Burning Through Power: Linux Regressions Found

[DanL]

xine-lib VAAPI support.

http://crystalhd.svn.sourceforge.net...lib-1.2-vaapi/

Phoronix doesn't tell you that there is also VAAPI support for xine-lib. Thx to this censored board where to information depends on phoronix, they wanna let you know or not.

This is completely unrelated. At any rate, it looks like what you're linking to is crystalhd or Broadcom graphics. We already knew they support VA-API (and it was mentioned on this site.)

[marwi509]

Awesome.

[crumja]

No, Michael is definitely not the only person to work on an issue that affects lots of people. But he behaves like he is. He also forgets that his whole "enterprise" is based on free software and he didn't pay them anything back, he only tries to make money out of it.

It only appears that he is because he's reaping the benefits of a huge community but can only contribute what he can with limited time. Other companies such as Oracle and Red Hat can contribute much more because of money and manpower.

With that said, what is wrong with trying to make money out of tech reporting? LWN does it. Anandtech does it. Subscribe or sponsor Michael if you want to cut down on hyping headlines for revenue.

[RealNC]

With that said, what is wrong with trying to make money out of tech reporting? LWN does it. Anandtech does it. Subscribe or sponsor Michael if you want to cut down on hyping headlines for revenue.

Do you realize that this sounds like a threat?

[birdie]

Additionally, hell yes, to maximize traffic and ad revenues off of this work due to the amount of time and resources I've spent tracking down these Linux kernel power bugs and other problems on top of my already maxed workload;

Michael, I like you!

[Draconiak]

Wow , >10% more of battery is enought reason for compiling a kernel, when 3.0 become stable, will be a pleasure to test the behavior of your patchs if it help somehow, (opensuse 11.4 - i3 notebooks), thanks for the great job Michel.

[crumja]

Do you realize that this sounds like a threat?

Well, you obviously have a different interpretation of tone than most individuals I've interacted with. It's a simple statement of facts. No one is pointing a gun at your head forcing you to subscribe to phoronix. Also, no one can compel Michael to do something other than what he wants. However, if you want Michael to do something, you have to incentivize him, with money. That's mutually agreed upon exchange of goods and/or services - just the way the world works.

Stated another way, Linux is available for free. You, or Oracle or CentOS, can package up Red Hat's version and distribute/sell it. You have to release all the changes that the license stipulates but you don't have to go a step further. You don't have to pay the community back anything. Doing so should be done on a cost-benefit analysis. Ask yourself: the "goodwill" gained in the community from that work - is it enough to compensate for losing revenue. Keeping a good balance is the difference between ending up like Google vs ending up like Novell.

[bug77]

Michael has no clue that a site should place its readers before their adds sponsors?

If you treat your readers like this, what the hell do you expect from them in return? You want add revenue? Be a good website then that people visit often and want to read. If you suck, you don't deserve revenue.

I too was surprised Michael didn't realize your world will end if you get the details about this regression right now.
Come on, it's not like the delay will cause the fix will miss a merge window or something.

[crumja]

I too was surprised Michael didn't realize your world will end if you get the details about this regression right now.
Come on, it's not like the delay will cause the fix will miss a merge window or something.

Another thing to note is that Michael is getting heavy flak from the readership for "sensational" postings without much substance (so far). AFAICT, the kernel development community has also chosen to ignore this issue, not even opening up discussion about it, except perhaps to say, "another piece of Phoronix junk".

Thus, there is a lot of credibility on the line here. Either Michael discovers a regression and submits patches/convinces the subarchitecture author to fix the issue or he will lose much readership over hyping up something that doesn't pan out.

So far, the evidence seems to be weighted towards Michael's side. Tomshardware has confirmed battery run time reductions. Cyanogen is apparently staying put on 2.6.37 for now. Michael has apparently found the regression (putting even more credibility on the line) and is following up on it.

Yes, Phoronix does hype up headlines sometimes. Maybe some benchmarks are biased and not all tweaks are used. However, it has not, in my experience, published something categorically untrue or been intentionally deceptive.

Now, a good piece of investigational journalism, and something I'm dying to know, is why the issue has gotten so much press and mindshare in the user base (see bug reports in distributions, phoronix) but has almost no traction with developers. It seems suspiciously like the slowdown in responsiveness under heavy HD use that was denied and ignored for years, despite heavy user testimonial.

[Michael]

The 'workaround' I have come up with for this issue is very easy to do and tomorrow morning anyone who's affected can quickly and easily try it out... Then just see your battery life. If your system was affected, you should quickly see the difference. Only if you have a really bad BIOS may you still have problems as so far it's working on every one of my affected systems. Also forming a queue of other likely Linux power optimizations.