[uml-devel] host cpu frequency scaling and guest bogomips

[uml-devel] host cpu frequency scaling and guest bogomips

[Victor Condino]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

just noticed something interesting while testing out my gentoo/uml
root-fs generation script on my laptop. my l/t is an thinkpad T42p,
which has a p3 centrino with speedstep. i use cpufreqd to throttle the
cpu when not on ac between 600 and 1800MHz based on load. what i noticed
is that when I fired up UML while the host's core frequency was at
600mhz, /proc/cpuinfo reported ~600 bogomips in the guest, but when i
put some load on the host and cpufreq stepped up the core frequency, UML
still reports ~600 bogomips. likewise, when I fired up UML while the
host was under load and it's core at 1200MHz, UML reports higher
bogomips, but then when the host drops back down to the lower core freq,
UML still reports the same higher bogomips.

Not that this actually affects anything I do or would affect most
people... just thought you guys might want to know =D

- - V. Condino

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFC01SbW5g8mX2/4TYRAjs0AKCKRXwTy6g/UK5w5ioA8WxpgcB/3QCfRZyj
H9HTVRNl0FzeqUiYRxwrTAg=
=VsPI
-----END PGP SIGNATURE-----


-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA.  To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] host cpu frequency scaling and guest bogomips

[Blaisorblade]

On Tuesday 12 July 2005 07:26, Victor Condino wrote:
> just noticed something interesting while testing out my gentoo/uml
> root-fs generation script on my laptop. my l/t is an thinkpad T42p,
> which has a p3 centrino with speedstep. i use cpufreqd to throttle the
> cpu when not on ac between 600 and 1800MHz based on load. what i noticed
> is that when I fired up UML while the host's core frequency was at
> 600mhz, /proc/cpuinfo reported ~600 bogomips in the guest, but when i
> put some load on the host and cpufreq stepped up the core frequency, UML
> still reports ~600 bogomips. likewise, when I fired up UML while the
> host was under load and it's core at 1200MHz, UML reports higher
> bogomips, but then when the host drops back down to the lower core freq,
> UML still reports the same higher bogomips.

> Not that this actually affects anything I do or would affect most
> people... just thought you guys might want to know =D
Hmm, bogomips are calculated at startup and that won't change easily. However, 
they become inaccurate in a much wider variety of situation, for instance 
when there's some load on the host; so, using CONFIG_UML_REAL_TIME_CLOCK 
(IIRC) uses the host clock for time calculation.

When you disable this option, I guess that a "sleep 10" when host is at 600, 
if UML was started at 1800, might last 30 seconds more or less...
> - V. Condino

-- 
Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!".
Paolo Giarrusso, aka Blaisorblade (Skype ID "PaoloGiarrusso", ICQ 215621894)
http://www.user-mode-linux.org/~blaisorblade

		
___________________________________ 
Yahoo! Messenger: chiamate gratuite in tutto il mondo 
http://it.beta.messenger.yahoo.com



-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA.  To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] host cpu frequency scaling and guest bogomips

[Victor Condino]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Blaisorblade wrote:
> Hmm, bogomips are calculated at startup and that won't change easily. However, 
> they become inaccurate in a much wider variety of situation, for instance 
> when there's some load on the host; so, using CONFIG_UML_REAL_TIME_CLOCK 
> (IIRC) uses the host clock for time calculation.
> 
> When you disable this option, I guess that a "sleep 10" when host is at 600, 
> if UML was started at 1800, might last 30 seconds more or less...

hmm... so the moral is, CONFIG_UML_REAL_TIME_CLOCK is a "good thing"(tm)
if your host is going to vary it's core freq?

- - V.C.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFC0+73W5g8mX2/4TYRAvv3AJ9uA3HzbsM27evJHwUofX+1WOViMACfRL3J
AqZgxrkfanj50IaQxjdcyec=
=K/Cr
-----END PGP SIGNATURE-----


-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA.  To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] host cpu frequency scaling and guest bogomips

[Geert Uytterhoeven]

On Tue, 12 Jul 2005, Blaisorblade wrote:
> On Tuesday 12 July 2005 07:26, Victor Condino wrote:
> > just noticed something interesting while testing out my gentoo/uml
> > root-fs generation script on my laptop. my l/t is an thinkpad T42p,
> > which has a p3 centrino with speedstep. i use cpufreqd to throttle the
> > cpu when not on ac between 600 and 1800MHz based on load. what i noticed
> > is that when I fired up UML while the host's core frequency was at
> > 600mhz, /proc/cpuinfo reported ~600 bogomips in the guest, but when i
> > put some load on the host and cpufreq stepped up the core frequency, UML
> > still reports ~600 bogomips. likewise, when I fired up UML while the
> > host was under load and it's core at 1200MHz, UML reports higher
> > bogomips, but then when the host drops back down to the lower core freq,
> > UML still reports the same higher bogomips.
> 
> > Not that this actually affects anything I do or would affect most
> > people... just thought you guys might want to know =D
> Hmm, bogomips are calculated at startup and that won't change easily. However, 
> they become inaccurate in a much wider variety of situation, for instance 
> when there's some load on the host; so, using CONFIG_UML_REAL_TIME_CLOCK 
> (IIRC) uses the host clock for time calculation.
> 
> When you disable this option, I guess that a "sleep 10" when host is at 600, 
> if UML was started at 1800, might last 30 seconds more or less...

I don't think user space sleep uses the kernel's udelay() or mdelay()...

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds


-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA.  To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] host cpu frequency scaling and guest bogomips

[Blaisorblade]

On Tuesday 12 July 2005 18:35, Geert Uytterhoeven wrote:
> On Tue, 12 Jul 2005, Blaisorblade wrote:
> > On Tuesday 12 July 2005 07:26, Victor Condino wrote:
> > > just noticed something interesting while testing out my gentoo/uml
> > > root-fs generation script on my laptop. my l/t is an thinkpad T42p,
> > > which has a p3 centrino with speedstep. i use cpufreqd to throttle the
> > > cpu when not on ac between 600 and 1800MHz based on load. what i
> > > noticed is that when I fired up UML while the host's core frequency was
> > > at 600mhz, /proc/cpuinfo reported ~600 bogomips in the guest, but when
> > > i put some load on the host and cpufreq stepped up the core frequency,
> > > UML still reports ~600 bogomips. likewise, when I fired up UML while
> > > the host was under load and it's core at 1200MHz, UML reports higher
> > > bogomips, but then when the host drops back down to the lower core
> > > freq, UML still reports the same higher bogomips.
> > >
> > > Not that this actually affects anything I do or would affect most
> > > people... just thought you guys might want to know =D

> > Hmm, bogomips are calculated at startup and that won't change easily.
> > However, they become inaccurate in a much wider variety of situation, for
> > instance when there's some load on the host; so, using
> > CONFIG_UML_REAL_TIME_CLOCK (IIRC) uses the host clock for time
> > calculation.

> > When you disable this option, I guess that a "sleep 10" when host is at
> > 600, if UML was started at 1800, might last 30 seconds more or less...

> I don't think user space sleep uses the kernel's udelay() or mdelay()...

Sorry, right... they're the only one to use bogomips (just learned), still 
UML_REAL_TIME_CLOCK makes "sleep(1)" in user-space precise.

That option drops the assumption that SIGALRM (or VTALRM) is exact, and 
rather, at ALRM delivery, uses gettimeofday() to discover how long passed 
from last interrupt.
-- 
Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!".
Paolo Giarrusso, aka Blaisorblade (Skype ID "PaoloGiarrusso", ICQ 215621894)
http://www.user-mode-linux.org/~blaisorblade

	

	
		
___________________________________ 
Yahoo! Mail: gratis 1GB per i messaggi e allegati da 10MB 
http://mail.yahoo.it



-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA.  To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel