Description

The PFND3D12DDI_VIDEO_ENCODE_RESOLVE_OUTPUT_METADATA_0082_0 callback function records a command to resolve the output metadata of an encoding operation into a readable format.

Parameters

hDrvCommandList [in]

Handle to the driver's data for the command list. The driver uses this region of memory to store internal data structures that are related to its command list.

pInputArguments [in]

Pointer to a D3D12DDI_VIDEO_ENCODER_RESOLVE_METADATA_INPUT_ARGUMENTS_0080_2 structure that contains input arguments for this callback.

pOutputArguments [out]

Pointer to a D3D12DDI_VIDEO_ENCODER_RESOLVE_METADATA_OUTPUT_ARGUMENTS_0082_0 structure in which to store output from this callback.

Remarks

For all encoders, the driver must position all layout information contiguously in memory in the ResolvedLayoutEncoderMetadataDrvBuffer. The API Client will parse this metadata accordingly depending on the encoder type (D3D12DDI_VIDEO_ENCODER_RESOLVE_METADATA_INPUT_ARGUMENTS_0080_2.EncoderCodec at the driver level).

ResolvedLayoutEncoderMetadataDrvBuffer for H264/HEVC codecs

For H264 and HEVC encoders (types supported before the AV1 extension was added), the resolved buffer memory layout isn't changed.

• D3D12DDI_VIDEO_ENCODER_OUTPUT_METADATA_0083_0 structure

• WrittenSubregionsCount elements of type D3D12DDI_VIDEO_ENCODER_FRAME_SUBREGION_METADATA_0083_0 indicating each tile in the same order they're written in the compressed output bitstream.

• The caller can infer the maximum size of the resolved metadata buffer by the H264/HEVC layout for ResolveEncoderOutputMetadata:

  • maxSliceNumber = D3D12DDI_VIDEO_ENCODER_RESOLUTION_SUPPORT_LIMITS_0080_2.MaxSubregionsNumber;
  • MaxEncoderOutputMetadataResolvedBufferSize = sizeof(D3D12DDI_VIDEO_ENCODER_OUTPUT_METADATA_0083_0) + (maxSliceNumber * sizeof(D3D12DDI_VIDEO_ENCODER_FRAME_SUBREGION_METADATA_0083_0))

ResolvedLayoutEncoderMetadataDrvBuffer for AV1 codec

For AV1 encoding (pInputArguments->EncoderCodec equal to D3D12DDI_VIDEO_ENCODER_CODEC_0095_AV1), which was added in Windows 11, version 24H2 (WDDM 3.2), the driver should use the added AV1-specific resolved buffer layout.

• The memory layout is defined as:

  • D3D12DDI_VIDEO_ENCODER_OUTPUT_METADATA_0083_0 structure

  • WrittenSubregionsCount elements of type D3D12DDI_VIDEO_ENCODER_FRAME_SUBREGION_METADATA_0083_0 indicating each tile in the same order they're written in the compressed output bitstream.

    • bSize = tile_size_minus_1 + 1 + bStartOffset
    • bStartOffset = Bytes to skip relative to this tile, the actual bitstream coded tile size is tile_size_minus_1 = (bSize - bStartOffset - 1).
    • bHeaderSize = 0
    • The i-th tile is read from the compressed_bitstream[offset] with offset = [sum j = (0, (i-1)){ tile[j].bSize }] + tile[i].bStartOffset
  • The D3D12DDI_VIDEO_ENCODER_AV1_PICTURE_CONTROL_SUBREGIONS_LAYOUT_DATA_TILES_0095 structure indicating the encoded frame tile grid structure
  • D3D12DDI_VIDEO_ENCODER_AV1_POST_ENCODE_VALUES_0095 structure indicating encoding metadata values that are only obtained post-execution of EncodeFrame on the GPU.

• The caller can infer the maximum size of the resolved metadata buffer by the layout for AV1 in ResolveEncoderOutputMetadata.

  • MaxTiles = D3D12DDI_VIDEO_ENCODER_RESOLUTION_SUPPORT_LIMITS_0080_2.MaxSubregionsNumber;
  • MaxEncoderOutputMetadataResolvedBufferSize = sizeof(D3D12DDI_VIDEO_ENCODER_OUTPUT_METADATA_0083_0) + (MaxTiles * sizeof(D3D12DDI_VIDEO_ENCODER_FRAME_SUBREGION_METADATA_0083_0));
• sizeof(D3D12DDI_VIDEO_ENCODER_AV1_PICTURE_CONTROL_SUBREGIONS_LAYOUT_DATA_TILES_0095)
• sizeof(D3D12DDI_VIDEO_ENCODER_AV1_POST_ENCODE_VALUES_0095)

See D3D12 video encoding for general information.

See also

PFND3D12DDI_VIDEO_ENCODE_FRAME_0082_0