Re: Performance degradation in commit ac1d794

Andres Freund <andres(at)anarazel(dot)de> writes:
> 0004 currently contains one debatable optimization, which I'd like to
> discuss: Currently the 'sock' passed to WaitLatchOrSocket is not
> removed/added to the epoll fd, if it's the numerically same as in the
> last call. That's good for performance, but would be wrong if the socket
> were close and a new one with the same value would be waited on. I
> think a big warning sign somewhere is sufficient to deal with that
> problem - it's not something we're likely to start doing. And even if
> it's done at some point, we can just offer an API to reset the last used
> socket fd.

Perhaps a cleaner API solution would be to remove the socket argument per
se from the function altogether, instead providing a separate
SetSocketToWaitOn() call.

(Also, if there is a need for it, we could provide a function that still
takes a socket argument, with the understanding that it's to be used for
short-lived sockets where you don't want to change the process's main
epoll state.)

regards, tom lane