Struct WorkgroupTuner 

pub struct WorkgroupTuner {
    pub raster_2d: (u32, u32),
    pub reduction: u32,
    pub fft: u32,
}

Optimal workgroup sizes derived from GPU adapter limits.

Workgroup sizes are clamped to the adapter’s reported maximums, so kernels compile and dispatch correctly on every device.

Derivation

Call WorkgroupTuner::derive_from_limits after obtaining the adapter to get device-specific sizes. Use WorkgroupTuner::unlimited (or Default) when no real device is available (e.g. in unit tests).

Fields

raster_2d: (u32, u32)

2D workgroup size for raster/image kernels (e.g. 16×16 or 8×8).

reduction: u32

1D workgroup size for reduction kernels (e.g. 256 or 64).

fft: u32

1D workgroup size for FFT kernels (e.g. 64 or 32).

Implementations

impl WorkgroupTuner

pub fn derive_from_limits(limits: &Limits) -> Self

Derive optimal workgroup sizes from the adapter’s Limits.

The function tries preferred sizes in descending order of quality and falls back to smaller values when the adapter cannot accommodate the larger ones.

Class	Preferred	Fallback 1	Fallback 2
raster_2d	16 × 16	8 × 8	4 × 4
reduction	256	—	—
fft	64	—	—

reduction and fft are clamped to min(preferred, max_x, max_total).

pub fn unlimited() -> Self

Default tuner assuming unbounded adapter limits (production default).

Equivalent to calling derive_from_limits with Limits::default(), which reports max_compute_workgroup_size_x = 256, max_compute_workgroup_size_y = 256, and max_compute_invocations_per_workgroup = 256.

Trait Implementations

impl Clone for WorkgroupTuner

fn clone(&self) -> WorkgroupTuner

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for WorkgroupTuner

impl Debug for WorkgroupTuner

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

Formats the value using the given formatter.

impl Default for WorkgroupTuner

fn default() -> Self

Returns the “default value” for a type.

impl Eq for WorkgroupTuner

impl PartialEq for WorkgroupTuner

fn eq(&self, other: &WorkgroupTuner) -> 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 StructuralPartialEq for WorkgroupTuner