The funny thing is that I'm Arch user as well, but build my kernel with MNATIVE option as I guess it's up to compiler to decide what optimizations to use.
Anyway, if people's lament persists, I may update graysky's patch and break Intel-specific optimizations for older GCC.
I followed versioning voting inspired by Linus, and I guess I should do the same. So, vote please: https://plus.google.com/105187899411277204747/posts/HNaTndmBCyr
Search Criteria
Package Details: linux-pf-headers 6.11.pf4-1
Package Actions
Git Clone URL: | https://aur.archlinux.org/linux-pf.git (read-only, click to copy) |
---|---|
Package Base: | linux-pf |
Description: | Headers and scripts for building modules for the pf-kernel |
Upstream URL: | https://pfkernel.natalenko.name |
Keywords: | bbr bbr3 kernel ksm linux linux-pf pf-kernel uksm uksmd v4l2loopback zstd |
Licenses: | GPL-2.0-only |
Provides: | linux-pf-headers |
Submitter: | nous |
Maintainer: | post-factum |
Last Packager: | post-factum |
Votes: | 210 |
Popularity: | 0.27 |
First Submitted: | 2011-07-24 12:01 (UTC) |
Last Updated: | 2024-11-06 15:49 (UTC) |
Dependencies (10)
- pahole (pahole-gitAUR)
- bc (bc-ghAUR) (make)
- cpio (cpio-gitAUR) (make)
- gettext (gettext-gitAUR) (make)
- libelf (elfutils-gitAUR) (make)
- pahole (pahole-gitAUR) (make)
- perl (perl-gitAUR) (make)
- python (python37AUR, python311AUR, python310AUR) (make)
- tar (tar-gitAUR, busybox-coreutilsAUR) (make)
- xz (xz-gitAUR) (make)
Required by (0)
Sources (2)
Latest Comments
« First ‹ Previous 1 .. 20 21 22 23 24 25 26 27 28 29 30 .. 106 Next › Last »
post-factum commented on 2015-02-25 20:42 (UTC)
pavbaranov commented on 2015-02-25 14:45 (UTC)
No, certainly not. What Natalenko support in his patch it's his choice. But we are in Arch Linux. In Arch Linux we have gcc 4.9+. You are mainteiner of the most important part of system - kernel and offers it with not proper patch. See: https://github.com/graysky2/kernel_gcc_patch/blob/master/outdated_versions/linux-3.15%2B/gcc-4.2%2B/enable_additional_cpu_optimizations_for_gcc_kernel_v3.15%2B.patch for REQUIREMENTS. Since gcc 4.9 names of flags has been changed. It's impossible to build optimized kernel (any binary apps) with wrong compiler's flags. Then, maybe it's possible to build kernel with wrong flags but this kernel might be unstable or doing some things in a wrong way.
Thaodan commented on 2015-02-25 11:57 (UTC)
it if he wan'ts to support gcc 4.9+ it is his problem.
It does build for intel cpus with gcc optimizations just fine.
pavbaranov commented on 2015-02-25 07:59 (UTC)
It's not a problem of Natalenko. His patch works well i.e. in Chakra, KaOS or Debian. Everywhere where GCC=<4,9 is. But not on Arch with GCC 4.9. This is a problem only with PKGBUILD not with patch. Currently the kernel builds fine with GCC optimizations only for AMD processors. For Intel's processors we can build kernel only without GCC optimizations.
Thaodan commented on 2015-02-25 01:07 (UTC)
Please ping Natalenko for this , to update the patch.
I'll add the patch and fix the pkgbuild when I have time.
Currently the kernel builds just fine using the old patch for haswell, sandybridge etc.
pavbaranov commented on 2015-02-25 00:56 (UTC)
Natalenko's patch uses graysky's GCC patch for GCC prior then 4.9. It means, that it use old -march CFF terminology. Just take a look at: https://wiki.archlinux.org/index.php/Repo-ck#Selecting_the_correct_CPU_optimized_package or https://github.com/graysky2/kernel_gcc_patch for new ones and inspect patch-3.19-pf1 for -march (GCC patch started from 2058 lines of Natalenko's patch). For example, for gcc 4.9 and Intel Nehalem platform proper -march flag is nehalem, but in pf patch it's old, but proper for gcc prior then 4.9, corei7. Wrong CFF flags for gcc 4.9 are for IvyBridge, SandyBridge, Haswell... etc.
When PKGBUILD for your's linux-pf is started, we can choose special optimizations for specific processor. But if somebody has got i.e. SandyBridge, Natalenko's patch try to build kernel with -march=corei7-avx. It's unknown flag for GCC 4.9 and kernel will be incorectly built. In extremally causes it will result in kernel panic.
Then it's impossible to build corectly kernel for (i.e.) Sandy Bridge with GCC optimizations in Arch with this PKGBUILD. I think, that your PKGBUILD should use, for example, sed command for correct this.
Thaodan commented on 2015-02-24 23:04 (UTC)
Why please explain? I you have something to add/complain please explain it in detail I can't do anything whithout valid infos.
pavbaranov commented on 2015-02-24 22:56 (UTC)
Natalenko's pf-patch uses old graysky's GCC patch (for GCC prior then 4.9). In Arch we have GCC 4.9 now, then it's impossible to build proper kernel enabling CPU-specific compile option in the Linux kernel for some Intel processors.
The PKGBUILD should changes pf-patch before it's implemented.
Thaodan commented on 2015-02-24 11:53 (UTC)
nous: thanks, I added a workaround from gravsky for this, without dvb it builds just fine now.
Now I just runned qconf and saved the new config. I think I have missed the changes in ARCH config .
nous commented on 2015-02-24 09:53 (UTC)
@Thaodan: About the security stuff, perhaps you accidentally copied over the .config from 3.18*? In the diff @AnAkkk posted seems so.
Pinned Comments
post-factum commented on 2023-09-25 20:30 (UTC) (edited on 2024-10-08 14:21 (UTC) by post-factum)
Official binary builds for various x86_64 ISA levels are available here.