[FFmpeg-devel] [PATCH 0/2][RFC] VC-1 Interlaced Hardware decoding.

[Hendrik Leppkes]

On Sat, Oct 6, 2012 at 9:55 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
> On Sun, Sep 23, 2012 at 02:41:37PM +0200, Hendrik Leppkes wrote:
>> Hi,
>>
>> these patches are a reworked version of the patchset for VC-1 hardware decoding i posted a few weeks ago.
>> I addressed the concerns from Gwenole, and made the patch far less intrusive.
>>
>> This patch should have zero impact on progressive decoding, and only add a new code path for field-interlaced content.
>>
>> The first patch actually implements the decoding, and has been tested on NVIDIA and AMD hardware, however only with DXVA2 on Windows, i do not have access to a VA-API setup, therefor i consider this patchset a RFC until someone can report how it works on VA-API, if at all.
>
> I think you should explain how it can be tested, it will increase the
> chances that people who have the right hw will actually test it
>

Testing is really simple. If you have a working VA-API setup, test
without the patch to confirm it works, apply patch, confirm it still
works.
Behaviour should hopefully not change at all for progressive or
frame-interlaced files, and decoding should work just fine.

As a bonus, test a field interlaced file with the patch applied, to
check if it happens to work as well.

I can provide simple progressive, frame interlaced and field
interlaced sample streams, if anyone needs them for testing.

> also what should be done with these patches if noone comes forth and
> tests them with VA-API ? i think letting them rot is not a good idea
>

I could always somehow add a check in there to only make it use the
new code with DXVA... I'm sure VLC and other players that also use
DXVA would be happy about the field-interlaced DXVA support.

- Hendrik