Enum CmrKernel 

pub enum CmrKernel {
    Rank,
    Bitonic,
    Vectorized {
        width: usize,
    },
}

How execute_common_median finds each sample’s median over channels. All are exact and give the same output; which is faster depends on the device, so the autotuner picks.

Variants

Rank

Rank counting ([common_median_kernel]): each value compared with every channel, no barriers.

Bitonic

Bitonic sort in shared memory ([common_median_bitonic_kernel]): about log₂² C / 2 steps per sample, one barrier each.

Vectorized

Rank counting on width consecutive samples per unit ([common_median_vector_kernel]; SIMD on the CPU runtime). Runs the widest width up to width that the runtime offers (else Rank); when it does not divide the sample count, [common_median_vector_unaligned_kernel] (lane by lane loads and stores).

Fields

width: usize

Trait Implementations

impl Clone for CmrKernel

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for CmrKernel

impl Debug for CmrKernel

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for CmrKernel

impl PartialEq for CmrKernel

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for CmrKernel