H264Metadata> {
                    media::H264Metadata* out_metadata);
 };
 
-template <>
-class StructTraits<media::mojom::H265MetadataDataView, media::H265Metadata> {
- public:
-  static uint8_t temporal_idx(const media::H265Metadata& h265) {
-    return h265.temporal_idx;
-  }
-
-  static bool Read(media::mojom::H265MetadataDataView data,
-                   media::H265Metadata* out_metadata);
-};
-
 template <>
 class StructTraits<media::mojom::Vp8MetadataDataView, media::Vp8Metadata> {
  public:
@@ -356,17 +328,6 @@ class StructTraits<media::mojom::Vp9MetadataDataView, media::Vp9Metadata> {
                    media::Vp9Metadata* out_metadata);
 };
 
-template <>
-class StructTraits<media::mojom::Av1MetadataDataView, media::Av1Metadata> {
- public:
-  static uint8_t temporal_idx(const media::Av1Metadata& av1) {
-    return av1.temporal_idx;
-  }
-
-  static bool Read(media::mojom::Av1MetadataDataView data,
-                   media::Av1Metadata* out_metadata);
-};
-
 template <>
 class StructTraits<media::mojom::SVCGenericMetadataDataView,
                    media::SVCGenericMetadata> {

media/video/video_encode_accelerator.cc

@@ -20,10 +20,6 @@ Vp9Metadata::Vp9Metadata() = default;
 Vp9Metadata::~Vp9Metadata() = default;
 Vp9Metadata::Vp9Metadata(const Vp9Metadata&) = default;
 
-Av1Metadata::Av1Metadata() = default;
-Av1Metadata::~Av1Metadata() = default;
-Av1Metadata::Av1Metadata(const Av1Metadata&) = default;
-
 BitstreamBufferMetadata::BitstreamBufferMetadata()
     : payload_size_bytes(0), key_frame(false) {}
 BitstreamBufferMetadata::BitstreamBufferMetadata(
@@ -289,10 +285,6 @@ bool operator==(const H264Metadata& l, const H264Metadata& r) {
   return l.temporal_idx == r.temporal_idx && l.layer_sync == r.layer_sync;
 }
 
-bool operator==(const H265Metadata& l, const H265Metadata& r) {
-  return l.temporal_idx == r.temporal_idx;
-}
-
 bool operator==(const Vp8Metadata& l, const Vp8Metadata& r) {
   return l.non_reference == r.non_reference &&
          l.temporal_idx == r.temporal_idx && l.layer_sync == r.layer_sync;
@@ -309,10 +301,6 @@ bool operator==(const Vp9Metadata& l, const Vp9Metadata& r) {
          l.p_diffs == r.p_diffs;
 }
 
-bool operator==(const Av1Metadata& l, const Av1Metadata& r) {
-  return l.temporal_idx == r.temporal_idx;
-}
-
 bool operator==(const SVCGenericMetadata& l, const SVCGenericMetadata& r) {
   return l.follow_svc_spec == r.follow_svc_spec &&
          l.temporal_idx == r.temporal_idx && l.spatial_idx == r.spatial_idx &&
@@ -325,7 +313,7 @@ bool operator==(const BitstreamBufferMetadata& l,
   return l.payload_size_bytes == r.payload_size_bytes &&
          l.key_frame == r.key_frame && l.timestamp == r.timestamp &&
          l.vp8 == r.vp8 && l.vp9 == r.vp9 && l.h264 == r.h264 &&
-         l.av1 == r.av1 && l.h265 == r.h265 && l.svc_generic == r.svc_generic;
+         l.svc_generic == r.svc_generic;
 }
 
 bool operator==(const VideoEncodeAccelerator::Config::SpatialLayer& l,

media/video/video_encode_accelerator.h

@@ -53,12 +53,6 @@ struct MEDIA_EXPORT H264Metadata final {
   bool layer_sync = false;
 };
 
-// Metadata for H265 bitstream buffer.
-//  |temporal_idx|  indicates the temporal index of this frame.
-struct MEDIA_EXPORT H265Metadata final {
-  uint8_t temporal_idx = 0;
-};
-
 //  Metadata for a VP8 bitstream buffer.
 //  |non_reference| is true iff this frame does not update any reference buffer,
 //                  meaning dropping this frame still results in a decodable
@@ -119,15 +113,6 @@ struct MEDIA_EXPORT Vp9Metadata final {
   std::vector<uint8_t> p_diffs;
 };
 
-// Metadata for an AV1 bitstream buffer.
-struct MEDIA_EXPORT Av1Metadata final {
-  Av1Metadata();
-  ~Av1Metadata();
-  Av1Metadata(const Av1Metadata&);
-  // The temporal index for this frame.
-  uint8_t temporal_idx = 0;
-};
-
 // Metadata for filling webrtc::CodecSpecificInfo.generic_frame_info.
 struct MEDIA_EXPORT SVCGenericMetadata final {
   // True iff the reference dependency follows any of the scalability modes
@@ -180,15 +165,13 @@ struct MEDIA_EXPORT BitstreamBufferMetadata final {
   bool dropped_frame() const;
   std::optional<uint8_t> spatial_idx() const;
 
-  // |drop|, |h264|, |vp8|, |vp9|, |av1| and |h265| may be set, but not multiple
-  // of them. Presumably, it's also possible for none of them to be set.
+  // |drop|, |h264|, |vp8| and |vp9| may be set, but not multiple of them.
+  // Presumably, it's also possible for none of them to be set.
   // |drop| is set if and only if the frame is dropped.
   std::optional<DropFrameMetadata> drop;
   std::optional<H264Metadata> h264;
   std::optional<Vp8Metadata> vp8;
   std::optional<Vp9Metadata> vp9;
-  std::optional<Av1Metadata> av1;
-  std::optional<H265Metadata> h265;
 
   // Metadata for SVC encoding is expected to be set in |svc_generic|.
   // TODO: Deprecate the above legacy codec specific medadata and replace them
@@ -536,7 +519,6 @@ MEDIA_EXPORT bool operator==(const VideoEncodeAccelerator::SupportedProfile& l,
                              const VideoEncodeAccelerator::SupportedProfile& r);
 MEDIA_EXPORT bool operator==(const Vp8Metadata& l, const Vp8Metadata& r);
 MEDIA_EXPORT bool operator==(const Vp9Metadata& l, const Vp9Metadata& r);
-MEDIA_EXPORT bool operator==(const Av1Metadata& l, const Av1Metadata& r);
 MEDIA_EXPORT bool operator==(const BitstreamBufferMetadata& l,
                              const BitstreamBufferMetadata& r);
 MEDIA_EXPORT bool operator==(

media/video/video_encode_accelerator_adapter.cc

@@ -826,10 +826,10 @@ void VideoEncodeAcceleratorAdapter::BitstreamBufferReady(
     result.temporal_id = metadata.vp9.value().temporal_idx;
   else if (metadata.vp8.has_value())
     result.temporal_id = metadata.vp8.value().temporal_idx;
-  else if (metadata.av1.has_value())
-    result.temporal_id = metadata.av1.value().temporal_idx;
-  else if (metadata.h265.has_value())
-    result.temporal_id = metadata.h265.value().temporal_idx;
+
+  if (metadata.svc_generic.has_value()) {
+    result.temporal_id = metadata.svc_generic->temporal_idx;
+  }
 
   if (metadata.encoded_size)
     result.encoded_size = metadata.encoded_size;

media/video/video_encode_accelerator_adapter_test.cc

@@ -300,8 +300,6 @@ TEST_F(VideoEncodeAcceleratorAdapterTest, TemporalSvc) {
           EXPECT_EQ(output.temporal_id, 2);
         else if (output.timestamp == base::Milliseconds(4))
           EXPECT_EQ(output.temporal_id, 2);
-        else
-          EXPECT_EQ(output.temporal_id, 2);
 
         EXPECT_EQ(output.color_space, expected_color_space);
         outputs_count++;
@@ -320,11 +318,8 @@ TEST_F(VideoEncodeAcceleratorAdapterTest, TemporalSvc) {
           result.vp9 = Vp9Metadata();
           result.vp9->temporal_idx = 2;
         } else if (frame->timestamp() == base::Milliseconds(4)) {
-          result.av1 = Av1Metadata();
-          result.av1->temporal_idx = 2;
-        } else {
-          result.h265 = H265Metadata();
-          result.h265->temporal_idx = 2;
+          result.svc_generic = SVCGenericMetadata();
+          result.svc_generic->temporal_idx = 2;
         }
         return result;
       }));
@@ -339,8 +334,7 @@ TEST_F(VideoEncodeAcceleratorAdapterTest, TemporalSvc) {
       CreateGreenFrame(options.frame_size, pixel_format, base::Milliseconds(3));
   auto frame4 =
       CreateGreenFrame(options.frame_size, pixel_format, base::Milliseconds(4));
-  auto frame5 =
-      CreateGreenFrame(options.frame_size, pixel_format, base::Milliseconds(5));
+
   adapter()->Encode(frame1, VideoEncoder::EncodeOptions(true),
                     ValidatingStatusCB());
   RunUntilIdle();
@@ -353,10 +347,7 @@ TEST_F(VideoEncodeAcceleratorAdapterTest, TemporalSvc) {
   adapter()->Encode(frame4, VideoEncoder::EncodeOptions(true),
                     ValidatingStatusCB());
   RunUntilIdle();
-  adapter()->Encode(frame5, VideoEncoder::EncodeOptions(true),
-                    ValidatingStatusCB());
-  RunUntilIdle();
-  EXPECT_EQ(outputs_count, 5);
+  EXPECT_EQ(outputs_count, 4);
 }
 
 TEST_F(VideoEncodeAcceleratorAdapterTest, FlushDuringInitialize) {