[issue8299] Improve GIL in 2.7

[Kristján Valur Jónsson]

Kristján Valur Jónsson <kristjan at ccpgames.com> added the comment:

Martin, I´ve explained it in my other dissue, issue 8411, with a step by step example.
It is unfair because a thread can _bypass_ the condition variable.  A thread just woken up from the condition variable has to race to get the lock, and it is a race that it will invariably loose if the other thread is doing a release/acquire (yielding the GIL as happens in ceval.py)  The ConditionVariable can only endow the lock with its fairness property if all the threads play by the same rules.

This was a design decision made by Tim (according to the comment) but a misguided one.  It is a good stragegy for resources that are held for a short time to avoid lock convoying, but not appropriate in this case.

You also don't have to take my word for it.  Just try it out.  Notice that 99% of all yields between threads fail, causing starvation of a thread which is the definition of "unfairness."

Anyway, this is the last time I explain why the "emulated" semaphore is unfair.  I think I´ve done so on at least four or five different occasions and it would be helpful if people would actually bother to read my comments.

----------

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue8299>
_______________________________________