From: Eric Dumazet <eric.dumazet@gmail.com>
To: Andi Kleen <andi@firstfloor.org>
Cc: Thomas Gleixner <tglx@linutronix.de>,
john stultz <johnstul@us.ibm.com>,
lkml <linux-kernel@vger.kernel.org>,
Paul Mackerras <paulus@samba.org>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
Anton Blanchard <anton@samba.org>, Ingo Molnar <mingo@elte.hu>
Subject: Re: [RFC] time: xtime_lock is held too long
Date: Fri, 06 May 2011 03:41:43 +0200 [thread overview]
Message-ID: <1304646103.3131.3.camel@edumazet-laptop> (raw)
In-Reply-To: <20110505210118.GI2925@one.firstfloor.org>
Le jeudi 05 mai 2011 à 23:01 +0200, Andi Kleen a écrit :
> > > Another idea would be to prime cache lines to be dirtied in cpu cache
> > > before taking locks, and better pack variables to reduce number of cache
> > > lines.
> >
> > Most variables are packed already in struct timekeeper, which should
> > be pretty cache hot anyway, so I don't know whether we gain much.
>
> There's actually some potential here. I got a moderate speedup in a
> database benchmark with this patch recently. The biggest win
> was in the timer interrupt.
>
> All those variables are on separate cache lines.
>
> (needs some cleanups, just for illustration)
>
> -Andi
>
> From: Andi Kleen <ak@linux.intel.com>
> Subject: [PATCH] do prefetches for timer state in timer interrupt
>
> diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
> index 76b96d7..2704267 100644
> --- a/arch/x86/kernel/apic/apic.c
> +++ b/arch/x86/kernel/apic/apic.c
> @@ -856,6 +856,17 @@ void __irq_entry smp_apic_timer_interrupt(struct pt_regs *regs)
> {
> struct pt_regs *old_regs = set_irq_regs(regs);
>
> + extern struct timespec xtime;
> + extern struct timespec wall_to_monotonic;
> +
> +
> + /* Prefetch here to make the ktime_get later faster.
> + * XXX move them all on the same cache line.
> + */
> + prefetch(&xtime_lock);
> + prefetch(&xtime);
> + prefetch(&wall_to_monotonic);
> +
> /*
> * NOTE! We'd better ACK the irq immediately,
> * because timer handling can be slow.
> diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
> index d27c756..0002e9e 100644
> --- a/kernel/time/timekeeping.c
> +++ b/kernel/time/timekeeping.c
> @@ -157,8 +157,8 @@ __cacheline_aligned_in_smp DEFINE_SEQLOCK(xtime_lock);
> * - wall_to_monotonic is no longer the boot time, getboottime must be
> * used instead.
> */
> -static struct timespec xtime __attribute__ ((aligned (16)));
> -static struct timespec wall_to_monotonic __attribute__ ((aligned (16)));
> +struct timespec xtime __attribute__ ((aligned (16)));
> +struct timespec wall_to_monotonic __attribute__ ((aligned (16)));
> static struct timespec total_sleep_time;
>
> /*
>
Did you tried prefetchw() by any chance ?
next prev parent reply other threads:[~2011-05-06 1:41 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-05-04 3:11 [PATCH] time: Add locking to xtime access in get_seconds() John Stultz
2011-05-04 3:52 ` Andi Kleen
2011-05-05 2:54 ` john stultz
2011-05-05 5:44 ` Eric Dumazet
2011-05-05 6:21 ` john stultz
2011-05-05 6:50 ` Eric Dumazet
2011-05-05 8:14 ` Paul E. McKenney
2011-05-05 18:51 ` john stultz
2011-05-05 14:04 ` [RFC] time: xtime_lock is held too long Eric Dumazet
2011-05-05 14:39 ` Thomas Gleixner
2011-05-05 15:08 ` Eric Dumazet
2011-05-05 15:59 ` Thomas Gleixner
2011-05-05 21:01 ` Andi Kleen
2011-05-06 1:41 ` Eric Dumazet [this message]
2011-05-06 6:55 ` Andi Kleen
2011-05-06 10:18 ` Thomas Gleixner
2011-05-06 10:22 ` Ingo Molnar
2011-05-06 16:53 ` Andi Kleen
2011-05-07 8:20 ` Ingo Molnar
2011-05-06 16:59 ` Andi Kleen
2011-05-06 17:09 ` Eric Dumazet
2011-05-06 17:17 ` Andi Kleen
2011-05-06 17:42 ` Eric Dumazet
2011-05-06 17:50 ` Andi Kleen
2011-05-06 19:26 ` Eric Dumazet
2011-05-06 20:04 ` Eric Dumazet
2011-05-06 20:24 ` john stultz
2011-05-06 22:30 ` Eric Dumazet
2011-05-06 22:46 ` john stultz
2011-05-06 23:00 ` Eric Dumazet
2011-05-06 23:28 ` john stultz
2011-05-07 5:02 ` Eric Dumazet
2011-05-07 7:11 ` Henrik Rydberg
2011-05-09 8:40 ` Thomas Gleixner
2011-05-12 9:13 ` [PATCH] seqlock: don't smp_rmb in seqlock reader spin loop, [PATCH] seqlock: don't smp_rmb in seqlock reader spin loop Milton Miller
2011-05-12 9:35 ` Eric Dumazet
2011-05-12 14:08 ` Andi Kleen
2011-05-06 20:18 ` [RFC] time: xtime_lock is held too long john stultz
2011-05-05 17:57 ` [PATCH] time: Add locking to xtime access in get_seconds() Andi Kleen
2011-05-05 20:17 ` john stultz
2011-05-05 20:24 ` Eric Dumazet
2011-05-05 20:40 ` john stultz
2011-05-05 20:43 ` Eric Dumazet
2011-05-05 20:56 ` Andi Kleen
2011-05-04 16:51 ` Max Asbock
2011-05-04 21:05 ` Andi Kleen
2011-05-04 23:05 ` john stultz
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1304646103.3131.3.camel@edumazet-laptop \
--to=eric.dumazet@gmail.com \
--cc=andi@firstfloor.org \
--cc=anton@samba.org \
--cc=johnstul@us.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=paulmck@linux.vnet.ibm.com \
--cc=paulus@samba.org \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox