[FFmpeg-devel] [PATCH] vp9: 16bpp tm/dc/h/v intra pred simd (mostly sse2) functions.

Ronald S. Bultje rsbultje at gmail.com
Fri Sep 25 23:13:13 CEST 2015


Hi,

On Fri, Sep 25, 2015 at 5:09 PM, Christophe Gisquet <
christophe.gisquet at gmail.com> wrote:

> Hi,
>
> 2015-09-25 22:36 GMT+02:00 Ronald S. Bultje <rsbultje at gmail.com>:
> > +pd_16: times 8 dd 16
>
> In h264_intrapred_10bit.asm
>
> > +pd_32: times 8 dd 32
>
> In h264_idct_10bit.asm
>
> And that's about anything remotely useful I have to say, as you were
> careful about those duplications.


Will move into constants.c, thanks for noticing.

> +    pmaxsw                  m0, m4
> > +    pmaxsw                  m1, m4
> > +    pmaxsw                  m2, m4
> > +    pmaxsw                  m3, m4
> > +    pminsw                  m0, m5
> > +    pminsw                  m1, m5
> > +    pminsw                  m2, m5
> > +    pminsw                  m3, m5
>
> That was the only thing that surprised me, but then again I didn't
> look much at the tm mode.


Hm, right, so the reason tm needs to clip (and others don't) is because
this is the only one applying a non-positive filter. Most other filters
(like in directional intra pred, but also dc) are a+b*2+c+2>>2 or a+b+1>>1
or just simple pixel copies (like v/h), but this one is a+b-c, so it can
flip out of the allowed bitrange and thus needs clipping.

Ronald


More information about the ffmpeg-devel mailing list