Get your high perfromance Intel compiled kernel here.

[Thaidog]

In the past there was an effort to use the Intel C++ compiler on Red Hat to compile the 2.6.18 kernel. For the last few months I have been working (along with some friends) with Intel development to get a more recent kernel source going that will compile under Intel's high performance C++ compiler. We have finally got it to compile, Still a work in progress but I have put up a mirror with the modified source and the special wrapper needed to compile it with:

www.linuxdna.com

Feel free to download it and test it yourself!

Right now we are using version 10.0.018 of the Intel compiler to compile with so you will need at least this version or higher to compile with.

***If you are using a higher version of the compiler to work with you will need to make the following changes to a few lines in the source:

1.) Makefile, add libirc_s.a as follows:

--start-group $(vmlinux-main) /opt/intel/cc/10.1.021/lib/libirc_s.a --end-group \

(in the source now it is 10.1.018 - replace with whatever version you are using... in this case 10.1.021)

2.) Edit arch/x86/boot/Makefile,find these lines:

LDFLAGS_setup.elf := -T
$(obj)/setup.elf: $(src)/setup.ld $(SETUP_OBJS) FORCE
$(call if_changed,ld)

Then add: /opt/intel/cc/10.1.021/lib/libirc_s.a after "FORCE".

(in the source now it is 10.1.018 - replace with whatever version you are using... in this case 10.1.021)

3.) In ./scripts/Makefile.modpost,

quiet_cmd_ld_ko_o = LD [M] $@
cmd_ld_ko_o = $(LD) $(LDFLAGS) $(LDFLAGS_MODULE) -o $@ \
- $(filter-out FORCE,$^)
+ $(filter-out FORCE,$^) /opt/intel/cc/10.1.018/lib/libirc_s.a

[Thaidog]

The source is now ready to try! The wiki is for Gentoo but if you know how to compile a kernel you can do this easily

[Thaidog]

Update; Our project just got an article in Linux journal:

http://www.linuxjournal.com/content/...el-cc-compiler

[Thaidog]

Update: We now have a 64bit kernel that is very fast for desktops and multitasking - 40 to over 150% faster than the same gcc kernel!

wget http://www.linuxdna.com/dna-2.6.32.7-intel64-3.patch

wget http://www.linuxdna.com/dna-2.6.33-r...ntel64-2.patch

wget http://www.kernel.org/pub/linux/kern...6.32.7.tar.bz2

It has been tested to work with the latest version 11.1.064 and .059 compilers and kernel 2.6.32.7. earlier .32 kernels *do not* work, fyi!

The patch has been made much easier to use... simply apply the patch and build your kernel like normal. This patch also fixes various icc module problems which should be a big relief to Red Hat type users.

[leigh123linux]

Update: We now have a 64bit kernel that is very fast for desktops and multitasking - 40 to over 150% faster than the same gcc kernel!

wget http://www.linuxdna.com/dna-2.6.32.7-intel64-3.patch

wget http://www.linuxdna.com/dna-2.6.33-r...ntel64-2.patch

wget http://www.kernel.org/pub/linux/kern...6.32.7.tar.bz2

It has been tested to work with the latest version 11.1.064 and .059 compilers and kernel 2.6.32.7. earlier .32 kernels *do not* work, fyi!

The patch has been made much easier to use... simply apply the patch and build your kernel like normal. This patch also fixes various icc module problems which should be a big relief to Red Hat type users.

Your claims are unfounded and the benchmark tests back this up.

http://www.linuxdna.com/benchmarks.pdf

"... boost up to 40% for certain kernel parts and an average boost of 8-9% possible"

Even this claim is optimistic.

Currently, closed source drivers will not install out of the box, though the developers believed this issue should be easily fixable.

And you loose compatibility with other software

[Thaidog]

There are no compatibility problems anymore... And if you look at context switching in the becnhmarks pdf you will see that it is at least over 40% faster than gcc on each trial - that is why it is faster for multitasking. In all of the other tests icc is slightly faster with the exception of one test.

[leigh123linux]

There are no compatibility problems anymore... And if you look at context switching in the becnhmarks pdf you will see that it is at least over 40% faster than gcc on each trial - that is why it is faster for multitasking. In all of the other tests icc is slightly faster with the exception of one test.

I did note that.

[Thaidog]

I did note that.

Thanks Give it a try! If you have a netbook it will make thing's better I guarantee!

[tripleninez]

Update: We now have a 64bit kernel that is very fast for desktops and multitasking - 40 to over 150% faster than the same gcc kernel!

wget http://www.linuxdna.com/dna-2.6.32.7-intel64-3.patch

wget http://www.linuxdna.com/dna-2.6.33-r...ntel64-2.patch

wget http://www.kernel.org/pub/linux/kern...6.32.7.tar.bz2

It has been tested to work with the latest version 11.1.064 and .059 compilers and kernel 2.6.32.7. earlier .32 kernels *do not* work, fyi!

The patch has been made much easier to use... simply apply the patch and build your kernel like normal. This patch also fixes various icc module problems which should be a big relief to Red Hat type users.

i have been reading alot about linuxDNA recently, this is your project?? very interesting stuff.
i had tried to compile a kernel using ICC and didn't have much luck, i couldn't get it to build properly, but maybe i will give it a shot again, sometime soon...

do know if anyone has taken it a step further, patching linuxDNA with ck - BFS scheduler or rt-kernel patches??? i would be curious to know....

in fedora 12, by far the best kernel i use is actually a Zen-kernel using the BFS scheduler. Zen is designed for gentoo and source mage and is really customizable, it comes with alot of extra options in menuconfig and seems to have well-thought out optimization and design...
i have done some comparisons/benchmarks on my various kernels (planet-ccrma, stock fedora-kernel, Zen-kernel-BFS) and the Zen kernel wins everytime...

it would be interesting to see how ICC would improve this kernel.....even an extra 8-9% would be a big gain. as i am trying to get the best performance possible. i suppose when i get my gentoo machine up and running i might try linuxDNA again, with some extra patches - if i can do that without having huge problems!

i use linux for mostly for proaudio, some programming(i am learning C++) and a whole bunch of compiling too, pretty much most things i use, i compile... so the more power/optimization the more i can do!

[kazysmaster]

Is there somewhere a yum repo with Zen-kernel-BFS rpms ? I' too lazy to compile myself

[tripleninez]

Is there somewhere a yum repo with Zen-kernel-BFS rpms ? I' too lazy to compile myself

nope. there isn't. this one you have to compile yourself, even if you were running the same ARCH as me,
i only have kernel modules, drivers, optimization, etc for my machine...even though i do have .rpms of the kernel compiled. so if you happen to own a dell inspiron 6400, with an ATI X1300.....your in luck! lol..

..and Zen-kernel's aren't supported by red-hat or fedora, or most distro's for that matter. just Gentoo, source mage, and yoper linux...

it's not that hard to compile though, the tricky part is figuring out what you need and don't need
in the "menuconfig" or .config before compiling....

i think i will write a guide once i get it perfect!


PS: your too lazy to ever compile??? (if so, your missing out on having real performance)

[Gödel]

What actually will be faster (practical apps) on an intel atom netbook with this kernel?

Will full-screen flash video be smooth?

[leigh123linux]

What actually will be faster (practical apps) on an intel atom netbook with this kernel?

Will full-screen flash video be smooth?

That would qualify as a miracle

[kazysmaster]

Sorry if that's off topic but I found something fedora 11 users might want to try

http://www.vacuumtube.org.uk/folding...esting/x86_64/

[tripleninez]

Sorry if that's off topic but I found something fedora 11 users might want to try

http://www.vacuumtube.org.uk/folding...esting/x86_64/

you are referring to folding forums BFS kernel???

that kernel, i believe is designed for thier purposes - being distributed computing, analyzing genetic data....so i doubt it is ideal for us(depedning on what you use your box for. also, what benefits will you gain from having an un-optimized kernel???

you also should look in thier site as you can get the source code - so posting an x86_64 - fedora 11 only
rpm's isn't very ideal...

no matter what you are still better off making your own kernel and patching it....

...although once i get it perfect - i will be sticking with zen.

no kernel i have used beats my zen-kernel. noticable performance, and banchmarking
don't lie. far better than fedora's or planet ccrma's kernels...