If it doesn't require a huge quantity of work, I plan to port my 3D simulation/game
engine from windoze to fedora8 Linux. My code totally assumes and requires
OpenGL 2.0 and GLSL shaders. My code does everything via IBOs/VBOs and
GLSL shaders, and nothing via "the old ways", so it is utterly pointless to attempt
a port without OpenGL2 + GLSL shaders --- or an OpenGL2/GLSL clone that works.

What is the status of OpenGL2+ and GLSL shaders on modern Linux?

Look into using GLUT, that may do what you need.

http://www.opengl.org/resources/libraries/glut/

Thanks for the suggestion, but decades of trials and tribulations (pain and misery) has taught me to avoid every "helper" package that can be avoided - and program to the lowest level library that offers the features I need (namely, hardware support of OpenGL2 and GLSL). Interestingly, the link you mentioned also agrees with my attitude for all but "small and medium size" applications (which mine surely is not). Thanks for the tip, though. For most people/applications, it's probably good advice. I guess the headers installed with the nvidia driver, plus the glx part of xwindows looks like the alternative. Thanks again.

If all you need is just something to do the rendering, and already have it ported, then GLX should be sufficient.

Give it a try and post back, and we can take it from there.

I'll ask one of my friends to post on here if he has any ideas, as he has been known to write a lot of OpenGL stuff in the past, but has commented on the lack of resources and help within the linux community. He will have more idea of what you require.

I'll be happy to hear what he has to say. I am rather surprised to find so little written about getting OpenGL+GLSL programs running on Linux, especially if the nvidia drivers deliver performance only ~10% slower than windoze.

Perhaps this means *I* must make a simple nvidia-drivers+OpenGL+GLSL+Linux program work properly, and post it somewhere. I'd much rather find out where someone else has done this already, though. Let them take the credit! :-)

Look for the Red book and Blue book, apparently they have a lot of information that you need to read.

Also the site below contains more info that will be useful.

http://nehe.gamedev.net/

I do have the red book, and I will look for any Linux/UNIX information in it.

I found one section in a book named "the OpenGL superbible 3rd edition" that looks like it has some hints for me. And I found a couple very old articles from kilgard that might help.

Of course, I do have a huge program I wrote 20 years ago on top of XWindows and xlib - a software development environment with compiler/debugger/GuiDesigner, so I can get some first order hints from that. But that old program contains zero OpenGL, so the role of GLX seems very important in my current OpenGL project.

Somehow I need to make sure the appropriate headers and libraries get included and linked in --- because I see files from many organizations (some of which are not complementary). Just for example, here are some file-groups:


##### nvidia #####
/usr/include/nvidia/GL/gl.h ----- include this instead of other gl.h ???
/usr/include/nvidia/GL/glext.h ----- include this instead of other glext.h ???
/usr/include/nvidia/GL/glx.h ----- include this instead of other glx.h ???
/usr/include/nvidia/GL/glxext.h ----- include this instead of other glxext.h ???
/usr/include/nvidia/cuda/cuda.h ----- include this for GPGPU in video-card ???
/usr/include/nvidia/cuda/cudaGL.h ----- include this for GPGPU via OpenGL ???


##### OpenGL == KhronosGroup #####
/usr/include/GL/glext.h ----- do not include this due to nvidia version ???
/usr/include/GL/glxext.h ----- do not include this due to nvidia version ???


##### XFree86 #####
/usr/include/GL/glxint.h ----- include to make XWindows bind to OpenGL ???
/usr/include/GL/glxmd.h ----- include to make XWindows bind to OpenGL ???
/usr/include/GL/glxproto.h ----- include to make XWindows bind to OpenGL ???
/usr/include/GL/glxtokens.h ----- include to make XWindows bind to OpenGL ???


##### SiliconGraphics #####
/usr/include/GL/internal/glcore.h ----- ?????
/usr/include/GL/glu.h ----- ????? do not include unless we want GLUT ???
/usr/include/GL/GLwDrawA.h ----- ???? ignore this ??
/usr/include/GL/GLwDrawAP.h ----- ????? ignore this ??
/usr/include/GL/GLwMDrawA.h ----- ????? ignore this ??
/usr/include/GL/GLwMDrawAP.h ----- ????? ignore this ??


##### precision insight #####
/usr/include/GL/internal/dri_interface.h ----- ????? ignore this ??
/usr/include/GL/xmesa_x.h --- for building normal ----- ????? ignore this ??
/usr/include/GL/xmesa_xf86.h --- for building server ----- ????? ignore this ??


##### Mesa3D == OpenGL clone #####
/usr/include/GL/gl.h ----- ????? ignore this ??
/usr/include/GL/gl_mangle.h ----- ????? ignore this ??
/usr/include/GL/glu_mangle.h ----- ????? ignore this ??
/usr/include/GL/glx.h ----- ????? ignore this ??
/usr/include/GL/glx_mangle.h ----- ????? ignore this ??
/usr/include/GL/osmesa.h ----- ????? ignore this ??
/usr/include/GL/xmesa.h ----- ????? ignore this ??

----------

Well, those are the .h include files that I noticed off hand. Who knows how many more may be lurking in some other directories and subdirectories? Surely not me!

And this completely ignores the problem --- ???where??? are all the appropriate pieces that need to be linked into the executable during the build process?????

Got me batman.

Nonetheless, perhaps somebody out there in nvidia linux land can immediately stare at my guesses above --- and set me straight! To clarify what I think I understand:

1: XWindows includes and libs are from XFree86.
2: OpenGL includes and libs are from nvidia.
3: Mesa files are for computers with video cards that come without OpenGL files.
4: SiliconGraphics files are for unknown-to-me.
5: PrecisionInsight has some product that interacts with Mesa - sometimes.

----------

I shall continue my reading, experimenting, and reporting to this forum.

Thanks for any advice and information any of you choose to give.

You could try finding them with either locate or find.

Either:

The next suggestion is to try posting at the links below:

http://www.nvnews.net/vbulletin/forumdisplay.php?f=14
http://forums.nvidia.com/index.php?showforum=23

They may be able to suggest something else.

---

Silicon Graphics is probably a reference to SGI workstations which are/were high end (read expensive ) graphics editing workstation back in the day. I have no idea what the state of affairs is with regard the company nowadays, but they still have a website.

OpenGL libs and includes could be from Nvidia, if you have a Nvidia card. From ATI if you have one of their cards.

Mesa3d I havent come across in a long time, since i used 3dfx cards, and as such you probably know/understand more about their drivers than I do.

DRI tends to be used by ATI and Intel Graphic card users.

As voodoo states, your bets bet is nehe (look at the bottom of each tutorial for all the different implementations of libraries and OS's).

It's been a while but here's the includes of the first example I could find:
And gcc use looks a bit like this:
Hope this enables you to get going?



Oh, these are some of the references I found on my fed8...(using $ locate GL | grep -v home | grep -v python | grep ".h")
There were some under the SDL directories but I presume you don't want to use that. If your really stuck then theres loads of game libraries that use whatever drawing libraries are available which are generally OS independent and fallback to available capabilities.

Since I promised to report my experiences, here are my first results.

I wrote a super short program to display a 3D cube (the old gross way, no VBOs), and got that to work with very little effort.

It contains the following includes:

#include <stdio.h>
#include <stdlib.h>
#include <nvidia/GL/gl.h> // file is "/usr/include/nvidia/GL/gl.h"
#include <nvidia/GL/glx.h> // file is "/usr/include/nvidia/GL/glx.h"
#include <nvidia/GL/glext.h> // file is "/usr/include/nvidia/GL/glext.h"
#include <nvidia/GL/glxext.h> // file is "/usr/include/nvidia/GL/glxext.h"

However, this tells me/us very little so far. But hey, it is something.

I am programming in the eclipse environment for the first time, so I don't really know what I'm doing. For example, I'm not sure exactly what libraries it is linking into the project. In one of the 999 settings dialogs, under linker, I entered "X11" and "GL" under the "Libraries (-I)" heading, and I entered "/usr/lib/nvidia" under the "Library search path (-L)" heading. However, I don't know whether those settings are doing anything, and I certainly don't know whether it includes or links-in files from its default paths (whatever they are) in preference to these settings (if it finds files in both). For that matter, I wonder how it knows which of the several linkable library files in "/usr/lib/nvidia/" to link into my executable. I guess I'll figure these things out as I go along.

When I tried to compile my whole 3D engine, I ran into zillions of syntax errors. Most of them I understand, and will be trivial to fix. A few I am scratching my head about, and may take some research to extract. At least I have already tested the Linux/XWindows/xlib alternatives to the windoze specific code in my engine. Turns out that was the easy part, I suspect.

More reports as I plow through the muck. I know the report most of you want to hear is "how much slower (or faster :-) perhaps) does the engine run on Linux". Though I will be happy with any result between -20% to +100%, I will not be so happy if the speed is more than 25% slower. We shall see. Probably we all should place our bets. :-) My bet: "the Linux version will run 12% slower", due to the speed of the nvidia linux drivers/OpenGL vs nvidia windoze drivers/OpenGL. Your turn.

Max

What about taking a look at the source for glxgears? glxgears is the "standard" 3d test app for Linux (at least Fedora). A quick look at their code should go a long way.

That sounds like a good idea to me. Unfortunately, after wandering around pages generated by my google searches, I cannot find the source code for glxgears - only millions of comments about how fast glxgears runs on various computers or video-cards.

Can you provide me a link (or the file)? Thanks.

Max

I had to figure out what package glxgears comes from. It is in glx-utils. Do you know about srpms? All the Fedora rpms have matching source rpms (srpms). You can get the srpms through yum. You just have to set the enable=0 to enable=1 in your /etc/yum.repos.d/fedora.repo file. Here is the section I am talking about:

Strange. When I select "add/remove software" under "applications" and search through installed packages, it shows glx-utils as installed (a check mark). Specifically, it lists it as "glx-utils - 7.0.2-2.fc8.i386".

Unfortunately, when i search my entire filesystem for "glx-utils" and "glxgears" with the "find" command, it outputs only the "glxgears" binary executable - nothing else. I have not been able to find glxgears anywhere on the internet, though I have encountered many people discussing how they compiled and ran glxgears. But the glx-utils packages I have found so far do not include glxgears source code. And the "add/remove software" utility on fedora8 does not list a separate package called "glx-utils development" or similar. I hate to ask "where is it hiding"... but where is it hiding? Thanks.

By default the srpms repos are shut off. You have to turn them on as described in post #13. Most (not all) of the graphical ways to install stuff on Fedora are just front ends for yum (which in turn is just a front end for rpm, kinda).