speed increment in kde 2.2

speed increment in kde 2.2

Post by Marco Vimercat » Fri, 05 Oct 2001 02:48:15



Hello

I read that in kde2.2.1 there're some optimization for speed of launch
application. (pre binding...... I don't know how it works)
Did I must compile kde2.2.1 with some (configure) option to obtain
these optimization?

bye Vime

 
 
 

speed increment in kde 2.2

Post by Brett I. Holcom » Fri, 05 Oct 2001 06:44:15


It's object prelinking and you have to compile with the switch for it.


> Hello

> I read that in kde2.2.1 there're some optimization for speed of launch
> application. (pre binding...... I don't know how it works)
> Did I must compile kde2.2.1 with some (configure) option to obtain
> these optimization?

> bye Vime

--
Brett I. Holcomb

AKA Grunt <><
Remove R777 to email

 
 
 

speed increment in kde 2.2

Post by A.O. » Fri, 05 Oct 2001 14:10:49


http://www.research.att.com/~leonb/objprelink/howto.html has everything
you need.  For KDE 2.2.1 you don't need the KDE patch since it's already
integrated (You will need it, though, if you want to recompile KOffice
1.1).  A couple of tips:  don't use -march=i686 when you compile because
it's known to mess up KDE2 (-mcpu=i686 works fine for me) and be sure to
configure each package with '--enable-final' - it wil cut your compile
times by about 30%.
 
 
 

speed increment in kde 2.2

Post by Marco Vimercat » Fri, 05 Oct 2001 20:13:58




Quote:>http://www.research.att.com/~leonb/objprelink/howto.html has everything

ok! Thank you.

I start to compile QT2.3.1 with the patch for objprelink, but not all
works fine. I get segmentation fault from "moc" tool.
I can't compile QT examples and if I skip them I can't still compile
KDELIBS (always moc segmentation fault)

Quote:>you need.  For KDE 2.2.1 you don't need the KDE patch since it's already

ok, I've 2.2.1

Quote:>1.1).  A couple of tips:  don't use -march=i686 when you compile because
>it's known to mess up KDE2 (-mcpu=i686 works fine for me) and be sure to

I've a K6-2 , so my -march parameter is "k6".
Did I must modify the Makefile manually to set this parameter???

Quote:>configure each package with '--enable-final' - it wil cut your compile
>times by about 30%.

ok! thankyou for the tips.
I also use --disable-debug option

Vime

 
 
 

speed increment in kde 2.2

Post by A.O. » Sat, 06 Oct 2001 01:54:28






>>http://www.research.att.com/~leonb/objprelink/howto.html has everything

> ok! Thank you.

> I start to compile QT2.3.1 with the patch for objprelink, but not all
> works fine. I get segmentation fault from "moc" tool. I can't compile QT
> examples and if I skip them I can't still compile KDELIBS (always moc
> segmentation fault)

First, make sure you have the latest version of binutils (run ld -V to
check, the latest version is 2.11.2).  If you already have the latest
binutils, then I'm afraid there may be no way to fix it.  I had the same
problem before, and it only worked when I rebuilt my Linux system from
the new version of Linux From Scratch.  

Quote:> I've a K6-2 , so my -march parameter is "k6". Did I must modify the
> Makefile manually to set this parameter???

Optimization flags are controlled by the CXXFLAGS environment variable.
Just run

export CXXFLAGS='-O3 -march=k6'

before you run the configure scripts and CXXFLAGS will be automatically
added to all of the makefiles.

 
 
 

speed increment in kde 2.2

Post by Marco Vimercat » Sat, 06 Oct 2001 04:33:30






>First, make sure you have the latest version of binutils (run ld -V to
>check, the latest version is 2.11.2).  If you already have the latest
>binutils, then I'm afraid there may be no way to fix it.  I had the same
>problem before, and it only worked when I rebuilt my Linux system from
>the new version of Linux From Scratch.  

yes, with binutils update it works.
but tokay I decide to upgrade my system (RH6.1) to a new Slackware 8

With this I can't compile objprelink..
(see my other post about it)

Quote:>Optimization flags are controlled by the CXXFLAGS environment variable.
>Just run

>export CXXFLAGS='-O3 -march=k6'

>before you run the configure scripts and CXXFLAGS will be automatically
>added to all of the makefiles.

thanks
 
 
 

speed increment in kde 2.2

Post by John Daile » Sun, 07 Oct 2001 12:44:20


Quote:> yes, with binutils update it works.
> but tokay I decide to upgrade my system (RH6.1) to a new Slackware 8

> With this I can't compile objprelink..
> (see my other post about it)

I had this problem too -- there's a problem with the binutils that Slack
8.0 comes with.  I just down loaded the 7.1 binutils package (it's in the
'd1' folder), installed it and compiled the objprelink program, then
reinstalled the 8.0 version.  Took me about 3 or 4 minutes -- mostly
downloading the package  :-)

I saw in a post that someone else had success by just downloading the
newest version of binutils and compiling/installing it first (apparently
there's a new version since the slack8 distro was put together).  I never
tried that but I guess that would probably work pretty easily too.

--
John Dailey

 
 
 

speed increment in kde 2.2

Post by akar » Sun, 07 Oct 2001 15:37:07



> (-mcpu=i686 works fine for me)

I bet it does, it wont do much though, you are still doing a 386 compile.
mcpu means absolutely nothing unless march is set. march also infers mcpu
so you only have to set it but that does not work back the other way.
Refer to your info pages under compiler switches.
BTW 686 is fine as long as you have a 686 capable chip which does not
include pentium. For athlon/duron you sometimes have to use march=i686
instead of march=athlon especially when compiling SRPMs where you cant
easily modify the scripts written by people that only test for iX86.
--
regards,
andrew
 
 
 

speed increment in kde 2.2

Post by A.O. » Mon, 08 Oct 2001 00:29:40





>> (-mcpu=i686 works fine for me)
> I bet it does, it wont do much though, you are still doing a 386
> compile. mcpu means absolutely nothing unless march is set. march also

-mcpu controls which submodel the instructions are optimally scheduled
for.  -march controls what instructions are generated.  So -march=i386
-mcpu=i686 generates code that will work on a 386, but is optimized for
execution on a Pentium Pro or higher.  -march=i386 -mcpu=i686 is what Red
Hat uses to compile all of their packages.