From: Ingo Molnar <mingo@kernel.org>
To: Len Brown <lenb@kernel.org>
Cc: "Jan H. Schönherr" <jschoenh@amazon.de>,
"Thomas Gleixner" <tglx@linutronix.de>, "X86 ML" <x86@kernel.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"Anthony Liguori" <aliguori@amazon.com>,
"Ingo Molnar" <mingo@redhat.com>,
"H. Peter Anvin" <hpa@zytor.com>, "Tim Deegan" <tim@xen.org>,
"Gang Wei" <gang.wei@intel.com>,
"Linus Torvalds" <torvalds@linux-foundation.org>
Subject: Re: [PATCH] x86: skip delays during SMP initialization similar to Xen
Date: Tue, 12 May 2015 11:43:49 +0200 [thread overview]
Message-ID: <20150512094348.GA11477@gmail.com> (raw)
In-Reply-To: <CAJvTdKk9NZ7-jcJARBNhfTh1O_mV=xgZkN0e84ajPVq=4pciiQ@mail.gmail.com>
* Len Brown <lenb@kernel.org> wrote:
> On Thu, May 7, 2015 at 6:23 AM, Ingo Molnar <mingo@kernel.org> wrote:
>
> These numbers do not look right.
>
> > Here's what the boot time looks like on a 120 CPUs system, with the
> > patch applied:
> >
> > [ 0.558947] x86: Booting SMP configuration:
> > [ 0.563375] .... node #0, CPUs: #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 #14
> > [ 0.644851] .... node #1, CPUs: #15 #16 #17 #18 #19 #20 #21 #22 #23 #24 #25 #26 #27 #28 #29
> > [ 0.830474] .... node #2, CPUs: #30 #31 #32 #33 #34 #35 #36 #37 #38 #39 #40 #41 #42 #43 #44
> > [ 1.016357] .... node #3, CPUs: #45 #46 #47 #48 #49 #50 #51 #52 #53 #54 #55 #56 #57 #58 #59
> > [ 1.202342] .... node #0, CPUs: #60 #61 #62 #63 #64 #65 #66 #67 #68 #69 #70 #71 #72 #73 #74
> > [ 1.283864] .... node #1, CPUs: #75 #76 #77 #78 #79 #80 #81 #82 #83 #84 #85 #86 #87 #88 #89
> > [ 1.397131] .... node #2, CPUs: #90 #91 #92 #93 #94 #95 #96 #97 #98 #99 #100 #101 #102 #103 #104
> > [ 1.510417] .... node #3, CPUs: #105 #106 #107 #108 #109 #110 #111 #112 #113 #114 #115 #116 #117 #118 #119
> > [ 1.620967] x86: Booted up 4 nodes, 120 CPUs
> > [ 1.625928] smpboot: Total of 120 processors activated (672866.16 BogoMIPS)
> >
> > 1.1 seconds to boot 120 CPUs, 10.8 seconds to hit init, that's an
> > entirely reasonable runtime I think.
> >
> > It was 20+ seconds before that, 10+ seconds for the SMP bootup
> > sequence.
>
> (1.625928-0.558947) = 1.07 seconds to online 119 additional cpus.
> /119 = .0089662268 each, lets call it 9ms.
>
> Here is my ivb-ex running stock fedora 21's Linux-3.19 (no patch applied):
>
> [ 0.404369] x86: Booting SMP configuration:
> [ 0.409492] .... node #0, CPUs: #1
> [ 0.439900] NMI watchdog: enabled on all CPUs, permanently consumes
> one hw-PMU counter.
> [ 0.450533] #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 #14
> [ 0.652409] .... node #1, CPUs: #15 #16 #17 #18 #19 #20
> #21 #22 #23 #24 #25 #26 #27 #28 #29
> [ 0.999645] .... node #2, CPUs: #30 #31 #32 #33 #34 #35
> #36 #37 #38 #39 #40 #41 #42 #43 #44
> [ 1.346991] .... node #3, CPUs: #45 #46 #47 #48 #49 #50
> #51 #52 #53 #54 #55 #56 #57 #58 #59
> [ 1.694171] .... node #0, CPUs: #60 #61 #62 #63 #64 #65
> #66 #67 #68 #69 #70 #71 #72 #73 #74
> [ 1.928248] .... node #1, CPUs: #75 #76 #77 #78 #79 #80
> #81 #82 #83 #84 #85 #86 #87 #88 #89
> [ 2.198370] .... node #2, CPUs: #90 #91 #92 #93 #94 #95
> #96 #97 #98 #99 #100 #101 #102 #103 #104
> [ 2.468574] .... node #3, CPUs: #105 #106 #107 #108 #109 #110
> #111 #112 #113 #114 #115 #116 #117 #118 #119
> [ 2.737884] x86: Booted up 4 nodes, 120 CPUs
> [ 2.743758] smpboot: Total of 120 processors activated (671097.18 BogoMIPS)
>
> (2.743758-0.404369) = 2.339389 for all 119 processors
> /119 = .01965873109243697478 - lets call it 19ms each
>
> so this baseline case is 19ms/processor -- which matches above, where
> we delete 10ms/processor.
> But even at 19ms each, this is totals only 2.3 seconds for all 119
> processors. So I don't understand your reference to 10+ seconds,
> above.
Hm, so what I was referring to is:
[ 0.205318] x86: Booting SMP configuration:
[ 0.206005] .... node #0, CPUs: #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 #14 #15 #16 #17 #18 #19 #20 #21 #22 #23 #24 #25 #26 #27 #28 #29 #30 #31 #32 #33 #34 #35 #36 #37 #38 #39 #40 #41 #42 #43 #44 #45 #46 #47 #48 #49 #50 #51 #52 #53 #54 #55 #56 #57 #58 #59 #60 #61 #62 #63 #64 #65 #66 #67 #68 #69 #70 #71 #72 #73 #74 #75 #76 #77 #78 #79 #80 #81 #82 #83 #84 #85 #86 #87 #88 #89 #90 #91 #92 #93 #94 #95 #96 #97 #98 #99 #100 #101 #102 #103 #104 #105 #106 #107 #108 #109 #110 #111 #112 #113 #114 #115 #116 #117 #118 #119
[ 9.021191] x86: Booted up 1 node, 120 CPUs
[ 9.021803] smpboot: Total of 120 processors activated (672102.75 BogoMIPS)
[ 9.027282] devtmpfs: initialized
So it's 8.9 seconds - but still much more than you'd expect. The delay
is evenly distributed amongst CPUs.
> ps. I wish the BIOS boot time on this box were as fast as Linux:-)
LOL! :-)
Thanks,
Ingo
next prev parent reply other threads:[~2015-05-12 9:43 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-05-04 9:42 [PATCH] x86: skip delays during SMP initialization similar to Xen Jan H. Schönherr
2015-05-06 8:27 ` Ingo Molnar
2015-05-07 1:27 ` Len Brown
2015-05-07 10:23 ` Ingo Molnar
2015-05-12 8:03 ` Len Brown
2015-05-12 9:43 ` Ingo Molnar [this message]
2015-05-12 9:47 ` Ingo Molnar
2015-05-12 10:12 ` Borislav Petkov
2015-05-14 6:36 ` Len Brown
2015-05-14 6:44 ` Ingo Molnar
2015-05-16 8:46 ` Len Brown
2015-05-17 5:02 ` Ingo Molnar
2015-05-28 5:31 ` Len Brown
2015-05-29 2:16 ` Len Brown
2015-05-14 7:18 ` Len Brown
2015-05-14 14:25 ` "Jan H. Schönherr"
2015-05-14 17:57 ` Ingo Molnar
2015-05-16 9:07 ` Len Brown
2015-05-17 5:26 ` Ingo Molnar
2015-05-06 10:13 ` [tip:x86/apic] x86/smpboot: Skip " tip-bot for Jan H. Schönherr
2015-05-13 7:15 ` [tip:x86/apic] Revert f5d6a52f5111 ("x86/smpboot: Skip delays during SMP initialization similar to Xen") tip-bot for Ingo Molnar
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=20150512094348.GA11477@gmail.com \
--to=mingo@kernel.org \
--cc=aliguori@amazon.com \
--cc=gang.wei@intel.com \
--cc=hpa@zytor.com \
--cc=jschoenh@amazon.de \
--cc=lenb@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=tglx@linutronix.de \
--cc=tim@xen.org \
--cc=torvalds@linux-foundation.org \
--cc=x86@kernel.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.