[mythtv-users] Lowest power HD frontend?

[John Morris]

On Thu, 2013-06-06 at 00:48 +0000, Gary Buhrmaster wrote:
> Some thoughts:
> 
> - Few of these ARM devices can do hardware temporial/spatial
>   de-interlacing.  Interlaced content is common on many
>   locations.  None of the ARM devices with their limited
>   capability decoders (that I am aware of) can do High
>   Def content temporal/spatial de-interlacing in software.
>   That may be fine on a 4" screen, where either low
>   resolution bob/weave can be acceptable (or even throwing
>   away every other frame), but is may not be acceptable on
>   that 65" OLED (your tolerance to low quality will vary)
> 
> - Some of the hardware decoders have no linux driver
>   support.

Most have no proper Linux support.  Initial attempts at open source
drivers for a couple but nothing production ready yet.  I have read of
some people getting the binary blobs distributed with Android going on
some, but haven't heard about accelerated video decode, only EGL.  After
all, modern desktops are pretty much useless without 3D so that seems to
be the initial thrust of development.  Had hoped that at least one
vendor would have realized there was enough of a market from us Linux
folk they would have at least released an Nvidia style binary blob that
fully supported their device under X.  Alas.

But I suspect there ARE chips that can do quality video playback.  Too
many consumer video products are using many of the same SoC elements,
including BluRay players.  The cut down ultra low power parts that go to
phones might not produce perfect playback on a large TV but I was
talking about units marketed as settop boxes.  I'd expect them to have
pretty good hardware video since software decode, deinterlace, etc. is
pretty much out of the question on an ARM.

> - Mythfrontend has a lot of capabilities that simply
>   require a lot of processing power.  That means
>   large(r), and needing more resources.  While it
>   certainly could be slimmed down, one would first
>   have to agree on what one should throw out.
>   That might be an interesting conversation to
>   start.  What features should a frontend have?
>   Clearly (from other posts) some find the XBMC
>   feature set adequate.  While some others
>   want more and more features added to the
>   frontend.  Can the community agree on what
>   features you want to deprecate, or what content
>   types you are willing to abandon.

Not really.  Unless you are talking about non-TV plugins like MythGame,
the basic frontend shouldn't consume much in the way of resources.  Even
the browser wouldn't be a problem.  The use case these machines are sold
for includes running a browser after all... in an environment mostly
written in Java; Qt is slim compared to that sort of resource waste.
Modern ARM parts are more than fast enough for drawing a UI.  And would
make running a FE without moving parts very achievable without being
expensive.

But like I mentioned earlier, transcoding and commercial detection would
be an issue on a combined FE/BE.  It could probably manage commercial
detection, it would just take longer.  Transcoding might need hardware
assistance, and it certainly would if you wanted to stream via http.

All we need is somebody to sell hardware that a mainstream distro can
run on and actually use the full hardware feature set.  And that seems
to be a showstopper for now.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://www.mythtv.org/pipermail/mythtv-users/attachments/20130606/1cff4115/attachment.sig>