Why is Rust WASM slower than JavaScript in calculating prime numbers

Question

I wrote a very rudimentary piece of code to calculate if a number is prime in Rust (compiled to WASM) and in JavaScript to benchmark the arithmetic performance.

I was fully expecting Rust/WASM to blow away JavaScript. In all other arithmetic benchmarks I've done Rust/WASM seems to have the edge over JavaScript or at least match it. However in this test, JavaScript seems to heavily outperform WASM and I don't really have an explanation to why that is.

Rust Code:

pub fn calculate_is_prime(number: u64) -> bool {
    if number == 1 {
        return false;
    }
    if number == 2 {
        return true;
    }
    for i in 2..number {
        if number % i == 0 {
            return false;
        }
    }
    return true;
}

#[wasm_bindgen]
pub fn bench_rs(max: u64) -> u64 {
    (1..=max).map(|n| calculate_is_prime_rs(n) as u64).sum()
}

JavaScript code:

function calculateIsPrime(number) {
    if (number === 1) {
        return false;
    }
    if (number === 2) {
        return true;
    }
    for (let i = 2; i < number; i++) {
        if (number % i === 0) {
            return false;
        }
    }
    return true;
}

function bench_js(max) {
    let tot = 0;
    for (let n = 1; n <= max; n++) {
      tot += calculateIsPrime(n);
    }
    return tot;
}

let max = 200000;
console.log(`Amount of primes under ${max} is ${bench_js(max)}`);

Basic sample project: https://github.com/Mcluky/Stack-Overflow-Rust-Wasm-Performance-Example

Things I've already checked/done:

• I made sure to always set the --release flag while building the rust code.
• Ran the Rust code directly on my machine which is a lot faster than both JS and WASM so I suspect it has something to do with the WASM target.
• Replaced the integer type (u64) with a floating point (f64) in Rust since JavaScript is using floats but I got similar results.
• Tested over iterations to make sure it's consistent.
• Tried while instead of for-in in the Rust version in case it wasn't as optimized as you'd think.

Answer 1

I can not reproduce your results on a Ryzen Threadripper 2950x on Windows 10. I added the following functions:

#[wasm_bindgen]
pub fn bench_rs(max: u64) -> u64 {
    (1..=max).map(|n| calculate_is_prime_rs(n) as u64).sum()
}

function bench_js(max) {
    let tot = 0;
    for (let n = 1; n <= max; n++) {
        tot += calculateIsPrime(n);
    }
    return tot;
}

I then compiled with wasm-pack build --release --target web and evaluated in both Google Chrome:

> console.time("rs"); console.log(bench_rs(BigInt(200000))); console.timeEnd("rs");
17984n
rs: 6015.033935546875 ms

> console.time("js"); console.log(bench_js(200000)); console.timeEnd("js");
17984
js: 6017.426025390625 ms

And in Firefox:

> console.time("rs"); console.log(bench_rs(BigInt(200000))); console.timeEnd("rs");
17984n
rs: 6076ms - timer ended

> console.time("js"); console.log(bench_js(200000)); console.timeEnd("js");
17984
js: 6074ms - timer ended