Pay For Faster Linux Kernel Performance? There's Patches For That

[uid313]

Wouldn't surprise me if this work is inspired by http://www.linuxdna.com/
In which case the need for patches would be definitive. However, from what I can read in the comments, the kernel has built-in power management. Meaning that using TLP will conflict with existing solutions.
This is great work that is in need of a conflict list.

Pozdrowienia!

What is this Linux DNA thing?

[ext73]

I'm polish and I visit dobreprogramy.pl pretty often.

Those guys who make this kernel are well known among community for spamming forums and comment section with shameless advertisments. In general all of this is kinda suspicious and I honestly would avoid prebuild kernels from unknown source which said group is.

I'd say that until some main kernel developer or even Linus himself comment on this - avoid this shit.

I see that we are dealing with so-called. "expert specialist" ... from everything before you further write a such "nonsense" show us your work, what you can do ... nothing ?

[ext73]

I'm a native Polish speaker.

I've spent almost 10 minutes on the site, and didn't find anything interesting. Looks like a scam to me. Just prebuild kernels, with flags for certain CPUs. Also Noop is used by default for SSD drives. Can't find anything else. It's either a very bad website, or a strange scam.

So ... audit done in 10 minutes - it's called 'professionalism' ... despite my 14 years of experience in audit, I can not do this within such time - respect ! or an embarrassment ?

[V10lator]

I'll just quote

"What other linux will let you have a vector optimized wordprocessor?"

http://fun.irq.dk/funroll-loops.org/

That's not an answer to my question or do you want to say the patch wasn't merged cause it's optimized to much? With that attitude there should be no options for K8 and all the other AMD/Intel variants, too. Generic x86_64 should be enough... And why did they talk to the compiler guys then?

[Gloomwing]

Jeśli chodzi o kwestię techniczną, to sprawa, choć na podstawie TYLKO jednego źr?dła (co mam nadzieję się zmieni) jest jasna. W kwestii etycznej, też nie mam zastrzeżeń, skoro to jest dużo pracy, a źr?dła są wciąż dostępne (a przecież nie tylko o kernel się rozchodzi). Jeśli chodzi o płatność - 50 zł, to nie jest dużo jak za rok subskrypcji, a i bez tego pozostaje opcja wersji bezpłatnych. Płatność de facto powinna być rozpatrywana jako płatność za lenistwo - nie każdemu się chce zagłębiać w kwestie co i jak. Proponuję zarzucić tę jałową rozmowę i wykonać więcej test?w, uwzględniających darmowe też wersje.

From technical point of view the case is closed - unfortunetly based on only one test made by dobreprogramy.pl (hopefully it will soon be changed). In ethical way, in my opinion there is nothing wrong still - it is huge amount of work, and sources are still available (and kernel work is not the only done). Case of payment - 50 PLN is not too much for one-year subscription, anyway there is always free of charge option. Payment should be seen as payment for being lazy - not every one wants to learn how to do all of this stuff. I suggest to forget about all this useless talk and do more test, including free of charge ones.

[SSIV]

What is this Linux DNA thing?

Linux DNA is a set of linux kernel patches which rid vanilla 2.6 kernel of incompatibilities with Intel's Compiler.
The need for this is because some kernel code is written syntax that only GCC recognizes.

If someone did a "Linux DNA" for linux kernel 3.17+ that'd be great and profitable for everybody.
What remains to see is whether the source ext73 gave us (link here) compiler with ICC.
If it does, then this is most certainly a Linux DNA spin-off.

[Guest]

That's not an answer to my question or do you want to say the patch wasn't merged cause it's optimized to much? With that attitude there should be no options for K8 and all the other AMD/Intel variants, too. Generic x86_64 should be enough... And why did they talk to the compiler guys then?

Most modern cpu specific optimisations in kernel are pointless. IIRC kernel developers are encouraged to not use event floating-point.

You can get best optimisations by changing code algorithms and use better designed data structures - not by adding some magic switch to a compiler.

I would consider buying a kernel with better cpu scheduler, better i/o scheduler, improved network stack. Compiler magic is for gentoo funrollops folks.

[tuxd3v]

... Generic x86_64 should be enough... And why did they talk to the compiler guys then?

Well,
is you talk about that x86_64 is not generic... amd64 is gen?ric!!Since the 64bits in x86 cores is a AMD thing...

What I want to know is if I will benefict from having a kernel from this guys...I have a AMd phenom II x3 N830 in my laptop...

[gutigen]

I see that we are dealing with so-called. "expert specialist" ... from everything before you further write a such "nonsense" show us your work, what you can do ... nothing ?

Show me your upstream kernel patches then we we'll talk. I'm serious, if you got some work upstreamed, then I may have change my opinion about your work (or at least it would be a start). Otherwise your response only confirms my suspicion that you are nothing else than a scam or in best case scenario - a security risk.

Put yourself in other Linux users position - would you trust a kernel coming from some guy no one knows?

And no, I'm no expert or even a developer as a matter of fact, but careful about where my software comes from. So far you, judging by dobreprogramy.pl comment posts and website itself, have not given me a single reason to trust you.

[SSIV]

So ... audit done in 10 minutes - it's called 'professionalism' ... despite my 14 years of experience in audit, I can not do this within such time - respect ! or an embarrassment ?

I see that we are dealing with so-called. "expert specialist" ... from everything before you further write a such "nonsense" show us your work, what you can do ... nothing ?

If I were you I'd value my customer relations better. Congratulations on being able to write cynical comments.
Also, where are the patches? If your work is a patched vanilla/debian kernel, then you're bound to have patches. Given that you obviously have a version control system like git, because you're very professional, then providing such patches should be a piece of cake.

If you're using the ICC then the patches will modify the core out of necessity. In which case it is important that an experienced 3rd party evaluates the kernel - that LTP and Autotest results are published.
Hopefully the code doesn't resemble the Indian electrical grid.