Struct CarrierVideoStreamDecoder 

pub struct CarrierVideoStreamDecoder<C: FfmpegCarrier> { /* private fields */ }

mediadecode::VideoStreamDecoder impl with transparent HW → SW fallback.

Implementations

impl CarrierVideoStreamDecoder<View>

pub fn open( parameters: Parameters, time_base: Timebase, limits: DecoderLimits, ) -> Result<Self, Error>

Opens a video decoder for parameters, probing hardware backends in order and falling back to software.

open_as(.., DecodePath::Auto), which is what this has always done.

pub fn open_as( parameters: Parameters, time_base: Timebase, limits: DecoderLimits, path: DecodePath, ) -> Result<Self, Error>

Opens a video decoder on a named decode path.

DecodePath::Auto is open exactly; the other two arms pin the session to hardware or to software for its whole life. See DecodePath for what a pin promises and what it costs.

Everything else about the session is unchanged — the same VideoStreamDecoder face, the same frames, the same is_hardware / is_software readings. The choice is which decoder is behind them, which is what a determinism comparison and a deployment policy each need and neither could reach.

Errors

DecodePath::Hardware fails here when the named backend cannot be opened for the stream — where DecodePath::Auto would have gone on to software. DecodePath::Software fails only where libavcodec has no decoder for the stream, or the context cannot be built.

Examples

use mediadecode_ffmpeg::{DecodePath, DecoderLimits, FfmpegVideoStreamDecoder};
// The same stream, decoded without a GPU anywhere in the story.
let decoder = FfmpegVideoStreamDecoder::open_as(
  parameters,
  time_base,
  DecoderLimits::default(),
  DecodePath::Software,
)?;
assert!(decoder.is_software());

pub const fn is_software(&self) -> bool

Whether this decoder is currently running on software.

pub const fn is_hardware(&self) -> bool

Whether this decoder is currently running on hardware.

pub fn scaled_output_capability(&self) -> ScaledOutputCapability

Whether this session can currently emit pictures at a caller-requested output size. See ScaledOutputCapability and Self::request_scaled_output.

Supported on a live VideoToolbox session on an Apple target, Unsupported everywhere else — including on a session that has degraded to software, which is why this reads the session’s live state rather than a fact recorded once at open.

pub fn request_scaled_output( &mut self, size: (u32, u32), ) -> ScaledOutputCapability

Requests that this session emit pictures at size (width, height) from the next frame on, and reports whether the request was recorded. See VideoStreamDecoder::request_scaled_output for the full contract (never an error) and Self::scaled_output_capability’s documentation for which road can honor one, what a mid-stream request means, and the zero / upscale refusals this seat mints itself.

pub fn hardware_inner(&self) -> Option<&VideoDecoder>

The hardware wrapper, when one is in use.

pub const fn time_base(&self) -> Timebase

The stream timebase every produced timestamp is stamped with.

impl CarrierVideoStreamDecoder<Owned>

pub fn open( parameters: Parameters, time_base: Timebase, limits: DecoderLimits, ) -> Result<Self, Error>

Opens a video decoder for parameters, probing hardware backends in order and falling back to software.

open_as(.., DecodePath::Auto), which is what this has always done.

pub fn open_as( parameters: Parameters, time_base: Timebase, limits: DecoderLimits, path: DecodePath, ) -> Result<Self, Error>

Opens a video decoder on a named decode path.

DecodePath::Auto is open exactly; the other two arms pin the session to hardware or to software for its whole life. See DecodePath for what a pin promises and what it costs.

Everything else about the session is unchanged — the same VideoStreamDecoder face, the same frames, the same is_hardware / is_software readings. The choice is which decoder is behind them, which is what a determinism comparison and a deployment policy each need and neither could reach.

Errors

DecodePath::Hardware fails here when the named backend cannot be opened for the stream — where DecodePath::Auto would have gone on to software. DecodePath::Software fails only where libavcodec has no decoder for the stream, or the context cannot be built.

Examples

use mediadecode_ffmpeg::{DecodePath, DecoderLimits, FfmpegVideoStreamDecoder};
// The same stream, decoded without a GPU anywhere in the story.
let decoder = FfmpegVideoStreamDecoder::open_as(
  parameters,
  time_base,
  DecoderLimits::default(),
  DecodePath::Software,
)?;
assert!(decoder.is_software());

pub const fn is_software(&self) -> bool

Whether this decoder is currently running on software.

pub const fn is_hardware(&self) -> bool

Whether this decoder is currently running on hardware.

pub fn scaled_output_capability(&self) -> ScaledOutputCapability

Whether this session can currently emit pictures at a caller-requested output size. See ScaledOutputCapability and Self::request_scaled_output.

Supported on a live VideoToolbox session on an Apple target, Unsupported everywhere else — including on a session that has degraded to software, which is why this reads the session’s live state rather than a fact recorded once at open.

pub fn request_scaled_output( &mut self, size: (u32, u32), ) -> ScaledOutputCapability

Requests that this session emit pictures at size (width, height) from the next frame on, and reports whether the request was recorded. See VideoStreamDecoder::request_scaled_output for the full contract (never an error) and Self::scaled_output_capability’s documentation for which road can honor one, what a mid-stream request means, and the zero / upscale refusals this seat mints itself.

pub fn hardware_inner(&self) -> Option<&VideoDecoder>

The hardware wrapper, when one is in use.

pub const fn time_base(&self) -> Timebase

The stream timebase every produced timestamp is stamped with.

Trait Implementations

impl VideoStreamDecoder for CarrierVideoStreamDecoder<View>

type Adapter = Ffmpeg

Backend-specific vocabulary.

type Buffer = <View as FfmpegCarrier>::Buffer

Buffer type held by the packets and frames this decoder produces or accepts.

type Error = VideoDecodeError

Decoder-specific error type.

fn send_packet( &mut self, packet: &VideoPacket<VideoPacketExtra, Self::Buffer>, ) -> Result<Sent, Self::Error>

Submits one compressed packet.

fn receive_frame( &mut self, dst: &mut VideoFrame<PixelFormat, VideoFrameExtra, Self::Buffer>, ) -> Result<Received, Self::Error>

Drains one decoded frame into dst.

fn send_eof(&mut self) -> Result<Sent, Self::Error>

Signals end-of-stream.

fn flush(&mut self) -> Result<(), Self::Error>

Flushes internal state. Not a submission — nothing is offered, so there is nothing to be back-pressured.

fn scaled_output_capability(&self) -> ScaledOutputCapability

Whether this session can currently emit pictures at a caller- requested output size instead of full coded size. See ScaledOutputCapability for what a Self::Supported answer costs a caller who acts on it.

fn request_scaled_output(&mut self, size: (u32, u32)) -> ScaledOutputCapability

Requests that this session emit decoded pictures at size (width, height) from here on, and reports whether the request was honored.

impl VideoStreamDecoder for CarrierVideoStreamDecoder<Owned>

type Adapter = Ffmpeg

Backend-specific vocabulary.

type Buffer = <Owned as FfmpegCarrier>::Buffer

Buffer type held by the packets and frames this decoder produces or accepts.

type Error = VideoDecodeError

Decoder-specific error type.

fn send_packet( &mut self, packet: &VideoPacket<VideoPacketExtra, Self::Buffer>, ) -> Result<Sent, Self::Error>

Submits one compressed packet.

fn receive_frame( &mut self, dst: &mut VideoFrame<PixelFormat, VideoFrameExtra, Self::Buffer>, ) -> Result<Received, Self::Error>

Drains one decoded frame into dst.

fn send_eof(&mut self) -> Result<Sent, Self::Error>

Signals end-of-stream.

fn flush(&mut self) -> Result<(), Self::Error>

Flushes internal state. Not a submission — nothing is offered, so there is nothing to be back-pressured.

fn scaled_output_capability(&self) -> ScaledOutputCapability

Whether this session can currently emit pictures at a caller- requested output size instead of full coded size. See ScaledOutputCapability for what a Self::Supported answer costs a caller who acts on it.

fn request_scaled_output(&mut self, size: (u32, u32)) -> ScaledOutputCapability

Requests that this session emit decoded pictures at size (width, height) from here on, and reports whether the request was honored.