> On Dec 15, 2015, at 6:01 PM, Greg Parker via swift-dev <swift-dev@swift.org> > wrote: >> On Dec 15, 2015, at 5:32 PM, Kevin Ballard <ke...@sb.org> wrote: >> >>> On Tue, Dec 15, 2015, at 01:38 PM, Greg Parker via swift-dev wrote: >>> >>> Another solution is to use a handoff lock algorithm. This is what libobjc >>> does now. The lock owner stores its thread ID in the lock. Each lock waiter >>> yields to the owner thread specifically, donating its priority and >>> resolving the inversion. This scheme has theoretical holes when multiple >>> locks are involved, but in practice we haven't seen any problems. >>> >>> As far as I know the spinlock that libobjc now uses is not API. >> >> It's not. I checked earlier today against the copy of the objc runtime I >> have (from opensource.apple.com; my copy is slightly out of date, at version >> 647, but it's new enough). The runtime defines a type spinlock_t that's >> backed by something called os_lock_handoff_s. It also imports a header >> <os/lock_private.h>, which is presumably where this type is defined. I >> already updated my radar to suggest that perhaps this type would be a good >> candidate to expose. >> >> Incidentally, if I wanted to implement my own spinlock like this, how can >> you yield to a specific thread? I'm only aware of pthread_yield_np(), which >> doesn't take a target thread. > > It uses Mach's thread_switch() and passes a Mach thread port. It also uses a > private option flag; I don't know if any of the public ones are good enough > to solve the problem.
So, just to complete the loop here: absent Darwin granting public and backwards-compatible access to an internal API, we need to write this in a way that falls back on using a heavyweight lock in the presence of contention. I’m fine with that being a global lock. I’m also fine with using a spin lock on other platforms. Note that Darwin platforms need this to interoperate with the unknownWeak entrypoints. John. _______________________________________________ swift-dev mailing list swift-dev@swift.org https://lists.swift.org/mailman/listinfo/swift-dev
