[MPlayer-dev-eng] [PATCH] autoq support for control()

[Juan J. Sierralta P.]

On Sun, 2002-02-10 at 19:02, Arpi wrote:

> i think the bitstream writer optim should be the first, as it's relative
> simple, and doesn't need architectural changes.

	Agree ...
 
> instead of doing template stuff, we may could try to localize most critical
> if()'s, and try to separate only them. yes, it isn't 100% solution yet, but
> may help a bit without big mess.
> 
> as Juanjo alerady did with MC slices, we could do the same to quant/DCT,
> i mean doing DCT for whole slice/frame first, and then doing quantization.
> maybe detecting and skipping black blocks also help a bit.

	I did that on 0.3.4 and Gerard reject it. Actually I detected on the MC
part if a MB has little or zero diferential variance and marked it as
skipped to avoid DCT/Q/IDCT/Q. But I'm thinking on bring that code
again. Maybe not based on the MB diferences variance (has problems) but
based on the SAD of the MB. But this could help on non intra since the
MC is performed on that case only. I'm not sure if there are a routine
that detect all zero coefficients on intra MB (it`s easy on the current
implementation) to avoid the IDCT/Q right now.

> (or even combining with partial intra refresh, so do not re-code all blocks
> at every frame, just teh changed blocks. dunno if it's allowed by mpeg1, but
> may help in both output bitrate and coding speed)

	You mean the old trick of making P frames with just intra MB, as VIC
does.

-- 
Juanjo sin .sig :(