[FFmpeg-devel] [PATCH] x86/exrdsp: optimize ff_reorder_pixels_avx2()

[James Almer]

On 9/18/2017 4:35 AM, Martin Vignali wrote:
> 2017-09-18 3:52 GMT+02:00 James Almer <jamrial at gmail.com>:
> 
>> From: Henrik Gramner <henrik at gramner.com>
>>
>> Tested with "checkasm --test=exrdsp -bench"
>>
>> Before:
>> reorder_pixels_c: 5187.8
>> reorder_pixels_sse2: 377.0
>> reorder_pixels_avx2: 331.3
>>
>> After:
>> reorder_pixels_c: 5181.5
>> reorder_pixels_sse2: 377.0
>> reorder_pixels_avx2: 313.8
>>
>> I don't have the same result using a start/stop timer,
> but your testing approach is probably better than mine.

I also had a hard time getting to notice a difference with
star/stop_timer, and it's clear why seeing how little difference this
change truly makes.

You can build the above tool with "make checkasm", and the executable
will be in the tests/checkasm folder. The results tend to be less
variable and it's better detecting small differences between functions.

> And like, you both think it's a better way to do it, it's ok for me !
> 
> Thanks
> 
> Martin
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>