From: Peter Zijlstra <a.p.zijlstra@chello.nl>
To: David Howells <dhowells@redhat.com>
Cc: torvalds@osdl.org, akpm@linux-foundation.org, nico@cam.org,
linux-am33-list@redhat.com, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 2/2] MN10300: Make sched_clock() report time since boot
Date: Fri, 26 Sep 2008 12:58:39 +0200 [thread overview]
Message-ID: <1222426720.16700.262.camel@lappy.programming.kicks-ass.net> (raw)
In-Reply-To: <20080924164831.14867.11054.stgit@warthog.procyon.org.uk>
On Wed, 2008-09-24 at 17:48 +0100, David Howells wrote:
> Make sched_clock() report time since boot rather than time since last timer
> interrupt.
>
> Make sched_clock() expand and scale the 32-bit TSC value running at IOCLK
> speed (~33MHz) to a 64-bit nanosecond counter, using cnt32_to_63() acquired
> from the ARM arch and without using slow DIVU instructions every call.
Right, so since you have to up-scale the 63 bit counter obtained using
the smarty pants cnt32_to_63 code, you actually end up with a genuine
64bit counter.. sweet.
I'm not going to pretend to understand the NM10300 details, but the
generic idea makes sense.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> Signed-off-by: David Howells <dhowells@redhat.com>
> ---
>
> arch/mn10300/kernel/time.c | 52 +++++++++++++++++++++++++++++++++++---------
> 1 files changed, 41 insertions(+), 11 deletions(-)
>
>
> diff --git a/arch/mn10300/kernel/time.c b/arch/mn10300/kernel/time.c
> index babb7c2..e460658 100644
> --- a/arch/mn10300/kernel/time.c
> +++ b/arch/mn10300/kernel/time.c
> @@ -1,6 +1,6 @@
> /* MN10300 Low level time management
> *
> - * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
> + * Copyright (C) 2007-2008 Red Hat, Inc. All Rights Reserved.
> * Written by David Howells (dhowells@redhat.com)
> * - Derived from arch/i386/kernel/time.c
> *
> @@ -16,6 +16,7 @@
> #include <linux/init.h>
> #include <linux/smp.h>
> #include <linux/profile.h>
> +#include <linux/cnt32_to_63.h>
> #include <asm/irq.h>
> #include <asm/div64.h>
> #include <asm/processor.h>
> @@ -40,27 +41,54 @@ static struct irqaction timer_irq = {
> .name = "timer",
> };
>
> +static unsigned long sched_clock_multiplier;
> +
> /*
> * scheduler clock - returns current time in nanosec units.
> */
> unsigned long long sched_clock(void)
> {
> union {
> - unsigned long long l;
> - u32 w[2];
> - } quot;
> + unsigned long long ll;
> + unsigned l[2];
> + } tsc64, result;
> + unsigned long tsc, tmp;
> + unsigned product[3]; /* 96-bit intermediate value */
> +
> + /* read the TSC value
> + */
> + tsc = 0 - get_cycles(); /* get_cycles() counts down */
>
> - quot.w[0] = mn10300_last_tsc - get_cycles();
> - quot.w[1] = 1000000000;
> + /* expand to 64-bits.
> + * - sched_clock() must be called once a minute or better or the
> + * following will go horribly wrong - see cnt32_to_63()
> + */
> + tsc64.ll = cnt32_to_63(tsc) & 0x7fffffffffffffffULL;
>
> - asm("mulu %2,%3,%0,%1"
> - : "=r"(quot.w[1]), "=r"(quot.w[0])
> - : "0"(quot.w[1]), "1"(quot.w[0])
> + /* scale the 64-bit TSC value to a nanosecond value via a 96-bit
> + * intermediate
> + */
> + asm("mulu %2,%0,%3,%0 \n" /* LSW * mult -> 0:%3:%0 */
> + "mulu %2,%1,%2,%1 \n" /* MSW * mult -> %2:%1:0 */
> + "add %3,%1 \n"
> + "addc 0,%2 \n" /* result in %2:%1:%0 */
> + : "=r"(product[0]), "=r"(product[1]), "=r"(product[2]), "=r"(tmp)
> + : "0"(tsc64.l[0]), "1"(tsc64.l[1]), "2"(sched_clock_multiplier)
> : "cc");
>
> - do_div(quot.l, MN10300_TSCCLK);
> + result.l[0] = product[1] << 16 | product[0] >> 16;
> + result.l[1] = product[2] << 16 | product[1] >> 16;
>
> - return quot.l;
> + return result.ll;
> +}
> +
> +/*
> + * initialise the scheduler clock
> + */
> +static void __init mn10300_sched_clock_init(void)
> +{
> + sched_clock_multiplier =
> + __muldiv64u(NSEC_PER_SEC, 1 << 16, MN10300_TSCCLK);
> }
>
> /*
> @@ -128,4 +156,6 @@ void __init time_init(void)
> /* start the watchdog timer */
> watchdog_go();
> #endif
> +
> + mn10300_sched_clock_init();
> }
>
next prev parent reply other threads:[~2008-09-26 10:58 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-09-24 16:48 [PATCH 1/2] MN10300: Move asm-arm/cnt32_to_63.h to include/linux/ David Howells
2008-09-24 16:48 ` [PATCH 2/2] MN10300: Make sched_clock() report time since boot David Howells
2008-09-26 10:58 ` Peter Zijlstra [this message]
2008-09-26 10:56 ` [PATCH 1/2] MN10300: Move asm-arm/cnt32_to_63.h to include/linux/ Peter Zijlstra
2008-09-26 12:03 ` Nicolas Pitre
2008-09-26 12:08 ` Peter Zijlstra
2008-09-29 1:21 ` Nicolas Pitre
2008-09-26 12:20 ` Peter Zijlstra
2008-09-29 1:42 ` Nicolas Pitre
2008-10-01 14:34 ` David Howells
2008-10-01 15:36 ` Nicolas Pitre
2008-10-02 22:23 ` David Howells
2008-10-03 19:15 ` Nicolas Pitre
2008-10-06 10:44 ` David Howells
2008-10-06 15:06 ` Nicolas Pitre
2008-09-26 13:27 ` David Howells
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=1222426720.16700.262.camel@lappy.programming.kicks-ass.net \
--to=a.p.zijlstra@chello.nl \
--cc=akpm@linux-foundation.org \
--cc=dhowells@redhat.com \
--cc=linux-am33-list@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=nico@cam.org \
--cc=torvalds@osdl.org \
/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