VMD-L Mailing List

From: Massimiliano Stengel (stengel_at_dimca20.univ.trieste.it)
Date: Tue Aug 31 1999 - 06:47:27 CDT

On Mon, 30 Aug 1999, John Stone wrote:

> > Hi,
> >
> > I have Linux RedHat 6.0, a 3Dfx Voodoo 3 3000 card and I'm trying to
> > recompile VMD to use the latest Glide (3D accelerated) library. I have
> > one problem and one question. First the problem: if I build Mesa with
> > linux-glide option I get only *dynamic* .so libraries, and VMD needs
> > static ones.
>
> You can change the configure.options.LINUX.mesa flags and remove the
> STATIC flag. This should allow you to link against the dynamic link
> libraries. We normally build the Linux binaries statically linked in order
> to avoid the myriad problems with different people's linux boxes having
> different revs of libC, Mesa, etc...
>

Thanks very much, that worked! In only 10 minutes I had a new dynamic
executable, and ldd confirmed that *all* libraries were dynamically
linked. I executed it and worked fine. I set the env variable MESA_GLX_FX
to "fullscreen" in order to use hw acceleration, but the fullscreen mode,
as you told me, didn't work at all. The glide drivers are surely working:
at startup in the console I see the 'video memory unprotecting' message,
the video blinks for a moment just like if it were trying to swich to
fullscreen mode, but it suddenly returns to normal behavior, i.e. in
window rendering and no hardware acceleration.

I don't know anything about 3D drivers, I've heard around that this
version of glide uses a glx protocol: the program opens a window and the
drivers switch automatically to fullscreen, but don't ask me how :). The
only thing I know is that some Mesa-3.0 demos work with this trick (glutfx
for example), while others like gears don't and... yes, gears behaves
exactly like VMD.

I'm also working on another front: the past week I recompiled VMD using
glx drivers for the Matrox G200 card. This time I was able to get hardware
acceleration in a window and the performance was quite good, but due to a
problem with my X server (it was quite unstable with glx module loaded) I
did't test it for much time.

> > The solution would be to build the statc .a libraries or to
> > let VMD use the dynamic ones. Any advice is highly appreciated in both
> > tasks. Then the question: at the time the glide lib for V3 is in alpha
> > stage and can only work in fullscreen.
> > Is there any drawback in running VMD in fullscreen mode?
>
> VMD has no real concept of "fullscreen". You might
> be able to modify the OpenGLDisplayDevice.C code to make it work with
> your fullscreen mode, but its probably more trouble than its worth. It
> would be better if they fix the Voodoo drivers to work in a window.
> What happens when you try using a non-fullscreen window? Does it fail
> utterly, or does it switch to fullscreen, drawing to a clipped window, or to
> the whole screen anyway?
>
> > Is there a way to put GL display on local
> > machine and have the dialog windows on another computer (one meter apart
> > from the other). Well, to be honest I asked two questions :).
>
> Not at the present time, but this is something that we're interested in
> doing for various reasons, and it would certainly make VMD work better with
> boards that are fullscreen-only. This is also a nice setup if you have a
> dual screen system, or when you want to use VMD to give a presentation, but
> you only want the 3-D graphics to show up on the projector etc.
> I'm planning on having this capability for the next rev (after 1.4) of
> VMD.
>

Well, if the fullscreen mode worked there would be the possibility to
start VMD from a telnet session on another box and submit commands from
the console, and this would be a simple work-around to avoid complication
of sending dialog windows to another host... but I really don't know how
the glide glx fullscreen mode works, and I'm quite scared about the idea
of hand-hacking a C OpenGL code...

Regards,

  Max

> Let us know how the 3dfx drivers work out. I've got a Voodoo3 in an NT
> box here that I've been testing with the Windows version of VMD, and the
> 3dfx drivers have a few critical bugs that prevent a normal build of
> VMD from working, specifically, their driver causes applications to crash
> when the application draws certain kinds of lines and points. Drawing
> triangles and quads works out fine however. I've been waiting for 3dfx
> to get a new driver out that has this fixed.
>
> Let us know when you make any new discoveries on the Linux front,
> many people would LOVE to be able to get hardware acceleration on their
> Linux machines for running VMD.
>
> Thanks,
> John Stone
> vmd_at_ks.uiuc.edu
>
> --
>
> Theoretical Biophysics Group Email: johns_at_ks.uiuc.edu
> Beckman Institute http://www.ks.uiuc.edu/~johns/
> University of Illinois Phone: (217) 244-3349
> 405 N. Mathews Ave FAX: (217) 244-6078
> Urbana, IL 61801, USA Unix Is Good For You!!!
>