[mythtv-users] Intel Compute Stick as a frontend?

[Karl Newman]

On Wed, Aug 5, 2015 at 3:16 PM, Ozzy Lash <ozzy.lash at gmail.com> wrote:

> On Wed, Aug 5, 2015 at 11:34 AM, David Engel <david at istwok.net> wrote:
>
>> On Wed, Aug 05, 2015 at 02:52:41PM +0000, Gary Buhrmaster wrote:
>> > On Wed, Aug 5, 2015 at 2:31 PM, Stuart Auchterlonie
>> > <stuarta at squashedfrog.net> wrote:
>> > .....
>> > > Given that we have not yet implemented deinterlacing with the VAAPI
>> > > interface, i'm not surprised about this.
>> >
>> > I had thought that the current codes support bob (which mostly sucks).
>> >
>> > While the current codes are not capable of enabling the latest VAAPI
>> > deinterlacing capabilities (MADI (Motion Adaptive Deinterlacing), and
>> > MCDI (Motion Compensation Deinterlacing)), even when they do, I do
>> > not think any of the current generation "sticks" do MCDI, so only
>> > those with low expectations should get their hopes up.
>>
>> That is my understanding as well.
>>
>
> I'm not sure about the compute sticks, but isn't the Liva using an Ivy
> Bridge part?  If I look at the code at
> http://code.metager.de/source/xref/freedesktop/vaapi/driver/intel/src/i965_device_info.c
> and go down to the section on the ivy bridge parts, I see
>
> extern struct hw_context *gen7_dec_hw_context_init(VADriverContextP,
> struct object_config *);
> 136extern struct hw_context *gen7_enc_hw_context_init(VADriverContextP,
> struct object_config *);
> 137static struct hw_codec_info ivb_hw_codec_info = {
>
> .....
> 161    .has_accelerated_getimage = 1,
> 162    .has_accelerated_putimage = 1,
> 163    .has_tiled_surface = 1,
> 164    .has_di_motion_adptive = 1,
> 165    .has_di_motion_compensated = 1,
>
> Which would seem to indicate that it would support motion compensated
> deinterlacing.  I thought that the compute stick used a Bay Trail part,
> which if I understand correctly is an ivy bridge with fewer processing
> units.  I don't see that platform called out in the file, so I would have
> assumed that it also supported MCDI.
>
> I use a Liva, and while it looked a bit different that the old ION1 system
> it replaced, I don't really notice any really bad deinterlacing issues, but
> I don't tend to record a lot of sports.  If at some point the picture gets
> even better, it will be a bonus in my book.
>
> Bill
>

/proc/cpuinfo on my Liva (one of the older models) identifies it as a
Celeron N2807, which is apparently a Bay Trail-M processor. Not sure about
its specific deinterlacing capabilities.

Karl
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mythtv.org/pipermail/mythtv-users/attachments/20150805/c22fa9dd/attachment.html>