Module alias 

Vose’s alias method for O(1) sampling from discrete distributions.

Given N outcomes with associated weights, an alias table pre-computes a structure that allows selecting an outcome in constant time from a single uniform u64 input.

Two variants:

• AliasTable<T>: generic, works with any Clone outcome type.
• AliasTableU64: flat parallel arrays, optimized for the Phase 2 compiled kernel path where outcomes are indices 0..N.

Structs

AliasSample

Sample an outcome index from a pre-built alias table over weights. The input is a uniform u64 (hash upstream for pseudo-random dispersion); the output is the chosen outcome index.

AliasTable

Generic alias table for O(1) weighted sampling.

AliasTableU64

Alias table optimized for u64 outcomes (indices 0..N).