On Sat, 2009-11-28 at 08:08 +0100, Gour wrote:
Hi,

> On #transcode I was told about the workaround to use '-V rgb24' and it
> really helps, but the man says 'rgb24' is the older internal format
> and it should not be used. The other format is 'yuv420p' as default
> standard.

the rgb24 internal format should'nt be used if the source material is
already in yuv420 colorspace because doing the conversion is slower and
(slightly) degrades the quality, that's the only reason.

> By more careful inspection of log file the following is found:
> 
> > transcode       V: IA32/AMD64 accel | sse3 sse2 sse mmx cmove asm 
> > decode_mpeg2.c  libmpeg2 acceleration: mmxext
> > /usr/bin/dvdcpics: line 263: 17694 Segmentation fault ...  
> transcode wants to use mmxext instruction set while my cpu (i7 860)
> does not provide it

More precisely, libmpeg2 wants to.
If that's the root of the problem, then it's a libmpeg2 problem and
there is little than we (as transcode team) we can do.

However. I must admit than I know the i7 cpus only by reading docs, but
I'm an hardware junkie and so I'm pretty confident that your cpu
provides such instructions.

`mmxext' is deep jargon (mostly found on linux field AFAIK) for some MMX
extended instructions that aren't explicitly advertised; I can't recall
the exact details but those were added together with the first round of
SSE or something like that.

So, unless I missing something huge, the core i7 definitley support
them :)


>  and I built transcode-1.1.5 with:
> ./configure --prefix=/usr \
>     --enable-sse=no --enable-sse2=no --enable-altivec=no --enable-mmx \

Uhm, looks like an unconventional choice of flags to me; However, as
written above, the libmpeg2 asm acceleration is mostly out of control of
transcode.

> I've tried to set --accel C, but it does not help.
> Any hint?

Recompile libmpeg2.
The other (and independent) way to go is patching the transcode's mpeg2 
import module.

Bests,

-- 
Francesco Romani // Ikitt
http://fromani.exit1.org  ::: transcode homepage
http://tcforge.berlios.de ::: transcode experimental forge

Reply via email to