nanogui: Multiple famebuffer
Subject:
Re: Multiple famebuffer
From:
Alex Holden ####@####.####
Date:
21 Nov 2000 11:02:27 -0000
Message-Id: <Pine.LNX.4.04.10011211042150.3925-100000@www.linuxhacker.org>
On Mon, 20 Nov 2000, Marc Nijdam wrote:
> If I draw to one fb, which is currently displayed by the hw, then draw
> to the other one, and point the hw with a single instruction at the new
> fb mem area, does that make it faster? No copies required, right?
I think overall it works out slower because you have to draw everything to
both framebuffers. It tends to be (or at least it used to be) used a lot
in things like games and video players because it has the advantage that
the change between one frame and the next takes place instantaneously,
which makes the animation look a lot smoother. Another useful technique is
the so called "dirty rectangles". With this method, you have keep a local
copy of the video memory, and whenever you make a change, you make it on
the in memory copy rather than the video memory and make a note of the
area which was changed. At some point later, you blit all of the changed
areas (the "dirty rectangles") to the video memory in one go. The
advantage is that you only copy the changed areas, and you don't copy the
bits which were changed more than once since the last frame update (the
overlapping areas). Another technique is off screen storage, which is
effectively using part of the video memory which isn't displayed on the
screen as a pixmap (or sprite) storage area, and using the built in
blitting capability of the video hardware to copy from one to the other
when necessary.
--
------- Alex Holden -------
http://www.linuxhacker.org/
http://www.robogeeks.org/
