Re: nptl 0.30

"dada1" <[email protected]> schrieb am 19.03.03 09:18:08:
[...]
> I have a big threaded application that benefit a huge gain just by inverting
> the unlock/wake. This is because the linux kernel *always* preempt the
                                                    ^^^^^^^^

Well, you might also want to take at look at this thread:

http://groups.google.com/groups?threadm=3E6DB2FD.24679793%40web.de
(Subject: Re: "Dreadlock" Revisited)

> current thread in favor of the awaken thread(s). The number of context
> switchs came from 11000 to 3000 per second.
> 
> You already did the change I suggested in
> sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S
> 
> Why dont you do the same in sysdeps/pthread/pthread_cond_broadcast.c ?

I also have a few questions:

  - why do you think that holding the mutex in signal() [while 
    signaling] is nessasary?

  - why do you think that you need an "exit condition" in wait()?

  - why do you think that do-nothing-wrt-futex-overflow is just fine?

If possible, please drop some short example(s) illustrating your 
thoughts (BTW, it wouldn't hurt anyone if your DESIGN-condvar.txt 
would provide answers to such things, I think). Thank you.

regards,
alexander.

______________________________________________________________________________
Ihre Freunde sind in Italien? Schicken Sie ihnen trotzdem eine SMS
mit WEB.DE FreeMail http://freemail.web.de/features/?mc=021172