On 12/22/08, Antoine Brodin <antoine_at_freebsd.org> wrote: > On Mon, Dec 22, 2008 at 4:29 PM, Norikatsu Shigemura <nork_at_freebsd.org> > wrote: >> I think that performance improvement is significant(I don't >> think performance improved by my patch). But my patch is the >> first implementation of kqueue/kevent support for TTY. >> In fact, I could find a bug. I hope to commit my patch. But >> by performance reason, my patch is not accepted, I have no >> problem. > > Hi, > > Could you make it a compile time option, similar to the USE_KQUEUE > option in make(1) ? > I had to write something similar to script(1) for work recently, that > ran on linux, and the freebsd sources were quite useful. > > Cheers, > > Antoine Would partitioning out the select(2) code and #ifdef'ing it to mux between the kevent(2) and select(2) interfaces be a prudent measure to follow? The select(2) interface is essentially fixed now, so unless POSIX changes that section of code shouldn't change. Any and all advancements forward could be using the kevent(2) method, as long as it's time permitting and the win / cost ratio is sufficiently high enough to warrant it. The fun part will be regression testing -_-. -GarrettReceived on Mon Dec 22 2008 - 19:15:45 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:39 UTC