Il 23/08/2012 14:10, Jan Kiszka ha scritto: >> Can you expand on this? > > Well, this patch removes an indirection from timer event deliveries. So > it reduces overhead, though only noticeable if you have high-rate timers.
Actually, timers (and bottom halves) are always run after iohandlers. So the qemu_notify_event should already be completely useless for Unix, even if we leave the host_alarm_handler indirection. But this leaves out Windows, where your next task of (IIUC) having multiple instances of struct qemu_alarm_timer would be complicated by the qemu_notify_event. I guess this is the original reason for your patch. So, in order to remove the qemu_notify_event completely, what about not using signals anymore for timers? You could just tweak the select timeout and drop all the -clock madness. Zero syscalls, practically no overhead. If this is not precise enough, use timerfd on Linux only (BTW, switching to an absolute deadline would be useful too). >> The Win32 bits look fine, but it's a bit scary to make the Unix and >> Win32 paths so different. It works well until we have a BQL for timers, >> but would this complicate shrinking the scope of the BQL? > > Nope, not yet. We continue to hold the BQL across qemu_run_all_timers. > Under Unix, this happens as qemu_iohandler_poll->sigfd_handler-> > host_alarm_handler runs under BQL, under win32 due to explicit locking. > The plan is to only pull specifically marked alarm timers out of this > standard path, in the future. Ok, thanks for clarifying. Paolo
