[XviD-devel] in interpolate8x8_halfpel_hv_xmm

[Peter Maersk-Moller]

Hi Michael

Michael Militzer wrote:
> Hi Peter,
> personally, I have never come across such a problem. Also, I believe no one
> ever before reported such a fault.

Thanks. That is useful info. It helps me narrowing it down.

 > Have you tried to encode some sequence
> with the xvid_encraw example encoder, which is included in the XviD source
> package?

No. It will not be easily to reproduce with other setups. Since the
error depends on the source. Let me explain a bit.

I have a lot of identical systems (PCs running mp4live/MPEG4IP) realtime
encoding TV channels. As the source I use bt8x8 based card for analog
tv sources and Technotrends DVB cards for digital sources.

When I tune (any of the systems) in to one specific local TV channel
(which broadcast quality is reduced compared to the others), then depending
on what time of the day (or what the channel broadcast), then the
SIGSEGV is triggered within seconds or minutes of startup.
I then sleep for a second, and starts the encoder up again and
within seconds or minutes it fails again at the same point
in the code interpolate8x8_halfpel_hv_xmm.

 > Do you experience the same problems or is the segfault specific
> to mp4live?

You could say it is specidic to mp4live so the problem could very well
be in mp4live, but read on. I have many other identical systems (but using
other sources that that tv-channel) and they are running flawlessly.
When I try one of these systems to encode the specific tv-channel, I get the
same error (under the same circumstnaces).

The channel causing the problems is btw captured as 352x576 at 25fps
YUV frames, then filtered with a modified YUV-denoiser and then encoded
with XviD 1.0.1, all done live.

> On what processor/hardware were you running XviD btw?

P4 HT 3.6 GHz

> Have you tried on a different system as well? Since the fault seems not
> reproduceable and hasn't been reported ever before (afaict), this could also
> likely be a hardware fault, e.g. a faulty RAM module. This is also possible
> even if you never experienced any similar faults with other software.

Other motherboards, rams and cpus yes, but all P4 HT.

I will first try to upgrade to latest XviD before testing any further.
I will report back when I have any more useful info.

I will go back and change one parameter at a time and see when the error
goes away since I have many other mp4live/xvid systems running largely
flawlessly albeit a bit slow ;-)

Thanks for the hints so far.

Kind regards

--PMM