[FFmpeg-devel] [PATCH] swscale/x86/rgb2rgb: add AVX512ICL versions of shuffle_bytes

Ronald S. Bultje rsbultje at gmail.com
Sun Jan 26 00:27:33 EET 2025


Hi,

On Sat, Jan 25, 2025 at 9:26 AM Shreesh Adiga <16567adigashreesh at gmail.com>
wrote:

> @@ -64,6 +64,18 @@ cglobal shuffle_bytes_%1%2%3%4, 3, 5, 2, src, dst, w,
> tmp, x
>      add        dstq, wq
>      neg          wq
>
> +%if mmsize == 64
> +    and xq, mmsize-4
> +    shr xq, 2
> +    mov tmpd, -1
> +    shlx tmpd, tmpd, xd
> +    not tmpd
> +    kmovw k7, tmpw
> +    vmovdqu32 m1{k7}{z}, [srcq + wq]
> +    pshufb         m1, m0
> +    vmovdqu32 [dstq + wq]{k7}, m1
> +    lea wq, [wq + 4 * xq]
> +%else
>  ;calc scalar loop
>      and xq, mmsize-4
>      je .loop_simd
> @@ -80,6 +92,7 @@ cglobal shuffle_bytes_%1%2%3%4, 3, 5, 2, src, dst, w,
> tmp, x
>     add            wq, 4
>     sub            xq, 4
>     jg .loop_scalar
> +%endif


Would it be possible to fix up the indentation a little bit? This is quite
ugly.

Ronald


More information about the ffmpeg-devel mailing list