bimodal_array

bimodal_array provides phase-separated, dual-mode access to a contiguous array.

It enables a scatter–gather access pattern where:

• Independent workers can update elements concurrently.
• A coordinator can later acquire exclusive access to the entire array.

---

Access Model

Access is granted through two handle types:

• ElementHandle<T> — grants mutable access to a single element.
• ArrayHandle<T> — grants exclusive mutable access to the entire array.

Handles represent the ability to attempt a lock. The actual lock is held by guards:

• ElementHandle::lock() returns an ElementGuard<T>
• ArrayHandle::lock() returns an ArrayGuard<T>

Locking Rules

• Guards for different elements may be held concurrently.
• An ArrayGuard<T> provides exclusive access to the entire array.
• An ArrayGuard<T> cannot coexist with any ElementGuard<T>.
• Lock acquisition is non-blocking.

Internally, lock state is coordinated via a single AtomicUsize. As a result, bimodal_array may not scale well under heavy contention.

---

Example

use bimodal_array::bimodal_array;
use std::thread;

const WORKER_COUNT: usize = 8;

let data = vec![0usize; WORKER_COUNT];
let (mut array_handle, element_handles) = bimodal_array(data);

// Scatter phase
let threads: Vec<_> = element_handles
    .into_iter()
    .enumerate()
    .map(|(idx, mut element)| {
        thread::spawn(move || {
            let mut guard = element.lock().unwrap();
            *guard = 10 - idx;
        })
    })
    .collect();

for t in threads {
    t.join().unwrap();
}

// Gather phase
let mut array_guard = array_handle.lock().unwrap();
array_guard.sort();