Re: condvar performance in .59 vs .60

Ulrich Drepper wrote:
> The best solution is to get requeue working (although two syscalls are
> made).

Requeue should be fine now, with the patch that I just posted on top
of the larger patch a couple of days ago.

I don't know what requeue bug you had with the old futex code, because
you haven't described it, but it isn't important if the new code
works.  The new code is easily back-ported to Red Hat's 2.4 once it is
shown to be stable.

> Alternatively, if the woken just gets added to the runqueue but doesn't
> replace the waker that would be OK, too.

A synchronous wakeup is like that.  If you can use requeue now,
there's no need for it, although perhaps some userspace applications
would appreciate being able to unblock a mutex or whatever without
being preempted immediately?

-- Jamie