x264 refusing to use Hi422P with 10-bit

[WasF]

Using "x264-10bit --profile high422" has output "x264 [info]: profile High 10, level 4.2, 4:2:0 10-bit"

x264 --help says "high422: Support for bit depth 8-10", so it is supported.

Why is x264 refusing to use Hi422P with 10-bit ?

[poisondeathray]

did you try --output-csp i422 ?

http://mewiki.project357.com/wiki/X264_Settings#output-csp

[Atak_Snajpera]

your source is 4:2:0 (YV12) so x264 by default use that for output. You can also add ConvertToYUY2 or ConvertToYV16 at end of avisynth script if you want to force 4:2:2 encoding in x264.

[WasF]

Thank you guys. "--output-csp i422" does it.
Do you know how x264 works internally ? I mean what colorspace is used for frames internally during processing: input cls, RGB, output cls ? Same in 8-bit and 10-bit operation modes ?

[MasterNobody]

Do you know how x264 works internally ? I mean what colorspace is used for frames internally during processing: input cls, RGB, output cls ? Same in 8-bit and 10-bit operation modes ?
Detected (or specfied by --input-csp/--input-fmt for raw input) input csp ->
most similar supported csp (8- or 16-bitdepth) ->
filters processing ->
output csp [specified by --output-csp and default to YUV 4:2:0] ->
bitdepth conversion (8- or 10-bitdepth according to build) ->
encoding.

[WasF]

Thanks MasterNobody for your clear answer.
Got me thinking: if input-csp == output-csp, is there any benefit to going to a higher order cls/bitdepth during the encode, before going back down to output-csp for output ? I mean doing the hard part with more bits floating around, regardless of input/output requirements.

[MasterNobody]

if input-csp == output-csp then there is no conversions (may be except the case of 10->16->10 bitdepth).

[WasF]

No, if developers modified x264 to make it go to a higher order cls/bitdepth during the encode, would there be any benefit to that ?

[MasterNobody]

Converting 4:2:0 to 4:4:4 is pointless but encoding 8-bit with 10-bit build improve compression and lower probability of banding (if it wasn't in source already).