I don't see how that can be used to make performant code though. I usually want to make my locked regions as small as possible to avoid how long other threads are blocked for.

So this is solving the problem of reentrant deadlocks, with a non compile-time enforceable exclusion mechanism in the form of a threadkey.

It's quite different from solving deadlocks in general, and mostly moves the problem to getting the thread key. It's an interesting thought experiment, but unless I'm missing something else, it has no practical purpose over a try_lock.

But anyway, thanks for taking the time to explain, I should have read more carefully