Well, we take a lock less but I can't say if it'll be measureable. The
test program above is not a performance thing though, just some code to
verify multiple threads doing waits on the same epoll fd.
OK, so I ported your patch to 2.4 (+epoll-lt-0.22) because I have some
code using it right there. At first, I thought I was observing measuring
errors, but after about 6 reboots, I seem to observe a consistent 6.5%
increase in the number of sessions/s on my fake web server on my dual
athlon. It jumps from 14350 hits/s with epoll-lt-0.22 alone to 15300 with
your patch. It seems much to me, but I'm sure I'm not dreaming (yet).
I'll send you (offlist) an update to 2.4-epoll-lt-0.22 which incorporates
this patch.