Parallel write to array with an indices array

user_12698762 0
Viewed 879

I'm having trouble understanding the concurrency model in Rust coming from C++.

My array is to be accessed concurrently using another array that defines the indices. For example (Pseudocode):

let indices = [1, 2, 3, 4, 1, 2, 3, 2, 1, 1, 3, 2, 2];
let mut arr = [1, 2, 3, 4, 5, 6, 7, 8, 10];

indices.iter_par().for_each(|x| {
    arr[x] += x;
});

In C++, I would protect each index in arr with a lock or use atomic access. How could I do the same in Rust?

EDIT

I have another related question.

How could I pass a normal array as mutable into the parallel iterator, where I'm sure that no race conditions can occur?

let indices = [1, 2, 3, 4, 5, 6, 7, 8];
let mut arr = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10];

indices.iter_par().for_each(|x| {
    arr[x] = some_function(x);
});
rust parallel.foreach concurrency arrays
1 Answers

I don't know what the point of performing this operation in parallel is if you need lock for each item, but you can achieve this using a Mutex around the array to mutate:

use rayon::prelude::*;
use std::sync::Mutex;

fn main() {
    let indices = [1, 2, 3, 4, 1, 2, 3, 2, 1, 1, 3, 2, 2];
    let arr = Mutex::new([1, 2, 3, 4, 5, 6, 7, 8, 10]);

    indices.par_iter().for_each(|&x| {
        let mut arr = arr.lock().unwrap();
        arr[x] += x;
    });
}

playground

EDIT

Based on the comment, you can have each element be atomic:

use rayon::prelude::*;
use std::sync::atomic::{AtomicUsize, Ordering};

fn main() {
    let indices = [1, 2, 3, 4, 1, 2, 3, 2, 1, 1, 3, 2, 2];
    let arr = [1, 2, 3, 4, 5, 6, 7, 8, 10]
        .iter()
        .map(|&n| AtomicUsize::new(n))
        .collect::<Vec<_>>();

    indices.par_iter().for_each(|&x| {
        arr[x].fetch_add(x, Ordering::SeqCst);
    });
}
user_63484980
Related