Arc/Mutex Performance

How slow is using Arc<Mutex<T>> in Rust? Is there any cheaper way to share data between threads? I'm writing a NES emulator and in my project I have several bools that need to be shared. Some of them are for checking if the threads should break the loop.

Note: I have three threads overall: main window, debugger window and emulator.

Benchmark Edit: Each locking of the <Arc<Mutex<bool>> takes ~201 ns while the rest of the loop takes (with a single <Arc<Mutex<CPU>> lock) ~418 ns. If I remove the CPU lock, it drops to ~90 ns.

Benchmark Edit 2: Those are the results I got after benchmarking (three runs of each):

Arc<AtomicBool>:

117 ns
167 ns
146 ns

Arc<Mutex<bool>>:

261 ns
219 ns
237 ns

Apparently, Arc<AtomicBool> is ~67% faster in this case (average).
After replacing all Arc<Mutex<bool>> by Arc<AtomicBool>, those are the results for a whole loop cycle:

Arc<AtomicBool>

383 ns
437 ns
332 ns

<Arc<Mutex<bool>>

465 ns
535 ns
416 ns

Arc<AtomicBool> is ~23% faster than Arc<Mutex<bool>> (average).
After that, I've replaced Arc<Mutex<u64>> by `` and those are the results:

Arc<Mutex<u64>>:

(single lock):

184 ns
94 ns
99 ns

(loop):

964 ns
579 ns
515 ns

Arc<RwLock<u64>>:

(single write):

171 ns
181 ns
201 ns

(write + loop):

430 ns
407 ns
423 ns

Even though RwLock<T> is slightly slower for isolated write() locks, it's ~63% faster in a whole loop cycle and has more stable timings (average).