12.01.2017 12:10, Alex Peshkoff wrote: > On 01/12/17 13:00, Vlad Khorsun wrote: >> 12.01.2017 10:24, Alex Peshkoff wrote: >>> On 01/11/17 23:35, Vlad Khorsun wrote: >>> >>>>>> I do it easily. AST thread is blocked at >>>>>> Database::Sync::lock on syncMutex.enter() despite of a lot of checkouts >>>>>> in worker >>>>>> thread. >>>>> Hmm... From posix POV that's a bug but certainly windows calls may have >>>>> different rules. May be it makes sense to increase thread's priority >>>>> when it's going to deliver AST? >>>> I did it in AstContextHolder - not helps. I see blocked thread with >>>> highest priority... >>>> I plan to try to change syncMutex by Event or Mutex (currently it is >>>> Critical Section) >>>> and see if it helps >> Mutex seems to fix the issue > > May be Critical Section does not promise fair scheduling?
Looks like no sync objects promise fair scheduling on Windows https://msdn.microsoft.com/library/windows/desktop/ms682530.aspx Critical Section Objects ... Starting with Windows Server 2003 with Service Pack 1 (SP1), threads waiting on a critical section do not acquire the critical section on a first-come, first-serve basis. This change increases performance significantly for most code. However, some applications depend on first-in, first-out (FIFO) ordering and may perform poorly or not at all on current versions of Windows (for example, applications that have been using critical sections as a rate-limiter). ... Windows Server 2003 and Windows XP: Threads that are waiting on a critical section are added to a wait queue; they are woken and generally acquire the critical section in the order in which they were added to the queue. However, if threads are added to this queue at a fast enough rate, performance can be degraded because of the time it takes to awaken each waiting thread. https://msdn.microsoft.com/library/windows/desktop/ms684266.aspx Mutex Objects ... If more than one thread is waiting on a mutex, a waiting thread is selected. Do not assume a first-in, first-out (FIFO) order. External events such as kernel-mode APCs can change the wait order. http://joeduffyblog.com/2006/12/14/anticonvoy-locks-in-windows-server-2003-sp1-and-windows-vista/ ... If you remove the strict fairness policy, however, the system scales. And that, my friends, is why all of the locks in Windows are now unfair. Regards, Vlad ------------------------------------------------------------------------------ Developer Access Program for Intel Xeon Phi Processors Access to Intel Xeon Phi processor-based developer platforms. With one year of Intel Parallel Studio XE. Training and support from Colfax. Order your platform today. http://sdm.link/xeonphi Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
