Parallel computation in WASM

Computation: Local Search for TSP

Randomly create tours and locally optimize with two-opt Iterations times, and return the best tour.

Frontend

Vanilla Vite + TypeScript UI.

// random-number-generator to construct initial random tours
let mut rng = SmallRng::seed_from_u64(seed);

// we will construct & improve `iterations` tours
(0..iterations)

    // `create_tour`constructs a random tour and locally-optimizes within 2-opt
    .map(|_| create_tour(&mut rng, locations))

    // among all created tours, we pick the one with minimum distance
    .min_by_key(|x| OrderedFloat::from(x.distance))

let mut rng = SmallRng::seed_from_u64(seed);
(0..iterations)
    .map(|_| create_tour(&mut rng, locations))
    .min_by_key(|x| OrderedFloat::from(x.distance))

// we will construct & improve `iterations` tours
(0..iterations)

    // convert the iterator into parallel iterator
    .into_par()

    // `use_new` enables mutable variables in parallel computations
    // each thread will have its own random number generator
    // `t` here is the thread index, with value in (0..num_threads)
    .use_new(|t| SmallRng::seed_from_u64(seed + t as u64))

    // `create_tour`constructs a random tour and locally-optimizes within 2-opt
    .map(|rng, _| create_tour(rng, locations))

    // among all created tours, we pick the one with minimum distance
    .min_by_key(|x| OrderedFloat::from(x.distance))

(0..iterations)
    .into_par()
    .use_new(|t| SmallRng::seed_from_u64(seed + t as u64))
    .map(|rng, _| create_tour(rng, locations))
    .min_by_key(|x| OrderedFloat::from(x.distance))