Looking for help understanding Rust's performance vs C++

community
4

You can "hoist" bounds checks out of the loop by using slices, as long as LLVM can follow that the slice len and index variable are related:

let len = vec.len();
let slice = &vec[0..len];
for i in 0..len {
    slice[i] // no bounds check
}

And unsafe {get_unchecked()} is fine. You can (with care) build a safe matrix multiplication function on top of it, so the risk doesn't spread.