[FFmpeg-devel] [PATCH] swscale/x86/input: add AVX2 optimized RGB24 to YUV functions

[James Almer]

On 6/4/2024 4:32 PM, Andreas Rheinhardt wrote:
> James Almer:
>> On 6/4/2024 4:25 PM, Andreas Rheinhardt wrote:
>>> James Almer:
>>>> rgb24_to_uv_8_c: 39.3
>>>> rgb24_to_uv_8_sse2: 14.3
>>>> rgb24_to_uv_8_ssse3: 13.3
>>>> rgb24_to_uv_8_avx: 12.8
>>>> rgb24_to_uv_8_avx2: 14.3
>>>
>>> Worse than avx and ssse3
>>
>> Can't be disabled for small sizes.
>>
> 
> Why can't this be disabled?

The function pointer is for all sizes. And adding a check + jmp in the 
assembly is not going to be better. Also, this being for 8 pixel wide 
buffers, it's not going to be anyone's bottleneck.

> 
>>>
>>>> rgb24_to_uv_128_c: 582.8
>>>> rgb24_to_uv_128_sse2: 127.3
>>>> rgb24_to_uv_128_ssse3: 107.3
>>>> rgb24_to_uv_128_avx: 111.3
>>>
>>> Worse than ssse3
>>
>> I know, but it's not related to this patch.
>> Same for all the comments below.
> 
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".