Roland,
Roland Dreier <[EMAIL PROTECTED]> wrote on 05/24/2006 02:03:34 PM:
> Shirley> My observation is the atomic operation is not that
> Shirley> expensive.
>
> It's just about the worst thing to do. For example, on x86/x86-64 an
> instruction with the lock; prefix is quite slow. If you look at an
> instruction level profile you can see that quite clearly.
Compared to have a single thread handling AH, I don't think this atomic operation is expensive.
> Shirley> I thought the path holding another AH reference to
> Shirley> prevent it to be freed?
>
> If that were true then why would we want to reference count sends at
> all? The whole point is that a path might be destroyed before the
> send is executed.
>
> - R.
It is true for unicast, it has a reference count before ipoib_send(). I need to look at multicast.
Thanks
Shirley Ma
IBM Linux Technology Center
15300 SW Koll Parkway
Beaverton, OR 97006-6063
Phone(Fax): (503) 578-7638
_______________________________________________ openib-general mailing list openib-general@openib.org http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general