[FFmpeg-devel] [PATCH] lavfi/overlay: add support for partial overlaying

[Stefano Sabatini]

On date Tuesday 2013-03-05 11:23:31 +0000, Paul B Mahol encoded:
> On 3/5/13, Stefano Sabatini <stefasab at gmail.com> wrote:
> > On date Tuesday 2013-03-05 01:46:58 +0100, Stefano Sabatini encoded:
> >> On date Friday 2013-02-22 15:57:27 +0100, Stefano Sabatini encoded:
> >> > On date Tuesday 2013-02-19 20:10:42 +0100, Stefano Sabatini encoded:
> >> > > TODO: bump micro
> >> > > ---
> >> > >  libavfilter/vf_overlay.c |   60
> >> > > ++++++++++++++++++++++++++--------------------
> >> > >  1 file changed, 34 insertions(+), 26 deletions(-)
> >> >
> >> > Ping.
> >>
> >> Updated. This avoids pointless loops and slightly more readable logic.
> >
> > New update, faster (and a bit uglier).
> > --
> > FFmpeg = Fast Forgiving Merciful Pure Efficient Glue
> >
> 
> It should be fine if fate passes and valgrind is silent, no crash
> happens if very big values are given.... and filter is not overall
> slower than before.

Before patch:
7756140 decicycles in over1, 1 runs, 0 skips
7798680 decicycles in over1, 2 runs, 0 skips
7810207 decicycles in over1, 4 runs, 0 skips
7984676 decicycles in over1, 8 runs, 0 skips
7903886 decicycles in over1, 16 runs, 0 skips
7956860 decicycles in over1, 32 runs, 0 skips
7929557 decicycles in over1, 64 runs, 0 skips
7890497 decicycles in over1, 128 runs, 0 skips
7885514 decicycles in over1, 256 runs, 0 skips
7876744 decicycles in over1, 512 runs, 0 skips
7858536 decicycles in over1, 1024 runs, 0 skips
7847870 decicycles in over1, 2048 runs, 0 skips
7864835 decicycles in over1, 4096 runs, 0 skips

After patch:
7825500 decicycles in over2, 1 runs, 0 skips
9005955 decicycles in over2, 2 runs, 0 skips
8448600 decicycles in over2, 4 runs, 0 skips
8157817 decicycles in over2, 8 runs, 0 skips
7992945 decicycles in over2, 16 runs, 0 skips
7945018 decicycles in over2, 32 runs, 0 skips
7899676 decicycles in over2, 64 runs, 0 skips
7858572 decicycles in over2, 128 runs, 0 skips
7843465 decicycles in over2, 256 runs, 0 skips
7841009 decicycles in over2, 512 runs, 0 skips
7839361 decicycles in over2, 1024 runs, 0 skips
7895600 decicycles in over2, 2048 runs, 0 skips
7874947 decicycles in over2, 4096 runs, 0 skips

After a few more optimizations, is about 0.01 percent slower.
Pushed.
-- 
FFmpeg = Fostering and Fancy Most Picky Efficient Gladiator