[MPlayer-users] vo_vesa introducing

[Arpi]

Hi,

> > > > also tried scaled yv12, vcd sized mpg to 640x480x16bpp
> > > > but it was extremly slow, 210% cpu usage...
> > > I'm watching DVD in real-time...
> > hmm. maybe mtrr was not set - X 4.1.0 sets it but i had no X started.
> > i'll check again
> > 
> During tests - I didn't touch mtrrs.
> (It useless since we have fastmemcpy which performs same logic of memory writing)
mayeb it's true for amd...
on intel mtrr does matter, it speedup video 5 times.

> > > > something must be wrong there. under X11, i'm able to scale same video
> > > > to 1024x768x16bpp using 30% cpu...
> > > > 
> > > Every my test I performed with ffmpeg12 and ffdivx decoders.
> > bad. test with libmpeg2 too, it renders video by slices.
> > i think you ignore x,y in draw_slice()
> > 
> Yes! You are right. But I'm didn't understand what they mean.
> Well - I'll try to use them.
> I guessed that that draw_slice accepted full frame.
draw_slice updates a region of the screen. it can be a small
macroblock or even a whole frame.
x;y w*h  defines the region to be updated.

x currently not used - always 0
y is used only by libmpeg2 codec. other codecs draw whole frame.


A'rpi / Astral & ESP-team

--
mailto:arpi at thot.banki.hu
http://esp-team.scene.hu