Enum CodeGenDecision 

pub enum CodeGenDecision {
    UnrollLoop {
        factor: u32,
    },
    PredicateBranch,
    PrefetchMemory {
        distance: u32,
    },
    IncreaseOccupancy {
        target_blocks: u32,
    },
    UseLargerTiles {
        tile_m: u32,
        tile_n: u32,
    },
    SwitchToSharedMemory,
    EnableSplitK {
        k_slices: u32,
    },
}

A concrete optimisation decision derived from profiling data.

Variants

UnrollLoop

Unroll a hot loop by the given factor.

Fields

factor: u32

Number of iterations to unroll.

PredicateBranch

Convert a heavily biased branch to a predicated instruction.

PrefetchMemory

Insert prefetch instructions at the given distance (in iterations).

Fields

distance: u32

Prefetch lookahead distance.

IncreaseOccupancy

Increase occupancy by targeting the given number of blocks per SM.

Fields

target_blocks: u32

Desired concurrent blocks per SM.

UseLargerTiles

Use larger tile dimensions for a compute-bound GEMM.

Fields

tile_m: u32

Tile size in the M dimension.

tile_n: u32

Tile size in the N dimension.

SwitchToSharedMemory

Promote global memory loads to shared memory.

EnableSplitK

Enable split-K parallelism.

Fields

k_slices: u32

Number of K-dimension slices.

Trait Implementations

impl Clone for CodeGenDecision

fn clone(&self) -> CodeGenDecision

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CodeGenDecision

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

Formats the value using the given formatter.

impl Display for CodeGenDecision

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

Formats the value using the given formatter.

impl PartialEq for CodeGenDecision

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

Tests for self and other values to be equal, and is used by ==.

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

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Eq for CodeGenDecision

impl StructuralPartialEq for CodeGenDecision