Re: Performance samples for LibreOffice ...

> > Ah, right, good catch. Our XLIB_* "trick" from > include/{pre,post}x.h (which had been duplicated in > extensions/source/plugin/inc/plugin/unx/plugcon.hxx prior to > > "Use prex.h/postx.h"). S

Re: Performance samples for LibreOffice ...

> On 09/20/2014 04:39 AM, Jan Hubicka wrote: > >/usr/include/X11/Xlib.h:948:3: warning: type ‘struct XAnyEvent’ violates one > >definition rule [-Wodr] > > } XAnyEvent; > >^ > >/usr/include/X11/Xlib.h:948:3: note: a different type is defined in another >

Re: Performance samples for LibreOffice ...

> On 09/13/2014 05:43 PM, Jan Hubicka wrote: > >I updated to current tree and my ODR checker reports no more violations > >during of libmergedlo. > >Did you integrated more stuff to it? > > I at least didn't address any further issues from your list than the >

Re: Performance samples for LibreOffice ...

Hi, I updated to current tree and my ODR checker reports no more violations during of libmergedlo. Did you integrated more stuff to it? It seems to be significantly bigger than one I tested in April (so it now hits limit on number of sections in ELF library used by GCC, hopefully that will be fi

Re: Performance samples for LibreOffice ...

> > Completely different. ;) Hehe, probably good enough for my comparing code to panic and run away ;) > > While the "hand-written" class in > i18npool/inc/nativenumbersupplier.hxx has member variables (direct > ones as well as ones inherited from the base class template > instantiation) along

Re: Performance samples for LibreOffice ...

> On 09/11/2014 07:05 AM, Jan Hubicka wrote: > >Anyway those are warning I currently get with ODR type merging. The first > >three are > >quite puzzling - the type layout differs, but by abstract field so I am > >unable to print > >out any resonable messag

Re: Performance samples for LibreOffice ...

> On 10/09/14 07:45, Jan Hubicka wrote: > > > Yep, I am curious too. I just got my trees updated after vacation so > > hopefully > > I will get there soon (for sure libreoffice will trigger few GCC bugs first) > > yes we do have a certain reputation for that...

Re: Performance samples for LibreOffice ...

> > Here's the struct: > > struct NDXHeader > { > sal_uInt32 db_rootpage;/* Rootpage position > */ > sal_uInt32 db_pagecount; /* Page count > */ > sal_uInt8 db_frei[4]; /* Reserved

Re: Performance samples for LibreOffice ...

Hello, I finally had chance to put the data together: http://hubicka.blogspot.ca/2014/09/linktime-optimization-in-gcc-part-3.html any comments are welcome. > On Fri, Aug 22, 2014 at 09:06:24PM +0200, Jan Hubicka wrote: > > > Another one I noticed :) > > /aux/hubicka/libreo

Re: Performance samples for LibreOffice ...

ng ;) Honza > > Thanks ! > > Michael. > > > On Fri, 2014-08-22 at 17:05 +0200, Jan Hubicka wrote: > > Hi, > > this is another warning that may be worth fixing > > /aux/hubicka/libreoffice/sc/source/ui/vba/vbastyles.cxx:68:7: warnin

Re: Performance samples for LibreOffice ...

> Hi Matus, > > Any chance you can fix / commit this one ? :-) > > Jan - wow - that is a nice error =) are there any other ODR issues ? > they habitually bite us hard so ... great to get libmerged debugged even > more. CC'ing the list too. Another one I noticed :) /aux/hubicka/libreof

[Libreoffice-commits] core.git: include/tools tools/source

include/tools/zcodec.hxx |5 - tools/source/zcodec/zcodec.cxx |5 - 2 files changed, 4 insertions(+), 6 deletions(-) New commits: commit 3db6d2cb99f9559ba29759990675e469613a8bb0 Author: Jan Hubicka Date: Tue Feb 18 19:17:46 2014 +0100 GCC 4.9, LTO: libvcl uses

Re: [Libreoffice] [GSoC] link-time optimization

Hi, thanks to Michael I made some progress on building Libreoffice on my setup. The problem of configury was that -flto was not passed at linktime and also -fuse-linker-plugin was missing. My build currently dies on svx on GCC bug. I've created tracking PR http://gcc.gnu.org/bugzilla/show_bug.

Re: [Libreoffice] [GSoC] link-time optimization

Hi, so I see, it is tested on GCC 4.5. I am quite surprised that the build works with that compiler as there was a lot of bugs. I will give it a try on 4.6 and 4.7. 4.6 will do parallel linking that saves a lot of time and some memory and 4.7 saves a lot more memory + will eventually hav

Re: [Libreoffice] [GSoC] link-time optimization

Am Mon 22 Aug 2011 04:56:31 PM CEST schrieb Jan Holesovsky : Hi Honza, So we have some proof-of-concept LTO work pushed to the LibreOffice repository - care to have a look & try it how it copes with your changes in gcc etc.? ;-) git clone git://anongit.freedesktop.org/libreoffice/core sudo zyp