Re: [4.2, Regression] Queued spinlocks cause major XFS performance regression

[Dave Chinner <david@fromorbit.com>]

On Fri, Sep 04, 2015 at 01:32:33PM +0200, Peter Zijlstra wrote:
> On Fri, Sep 04, 2015 at 06:12:34PM +1000, Dave Chinner wrote:
> > You probably don't even need a VM to reproduce it - that would
> > certainly be an interesting counterpoint if it didn't....
> 
> Even though you managed to restore your DEBUG_SPINLOCK performance by
> changing virt_queued_spin_lock() to use __delay(1), I ran the thing on
> actual hardware just to test.
> 
> [ Note: In any case, I would recommend you use (or at least try)
>   PARAVIRT_SPINLOCKS if you use VMs, as that is where we were looking for
>   performance, the test-and-set fallback really wasn't meant as a
>   performance option (although it clearly sucks worse than expected).

FSUse%        Count         Size    Files/sec     App Overhead
     0      1600000            0     319431.5         10116018
     0      3200000            0     307824.5         10054299
     0      4800000            0     296971.5         10770197
     0      6400000            0     281653.6         11748423
....

PARAVIRT_SPINLOCKS seems to work OK these days, too. I'll leave that
set so I'll end up testing whatever comes along down that pipe...

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com