Jon Gjengset

Given that there aren't any locks, `CondVar` won't work. This is also the reason why we need to use `park_timeout` instead of just `park` (though note that we _are_ using...

@hutch can you give it a try with the new `SpinWait` now in place?

@brunoqc Try 1.1.1?

Well, it depends what you're doing really. If you're sending and receiving in a tight loop, I'd expect it to hit 100%. If you're doing other compute-heavy work, I'd expect...

@xobs I'm confused -- the default parking time in `bus` now is 100µs (`thread::park_timeout(Duration::new(0, 100000))`; 100000ns = 100µs). Your diff (https://github.com/xobs/bus/commit/c8bb287df261cf038ac431a71b416ae996325d1c) is using 100000000ns = 100000µs = 100ms.

So, the reason I'm hesitant to raise the timeout is that, as [the comment above](https://github.com/jonhoo/bus/blob/master/src/lib.rs#L396) alludes to, the timeout dictates how long after a free slot opens up a thread...

It's a great suggestion, though I'd say [`zipf`](https://docs.rs/zipf/) is probably a better choice for benchmarking as it's both stable and very fast. Would be happy to take a look at...

> I'll work on it when I find the time :) Couldn't ask for anything more :) > The zipf create requires rand 0.8 which changed the `SmallRng` and its...

The challenge with Zipf is that the chance of duplicate keys is actually very high :sweat_smile: In reality, these checks aren't there so much to see that the backing implementation,...

I worry that, especially once you introduce key skew, you'd end up deviating so far from the target that the results no longer measure what the mix is intended to...