[MPlayer-dev-eng] [PATCH]: vo_gl2 bug fixes

Reimar Döffinger someone-guy at gmx.de
Wed Nov 12 19:53:36 CET 2003


Hi,

 > hmm, ok. I didn't get any slowdown on the two machines I had tested it on
 > so i assumed it was ok ;)

glClear only needs the framebuffer to be overwritten with a fixed value.

I think this only causes a slowdown when either

a) When the graphics card's pixel fill rate is too small to write about 
2*screenwidth*screenheight*fps pixels per second (can't imagine that is 
a problem with a _PC_ graphics card, but I don't know who uses this 
driver - I have no idea what kind of fill rate a SGI workstation has...).

b) When there is no hardware OpenGL support.

Nevertheless it shouldn't be needed.
If it is, this probably means there's another bug somewhere that should 
be fixed.

> ok, this works reliable if i startup in fullscreen mode. However, I still
> have the problem when I switch from window to fullscreen. Adding the same
> to the else block too improves the situation but still isn't reliable.
> i.e. when I switch to fullscreen a few times in a row, its successfully 
> cleared in like 50% of the cases, in the other cases its not.. hmm 

I just saw that there's (imho) a really stupid bug. glClear of course 
should be called after glViewport!
Could you try this patch as well ;-)
Could you also try if it works without the two glDrawBuffer calls?
Normally glClear shouldn't be needed when we're not in fullscreen, as in 
windowed mode the whole screen is used by the video and thus "cleared" 
by the next frame.

Greetings,
Reimar Döffinger
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: vo_gl2.diff
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/attachments/20031112/d7ea6b55/attachment.asc>


More information about the MPlayer-dev-eng mailing list