Struct Arm 

pub(super) struct Arm {

    profile: VAProfile,
    tools: Tools,
    level_idc: u32,
    tier: u32,
    ctb_width: u32,
    ctb_height: u32,
    slices: (u32, u32),
    dpb: u32,
    pic_init_qp: u32,
    cu_qp_delta: bool,
    cu_qp_delta_depth: u32,
    align: u32,
    low_delay_b: bool,
    sequence: VAEncSequenceParameterBufferHEVC,
    headers: Vec<u8>,
}

Fields

profile: VAProfile

tools: Tools

level_idc: u32

tier: u32

ctb_width: u32

ctb_height: u32

slices: (u32, u32)

dpb: u32

pic_init_qp: u32

cu_qp_delta: bool

cu_qp_delta_depth: u32

The depth of the coding units that each carry a quantizer delta, 0 for one per tree block.

align: u32

The surface alignment, the smallest coding block but where new says otherwise.

low_delay_b: bool

Whether a predicted frame is coded as a B slice with the one reference in both lists.

sequence: VAEncSequenceParameterBufferHEVC

headers: Vec<u8>

Implementations

impl Arm

pub(super) fn new( profile: VAProfile, features: Option<u32>, block_sizes: Option<u32>, prediction: Option<u32>, unit_qp_deltas: bool, ) -> Self

The arm on profile, with the driver’s feature and block-size attributes where it reports them. Intel’s full encoder (unit_qp_deltas), which reports no block sizes, codes blocks from 8x8 to 32x32 (Skylake’s, as FFmpeg’s defaults say) on surfaces aligned to 16.

pub(super) fn alignment(&self) -> (u32, u32)

The surface alignment: the smallest coding block, or 16 on Intel’s full encoder.

pub(super) fn ctu_size(&self) -> u32

pub(super) fn slices(&self) -> (u32, u32)

pub(super) fn level(&self) -> u32

The general_level_idc the sequence names.

fn profile_idc(n: &Negotiated) -> u32

The profile a session declares: the range extensions for 4:4:4, else Main 10 or Main.

fn profile_tier_level(&self, w: &mut BitWriter, n: &Negotiated)

pub(super) fn configure( &mut self, n: &Negotiated, surface_width: u32, surface_height: u32, slices: (u32, u32), )

The stream’s sequence: level and tier from the ladder, the level never below the one the decoded picture buffer was sized for, the picture in coding tree blocks, and the VPS, SPS, and PPS the key frames carry.

pub(super) fn sequence(&mut self, n: &Negotiated, out: &mut Buffers)

pub(super) fn picture( &mut self, n: &Negotiated, frame: &Frame<'_>, out: &mut Buffers, ) -> Result<(), String>

The picture: its order count since the key frame, the frames the client still has, newest first, as the driver’s reference list and the slice header’s reference picture set, and every slice.