Hi,

On Tue, Jan 24, 2012 at 7:21 PM, Ronald S. Bultje <[email protected]> wrote:
> Also implement sse2/ssse3/avx versions.
> ---
>  libswscale/x86/input.asm          |  299 
> +++++++++++++++++++++++++++++++++++++
>  libswscale/x86/swscale_mmx.c      |   48 ++++---
>  libswscale/x86/swscale_template.c |  159 +-------------------
>  3 files changed, 328 insertions(+), 178 deletions(-)

<Dark_Shikari> BBB: oh, sorry.
<Dark_Shikari> Heee, I like the pshufb.
<Dark_Shikari> ; Jason FIXME use 4xpshufb  <-- but this only uses 2x pshufb
<Dark_Shikari> It's not the same as the other?
<Dark_Shikari> Oh I think I see it, nvm

(I forgot to remove some code under %if 0, so will remove that before pushing.)

<Dark_Shikari> A TODO, but not required: replace pmaddwd with pmadcswd
to avoid 4x paddd in the xop version.
<Dark_Shikari> Looks good to me.

(I'll add the TODO.) Seems to be an XOP instruction so I'll need SSH
access to test that. Looks fun though.

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

Reply via email to