Struct LcpDispatch 

pub struct LcpDispatch { /* private fields */ }

A function-pointer dispatch for byte-level LCP. The architecture- specific pointer is selected once at construction by feature detection; later calls reduce to a register-resident indirect call.

LcpDispatch is Copy, Send, and Sync (a function pointer is all three), so it threads freely through rayon boundaries.

The same byte-level function backs every symbol width: the Self::lcp method casts &[S] to a byte view, calls the function with byte-scale offsets, and divides the result by size_of::<S>() to recover the symbol-level LCP.

Implementations

impl LcpDispatch

pub fn detect() -> Self

Detect the best LCP implementation for this CPU. Cheap (a couple of is_*_feature_detected! checks) but does still touch the feature-detection cache, so call it once per top-level build.

pub fn scalar() -> Self

Forced scalar dispatch — useful for tests and for clients that want a deterministic baseline.

pub fn lcp<S: Symbol>( &self, text: &[S], p: usize, q: usize, max_ctx: usize, ) -> usize

Longest common prefix of text[p..] and text[q..] in symbols, bounded by max_ctx. For any S: Symbol of non-zero size, this dispatches to the byte-level SIMD path with byte-scaled offsets and returns byte_lcp / size_of::<S>().

pub fn suffix_cmp<S: Symbol>( &self, text: &[S], p: usize, q: usize, max_ctx: usize, ) -> Ordering

Total order on two suffixes of text. Uses Self::lcp for the shared prefix, then resolves the first differing symbol or — if both suffixes are exhausted within max_ctx — orders by remaining length (shorter is smaller, the convention SAIS and CaPS-SA use).

Zero-cost wrapper around Self::suffix_cmp_with for the non-segmented case.

pub fn suffix_cmp_with<S: Symbol, L: LimitProvider>( &self, text: &[S], lp: &L, p: usize, q: usize, max_ctx: usize, ) -> Ordering

Like Self::suffix_cmp but takes a LimitProvider so the suffix lengths used for the LCP-cap and the boundary-tie-break come from a segmented view of the text. With PlainText this matches Self::suffix_cmp exactly.

The boundary tie-break (when both suffixes hit their limit before any byte differs) is delegated to LimitProvider::boundary_order; the default “shorter-is-smaller” gives the generalised-SA convention, custom impls can flip it (e.g. STAR’s spacer-as-largest).

Trait Implementations

impl Clone for LcpDispatch

fn clone(&self) -> LcpDispatch

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for LcpDispatch