[MPlayer-dev-eng] Faster vo_svga, and DR question

Matan Ziv-Av matan at svgalib.org
Tue Apr 2 14:21:43 CEST 2002


Hello,

DR question:
How does DR actually work? You can't draw to the currently displayed
image in video ram, since that might cause tearing. Since the decoder
might draw only changed parts of the picture, I can't see how double
buffering can be used.

attached is a modified vo_svga.c which is a bit faster.

Example benchmark:
mplayer -benchmark -nosound hapoel/goals/02/uefa/lokomotiv/loko.mpg

V:  32.7  804  31%  0%  0.0% 0 0 0%
BENCHMARKs: VC:  10.050s VO:   0.006s A:   0.000s Sys:   0.207s = 10.263s
BENCHMARK%: VC: 97.9290% VO:  0.0538% A:  0.0000% Sys:  2.0172% = 100.0000%

old vo_svga:
V:  32.7  804  10% 29%  0.0% 0 0 0%
BENCHMARKs: VC:   3.345s VO:   9.495s A:   0.000s Sys:   0.429s = 13.269s
BENCHMARK%: VC: 25.2095% VO: 71.5575% A:  0.0000% Sys:  3.2330% = 100.0000%

(It seems to me that the VC/VO seperation might not be accurate, but the
total time shows the actual gains). 

The main difference in the new driver is:
Does not use vgagl.
Displays OSD per image size, and not screen size.
No software scaling - it does not make sense for svgalib anyway.

-- 
Matan Ziv-Av.                         matan at svgalib.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: vo_svga.c.gz
Type: application/octet-stream
Size: 3805 bytes
Desc: 
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/attachments/20020402/9715a257/attachment.obj>


More information about the MPlayer-dev-eng mailing list