[Mplayer-dev-eng] x11 vs. dga

Balazs Tibor tibcu at sch.bme.hu
Mon Jun 18 18:03:25 CEST 2001


Hi Acki!

> Hi Balazs,

Still Tibor :-)
(Balázs is my surname - yes I know, Hungarians are crazy ;-) )

> now I've made the doublebuffering feature switchable from the commandline.

thx!!! :)

I've made some experiments with DGA at home. Without this switch I had to
modify the source, (and of course compile twice). Now I can disable/enable
doublebuffering in commandline, that's great!

My experiments were successfull, finally I could watch Every Given Sunday
without -framedrop!!! (with it, the fast motion parts were just
slide-shows) (352x288 avi, CD-quality sound (apart from compression ;-) )

If you are interested in some test results (-benchmark) (they're really
interesting at least for me), I'll post them to the list tomorrow or
better upload them to incoming. (A'rpi, is incoming still at thot?)

> By default it is off (this unfortunately also disables OSD in the DGA

Good, so I don't have to specify osdlevel 0 :)

> driver, sorry, but it would flicker like hell if I did it anyway !!!).

but pressing 'o' still works, doesn't it? (just to see the position where
I am actually)

> The X Server does a BUSY wait if I use the XDGASetViewport() call which
> does the trick for doublebuffering. I always thought this would be handled
> within a interrupt ... I was so sure, I never bothered to check this ...
> :-(

:-))))

> Here are some speed values for ChickenRun, I've always been using this one
> as a reference:

Could you give me some megs of this file? (for testing purpose naturally)

> The percentage values are the sum of CPU usages of X11 and mplayer summed
> up during the same section of the movie.

I've got no terminal attached to my computer :-(

maybe -benchmark can help?

> bpp. For convenience the above values with -bpp 32: (the X11 driver
> ignores the -bpp switch, so its output is listed only once ...)

?? -- -bpp 16 doesn'r work for you?

> I don't quite understand the *BIG* differnce between -bpp 24 and -bpp 15
> but the result reproduces (at least with top).

Yes, -bpp 15 can be a great help,
I used -vo dga /cdrom/Sunday1.avi -v -osdlevel 0 -bpp 15 to be able to
watch it fluently. (X running in 16 bpp) (X3.3.6)
without -bpp 15 (so actually 16 bpp) it was so slow

I don't understand this, because it says:
vo_dga: Mode: depth=15, bpp=16, r=007c00, g=0003e0, b=00001f, conversion (slow), (-bpp 15)
vo_dga: Mode: depth=16, bpp=16, r=00f800, g=0007e0, b=00001f, native (fast), (-bpp 16) ...disabled by -bpp 15
So I believed that using -bpp 15 will slower the things, but fortunately
the contrary happened :-)

(Of course with doublebuffering it wouldn't have been enough)

> I *believe* speed penalty for doublebuffering depends on the selected
> refresh rate - the higher the vertical refresh rate is, the less time is
> needed on the average to wait for the vertical blank.

old monitor -> 56, 60 Hz and 87 Hz interlaced
(but in low res mode it must be higher:
vo_dga: Selected video mode  400 x  300 @ 112 Hz @ depth 16, bitspp 16,
video 352 x 288.)

> However higher refresh rates increase transfer volume on graphics memory
> and thus slow down the copying ...

EHH :-)))))))))

thanks again for the fixes and the answer

Best regards
Tibor Balázs (Tibcu)


_______________________________________________
Mplayer-dev-eng mailing list
Mplayer-dev-eng at lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/mplayer-dev-eng



More information about the MPlayer-dev-eng mailing list