[FFmpeg-devel] [PATCH v12 5/9] libavcodec/dnxucdec: DNxUncompressed decoder

[martin schitter]

On 21.10.24 23:50, James Almer wrote:
> On 10/21/2024 5:44 PM, James Almer wrote:
>> On 10/21/2024 4:57 PM, Martin Schitter wrote:
>>> +            break;
>>> +        case MKTAG('y','2','1','0'):
>>> +            ret = fmt_frame(avctx, frame, avpkt, 
>>> AV_PIX_FMT_YUV422P10LE, 20, unpack_y210);
>>
>> Y210 has no pixel format, and it's packed, not planar, so definitely 
>> not AV_PIX_FMT_YUV422P10LE.
> 
> Nevermind, there's AV_PIX_FMT_Y210.

Although both entities share a very similar FOURCC code, they are very 
different bit structures in practice!

Sure, both of them refer 'somehow' to 10bit 4:2:2 YUV data and can be 
lossless converted in both directions, but that's all they have in common.

But in fact I could have converted the DNxUncompressed 10 and 12 bit 
formats to any packed format as well. I just choose the planar variant, 
because byte ordering changes are easier to handle in this arrangement 
and there were more systematically ordered variants of this kind already 
defined. At the beginning I hopped, that this would make the reuse of 
already defined  pixformats more easily and also enable a more 
consistent code layout for all the required RGB / YUV422 / YUV444 
variants...

Searching the chaotic list of already defined pixel formats is another 
topic, where I better keep quiet, because I could otherwise hurt somebody...

Martin