in All availableThis forumThis topic
 GCC 4.3.0 testing! (part 2) View unanswered posts View posts from last 24 hours Goto page Previous  1, 2, 3, 4, 5  Next
Author Message
yoshi314
l33t

Joined: 30 Dec 2004
Posts: 848
Location: PL

Posted: Thu Sep 04, 2008 10:19 am    Post subject:

 Quote: i use it globally but then if something doesn't compile you can always use gcc-config to switch to another gcc -v
there could be problems when system libraries link to other libstdc++ than the application you're trying to build with different gcc. that sometimes happened with gcc3/gcc4 switching.
_________________
~amd64
devsk

Joined: 24 Oct 2003
Posts: 2870
Location: Bay Area, CA

 Posted: Sat Sep 06, 2008 6:13 pm    Post subject: where did you guys get 4.3.2 ebuild from?
Xake
Guru

Joined: 11 Feb 2004
Posts: 588
Location: Göteborg, the rainy part of scandinavia

Posted: Sat Sep 06, 2008 8:43 pm    Post subject:

 devsk wrote: where did you guys get 4.3.2 ebuild from?

Portage, namebump (minus one patch)?
Or if you like hardened, there is a nice overlay somewhere around...
devsk

Joined: 24 Oct 2003
Posts: 2870
Location: Bay Area, CA

Posted: Sat Sep 06, 2008 9:58 pm    Post subject:

Xake wrote:
 devsk wrote: where did you guys get 4.3.2 ebuild from?

Portage, namebump (minus one patch)?
Or if you like hardened, there is a nice overlay somewhere around...
which patch is that?
Naib
Watchman

Joined: 21 May 2004
Posts: 5712
Location: Removed by Neddy

 Posted: Sat Sep 06, 2008 10:12 pm    Post subject: libiberty.h-asprintf-glibc-2.8.patch but that a hardened overlay patch _________________The best argument against democracy is a five-minute conversation with the average voter Great Britain is a republic, with a hereditary president, while the United States is a monarchy with an elective king
devsk

Joined: 24 Oct 2003
Posts: 2870
Location: Bay Area, CA

 Posted: Sat Sep 06, 2008 11:30 pm    Post subject: I had to remove 75_all_gcc43-pr36533.patch
kernelOfTruth
Watchman

Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

Posted: Sun Sep 07, 2008 1:23 am    Post subject:

 devsk wrote: I had to remove 75_all_gcc43-pr36533.patch

afaik it was the same I omitted from the patchset,

gcc-4.3.2 seems to fix several issues compared to .3.1: some apps wouldn't work correctly with .3.1 whereas they do now with .3.2
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0

Hardcore Gentoo Linux user since 2004
Gentree
Watchman

Joined: 01 Jul 2003
Posts: 5350
Location: France, Old Europe

 Posted: Mon Sep 08, 2008 8:18 am    Post subject: Hi devsk, could you post the line you added to remove that patch? Presumably an epatch command but I'm not that confident on getting it 100% correct and this really needs to be 100%. TIA. actually may be safest if you could post the ebuild _________________Linux, because I'd rather own a free OS than steal one that's not worth paying for. Gentoo because I'm a masochist AthlonXP-M on A7N8X. Portage ~x86
devsk

Joined: 24 Oct 2003
Posts: 2870
Location: Bay Area, CA

Posted: Mon Sep 08, 2008 4:28 pm    Post subject:

Steps I followed:

 Code: mkdir /tmp/gcc cd /tmp/gcc tar xjpf /usr/portage/distfiles/gcc-4.3.1-patches-1.1.tar.bz2 cd patch \rm 75_all_gcc43-pr36533.patch cd .. tar cjpf /usr/portage/distfiles/gcc-4.3.2-patches-1.1.tar.bz2 patch/ cd /usr/portage/*/gcc cp gcc-4.3.1-r1.ebuild gcc-4.3.2.ebuild ebuild gcc-4.3.2.ebuild manifest cd .. mkdir /usr/local/portage/sys-devel/ cp -a gcc /usr/local/portage/sys-devel/ emerge -1v gcc

of course you will need to change parts of it for your overlay and DISTDIR.
Gentree
Watchman

Joined: 01 Jul 2003
Posts: 5350
Location: France, Old Europe

 Posted: Mon Sep 08, 2008 6:21 pm    Post subject: OK, many thanks for the clear instructions that make a lot more sense. I'd taken one of the above comments a bit too literally , I thought the ebuild had been bumped then tweeked to remove the failing patch. Now I have the fake 4.3.2 patches bz2 all is running smoothly. If only my hard drive knew what I'm about to throw at it! Many thanks. _________________Linux, because I'd rather own a free OS than steal one that's not worth paying for. Gentoo because I'm a masochist AthlonXP-M on A7N8X. Portage ~x86
Gentree
Watchman

Joined: 01 Jul 2003
Posts: 5350
Location: France, Old Europe

 Posted: Mon Sep 08, 2008 9:46 pm    Post subject: OK, I have my first build of 4.3.2 , thanks for the help. Before I get into rebuilding world any new features worth investigating for arch=athlon-xp? In particular is tree-vectorize stable and productive now? The last time I reviewed CFLAGS et al I recall it was not any faster and produced occassional hickcoughs. CFLAGS="-march=athlon-xp -O2 -pipe -fomit-frame-pointer -falign-functions=64" CXXFLAGS="${CFLAGS} -fvisibility-inlines-hidden" #CXXFLAGS="${CXXFLAGS} -ftree-vectorize" LDFLAGS="-Wl,-O1 -Wl,--sort-common -s" LDFLAGS="${LDFLAGS} -Wl,--hash-style=gnu" I don't want to get ricey about this but if I'm going to do a clean rebuild of the system nows the time to make any ajustments worth doing. Thanks. _________________Linux, because I'd rather own a free OS than steal one that's not worth paying for. Gentoo because I'm a masochist AthlonXP-M on A7N8X. Portage ~x86 devsk Advocate Joined: 24 Oct 2003 Posts: 2870 Location: Bay Area, CA Posted: Wed Sep 10, 2008 1:30 am Post subject:  Gentree wrote: Before I get into rebuilding world any new features worth investigating for arch=athlon-xp? In particular is tree-vectorize stable and productive now? I will post if you don't call me a ricer.... OK, what the heck! Here it is:  Code: CFLAGS="-O2 -march=k8 -fforce-addr -fomit-frame-pointer -pipe" CFLAGS="${CFLAGS} -ftree-vectorize" CFLAGS="${CFLAGS} -frename-registers" CFLAGS="${CFLAGS} -fweb" CXXFLAGS="${CFLAGS}" LDFLAGS="-Wl,-O1 -Wl,--enable-new-dtags -Wl,--as-needed" These are based on my own benchmarks that I did when I moved to 4.3 some time back. Of course I will say that they are the best combo... But your mileage may vary. My goal was fastest with a balance on generated code size. kernelOfTruth Watchman Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :) Posted: Wed Sep 10, 2008 9:34 am Post subject: devsk wrote:  Gentree wrote: Before I get into rebuilding world any new features worth investigating for arch=athlon-xp? In particular is tree-vectorize stable and productive now? I will post if you don't call me a ricer.... OK, what the heck! Here it is:  Code: CFLAGS="-O2 -march=k8 -fforce-addr -fomit-frame-pointer -pipe" CFLAGS="${CFLAGS} -ftree-vectorize" CFLAGS="${CFLAGS} -frename-registers" CFLAGS="${CFLAGS} -fweb" CXXFLAGS="\${CFLAGS}" LDFLAGS="-Wl,-O1 -Wl,--enable-new-dtags -Wl,--as-needed"
These are based on my own benchmarks that I did when I moved to 4.3 some time back. Of course I will say that they are the best combo... But your mileage may vary. My goal was fastest with a balance on generated code size.

afaik that is deprecated and not supported anymore with gcc-4.3* have a look at the gcc manual/handbook
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0

Hardcore Gentoo Linux user since 2004
gringo

Joined: 27 Apr 2003
Posts: 3793

 Posted: Wed Sep 10, 2008 11:38 am    Post subject: i´ve been using -ftree-vectorize too for a while now and it looks sane, although i haven´t checked what ebuilds are filtering it. I think i have read it is also enabled by default in -O3 now. -frename-registers still scares me, specially for my powerbook. The only one i added ( apart from -ftree-vectorize ) is -fdirectives-only which is only usefull if you use distcc/ccache : http://lists.samba.org/archive/distcc/2007q3/003536.html cheers_________________Error: Failing not supported by current locale
Gentree
Watchman

Joined: 01 Jul 2003
Posts: 5350
Location: France, Old Europe

Posted: Sun Sep 14, 2008 8:01 pm    Post subject:

Hi, I got a fairly clean run but a couple of issues are blocking me from completing.

wondered if any may be down to 4.3.2

gtk+ fails when checking pango libs. It works on my old 4.2.3 root partition.

 Code: checking X11/extensions/XShm.h... yes checking Pango flags... -D_REENTRANT -I/usr/include/pango-1.0 -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -I/usr/include/cairo -I/usr/include/freetype2 -I/usr/include/directfb -I/usr/include/libpng12 -I/usr/include/pixman-1   -lpangocairo-1.0 -lpango-1.0 -lcairo -lgobject-2.0 -lgmodule-2.0 -ldl -lglib-2.0  configure: error: *** Can't link to Pango. Pango is required to build *** GTK+. For more information see http://www.pango.org !!! Please attach the following file when seeking support: !!! /usr/portage/tmp/portage/x11-libs/gtk+-2.12.11/work/gtk+-2.12.11/config.log  *  * ERROR: x11-libs/gtk+-2.12.11 failed.

cf
 Code: checking X11/extensions/XShm.h... yes checking Pango flags... -D_REENTRANT -I/usr/include/pango-1.0 -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -I/usr/include/cairo -I/usr/include/freetype2 -I/usr/include/directfb -I/usr/include/libpng12 -I/usr/include/pixman-1   -lpangocairo-1.0 -lpango-1.0 -lcairo -lgobject-2.0 -lgmodule-2.0 -ldl -lglib-2.0  checking ATK flags... -I/usr/include/atk-1.0 -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include   -latk-1.0 -lgobject-2.0 -lgmodule-2.0 -ldl -lglib-2.0  checking for cups-config... /usr/bin/cups-config checking cups/cups.h usability... yes checking cups/cups.h presence... yes

any idea why the pango tests are failing?

Thx

_________________
Linux, because I'd rather own a free OS than steal one that's not worth paying for.
Gentoo because I'm a masochist
AthlonXP-M on A7N8X. Portage ~x86
ConnClark
n00b

Joined: 15 Aug 2007
Posts: 57

 Posted: Tue Sep 16, 2008 10:36 pm    Post subject: I was wondering if anybody else here on x86/amd64 has played with the -minline-strings-dynamically optimization flag that was introduced in gcc 4.3.x? I have used it on things like bash, make, sed, gawk, grep, perl, python, etc ... and building things seems to have a speed increase. I have also hacked the eclass and rebuilt firefox and it seems to have helped speed it up too. If I could find a way to sneak it into building glibc and gcc I would try that too ._________________In formal computer science advances are made by standing on the shoulders of giants. Linux has shown that, if there are enough of you, you can advance just as far by stepping on each others toes.
kernelOfTruth
Watchman

Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

Posted: Tue Sep 16, 2008 11:05 pm    Post subject:

 ConnClark wrote: I was wondering if anybody else here on x86/amd64 has played with the -minline-strings-dynamically optimization flag that was introduced in gcc 4.3.x? I have used it on things like bash, make, sed, gawk, grep, perl, python, etc ... and building things seems to have a speed increase. I have also hacked the eclass and rebuilt firefox and it seems to have helped speed it up too. If I could find a way to sneak it into building glibc and gcc I would try that too .

thanks for the tip, ConnClark,

you mean -minline-stringops-dynamically, correct ?

 Quote: By default GCC inlines string operations only when destination is known to be aligned at least to 4 byte boundary. This enables more inlining, increase code size, but may improve performance of code that depends on fast memcpy, strlen and memset for short lengths.

_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0

Hardcore Gentoo Linux user since 2004

Last edited by kernelOfTruth on Tue Sep 16, 2008 11:10 pm; edited 2 times in total
ConnClark
n00b

Joined: 15 Aug 2007
Posts: 57

Posted: Tue Sep 16, 2008 11:08 pm    Post subject:

 kernelOfTruth wrote: you mean -minline-stringops-dynamically, correct ?

yep, thats the one.

though this is what I found about its behaviour
 Quote: With this option string operations of unknown size are expanded such that small blocks are copied by in-line code, while for large blocks a library call is used. This results in faster code than -minline-all-stringops when the library implementation is capable of using cache hierarchy hints.

_________________
In formal computer science advances are made by standing on the shoulders of giants. Linux has shown that, if there are enough of you, you can advance just as far by stepping on each others toes.

Last edited by ConnClark on Tue Sep 16, 2008 11:11 pm; edited 1 time in total
kernelOfTruth
Watchman

Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

Posted: Tue Sep 16, 2008 11:10 pm    Post subject:

ConnClark wrote:
 kernelOfTruth wrote: you mean -minline-stringops-dynamically, correct ?

yep, thats the one.

I'm personally not using that setting, since I read that -mno-align-stringops would be a pretty good switch to improve performance with core 2 duo and gcc-4.3* and thought they would do the exact opposite

do -mno-align-stringops and minline-stringops-dynamically exclude each others ?

 Quote: -mno-align-stringops Do not align destination of inlined string operations. This switch reduces code size and improves performance in case the destination is already aligned, but GCC doesn’t know about it.

_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0

Hardcore Gentoo Linux user since 2004
ConnClark
n00b

Joined: 15 Aug 2007
Posts: 57

 Posted: Tue Sep 16, 2008 11:13 pm    Post subject: I don't know if they exclude each other._________________In formal computer science advances are made by standing on the shoulders of giants. Linux has shown that, if there are enough of you, you can advance just as far by stepping on each others toes.
kernelOfTruth
Watchman

Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

Posted: Tue Sep 16, 2008 11:15 pm    Post subject:

 ConnClark wrote: I don't know if they exclude each other.

ok, just thought about it a little (it's already pretty late here )

one of them inlines, the other one aligns, so both should be fine

thanks again for the tip, in the future I'll use both

another tip:

try -fipa-cp

that one's gonna be default on gcc-4.4 with all -O* afaik
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0

Hardcore Gentoo Linux user since 2004
ultimA
n00b

Joined: 17 Sep 2008
Posts: 4

 Posted: Wed Sep 17, 2008 7:50 pm    Post subject: Some more info on building 4.3.2 I'm too excited to wait any longer for an "offical" 4.3.2 ebuild :DD Two questions I have: 1) Where did you get "gcc-4.3.2-patches-1.1.tar.bz2" from? Is it just the official gcc diff files found on gcc mirrors bzipped into one? Or is found on some overlay? 2) The reason I need to delete patch 36533 from 4.3.1 is because it was included extra by the gentoo maintainers, but is not part of 4.3.1, but 4.3.2 instead, right? I mean, if I were to take the old 4.3.1 ebuild (without -r1) I wouldn't need to delete this patch, right? Thx
Gentree
Watchman

Joined: 01 Jul 2003
Posts: 5350
Location: France, Old Europe

 Posted: Wed Sep 17, 2008 8:11 pm    Post subject: read above this was already explained. _________________Linux, because I'd rather own a free OS than steal one that's not worth paying for. Gentoo because I'm a masochist AthlonXP-M on A7N8X. Portage ~x86
ultimA
n00b

Joined: 17 Sep 2008
Posts: 4

 Posted: Wed Sep 17, 2008 9:50 pm    Post subject: [quote="Gentree"]read above this was already explained.[/quote] Thx for your "help". The problem is, I didn't find that info, which is obviously why I'm asking. You might be referring to the post about "portage, namebump", only problem is I don't have an idea what a namebump is. I'm a developer but relatively new to gentoo.
kernelOfTruth
Watchman

Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

Posted: Wed Sep 17, 2008 10:16 pm    Post subject:

ultimA wrote:

Thx for your "help". The problem is, I didn't find that info, which is obviously why I'm asking. You might be referring to the post about "portage, namebump", only problem is I don't have an idea what a namebump is. I'm a developer but relatively new to gentoo.

https://forums.gentoo.org/viewtopic-p-5208486.html#5208486
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0

Hardcore Gentoo Linux user since 2004
 Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First
 All times are GMTGoto page Previous  1, 2, 3, 4, 5  Next Page 3 of 5

 Jump to: Select a forum Assistance----------------News & AnnouncementsFrequently Asked QuestionsInstalling GentooMultimediaDesktop EnvironmentsNetworking & SecurityKernel & HardwarePortage & ProgrammingGamers & PlayersOther Things GentooUnsupported Software Discussion & Documentation----------------Documentation, Tips & TricksGentoo ChatGentoo Forums FeedbackOff the WallDuplicate Threads International Gentoo Users----------------中文 (Chinese)DutchFinnishFrenchDeutsches Forum (German)  Diskussionsforum  Deutsche DokumentationGreekForum italiano (Italian)  Forum di discussione italiano  Risorse italiane (documentazione e tools)Polskie forum (Polish)  Instalacja i sprzęt  Polish OTWPortuguese  Documentação, Ferramentas e DicasRussianScandinavianSpanishOther Languages Architectures & Platforms----------------Gentoo on AMD64Gentoo on ARMGentoo on PPCGentoo on SparcGentoo on Alternative ArchitecturesGentoo for Mac OS X (Portage for Mac OS X)
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum