[Ffmpeg-devel] [PATCH] Chinese AVS video decoder

Stefan Gehrer stefan.gehrer
Thu Jun 29 08:43:14 CEST 2006


thanks for the input.

Von: Michael Niedermayer

> some of the above look similar to h.264, they shouldnt be duplicated if
> they
> are identical

Well vertical, horizontal, dc_128 and probably plane mode too are
functionally the same, but while in h264 they take the surrounding
pixels directly out of the framebuffer I get them out of arrays top and
left. This is due to the fact that the rules how to set pixels according
to which neighbours are available are a bit more complex 
(see load_intra_pred*)

> above could go into dsputil

Ok, so a file cavsdsp.c containing mocomp, deblocking and inverse
transform ?

> of course can it be done branchless (X= (col_mv->x>>31) and ^X-X should
> help)
> and when you are already at it, get rid of the division too

I was hoping to provoke an answer like that :)

> this is not acceptable, please use get/release_buffer() from
> AVCodecContext

I had this some time along the way, but it somehow failed me. I will
try to get back to that.

> this is wrong, you must output the not yet outputed delayed frames here

I will come up with a new patch, but don't expect anything before the weekend.

Stefan Gehrer

Echte DSL-Flatrate dauerhaft f?r 0,- Euro*!
"Feel free" mit GMX DSL! http://www.gmx.net/de/go/dsl

More information about the ffmpeg-devel mailing list