[FFmpeg-devel] [PATCH] improve yuv422p to rgb in libswscale

[Baptiste Coudurier]

On 11/30/2010 06:15 PM, Michael Niedermayer wrote:
> On Tue, Nov 30, 2010 at 05:20:42PM -0800, Baptiste Coudurier wrote:
>> On 11/30/2010 05:13 PM, Michael Niedermayer wrote:
>>> On Tue, Nov 30, 2010 at 04:07:20AM -0800, Baptiste Coudurier wrote:
>>>> Hi
>>>>
>>>> $subject, use full vertical data when convert 422p, improve quality a lot.
>>>>
>>>> --
>>>> Baptiste COUDURIER
>>>> Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
>>>> FFmpeg maintainer                                  http://www.ffmpeg.org
>>>
>>>>    x86/yuv2rgb_template.c |   25 ++++-------
>>>>    yuv2rgb.c              |  109 ++++++-------------------------------------------
>>>>    2 files changed, 26 insertions(+), 108 deletions(-)
>>>> 16f384c9b114c76572a539511c42267ce2942c67  yuv422p_to_rgb.patch
>>>
>>> this looks like it would make 420p->rgb quite a bit slower.
>>
>> Do you think changing>>1 to>>vshift would make is quite a bit slower ?
>
> no but that stuff:
> @@ -152,134 +144,102 @@
>   YUV2RGBFUNC(yuv2rgb_c_48, uint8_t, 0)
>       LOADCHROMA(0);
>       PUTRGB48(dst_1,py_1,0);
> -    PUTRGB48(dst_2,py_2,0);
>
>       LOADCHROMA(1);
> -    PUTRGB48(dst_2,py_2,1);
>       PUTRGB48(dst_1,py_1,1);
>
>       LOADCHROMA(2);
>       PUTRGB48(dst_1,py_1,2);
> -    PUTRGB48(dst_2,py_2,2);
>
>       LOADCHROMA(3);
> -    PUTRGB48(dst_2,py_2,3);
>       PUTRGB48(dst_1,py_1,3);
>   ENDYUV2RGBLINE(48)
>       LOADCHROMA(0);
>       PUTRGB48(dst_1,py_1,0);
> -    PUTRGB48(dst_2,py_2,0);
>
>       LOADCHROMA(1);
> -    PUTRGB48(dst_2,py_2,1);
>       PUTRGB48(dst_1,py_1,1);
>   ENDYUV2RGBFUNC()
>
> and then running the code twice

This is the C code, the mmx routine for yuv420 to rgb24 is already present.

I don't understand running the code twice, can you please clarify ?

The only change that the patch does so far is to use the _correct_ line 
for 4:2:2 instead of skipping them.

>>
>>> If so thats a problem.
>>> making 422p->rgb slower isnt a real problem as it can be changed to 420p
>>> by manipulating linesizes
>>
>> That's what I want to avoid, this produces crap.
>
> It produces crap when misused on interlaced material.
 >
> Of course the user should have the option to use the perfect slow variant and
> the faster (correct) optimized variant.

No, it produces crap on progressive material because it skips one U and 
one V line. This seems pretty obvious to me. I'll show you some 
screenshots later (it involves a test pattern with bars, the abrupt 
change in colors looks crap)

-- 
Baptiste COUDURIER
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
FFmpeg maintainer                                  http://www.ffmpeg.org