Struct UnsortedSegmentProdPlan 

pub struct UnsortedSegmentProdPlan<T: Element> { /* private fields */ }

unsorted_segment_prod plan. Phase 25.

out[s, d] = Π input[n, d] over n : segment_ids[n] == s (any order). atomicMul-emulated CAS retry loop.

When to use: forward unsorted segment-product. For sorted IDs prefer the deterministic SegmentProdPlan. BW shares the SegmentProdBackwardPlan shape.

Dtypes: {f32, f64} only — atomicCAS slot widths are 32 / 64 bit. Empty segments emit 1.0 (multiplicative identity).

Workspace: none.

Precision guarantee: non-deterministic — atomic ordering varies across launches.

Implementations

impl<T: Element> UnsortedSegmentProdPlan<T>

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

Pick a kernel.

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

Validate args.

pub fn workspace_size(&self) -> usize

Workspace size — zero.

pub fn sku(&self) -> KernelSku

Identity of the kernel.

pub fn precision_guarantee(&self) -> PrecisionGuarantee

Numerical guarantees.

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

Launch.