Re: Re: [RFC] arm: Add for atomic half word exchange

[Sarbojit Ganguly]

Yes, the main advantage of Qspinlock code can be observed in NUMA but when I tested in an embedded system, a slight advantage was observed.


------- Original Message -------
Sender : Peter Zijlstra<peterz@xxxxxxxxxxxxx>
Date : May 19, 2015 21:43 (GMT+09:00)
Title : Re: [RFC] arm: Add for atomic half word exchange

On Tue, May 19, 2015 at 11:20:13AM +0000, Sarbojit Ganguly wrote:
> On Tuesday 19 May 2015 09:39:33 Sarbojit Ganguly wrote:
> > Since 16 bit half word exchange was not there and MCS based
> > qspinlock by Waiman's xchg_tail() requires an atomic exchange on a
> > half word, here is a small modification to __xchg() code.

Can you actually see a performance improvement with the qspinlock code
on ARM ?

The real improvements on x86 were on NUMA systems; although there were
real improvements on light loads as well.


Note that ARM (or any load-store arch) could get rid of all the cmpxchg
loops in that code. Although I suppose we replaced the most common ones
with these unconditional atomics already -- like that xchg16 -- so
implementing those with ll/sc, as you did, should be near optimal.




?????
???   ??   ?? ??
----------------------------------------------------------------------+
The Tao lies beyond Yin and Yang. It is silent and still as a pool of water.      |
It does not seek fame, therefore nobody knows its presence.                       |
It does not seek fortune, for it is complete within itself.                       |
It exists beyond space and time.                                                  |
----------------------------------------------------------------------+