Struct ArgReducePlan 

pub struct ArgReducePlan<T: Element, const N: usize, I: IndexOutputElement = i64> { /* private fields */ }

Arg-reduce plan (argmax / argmin) — see module docs for dtypes, tie-breaking, and precision.

T: Element is the value (input) dtype; I: IndexOutputElement is the output index dtype (defaults to i64). const N: usize is the tensor rank (1..=8).

The I = i64 default preserves source-compat for pre-Phase-12.2 callers; new callers opt into narrower output dtypes via ArgReducePlan::<T, N, u32>::select(...) or <T, N, i32>.

Implementations

impl<T: Element, const N: usize, I: IndexOutputElement> ArgReducePlan<T, N, I>

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

Pick a kernel for desc.

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

Validate args.

pub fn workspace_size(&self) -> usize

Workspace size in bytes.

pub fn sku(&self) -> KernelSku

Identity of the kernel this plan picked.

pub fn precision_guarantee(&self) -> PrecisionGuarantee

Numerical guarantees.

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

Launch.