[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: too many deamons by default - F7 test 2 live cd



Rahul Sundaram wrote:
> Jarod Wilson wrote:
>>
>> Everyone please kindly read the cpuspeed initscript before continuing to
>> hypothesize about it. It does indeed already look for necessary support,
>> and silently exits without doing a thing it its not present.
>>
>> /me is the cpuspeed maintainer... :)
> 
> It would do that everytime on bootup though. Right?

Yes. And within a few lines of bash, it'll exit if support isn't found.
The second line of the start function is essentially:

if /sys/devices/system/cpu/cpu*/cpufreq/scaling_driver doesn't exist,
immediately exit silently.

> Can we make it
> disable itself if the support is not there on first run?

Not worth it, if you ask me. This adds unnecessary complexity. And there
are good reasons to try again the next boot. Sometimes things break or
are broken in the kernel or in the bios that prevent cpuspeed from
working one time, and they subsequently get fixed (this actually happens
quite often with newer laptops).

And seriously:

# time service cpuspeed start
Enabling ondemand cpu frequency scaling:                   [  OK  ]

real    0m0.194s
user    0m0.032s
sys     0m0.043s

This is on a dual dual-core system w/hyperthreading enabled (ie 8 "cpus"
to adjust the freq scaling bits on). On a system without support, we'll
be done even faster than that:

# time service cpuspeed start

real    0m0.042s
user    0m0.021s
sys     0m0.024s

Is something in the ballpark of 0.09 seconds really a concern? I know
its all cumulative, but cpu frequency scaling is a Good Thing (save the
planet, save on your energy bill, yadda, yadda, etc), we want to do
everything we can to make sure we enable it whenever possible without
requiring users to fiddle with things. And you can chkconfig it off it
you really don't want it.

-- 
Jarod Wilson
jwilson redhat com


Attachment: signature.asc
Description: OpenPGP digital signature


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]