2017-12-02 18:59 GMT+01:00 Martin Vignali :
>
>
>> requiring FFMIN() to be evaluated per iteration could be slower
>> if the compiler fails to factor it out
>>
>>
>>
>> New patchs in attach :
>
> 001 : unchanged
> 002 : add "int min_width = FFMIN(w, 32)" at the start of the func
> 003 : add "int m
>
> requiring FFMIN() to be evaluated per iteration could be slower
> if the compiler fails to factor it out
>
>
>
> New patchs in attach :
001 : unchanged
002 : add "int min_width = FFMIN(w, 32)" at the start of the func
003 : add "int min_width = FFMIN(w, 8)" at the start of the func
Pass fate
On Sun, Nov 26, 2017 at 07:07:41PM +0100, Martin Vignali wrote:
> Hello,
>
> in attach patchs
>
> 0001-avcodec-huffyuvenc-increase-scalar-loop-count
> and
> 0003-avcodec-huffyuvenc-sub_left_prediction_bgr32-call-ds
>
> like diff_bytes and diff_bytes16, have AVX2 version, increase the scalar
> lo
2017-11-26 19:07 GMT+01:00 Martin Vignali :
> Hello,
>
> in attach patchs
>
> 0001-avcodec-huffyuvenc-increase-scalar-loop-count
> and
> 0003-avcodec-huffyuvenc-sub_left_prediction_bgr32-call-ds
>
> like diff_bytes and diff_bytes16, have AVX2 version, increase the scalar
> loop
> to call the align
Hello,
in attach patchs
0001-avcodec-huffyuvenc-increase-scalar-loop-count
and
0003-avcodec-huffyuvenc-sub_left_prediction_bgr32-call-ds
like diff_bytes and diff_bytes16, have AVX2 version, increase the scalar
loop
to call the aligned version in most case
0002-avcodec-huffyuvenc-remove-code-d