mailing-list for TeXmacs Users

Text archives Help


Re: [TeXmacs] New TeXmacs package available for Slackware 10


Chronological Thread 
  • From: Joris van der Hoeven <address@hidden>
  • To: am29 <address@hidden>
  • Cc: TeXmacs Users Mailing list <address@hidden>
  • Subject: Re: [TeXmacs] New TeXmacs package available for Slackware 10
  • Date: Wed, 1 Dec 2004 15:10:20 +0100 (CET)


[Notice that technical discussions on the implementation of TeXmacs
should take place on the developers mailing list address@hidden]

On Wed, 1 Dec 2004, am29 wrote:
> > Yes. Due to a bug in early versions of g++ 3.*, we had to include
> > the flags -fno-default-inline -fno-inline. In the most recent versions,
> > these options can be suppressed in order to achieve a considerable gain
> > in speed. But this situation is recognized by the configure script.
> > What version of g++ do you use?
> > Best wishes, Joris
> I've read src/common.makefile to understand what kind of optimization
> are turned on and I've found that if I do
>
> $ ./configure
>
> the variable CXXOPTIMIZE is set to
> CXXOPTIMIZE = -O3 -fexpensive-optimizations -fno-default-inline
> -fno-inline -fno-rtti -fno-exceptions
> I'm using gcc 3.3.4 so -fno-default-inline -fno-inline should be
> actvated by default and -fexpensive-optimizations is included in -O3
> As tou suggested I've done

I added version 3.3.4 to the configure.in for optimization by default.
Since this version is ulterior to 3.3.3 this should not lead to
the same bugs as with early versions of g++ 3.*.

> $ make CXXOPTIMIZE="-O3 -fexpensive-optimizations -fno-default-inline
> -fno-inline -fno-rtti -march=i486 -mcpu=i686"
>
> to add architecture specific optimizations, and I've tried
>
> $ make CXXFLAGS="-O3 -fexpensive-optimizations -fno-default-inline
> -fno-inline -fno-rtti -march=i486 -mcpu=i686"
>
> to deactivate warnings.
> I try to do my best to obtain a good compilation: I'd like to provide
> fast and stable binaries so I hope this optimization to be good, but if
> I've done some mistakes, please report me.

No, this seems to be OK.
Do you get a faster binary when using "-march=i486 -mcpu=i686"?
I remember having tried once, but I did not notice any clear difference.
In order to test, you may measure the time it takes to compile and
open the full documentation:

Help -> Full manuals -> User manual

Don't forget run TeXmacs, open the manual and close TeXmacs once
for fake, since some fonts may need to be created the first time.

Thanks for your help, Joris




Archive powered by MHonArc 2.6.19.

Top of page