Just to echo David’s comments: OpenGL is going to give you better
framerates, and doubly so at higher resolutions and color depths. Writing
OpenGL code is very different from writing to a 2D linear framebuffer,
maybe i am new to linux, but i am writing software for oTHER ™ systems
for 4 years.
i wrote few apps for DX and one with OpenGL only used for 2d images, its
very fast! and other advantage of opengl =
very good image quality + it’s fast, just upload all graphics to video
memory and go…
but all things i wrote till now are for just few systems, so i am doing
setup of those machines at the end.
I decided to not use opengl for 2 reasons:
1st. i will need to write my own SDL for OpenGL, or just add another layer
to existing one.
2nd. none of my friends, that have installed linux out of box is not running
OpenGL acceleration on their nvidia cards.
They cannot setup drivers downloaded from network by them selfs, i wrote
that today - but again - it will be very ! nice to have graphical wizard for
that, that can handle this (this should not be difficoult), just like
windows has ! I don’t have box with nvidia card, but if somebody will give
me one - i will write this !
i hope nobody is angry, for my tons of questions. i am just curious &
inquiring guy…
Ack, speaking of dealing with broken OpenGL, I have had no time to finish
up the examples for DynGL 3. Sorry all. It’s on my todo list, but it’s
sharing that space with about seven other things.On Thu, May 30, 2002 at 09:01:50PM +0200, David Olofson wrote:
do that means that i have to write it for OpenGL to have real fast 2d
graphics ?
If your definition of “really fast” is similar to mine; yes, that seems
to be the only remotely reliable way. (Despite all these issues with
broken OpenGL features, I’d still say you’re much more likely to find a
working OpenGL setup than any other kind of accelerated video on Linux.)
It’s not much about portability, but about keeping the whole idea of X being
operated over the network. It doesn’t make much sense to provide the way to
allocate the VRAM area if the data for it has to be transferred over the
network anyway. Of course times had changed and currently X is usually
operated on the same machine as the client app, so there are several
extensions to expose it. Firstly you have shared memory extension that
allows for lessening the network traffic as the data remains in the system
memory accessible for both client and server. Secondly you have a brand new
XRender extension which is mostly praised for its alpha capabilities, but it
provides also the way to hardware acceleration. Who knows maybe soon we’ll
have also XRender support in SDL ? I don’t know whether there’s somebody
working on it, or not.
Regards,
JacekOn Thu, May 30, 2002 at 09:28:55PM +0200, Grzegorz Jaskiewicz wrote:
but X server runs with root priviliges, at least some parts. So i guess
graphics stuff under X windows should use it !
on the other hand, everything should be done way that doesn’t requires root
priviliges ( in modules fe ).But i guess guys from xfree.org cannot do that becouse of portability issues…
The funny thing is that some of my initial code worked slower in DGA than in
normal X. The problem was that the rendering took just a bit longer than
refreshing the screen and the video driver waited in SDL_Flip for the
next retrace doing nothing. The same code executed without DGA just didn’t
care about retrace and executed about 1.5x faster. After adding some
complexity to the code, the framerates falled miserably anyway, so I moved
to OpenGL after all.
Regards,
JacekOn Fri, May 31, 2002 at 01:15:17AM +0200, Grzegorz Jaskiewicz wrote:
So perhaps, “deal with it” isn’t the best response; “don’t worry about
it” would be better.
memory accessible for both client and server. Secondly you have a brand new
XRender extension which is mostly praised for its alpha capabilities, but it
provides also the way to hardware acceleration. Who knows maybe soon we’ll
have also XRender support in SDL ? I don’t know whether there’s somebody
working on it, or not.
Is there an implementation of the XRender extension that supports
hardware-accelerated alpha? I was under the impression that everyone raced
to implement the anti-aliasing stuff in XRender, and never did the rest,
but I haven’t really paid that much attention.
when you are done scribbling on the surface, send it in batch to the X server,
which will then do a hardware-accelerated blit to video RAM (assuming that
you have accelerated drivers, of course).
When I install my Mandrake distro, it asks me if I want to install the “Hardware
Accelerated” or “Software Only” version of X11 v4… what’s the exact difference
between them?
| >when you are done scribbling on the surface, send it in batch to the
| >X server, which will then do a hardware-accelerated blit to video RAM
| >(assuming that you have accelerated drivers, of course).
|
| When I install my Mandrake distro, it asks me if I want to install the
| “Hardware Accelerated” or “Software Only” version of X11 v4… what’s
| the exact difference between them?
|
Never heard of that before… maybe it installs some OpenGL support or
something? Or perhaps it loads the DGA module for X.On Fri, May 31, 2002 at 07:52:00PM +0200, CRV?ADER/KY wrote:
Don’t forget, SDL supports some hardware acceleration (hardware
accelerated blits at least) under Linux’s DirectFB. So, you don’t
absolutely need OpenGL to get hardware acceleration in Linux.On Thu, 2002-05-30 at 10:55, Niels Wagenaar wrote:
SDL supports hardware acceleration (only?) under OpenGL by
making use of the GLX libraries.
So, you don’t