Fabian Groffen
Fabian Groffen
this is with libevent right?
I am a bit hesitant to take the patch(es) because I don't understand them. Mostly don't get why it performs better. That is, I see there is more than just...
> Current dispatcher code have some problems: > > 1. read from all sockets (without poll) > > 2. full scan connections table when new connections established > > 3....
No, I don't have plans to merge this. I switched from a micro-sleep spin-lock approach to a semaphore-based approach (basically uses notifications), this doesn't require an external dep (libevent). I...
poll() is already used to check which socket has work to do, the semaphore is used to wake up the worker threads.
right, could look at using epoll()
this is a good point, I wasn't aware of that
currently not, this would require buffering (like aggregations), but it is an interesting use-case.
It seems to suggest the sleeps should be longer. The sleeps are a vital part in avoiding contention/thundering hurd behaviour, they are not used to synchronise or something.
Because poll is now used to see if there's something to read, the read() cost should go down. A semaphore is now used with a timeout, this is where the...