There's Another Linux Kernel Power Problem

[fewt]

Just like with the previous bug, your results are invalid for the general case, because you use some weird power thing, not the default configuration.

It so happens that most users will use the defaults, not the weird power thing. So the default tests carry much more value than tests with some custom config.

My tests aren't invalid - as I have said many times, the kernel isn't responsible for tuning itself for power management.

[fewt]

Results from one kernel, without comparison to an earlier kernel, and without full detailed hardware info, are meaningless. Also, what about the kernel bug linked twice already in this thread?

Specs:

Processor: Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz
Memory: 8192MB
Hard Drive: Seagate Momentus 7200 250GB (ST9250412AS)
Video: Intel(R) Sandybridge Mobile (GT2+)
Network: Intel Corporation 82579LM Gigabit Network Connection, Intel Corporation 6000 Series Gen2 (uses 1 watt)
OS: Fuduntu 2013.1
Desktop: GNOME 2
Display Driver: Intel 2.20.10
OpenGL: Mesa 8.0.4
Compiler: GCC 4.6.3
Filesystem: ext4
Resolution: 1600x900

...

[curaga]

My tests aren't invalid - as I have said many times, the kernel isn't responsible for tuning itself for power management.

Yes, it's not the kernel's responsibility to tune itself for any particular workload.

But if there's a clear, objective regression with the defaults, is that not a bug? If power usage increases, but performance does not, is that not a bug, even if you don't see it in a custom configuration?

[tizbac]

It should not be hard to bisect, maybe i can do it sometime in the week-end, it only takes some tens of full kernel builds

[fewt]

Yes, it's not the kernel's responsibility to tune itself for any particular workload.

But if there's a clear, objective regression with the defaults, is that not a bug? If power usage increases, but performance does not, is that not a bug, even if you don't see it in a custom configuration?

It is neither a regression, nor a bug - even if Michael says it is. You simply can't look at things that way - because it doesn't make sense.

Using this sort of logic it would be safe to say that if kernel 3.7 used more power than 1.0 that it was a regression - when it would just be due to the additional work the kernel needed to do to support new functionality in the kernel.

It is completely silly to perform these sorts of tests, they benefit no-one except Mr. Larabel.

[curaga]

I find it really hard to believe 3.6 enables some new functionality that takes 6-8 watts.

[fewt]

I find it really hard to believe 3.6 enables some new functionality that takes 6-8 watts.

There isn't any new functionality that takes 6-8 watts, nor is there any bug that takes 6-8 watts. Phoronix needed more dimes, hence this article.

[halo9en]

There isn't any new functionality that takes 6-8 watts, nor is there any bug that takes 6-8 watts. Phoronix needed more dimes, hence this article.

I wonder why he's still on this forum is all he does is complain about Larabel -- oh wait, it's called trolling.

Now, can we get back on topic?

Btw, links posted so far:
https://bbs.archlinux.org/viewtopic.php?id=149901
https://bbs.archlinux.de/viewtopic.php?id=22089 (German)
https://bbs.archlinux.org/viewtopic.php?id=150743
https://bugzilla.kernel.org/show_bug.cgi?id=48721

[fewt]

I wonder why he's still on this forum is all he does is complain about Larabel -- oh wait, it's called trolling.

Now, can we get back on topic?

Btw, links posted so far:
https://bbs.archlinux.org/viewtopic.php?id=149901
https://bbs.archlinux.de/viewtopic.php?id=22089 (German)
https://bbs.archlinux.org/viewtopic.php?id=150743
https://bugzilla.kernel.org/show_bug.cgi?id=48721

Disagreeing isn't trolling moron. Your links are unrelated to the Phoronix article, with one exception - probably the forum post that caused this article to be written.

[halo9en]

Disagreeing isn't trolling moron. Your links are unrelated to the Phoronix article, with one exception - probably the forum post that caused this article to be written.

Ok, I'll bite, troll. Complaining about ad revenues and repeating post after post that Larabel is <insert insult here>, spamming about your little distro and your websites, and resorting to personal insults qualifies as troll behavior that serves to shut down any civil dialogue. There are other websites if you don't like this one, you know.