View previous topic :: View next topic |
Author |
Message |
kaosone Guru
Joined: 01 Feb 2004 Posts: 446
|
Posted: Wed Dec 19, 2007 11:19 am Post subject: |
|
|
i'm not a big fan of cflags tuning, someone benchmarked various cflags configurations, and the thing is some program run better with some cflags, others run slower with the same configuration
so imho it's better keep it standard |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
rmh3093 Advocate
Joined: 06 Aug 2003 Posts: 2138 Location: Albany, NY
|
Posted: Wed Dec 19, 2007 4:05 pm Post subject: |
|
|
kernelOfTruth wrote: | Quote: | lib/genetic.c:383: warning: ‘genetic_ns_award_top_parents’ defined but not used |
waninkoko, could you please have a look at the following 'undefined' warning whether it hurts performance, please ?
it would be bad if the top parents would not be awarded - would it ? |
that means nothing, i've been very intimate with genetics in the past, the only cpuscheduler that has ever "awarded its top parents" was when I had genetic RSDL working......
.... it actually wouldnt be to hard to implement genetic_ns_award_top_parents in CFS.... if you understand how genetic libs works, basically what that command does it if a set of phenotypes performs well it tells them to make more children, so future phenotypes are generated based on a known working phenotype _________________ Do not meddle in the affairs of wizards, for they are subtle and quick to anger. |
|
Back to top |
|
|
Spaulding Apprentice
Joined: 16 Apr 2006 Posts: 159 Location: /dev/vagina
|
Posted: Wed Dec 19, 2007 4:54 pm Post subject: |
|
|
Code: |
CC kernel/sched.o
kernel/sched.c:387:2: error: #endif without #if
make[1]: *** [kernel/sched.o] Błąd 1
make: *** [kernel] Błąd 2
|
please fix it |
|
Back to top |
|
|
AYBABTU n00b
Joined: 17 Jun 2006 Posts: 31
|
Posted: Wed Dec 19, 2007 5:06 pm Post subject: |
|
|
Hi all!
I've got a trouble: ati-drivers fails to emerge
Code: | *
* ERROR: x11-drivers/ati-drivers-8.433 failed.
* Call stack:
* ebuild.sh, line 1701: Called dyn_compile
* ebuild.sh, line 1039: Called qa_call 'src_compile'
* ebuild.sh, line 44: Called src_compile
* ati-drivers-8.433.ebuild, line 191: Called linux-mod_src_compile
* linux-mod.eclass, line 518: Called die
* The specific snippet of code:
* emake HOSTCC="$(tc-getBUILD_CC)" CC="$(get-KERNEL_CC)" LDFLAGS="$(get_abi_LDFLAGS)" \
* ${BUILD_FIXES} ${BUILD_PARAMS} ${BUILD_TARGETS} \
* || die "Unable to make ${BUILD_FIXES} ${BUILD_PARAMS} ${BUILD_TARGETS}."
* The die message:
* Unable to make GCC_VER_MAJ=4 KVER=2.6.24-rc5-zen1ksn-g3a76be92 KDIR=/usr/src/linux kmod_build.
*
* If you need support, post the topmost build error, and the call stack if relevant.
* A complete build log is located at '/var/tmp/portage/x11-drivers/ati-drivers-8.433/temp/build.log'. |
Any idea?
P.s: thanks kaosone for .config, i've used parts for my config |
|
Back to top |
|
|
Waninkoko Guru
Joined: 13 May 2005 Posts: 549
|
Posted: Wed Dec 19, 2007 5:15 pm Post subject: |
|
|
CzErYnA wrote: | Code: |
CC kernel/sched.o
kernel/sched.c:387:2: error: #endif without #if
make[1]: *** [kernel/sched.o] Błąd 1
make: *** [kernel] Błąd 2
|
please fix it |
Fixed. |
|
Back to top |
|
|
oRDeX Veteran
Joined: 19 Oct 2003 Posts: 1325 Location: Italy
|
Posted: Wed Dec 19, 2007 5:19 pm Post subject: |
|
|
AYBABTU wrote: | Hi all!
I've got a trouble: ati-drivers fails to emerge
Code: | *
* ERROR: x11-drivers/ati-drivers-8.433 failed.
* Call stack:
* ebuild.sh, line 1701: Called dyn_compile
* ebuild.sh, line 1039: Called qa_call 'src_compile'
* ebuild.sh, line 44: Called src_compile
* ati-drivers-8.433.ebuild, line 191: Called linux-mod_src_compile
* linux-mod.eclass, line 518: Called die
* The specific snippet of code:
* emake HOSTCC="$(tc-getBUILD_CC)" CC="$(get-KERNEL_CC)" LDFLAGS="$(get_abi_LDFLAGS)" \
* ${BUILD_FIXES} ${BUILD_PARAMS} ${BUILD_TARGETS} \
* || die "Unable to make ${BUILD_FIXES} ${BUILD_PARAMS} ${BUILD_TARGETS}."
* The die message:
* Unable to make GCC_VER_MAJ=4 KVER=2.6.24-rc5-zen1ksn-g3a76be92 KDIR=/usr/src/linux kmod_build.
*
* If you need support, post the topmost build error, and the call stack if relevant.
* A complete build log is located at '/var/tmp/portage/x11-drivers/ati-drivers-8.433/temp/build.log'. |
Any idea?
P.s: thanks kaosone for .config, i've used parts for my config |
you should report the part of message before this.
where the error is |
|
Back to top |
|
|
hoacker Guru
Joined: 04 Aug 2007 Posts: 505 Location: Bürstadt, Germany
|
Posted: Wed Dec 19, 2007 5:28 pm Post subject: |
|
|
Next error:
Code: | CC [M] sound/pci/hda/patch_realtek.o
sound/pci/hda/patch_realtek.c:7164: error: 'ALC883_MITAC' undeclared here (not in a function)
make[3]: *** [sound/pci/hda/patch_realtek.o] Error 1
make[2]: *** [sound/pci/hda] Error 2
make[1]: *** [sound/pci] Error 2
make: *** [sound] Error 2
|
|
|
Back to top |
|
|
AYBABTU n00b
Joined: 17 Jun 2006 Posts: 31
|
Posted: Wed Dec 19, 2007 5:45 pm Post subject: |
|
|
oRDeX wrote: |
you should report the part of message before this.
where the error is |
Ops, sorry
Code: | /var/tmp/portage/x11-drivers/ati-drivers-8.433/work/common/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:3047: warning: ‘return’ with a value, in function returning void
/var/tmp/portage/x11-drivers/ati-drivers-8.433/work/common/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c: In function ‘KAS_SlabCache_Initialize’:
/var/tmp/portage/x11-drivers/ati-drivers-8.433/work/common/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:5544: warning: cast from pointer to integer of different size
make[2]: *** [/var/tmp/portage/x11-drivers/ati-drivers-8.433/work/common/lib/modules/fglrx/build_mod/2.6.x/firegl_public.o] Error 1
make[1]: *** [_module_/var/tmp/portage/x11-drivers/ati-drivers-8.433/work/common/lib/modules/fglrx/build_mod/2.6.x] Error 2
make[1]: Leaving directory `/usr/src/zen-sources'
make: *** [kmod_build] Error 2
|
|
|
Back to top |
|
|
Waninkoko Guru
Joined: 13 May 2005 Posts: 549
|
Posted: Wed Dec 19, 2007 6:10 pm Post subject: |
|
|
hoacker wrote: | Next error:
Code: | CC [M] sound/pci/hda/patch_realtek.o
sound/pci/hda/patch_realtek.c:7164: error: 'ALC883_MITAC' undeclared here (not in a function)
make[3]: *** [sound/pci/hda/patch_realtek.o] Error 1
make[2]: *** [sound/pci/hda] Error 2
make[1]: *** [sound/pci] Error 2
make: *** [sound] Error 2
|
|
Fixed. |
|
Back to top |
|
|
Veldrin Veteran
Joined: 27 Jul 2004 Posts: 1945 Location: Zurich, Switzerland
|
Posted: Wed Dec 19, 2007 6:13 pm Post subject: |
|
|
another error
Code: | make[2]: `scripts/unifdef' is up to date.
make[3]: *** No rule to make target `/usr/src/linux-2.6.24-rc5-zen1/include/linux/ticable.h', needed by `/usr/src/linux-2.6.24-rc5-zen1/usr/include/linux/ticable.h'. Stop.
make[2]: *** [linux] Error 2
make[1]: *** [headers_install] Error 2
make: *** [vmlinux] Error 2
|
there is no file called ticable.h...
cheers
V. |
|
Back to top |
|
|
hoacker Guru
Joined: 04 Aug 2007 Posts: 505 Location: Bürstadt, Germany
|
Posted: Wed Dec 19, 2007 6:18 pm Post subject: |
|
|
Waninkoko wrote: | hoacker wrote: | Next error:
Code: | CC [M] sound/pci/hda/patch_realtek.o
sound/pci/hda/patch_realtek.c:7164: error: 'ALC883_MITAC' undeclared here (not in a function)
|
|
Fixed. |
Many thanks. Compiles fine... |
|
Back to top |
|
|
Waninkoko Guru
Joined: 13 May 2005 Posts: 549
|
Posted: Wed Dec 19, 2007 6:43 pm Post subject: |
|
|
Veldrin wrote: | another error
Code: | make[2]: `scripts/unifdef' is up to date.
make[3]: *** No rule to make target `/usr/src/linux-2.6.24-rc5-zen1/include/linux/ticable.h', needed by `/usr/src/linux-2.6.24-rc5-zen1/usr/include/linux/ticable.h'. Stop.
make[2]: *** [linux] Error 2
make[1]: *** [headers_install] Error 2
make: *** [vmlinux] Error 2
|
there is no file called ticable.h...
cheers
V. |
It isn't even in linux-2.6.git tree...
Which command did you execute? Post your .config |
|
Back to top |
|
|
hoacker Guru
Joined: 04 Aug 2007 Posts: 505 Location: Bürstadt, Germany
|
Posted: Wed Dec 19, 2007 6:48 pm Post subject: |
|
|
Veldrin wrote: |
Code: |
`/usr/src/linux-2.6.24-rc5-zen1/usr/include/linux/ticable.h'. Stop.
|
there is no file called ticable.h...
|
There is /usr/include/linux/ticable.h on my machine. Wrong include path?
//EDIT: that file exists in rc4-zen1
Last edited by hoacker on Wed Dec 19, 2007 6:52 pm; edited 1 time in total |
|
Back to top |
|
|
Waninkoko Guru
Joined: 13 May 2005 Posts: 549
|
Posted: Wed Dec 19, 2007 6:52 pm Post subject: |
|
|
Veldrin wrote: | another error
Code: | make[2]: `scripts/unifdef' is up to date.
make[3]: *** No rule to make target `/usr/src/linux-2.6.24-rc5-zen1/include/linux/ticable.h', needed by `/usr/src/linux-2.6.24-rc5-zen1/usr/include/linux/ticable.h'. Stop.
make[2]: *** [linux] Error 2
make[1]: *** [headers_install] Error 2
make: *** [vmlinux] Error 2
|
there is no file called ticable.h...
cheers
V. |
According to this http://repo.or.cz/w/linux-2.6/zen-sources.git?a=commit;h=755271358cc401eb3db0db52b2c8fb8d71ae4d8f this error is already fixed. |
|
Back to top |
|
|
Veldrin Veteran
Joined: 27 Jul 2004 Posts: 1945 Location: Zurich, Switzerland
|
Posted: Wed Dec 19, 2007 7:01 pm Post subject: |
|
|
I disabled CONFIG_HEADERS_CHECK, and it builds.
I cannot upload my config to my webspace - they are doing some maintenance...
Leave it for now, I check later, if it does not work...
cheers
V.
PS. I have a ticable.h in /usr/include/linux too, but this comes AFAIK form linux-headers, and is not directly related to the current kernel |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Wed Dec 19, 2007 7:04 pm Post subject: |
|
|
Quote: | * Preparing fglrx module
make -j5 HOSTCC=i686-pc-linux-gnu-gcc CC=i686-pc-linux-gnu-gcc LDFLAGS= GCC_VER_MAJ=4 KVER=2.6.24-rc5-zen1-dirty KDIR=/usr/src/linux kmod_build
/usr/bin/make -C /usr/src/linux M=/var/tmp/paludis/x11-drivers/ati-drivers-8.433-r10/work/common/lib/modules/fglrx/build_mod/2.6.x modules
make[1]: Entering directory `/usr/src/linux-2.6.24-rc5-zen1'
CC [M] /var/tmp/paludis/x11-drivers/ati-drivers-8.433-r10/work/common/lib/modules/fglrx/build_mod/2.6.x/firegl_public.o
LD [M] /var/tmp/paludis/x11-drivers/ati-drivers-8.433-r10/work/common/lib/modules/fglrx/build_mod/2.6.x/fglrx.o
Building modules, stage 2.
MODPOST 1 modules
WARNING: could not find /var/tmp/paludis/x11-drivers/ati-drivers-8.433-r10/work/common/lib/modules/fglrx/build_mod/2.6.x/.libfglrx_ip.a.GCC4.cmd for /var/tmp/paludis/x11-drivers/ati-drivers-8.433-r10/work/common/lib/modules/fglrx/build_mod/2.6.x/libfglrx_ip.a.GCC4
FATAL: modpost: GPL-incompatible module fglrx.ko uses GPL-only symbol '__rcu_read_lock'
make[2]: *** [__modpost] Error 1
make[1]: *** [modules] Error 2
make[1]: Leaving directory `/usr/src/linux-2.6.24-rc5-zen1'
make: *** [kmod_build] Error 2
!!! ERROR in x11-drivers/ati-drivers-8.433-r10:
!!! In linux-mod_src_compile at line 3433
!!! Unable to make GCC_VER_MAJ=4 KVER=2.6.24-rc5-zen1-dirty KDIR=/usr/src/linux kmod_build.
!!! Call stack:
!!! * linux-mod_src_compile (/var/tmp/paludis/x11-drivers/ati-drivers-8.433-r10/temp/loadsaveenv:3433)
!!! * src_compile (/var/tmp/paludis/x11-drivers/ati-drivers-8.433-r10/temp/loadsaveenv:4302)
!!! * ebuild_f_compile (/usr/libexec/paludis/src_compile.bash:51)
!!! * ebuild_main (/usr/libexec/paludis/ebuild.bash:431)
!!! * main (/usr/libexec/paludis/ebuild.bash:449)
diefunc: making ebuild PID 2111 exit with error
die trap: exiting with error.
Install error:
* In program paludis --log-level qa --dl-deps-default discard --install x11-drivers/ati-drivers --add-to-world-spec ( x11-drivers/ati-drivers ):
* When performing install action from command line:
* When executing install task:
* When installing 'x11-drivers/ati-drivers-8.433-r10:0::sabayon':
* Install error: Install failed for 'x11-drivers/ati-drivers-8.433-r10:0::sabayon'
Resume command: paludis --log-level qa --install --add-to-world-spec '( x11-drivers/ati-drivers )' --dl-deps-default discard '=x11-drivers/ati-drivers-8.433-r10:0::sabayon' |
preemptible rcu seems to be incompatible with ati-drivers
these f** ati-drivers are a pain in the ass - when will they be at the level of nvidia-drivers finally ?
update:
solution: http://gentoo-wiki.com/HOWTO_ATI_Drivers#Build_ati-drivers_on_rt-kernels_failed
sorry for those forceful words, btw _________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
AYBABTU n00b
Joined: 17 Jun 2006 Posts: 31
|
Posted: Wed Dec 19, 2007 8:29 pm Post subject: |
|
|
Ok solved |
|
Back to top |
|
|
ilikenwf Tux's lil' helper
Joined: 04 Oct 2007 Posts: 119
|
Posted: Thu Dec 20, 2007 3:19 am Post subject: |
|
|
kaosone wrote: | i'm not a big fan of cflags tuning, someone benchmarked various cflags configurations, and the thing is some program run better with some cflags, others run slower with the same configuration
so imho it's better keep it standard |
Ummm...I just tried the AMD64 build I made with those flags...and...wow...it's amazing. SO FAST! I did install preload, which makes it feel faster (didn't test without preload, so that is probably why firefox launches in half the time, but hey...) The kernel feels good, too. _________________ Make The Fonts Pretty In KDE
Archlinux User |
|
Back to top |
|
|
gimpel Advocate
Joined: 15 Oct 2004 Posts: 2720 Location: Munich, Bavaria
|
Posted: Thu Dec 20, 2007 8:47 am Post subject: |
|
|
ilikenwf wrote: | kaosone wrote: | i'm not a big fan of cflags tuning, someone benchmarked various cflags configurations, and the thing is some program run better with some cflags, others run slower with the same configuration
so imho it's better keep it standard |
Ummm...I just tried the AMD64 build I made with those flags...and...wow...it's amazing. SO FAST! I did install preload, which makes it feel faster (didn't test without preload, so that is probably why firefox launches in half the time, but hey...) The kernel feels good, too. |
You really used -mfpmath=387 on an amd64? _________________ http://proaudio.tuxfamily.org/wiki - pro-audio software overlay
|
|
Back to top |
|
|
ilikenwf Tux's lil' helper
Joined: 04 Oct 2007 Posts: 119
|
Posted: Thu Dec 20, 2007 8:54 am Post subject: |
|
|
gimpel wrote: | ilikenwf wrote: | kaosone wrote: | i'm not a big fan of cflags tuning, someone benchmarked various cflags configurations, and the thing is some program run better with some cflags, others run slower with the same configuration
so imho it's better keep it standard |
Ummm...I just tried the AMD64 build I made with those flags...and...wow...it's amazing. SO FAST! I did install preload, which makes it feel faster (didn't test without preload, so that is probably why firefox launches in half the time, but hey...) The kernel feels good, too. |
You really used -mfpmath=387 on an amd64? |
Yes...the reasoning (from the guy who made a suggestion...his name: Martigen...) is below...one thing: The only flag that wouldn't work is -ftracer, and I am thinking I should try gcc 4.3 and see if it fixes this...
Martigen's Reasoning (he's been doing this stuff a long time btw):
Quote: | Hehe, saw the Gentoo thread on the cflags -- you will always get multiple sides! Some comments:
buddabrod said:
OMG ricer flags O.O..
-mfpath=387 is contrapoductive... for x86_64 sse is standard since 387 would slow things down.
Wrong... FP math isn't used at all in the kernel, and so SSE isn't used. If you check the flags passed to the kernel (make V=1) you'll see SSE, SSE2, SSE3 and MMX are all explicitly disabled to ensure these aren't used by mistake. Essentially, it should only leave 387 but the -mfpmath=387 flag makes explicitly sure (just as SSE etc are explicitly disabled). And yes, -ffast-math works in tandem with it
kaosone said:
the kernel size would increase, i personally i don't know all the flag mentioned, i know for sure that -funroll-all-loops is not a really good idea (increased kernel size) and same thing for -maccumulate-outgoing-args
-funroll-all-loops is a bad idea, but that's not in the flag I recommended I recommended -funroll-loops, which selectively unrolls those loops that can benefit from being unrolled, and won't unroll those that won't (which -funroll-all-loops forces). Tranquicool backs this up too. Yes the kernel size will increase with -funroll-loops and -maccumulate-outgoing-args, and the tradeoff is speed.
kaosone said:
i'm not a big fan of cflags tuning, someone benchmarked various cflags configurations, and the thing is some program run better with some cflags, others run slower with the same configuration
so imho it's better keep it standard
*sigh* yes this can be true if you use the same set of flags for your entire system, where you blanket flags for programs that may or may not benefit from them, but that's not what we're doing here. These flags are chosen specifically for the kernel -- that's what optimising is about!
As to your question, which is better, smaller kernel or faster code? That depends on the application. Os is primarily used to make small code to run on embedded systems -- less memory, and CPUs with small caches. This doesn't apply to desktop systems, so at a bare minimum O2 should be used (and is default for the kernel). For those that advocated Os for *loading* speed since smaller binaries load faster well -- yes, if your entire system is Os you'll notice a small speed increase due to the hundreds of smaller binaries. But again, this is just the kernel -- and ask yourself in this age of 60MB/s drives -- do you think you'd notice if a 1.8M kernel really loaded slower than a 1.3M kernel?
Anyway, one way to see if you're comfortable with these flags is to run a kernel with them on for a week, and just see how you go! |
_________________ Make The Fonts Pretty In KDE
Archlinux User |
|
Back to top |
|
|
Gregoire Apprentice
Joined: 15 Apr 2006 Posts: 292
|
Posted: Thu Dec 20, 2007 9:17 am Post subject: |
|
|
So it could be better to use -funroll-loops and -maccumulate-outgoing-args to build the kernel ?
I never changed the default to compil a kernel (I only use gcc-4.2.2 -march=native) and I also didn't try to compil it with gcc-4.3.
Should I set flags before make for it ? |
|
Back to top |
|
|
ilikenwf Tux's lil' helper
Joined: 04 Oct 2007 Posts: 119
|
Posted: Thu Dec 20, 2007 9:37 am Post subject: |
|
|
Gregoire wrote: | So it could be better to use -funroll-loops and -maccumulate-outgoing-args to build the kernel ?
I never changed the default to compil a kernel (I only use gcc-4.2.2 -march=native) and I also didn't try to compil it with gcc-4.3.
Should I set flags before make for it ? |
Don't use gcc 4.3, 4.2 is fine for the time being....but yeah, try those flags, and keep march=native. If you're lazy, you can just use this patch I made that adds an option in the kernel config...: http://www.mattparnell.com/zen/mattpatches/special-kbuild_cflags.patch
If you use the patch, backup the Makefile beforehand so you can restore it whenever you update git!
The worst thing that can happen is a bad compile or a big kernel image...whatever the case, give it a shot and see what happens. If something breaks, open the makefile and go through and try commenting out each flag one by one until you find the one causing the problem. Then re-enable the ones not causing problems and build...
Should work like a charm. _________________ Make The Fonts Pretty In KDE
Archlinux User |
|
Back to top |
|
|
Gregoire Apprentice
Joined: 15 Apr 2006 Posts: 292
|
Posted: Thu Dec 20, 2007 9:57 am Post subject: |
|
|
How should I apply it ?
Code: |
patch --dry-run -p1 -l < ../special-kbuild_cflags.patch
patching file Makefile
patch unexpectedly ends in middle of line
patching file init/Kconfig
patch unexpectedly ends in middle of line
|
or
Code: |
patch --dry-run -p0 -l < ../special-kbuild_cflags.patch
patching file Makefile
patch unexpectedly ends in middle of line
can't find file to patch at input line 31
Perhaps you used the wrong -p or --strip option?
The text leading up to this was:
--------------------------
|
| Index: zen-sources/init/Kconfig
|===================================================================
|*** Kconfig 2007-12-14 19:08:43.000000000 -0600
|--- Kconfig.new 2007-12-18 23:35:05.000000000 -0600
--------------------------
File to patch:
|
Thank |
|
Back to top |
|
|
hoacker Guru
Joined: 04 Aug 2007 Posts: 505 Location: Bürstadt, Germany
|
|
Back to top |
|
|
|