On Thu, Nov 11, 2010 at 8:41 AM, Paolo Bonzini <pbonz...@redhat.com> wrote: > On 11/10/2010 02:45 PM, Stefan Hajnoczi wrote: >> >> I wonder if the condition variable has a measurable performance >> overhead. We unconditionally broadcast on paiocb completion. One >> idea would be to keep a counter of waiters (should only ever be 0 or >> 1) protected by aiocb_mutex and broadcast only when there is a waiter. > > This should be handled anyway by the pthreads library. If we are sure there > is only one waiter at most, you can change broadcast to signal but that > would already be microoptimization. > > Alternatively, you can use futexes to implement a completion signal (similar > to Win32 events). But it seems too much effort.
Yes, let's use standard cond vars for now. Stefan