XO Gen 1.5

[Peter Robinson]

>> The video decompression acceleration will be a huge value. The primary
>> test is of course YouTube which I think means Flash flv. I would put
>> that on an early test list and I hope there's no driver incompatible BS
>> like with Geode. The H. codecs could pay off in better video
>> conferencing. A live chat with Niue would have really warmed up my
>> winter :-)
>
> As a reality check, hardware video decompression support has been
> really hard to integrate in all of the chips I've looked at so far:
> they are usually monkey-patched into the framebufer in ways which make
> them hard to integrate into window managers, and it's hard to actually
> get adobe and/or gnash to support them properly.  I don't know the
> details of this particular chip -- maybe VIA has gotten it right and
> the gnash guys are motivated -- or maybe this email will encourage
> some budding hacker to give it a go...

My experience is that is varies greatly from GPU chipset to X driver
to video codec. Most will now do some sort of hardware based X-Video
Motion Compensation but in terms of hardware assisted mpeg or h.264
encoding/decoding it all becomes very cloudy due to licensing and
various other stuff so it ends up that very little is supported and
ends up being small subsets of the total feature set and varies on the
generation of chip or X driver. Harald Welte mentions in his blog [1]
that VIA haven't yet released the manuals/code for the video decoding
of the GPUs due to legal stuff (but it is due). But at least with
SSE/SSE2/SSE3 there will be still reasonable gains wrt to
encoding/decoding of instructions through the use of things like
liboil.

Peter

[1] http://laforge.gnumonks.org/weblog/2008/11/22/#20081122-via-openchrome