[MPlayer-users] configure and Cyrix C3

Andre Majorel aym-reyalpm at teaser.fr
Fri Nov 11 23:45:04 CET 2005


On 2005-11-11 23:19 +0100, Andre Majorel wrote:
> On 2005-11-11 21:50 +0100, Guillaume POIRIER wrote:
> > On 11/11/05, Andre Majorel <aym-reyalpm at teaser.fr> wrote:
> > > As of today, MPlayer HEAD does not build out of the box on
> > > Cyrix/VIA C3 systems. It compiles, but the resulting binary
> > > crashes with an "Illegal instruction" error.
> > >
> > > configure (correctly) detects a 686 CPU and (incorrectly) passes
> > > GCC the -march=i686 option. gcc -march=i686 produces code which
> > > uses the CMOV instruction, which is missing on the C3. This is
> > > true for all versions of GCC, as far as I know. Checked with 3.3.5
> > > and 4.0.2.
> > 
> > Could you please try the attached patch with gcc-3.3 and 4.0?
> 
> With your patch, ./configure sets TARGET_CPU to 686 and puts
> -march=c3-2 and -mtune=c3-2 in OPTFLAGS. I suspect this is
> incorrect. From gcc(1) :
> 
>   c3  Via C3 CPU with MMX and 3dNOW! instruction set support.  (No
>       scheduling is implemented for this chip.)
> 
>   c3-2
>       Via C3-2 CPU with MMX and SSE instruction set support.  (No
>       scheduling is implemented for this chip.)
> 
> >From /proc/cpuinfo :
> 
>   flags           : fpu de tsc msr cx8 mtrr pge mmx 3dnow
> 
> I'll report on the run-time behaviour when the compile is done.

Compiled with GCC 4.0.2, mplayer bombs with "Illegal instruction".
Interestingly, it says it was compiled for 3DNow, not SSE as c3-2
implies :

  MPlayer dev-CVS-051111-10:47-4.0.2 (C) 2000-2005 MPlayer Team
  CPU: IDT/Centaur/VIA C3 Samuel 2 (WinChip C5B core)/C3 Ezra (Family: 6, Stepping: 8)
  CPUflags:  MMX: 1 MMX2: 0 3DNow: 1 3DNow2: 0 SSE: 0 SSE2: 0
  Compiled for x86 CPU with extensions: MMX 3DNow
                                            ^^^^^

-- 
André Majorel <URL:http://www.teaser.fr/~amajorel/>
Do not use this account for regular correspondence.
See the URL above for contact information.




More information about the MPlayer-users mailing list