Struct ParallelFeasibilityChecker 

pub struct ParallelFeasibilityChecker { /* private fields */ }

Parallel batch feasibility checker.

Checks N points against M constraints simultaneously using rayon. The outermost parallelism is over points (each point is independent); within a point, constraints are checked sequentially for cache efficiency.

Implementations

impl ParallelFeasibilityChecker

pub fn new(config: ParallelConfig) -> Self

Create a new checker with the given configuration.

pub fn add_constraint(&mut self, constraint: Box<dyn FastConstraint>)

Add a constraint to the checker.

pub fn num_constraints(&self) -> usize

Number of registered constraints.

pub fn check_batch(&self, points: &Array2<f32>) -> Vec<bool>

Check feasibility of a batch of points.

points must be a (num_points, dim) matrix in row-major order. Returns one bool per point indicating whether all constraints are satisfied.

pub fn violation_matrix(&self, points: &Array2<f32>) -> Array2<f32>

Compute violation for each point against all constraints.

Returns a (num_points, num_constraints) matrix where entry [i, j] is the violation of point i against constraint j (0 if feasible).

pub fn project_batch( &self, points: &Array2<f32>, max_iter: usize, ) -> Array2<f32>

Project all points to the constraint-satisfying region.

Uses Dykstra’s alternating projections algorithm in parallel — each point is independent, so rayon parallelism is embarrassingly parallel here.