On 04/09/2013 06:57 PM, Christophe Gisquet wrote:
> No vector tests it.

Do any of the ISO reference vectors test it? It seems almost pointless
to optimize a function that is literally never used.

> ---
>  libavcodec/x86/sbrdsp.asm    | 20 ++++++++++++++++++++
>  libavcodec/x86/sbrdsp_init.c |  2 ++
>  2 files changed, 22 insertions(+)
> 
> diff --git a/libavcodec/x86/sbrdsp.asm b/libavcodec/x86/sbrdsp.asm
> index 5c21b11..62df005 100644
> --- a/libavcodec/x86/sbrdsp.asm
> +++ b/libavcodec/x86/sbrdsp.asm
> @@ -275,3 +275,23 @@ cglobal sbr_qmf_pre_shuffle, 1,4,7,z
>      mova       m2, [zq]
>      movq    [r2q], m2
>      REP_RET
> +
> +INIT_XMM sse
> +cglobal sbr_qmf_deint_neg, 2,3,3,v,src,vrev
> +    lea     vrevq, [vq + (64-4)*4]
> +    add      srcq, (64-8)*4
> +    mova       m2, [ps_neg]
> +.loop:
> +    mova       m0, [srcq +  0]
> +    mova       m1, [srcq + 16]
> +    shufps     m0, m1, q3131
> +    shufps     m1, m1, q2020

is pshufd with memory arg any faster here?

Otherwise looks ok.

-Justin

_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to