using OpenGL to implement a window manager

[Keith Packard]

Around 19 o'clock on Apr 23, Jon Smirl wrote:

> Another plus or minus. If transparency is not involved some of the cases can
> paint straight into the display buffer without the need for an intermediate
> pbuffer. That will lower the VRAM pressure for things like xterm windows.

Transparency is only one of the many features gained by drawing window 
contents off-screen.  Of more common importance is the lack of visible 
exposure damage when moving windows around, and the (possibilty of) 
synchronized frame/window content updates when resizing windows.  Both of 
these improve the "feel" of the window environment without affecting the 
actual display at all.

In a transformed environment, the boundary pixels of the window will need 
to be composited with the underlying image (or windows); partial pixel 
coverage being equivalent to translucency, after all.

Even if the application window pixel contents account for the full
transformation, we'll still generally want to paint them off-screen and
blend them into place.

I suggest that any system capable of presenting a transformed desktop in 
real-time likely has sufficient video memory (or AGP space) to store 
enough window contents for a reasonable display.

One nice thing about making all of these architectural controls external 
to the window system is that you can still just turn them off and get back 
to the classic (if somewhat grotty looking) clipping-based composition 
model.

-keith


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
Url : http://freedesktop.org/pipermail/xserver/attachments/20040423/566f85f3/attachment.pgp