Hi,

On Fri, Oct 9, 2015 at 12:44 PM, Paul B Mahol <one...@gmail.com> wrote:

> +    packssdw                     m0, m0
>
+    packuswb                     m0, m0
> +    packssdw                     m1, m1
> +    packuswb                     m1, m1
> +    SBUTTERFLY                   dq, 0, 1, 2
> +    movh               [out_pixelq], m0
>

This should be shorted:

packssdw m0, m1
packuswb m0, m0
movh [out_pixelq], m0

One thing to notice is that SBUTTERFLY only makes sense if you care about
the high and low parts of the register. Here, you pack'ed them away so
they're effectively empty, so even in the above construct, you'd only want
punpckldq, not SBUTTERFLY dq, the second expands to punpckldq+punpckhdq.

Rest of patch looked OK at a quick glance.

Ronald
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Reply via email to