Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-18 Thread Bernhard Reiter
On Wed, May 17, 2006 at 11:09:21PM +1000, Paul Collins wrote:
 Bernhard Reiter [EMAIL PROTECTED] writes:
 
  but my idea would be that an ondemand govenor would just work
  on a load average with some delay to even out spikes.
 
 The userspace governor combined with powernowd will give you this.
 And since the userspace governor defaults to the lowest CPU frequency,
 it all works out nicely.  I've using it happily on my PowerBook.

Yes I will try powernowd it sounds nice 
and it is good to know that it works well for people with powerbooks.

My point was that I would expect that the kernel itself
comes with a simplistic governor doing this.


pgpWRpxNz1eeN.pgp
Description: PGP signature


Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-17 Thread Michael Schmitz
  speeds? The code path on the 7447A models appears to have a settle delay
  of 1 ms after the feature call (which boils down to an OUTB, basically),
  another 0.1 ms after the low_choose_7447a_dfs call which shouldn't take
  very long either. Am I missing something, Ben?

 I'm being over conservative bcs of the voltage change

That's the 1 ms, after which the speed switch happens. Do you think the
voltage switch actually takes longer to settle than this?

  I'd not be surprised if the latency was as low as 2 ms total - can we
  profile this reliably in some way? I guess I'll just set the latency to 5
  ms and see what happens ...

 Last time I looked the on-demand governor was using a workqueue...
 anything that has a latency of more than a few dozen usec in a workqueue
 shall be shot...

That may be right - it's probably the reason for the latency cutoff at 10
ms. People might not be impressed if they try ondemand switching :-)

Anyway: is it safe to try, or will I fry my system board?

Michael


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-17 Thread Bernhard Reiter
Ben, Michael,

On Wed, May 17, 2006 at 08:15:42AM +1000, Benjamin Herrenschmidt wrote:
 On Tue, 2006-05-16 at 14:54 +0200, Michael Schmitz wrote:

   JoseJX write there that ondemand and conservative cannot be used
   on 7xx/74xx processors because of the time required to switch speeds.
  
  The reason seems to be a rather conservative assumption on the latency
  ('probably takes forever') - does it really take more than 10 ms to switch
  speeds? The code path on the 7447A models appears to have a settle delay
  of 1 ms after the feature call (which boils down to an OUTB, basically),
  another 0.1 ms after the low_choose_7447a_dfs call which shouldn't take
  very long either. Am I missing something, Ben?
 
 I'm being over conservative bcs of the voltage change
 
  I'd not be surprised if the latency was as low as 2 ms total - can we
  profile this reliably in some way? I guess I'll just set the latency to 5
  ms and see what happens ...
 
 Last time I looked the on-demand governor was using a workqueue...
 anything that has a latency of more than a few dozen usec in a workqueue
 shall be shot...

you are talking over my head here,
but my idea would be that an ondemand govenor would just work
on a load average with some delay to even out spikes.
Why not raise that delay and have no problem?

I my simplistic world I would expect the kernel to have such an option
as this is very close to the core. If the user wants more sofisticated
algorithms, this is very real user space daemons come in.

Am I completey off the mark here?





pgplaEu6GT0hI.pgp
Description: PGP signature


Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-17 Thread Benjamin Herrenschmidt
On Wed, 2006-05-17 at 10:45 +0200, Michael Schmitz wrote:
   speeds? The code path on the 7447A models appears to have a settle delay
   of 1 ms after the feature call (which boils down to an OUTB, basically),
   another 0.1 ms after the low_choose_7447a_dfs call which shouldn't take
   very long either. Am I missing something, Ben?
 
  I'm being over conservative bcs of the voltage change
 
 That's the 1 ms, after which the speed switch happens. Do you think the
 voltage switch actually takes longer to settle than this?

I don't have proper specs, it seems to depend on the machine. I'd rather
avoid playing wave generator with the voltage control chip though :)

 That may be right - it's probably the reason for the latency cutoff at 10
 ms. People might not be impressed if they try ondemand switching :-)
 
 Anyway: is it safe to try, or will I fry my system board?

I suppose at worse you'll lockup randomly but I can't promise :)

Ben.



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-17 Thread Paul Collins
Bernhard Reiter [EMAIL PROTECTED] writes:

 but my idea would be that an ondemand govenor would just work
 on a load average with some delay to even out spikes.

The userspace governor combined with powernowd will give you this.
And since the userspace governor defaults to the lowest CPU frequency,
it all works out nicely.  I've using it happily on my PowerBook.

-- 
Dag vijandelijk luchtschip de huismeester is dood


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-16 Thread Bernhard Reiter
On Mon, May 15, 2006 at 07:27:42PM +0200, Bin Zhang wrote:
 On 5/8/06, Bernhard Reiter [EMAIL PROTECTED] wrote:
 cpufreq-set -g ondemand
 fails on linux-2.6.17-rc3 PowerBook5,6, config see below.
 
 dmesg has:
   ondemand governor failed to load due to too long transition latency
 
 same with conservative.
 Both modules are loaded as lsmod shows.
 
 I will probably try to change the .config to make ondemand
 or conservative the default, but this might be a bug nevertheless,
 thus I am reporting.
 
 Any hints?
 
 Maybe this helps:
 http://forums.gentoo.org/viewtopic-t-462417.html

Yes, it does.

JoseJX write there that ondemand and conservative cannot be used
on 7xx/74xx processors because of the time required to switch speeds. 
Powernowd is recommended.

This explains it.
I guess the description or dependencies of those modules could be improved
so it is more clear that they are not useful to me.

Bernhard


pgpBBdbcpi2Qv.pgp
Description: PGP signature


Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-16 Thread Michael Schmitz
  Maybe this helps:
  http://forums.gentoo.org/viewtopic-t-462417.html

 Yes, it does.

 JoseJX write there that ondemand and conservative cannot be used
 on 7xx/74xx processors because of the time required to switch speeds.

The reason seems to be a rather conservative assumption on the latency
('probably takes forever') - does it really take more than 10 ms to switch
speeds? The code path on the 7447A models appears to have a settle delay
of 1 ms after the feature call (which boils down to an OUTB, basically),
another 0.1 ms after the low_choose_7447a_dfs call which shouldn't take
very long either. Am I missing something, Ben?

I'd not be surprised if the latency was as low as 2 ms total - can we
profile this reliably in some way? I guess I'll just set the latency to 5
ms and see what happens ...

 Powernowd is recommended.

Indeed.

Michael


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-16 Thread Benjamin Herrenschmidt
On Tue, 2006-05-16 at 14:54 +0200, Michael Schmitz wrote:
   Maybe this helps:
   http://forums.gentoo.org/viewtopic-t-462417.html
 
  Yes, it does.
 
  JoseJX write there that ondemand and conservative cannot be used
  on 7xx/74xx processors because of the time required to switch speeds.
 
 The reason seems to be a rather conservative assumption on the latency
 ('probably takes forever') - does it really take more than 10 ms to switch
 speeds? The code path on the 7447A models appears to have a settle delay
 of 1 ms after the feature call (which boils down to an OUTB, basically),
 another 0.1 ms after the low_choose_7447a_dfs call which shouldn't take
 very long either. Am I missing something, Ben?

I'm being over conservative bcs of the voltage change

 I'd not be surprised if the latency was as low as 2 ms total - can we
 profile this reliably in some way? I guess I'll just set the latency to 5
 ms and see what happens ...

Last time I looked the on-demand governor was using a workqueue...
anything that has a latency of more than a few dozen usec in a workqueue
shall be shot...

Ben



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-15 Thread Bernhard Reiter
On Mon, May 08, 2006 at 03:07:50PM +0200, Bernhard Reiter wrote:
 cpufreq-set -g ondemand 
 fails on linux-2.6.17-rc3 PowerBook5,6, config see below.
 
 dmesg has:
   ondemand governor failed to load due to too long transition latency

Nobody saw this.
It is all working for you?
Are people using the ondemand or conservative user land governor?

 same with conservative. 
 Both modules are loaded as lsmod shows.
 
 I will probably try to change the .config to make ondemand 
 or conservative the default, but this might be a bug nevertheless,
 thus I am reporting.
 
 Any hints?
 
 Relevant .config:
 
 
 CONFIG_CPU_FREQ=y
 CONFIG_CPU_FREQ_TABLE=y
 CONFIG_CPU_FREQ_DEBUG=y
 CONFIG_CPU_FREQ_STAT=y
 CONFIG_CPU_FREQ_STAT_DETAILS=y
 # CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set
 CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y
 CONFIG_CPU_FREQ_GOV_PERFORMANCE=m
 CONFIG_CPU_FREQ_GOV_POWERSAVE=m
 CONFIG_CPU_FREQ_GOV_USERSPACE=y
 CONFIG_CPU_FREQ_GOV_ONDEMAND=m
 CONFIG_CPU_FREQ_GOV_CONSERVATIVE=m
 CONFIG_CPU_FREQ_PMAC=y
 # CONFIG_PPC601_SYNC_FIX is not set
 



-- 
Professional Service for Free Software (intevation.net)  
The FreeGIS Project   (freegis.org)
FSFE(fsfeurope.org)


pgp40zoAuwCVa.pgp
Description: PGP signature


Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-15 Thread Bin Zhang

On 5/8/06, Bernhard Reiter [EMAIL PROTECTED] wrote:

cpufreq-set -g ondemand
fails on linux-2.6.17-rc3 PowerBook5,6, config see below.

dmesg has:
  ondemand governor failed to load due to too long transition latency

same with conservative.
Both modules are loaded as lsmod shows.

I will probably try to change the .config to make ondemand
or conservative the default, but this might be a bug nevertheless,
thus I am reporting.

Any hints?


Maybe this helps:
http://forums.gentoo.org/viewtopic-t-462417.html


Best regards,
Bin


Relevant .config:


CONFIG_CPU_FREQ=y
CONFIG_CPU_FREQ_TABLE=y
CONFIG_CPU_FREQ_DEBUG=y
CONFIG_CPU_FREQ_STAT=y
CONFIG_CPU_FREQ_STAT_DETAILS=y
# CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set
CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y
CONFIG_CPU_FREQ_GOV_PERFORMANCE=m
CONFIG_CPU_FREQ_GOV_POWERSAVE=m
CONFIG_CPU_FREQ_GOV_USERSPACE=y
CONFIG_CPU_FREQ_GOV_ONDEMAND=m
CONFIG_CPU_FREQ_GOV_CONSERVATIVE=m
CONFIG_CPU_FREQ_PMAC=y
# CONFIG_PPC601_SYNC_FIX is not set








Re: 2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-15 Thread Yves-Alexis Perez
On Mon, 2006-05-15 at 19:16 +0200, Bernhard Reiter wrote:
 On Mon, May 08, 2006 at 03:07:50PM +0200, Bernhard Reiter wrote:
  cpufreq-set -g ondemand 
  fails on linux-2.6.17-rc3 PowerBook5,6, config see below.
  
  dmesg has:
ondemand governor failed to load due to too long transition
 latency

[EMAIL PROTECTED]: sudo cpufreq-set -g ondemand [~]
Error setting new values. Common errors:
- Do you have proper administration rights? (super-user?)
- Is the governor you requested available and modprobed?
- Trying to set an invalid policy?
- Trying to set a specific frequency, but userspace governor is not
available,
   for example because of hardware which cannot be set to a specific
frequency
   or because the userspace governor isn't loaded?

[EMAIL PROTECTED]: dmesg | tail -1 [~]
ondemand governor failed to load due to too long transition latency

[EMAIL PROTECTED]: grep demand /boot/config-2.6.17-rc2 [~]
CONFIG_CPU_FREQ_GOV_ONDEMAND=y

[EMAIL PROTECTED]:
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
conservative ondemand powersave userspace performance

 
 Nobody saw this.
 It is all working for you?
 Are people using the ondemand or conservative user land governor? 
-- 
Yves-Alexis Perez


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



2.6.17-rc3: cpufreq-set -g ondemand does not work

2006-05-08 Thread Bernhard Reiter
cpufreq-set -g ondemand 
fails on linux-2.6.17-rc3 PowerBook5,6, config see below.

dmesg has:
  ondemand governor failed to load due to too long transition latency

same with conservative. 
Both modules are loaded as lsmod shows.

I will probably try to change the .config to make ondemand 
or conservative the default, but this might be a bug nevertheless,
thus I am reporting.

Any hints?

Relevant .config:


CONFIG_CPU_FREQ=y
CONFIG_CPU_FREQ_TABLE=y
CONFIG_CPU_FREQ_DEBUG=y
CONFIG_CPU_FREQ_STAT=y
CONFIG_CPU_FREQ_STAT_DETAILS=y
# CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set
CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y
CONFIG_CPU_FREQ_GOV_PERFORMANCE=m
CONFIG_CPU_FREQ_GOV_POWERSAVE=m
CONFIG_CPU_FREQ_GOV_USERSPACE=y
CONFIG_CPU_FREQ_GOV_ONDEMAND=m
CONFIG_CPU_FREQ_GOV_CONSERVATIVE=m
CONFIG_CPU_FREQ_PMAC=y
# CONFIG_PPC601_SYNC_FIX is not set



pgp50JtA6FeQW.pgp
Description: PGP signature