Struct SoftmaxPlan 

pub struct SoftmaxPlan<T: Element, const N: usize> { /* private fields */ }

Softmax forward plan — see the module-level docs for formulas, dtypes, workspace, and precision guarantees.

T: Element is the element type (f32 / f64 / f16 / bf16). const N: usize is the tensor rank (1..=8).

Implementations

impl<T: Element, const N: usize> SoftmaxPlan<T, N>

pub fn select( _stream: &Stream, desc: &SoftmaxDescriptor<N>, _pref: PlanPreference, ) -> Result<Self>

Pick a kernel for desc. Validates softmax_axis < N, the dtype is in the wired FP family, and tensor rank ≤ 8. Returns Error::Unsupported for cells outside the matrix and Error::InvalidProblem for malformed shapes / axes.

pub fn can_implement(&self, args: &SoftmaxArgs<'_, T, N>) -> Result<()>

Validate args.

pub fn workspace_size(&self) -> usize

Workspace size in bytes. Always zero — the kernel does its two-pass scan in registers.

pub fn sku(&self) -> KernelSku

Identity of the kernel this plan picked (for telemetry + autotuner cache keying).

pub fn precision_guarantee(&self) -> PrecisionGuarantee

Numerical guarantees for this plan’s kernel — deterministic, bit-stable on the same hardware, f32 accumulator for the FP-detour half / bf16 inputs and f32 / f64 native for those dtypes.

pub fn run( &self, stream: &Stream, _workspace: Workspace<'_>, args: SoftmaxArgs<'_, T, N>, ) -> Result<()>

Launch the kernel against args. Calls can_implement first; returns Ok(()) for empty tensors.