Re: [FFmpeg-devel] [PATCH v2] libavfilter/x86/vf_convolution: fix sobel swap issue on WIN64

[James Almer <jamrial@gmail.com>]

On 11/14/2022 12:20 PM, bin.wang-at-intel.com@ffmpeg.org wrote:
> From: "Wang, Bin" <bin.wang@intel.com>
> 
> Signed-off-by: Wang, Bin <bin.wang@intel.com>
> ---
>   libavfilter/x86/vf_convolution.asm | 11 ++++++-----
>   1 file changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/libavfilter/x86/vf_convolution.asm b/libavfilter/x86/vf_convolution.asm
> index c912d56752..9ac9ef5d73 100644
> --- a/libavfilter/x86/vf_convolution.asm
> +++ b/libavfilter/x86/vf_convolution.asm
> @@ -189,15 +189,16 @@ cglobal filter_sobel, 4, 15, 7, dst, width, matrix, ptr, c0, c1, c2, c3, c4, c5,
>   cglobal filter_sobel, 4, 15, 7, dst, width, rdiv, bias, matrix, ptr, c0, c1, c2, c3, c4, c5, c6, c7, c8, r, x
>   %endif
>   %if WIN64
> -    SWAP xmm0, xmm2
> -    SWAP xmm1, xmm3
> +    VBROADCASTSS m0, xmm2
> +    VBROADCASTSS m1, xmm3
>       mov  r2q, matrixmp
>       mov  r3q, ptrmp
>       DEFINE_ARGS dst, width, matrix, ptr, c0, c1, c2, c3, c4, c5, c6, c7, c8, r, x
> -%endif
> -    movsxdifnidn widthq, widthd
> +%else
>       VBROADCASTSS m0, xmm0
>       VBROADCASTSS m1, xmm1
> +%endif
> +    movsxdifnidn widthq, widthd
>       pxor  m6, m6
>       mov   c0q, [ptrq + 0*gprsize]
>       mov   c1q, [ptrq + 1*gprsize]
> @@ -281,7 +282,7 @@ cglobal filter_sobel, 4, 15, 7, dst, width, rdiv, bias, matrix, ptr, c0, c1, c2,
>       fmaddss xmm4, xmm5, xmm5, xmm4
>   
>       sqrtps    xmm4, xmm4
> -    fmaddss   xmm4, xmm4, xmm0, xmm1     ;sum = sum * rdiv + bias
> +    fmaddss   xmm4, xmm4, xm0, xm1     ;sum = sum * rdiv + bias
>       cvttps2dq xmm4, xmm4     ; trunc to integer
>       packssdw  xmm4, xmm4
>       packuswb  xmm4, xmm4

Should be ok.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".