From: Peter Zijlstra <peterz@infradead.org>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: john stultz <johnstul@us.ibm.com>, Torben Hohn <torbenh@gmx.de>,
LKML <linux-kernel@vger.kernel.org>,
hch@infradead.org, yong.zhang0@gmail.com
Subject: Re: [PATCH 01/18] move do_timer() from kernel/timer.c into kernel/time/timekeeping.c
Date: Tue, 25 Jan 2011 11:34:37 +0100 [thread overview]
Message-ID: <1295951677.28776.489.camel@laptop> (raw)
In-Reply-To: <1295946514.28776.476.camel@laptop>
On Tue, 2011-01-25 at 10:08 +0100, Peter Zijlstra wrote:
> On Mon, 2011-01-24 at 23:44 +0100, Thomas Gleixner wrote:
> > On Mon, 24 Jan 2011, Peter Zijlstra wrote:
> >
> > > On Mon, 2011-01-24 at 23:10 +0100, Thomas Gleixner wrote:
> > > > The call to calc_global_load() is there for hysterical raisins and we
> > > > really should get rid of it sooner than later. I'm quite sure that it
> > > > could be run from a timer callback as well. Peter ?
> > >
> > > calc_global_load() wants to be called on just one cpu, do we have a
> > > better place for that?
> >
> > Well, we can call it from a timer from a single CPU.
>
> Right that would work I guess.
Something like the (completely untested) below would do I guess:
---
Subject: sched: Move the calc_global_load() call into the scheduler
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date: Tue Jan 25 11:30:35 CET 2011
Remove the calc_global_load() call from the timekeeping code and make
it local to the scheduler.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
include/linux/sched.h | 2 --
kernel/sched.c | 22 ++++++++++++++++------
kernel/timer.c | 1 -
3 files changed, 16 insertions(+), 9 deletions(-)
Index: linux-2.6/include/linux/sched.h
===================================================================
--- linux-2.6.orig/include/linux/sched.h
+++ linux-2.6/include/linux/sched.h
@@ -142,8 +142,6 @@ extern unsigned long nr_iowait_cpu(int c
extern unsigned long this_cpu_load(void);
-extern void calc_global_load(unsigned long ticks);
-
extern unsigned long get_parent_ip(unsigned long addr);
struct seq_file;
Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -3192,7 +3192,7 @@ calc_load_n(unsigned long load, unsigned
* Once we've updated the global active value, we need to apply the exponential
* weights adjusted to the number of cycles missed.
*/
-static void calc_global_nohz(unsigned long ticks)
+static void calc_global_nohz(void)
{
long delta, active, n;
@@ -3209,11 +3209,13 @@ static void calc_global_nohz(unsigned lo
if (delta)
atomic_long_add(delta, &calc_load_tasks);
+
/*
* If we were idle for multiple load cycles, apply them.
*/
- if (ticks >= LOAD_FREQ) {
- n = ticks / LOAD_FREQ;
+ delta = jiffies - calc_load_update - 10;
+ if (delta >= LOAD_FREQ) {
+ n = delta / LOAD_FREQ;
active = atomic_long_read(&calc_load_tasks);
active = active > 0 ? active * FIXED_1 : 0;
@@ -3246,7 +3248,7 @@ static inline long calc_load_fold_idle(v
return 0;
}
-static void calc_global_nohz(unsigned long ticks)
+static void calc_global_nohz(void)
{
}
#endif
@@ -3266,15 +3268,20 @@ void get_avenrun(unsigned long *loads, u
loads[2] = (avenrun[2] + offset) << shift;
}
+static void calc_global_load(unsigned long __data);
+
+static struct timer_list global_load_timer =
+ TIMER_DEFERRED_INITIALIZER(calc_global_load, 0, 0);
+
/*
* calc_load - update the avenrun load estimates 10 ticks after the
* CPUs have updated calc_load_tasks.
*/
-void calc_global_load(unsigned long ticks)
+static void calc_global_load(unsigned long __data)
{
long active;
- calc_global_nohz(ticks);
+ calc_global_nohz();
if (time_before(jiffies, calc_load_update + 10))
return;
@@ -3287,6 +3294,7 @@ void calc_global_load(unsigned long tick
avenrun[2] = calc_load(avenrun[2], EXP_15, active);
calc_load_update += LOAD_FREQ;
+ mod_timer(&global_load_timer, calc_load_update + 10);
}
/*
@@ -8172,6 +8180,8 @@ void __init sched_init(void)
init_idle(current, smp_processor_id());
calc_load_update = jiffies + LOAD_FREQ;
+ global_load_timer.slack = 0;
+ mod_timer(&global_load_timer, calc_load_update + 10);
/*
* During early bootup we pretend to be a normal task:
Index: linux-2.6/kernel/timer.c
===================================================================
--- linux-2.6.orig/kernel/timer.c
+++ linux-2.6/kernel/timer.c
@@ -1303,7 +1303,6 @@ void do_timer(unsigned long ticks)
{
jiffies_64 += ticks;
update_wall_time();
- calc_global_load(ticks);
}
#ifdef __ARCH_WANT_SYS_ALARM
next prev parent reply other threads:[~2011-01-25 10:33 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-01-21 23:06 make arch code use xtime_update() instead of do_timer() Torben Hohn
2011-01-21 23:06 ` [PATCH 01/18] move do_timer() from kernel/timer.c into kernel/time/timekeeping.c Torben Hohn
2011-01-22 9:45 ` Thomas Gleixner
2011-01-24 20:32 ` john stultz
2011-01-24 22:10 ` Thomas Gleixner
2011-01-24 22:21 ` Peter Zijlstra
2011-01-24 22:44 ` Thomas Gleixner
2011-01-25 9:08 ` Peter Zijlstra
2011-01-25 10:34 ` Peter Zijlstra [this message]
2011-01-25 16:51 ` Peter Zijlstra
2011-01-26 5:56 ` Yong Zhang
2011-01-26 6:49 ` Yong Zhang
2011-01-26 10:03 ` Peter Zijlstra
2011-01-26 11:11 ` Yong Zhang
2011-01-21 23:06 ` [PATCH 02/18] provide xtime_update() which does not require holding xtime_lock like do_timer() Torben Hohn
2011-01-22 10:36 ` Thomas Gleixner
2011-01-21 23:06 ` [PATCH 03/18] alpha: change do_timer() to xtime_update() Torben Hohn
2011-01-21 23:06 ` [PATCH 04/18] arm: switch from " Torben Hohn
2011-01-21 23:06 ` [PATCH 05/18] arm/mach-clps711x: switch " Torben Hohn
2011-01-21 23:06 ` [PATCH 06/18] blackfin: switch from " Torben Hohn
2011-01-21 23:06 ` [PATCH 07/18] cris/arch-v10: switch " Torben Hohn
2011-01-21 23:06 ` [PATCH 08/18] cris/arch-v32: " Torben Hohn
2011-01-21 23:06 ` [PATCH 09/18] frv: " Torben Hohn
2011-01-22 10:01 ` Thomas Gleixner
2011-01-24 12:51 ` torbenh
2011-01-24 13:09 ` Thomas Gleixner
2011-01-21 23:06 ` [PATCH 10/18] h8300: " Torben Hohn
2011-01-21 23:06 ` [PATCH 11/18] ia64: " Torben Hohn
2011-01-21 23:06 ` [PATCH 12/18] m32r: switch from " Torben Hohn
2011-01-21 23:06 ` [PATCH 13/18] m68k: switch " Torben Hohn
2011-01-21 23:07 ` [PATCH 14/18] mn10300: switch do_timer() to xtimer_update() Torben Hohn
2011-01-21 23:07 ` [PATCH 15/18] parisc: switch do_timer() to xtime_update() Torben Hohn
2011-01-21 23:07 ` [PATCH 16/18] sparc: " Torben Hohn
2011-01-22 0:51 ` David Miller
2011-01-21 23:07 ` [PATCH 17/18] xtensa: " Torben Hohn
2011-01-21 23:07 ` [PATCH 18/18] make do_timer() and xtime_lock private to the timer code Torben Hohn
2011-01-22 10:14 ` make arch code use xtime_update() instead of do_timer() Thomas Gleixner
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=1295951677.28776.489.camel@laptop \
--to=peterz@infradead.org \
--cc=hch@infradead.org \
--cc=johnstul@us.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=tglx@linutronix.de \
--cc=torbenh@gmx.de \
--cc=yong.zhang0@gmail.com \
/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