On Thu, 5 Aug 1999, Jesse Kuang wrote:
>
> On Wed, 04 Aug 1999 09:39:27 +0200 (MET DST)
> Peter Putzer <[EMAIL PROTECTED]> wrote:
>
> > On Tue, 3 Aug 1999, Bernhard Rosenkraenzer wrote:
> >
> > > On Tue, 3 Aug 1999, Peter Putzer wrote:
> > >
> > > > I modified the RPM from Cooker to be compiled with PGCC and didn't notice
> > > > any problems with libfont.a (-O6 -fomit-frame-pointer -march=k6 -mcpu=k6
> > > > -funroll-all-loops -ffast-math -mieee-fp)
> > >
> > > The problems are specific to pgcc 1.1.3; any older or newer version
> > > doesn't cause them. (I'll try the 2.95 release tomorrow - the last
> > > snapshots worked.)
> > >
> > > Also, maybe your s/pentium/k6/ solved things... Does it still work if you
> > > compile for pentium optimizations rather than k6?
> >
> > PGCC 1.1.3
> >
> > Mhm, well, the pentium vs. k6 thingy could be the cause, I'll try with
> > -march=pentium -mcpu=pentium later
> I used -march=pentiumpro -mpentiumpro, X launched, so as gnome, I could
> open a xterm, but xfs crashes often and often, xfs always crashes when I
> open netscape. I replace libfont.so compiled with egcs, X works.
> Same as -march=pentium -mpentium
I compiled it with -mpentium, but as you experienced, it crashes xfs. BTW
this hasn't got anything to do with xfs because I replaced it with the
-march=k6 version and it still crashed.
So right now my machine at home is compiling XFree86 with my supposed
workaround enabled, maybe it will work this way... we'll see in the
evening. BTW -funroll-all-loops doesn't work with kdebase... but they say
it's largely untested it's our own fault ;)
Has anyone done some benchmarking regarding different compiler switches? I
once stumbled upon a page with some benchmarks, but unfortunately lost the
URL and haven't been able to find it ever since.
bye,
Peter
PS: I'll upload my kde*-RPMs now (where * currently ==
{base,libs,support,games,utils,admin})
