havent gotten to multitexturing yet, so i cant give you an example on that

thing with opengl is that there is old opengl and there is modern opengl
in modern you would probably use glActiveTexture to set it up
http://www.opengl.org/sdk/docs/man3/
i am speaking about 3.3 core
that is what i picked to learn
you can use calls from any version with any version

from what i got so far modern works kinda like this:

the driver is a state machine
it keeps track of all the vertices, pixels and shaders
like when you make a buffer using glGenBuffers it returns to you a handle (more like an ID)
you say you want to use that thing in later calls by making it current, by binding it
then when you fill it using glBufferData, that handle points to that data internally
note that it dosent care what is in the buffer, and you can use that data in shaders as you wish

the driver also keeps track in what memory that data is
it also compiles shaders as programs for the gpu to execute
then when tell it to actually draw something it starts the current active program on the gpu (glUseProgram)

modern shader based gpus are basically a bunch of compute units (they call em shaders in opegl context, or whatever)
so ye its a bunch of math processors grouped together in packs so they can share some cache memory
http://www.stanford.edu/~yifanz/images/research/GPUarchitecture_large.png
that is the compute part, there are couple parts to process vertices before it
kinda like http://www.techarp.com/article/NVIDIA/GeForce_8800_TR/8800_architecture.jpg
i remember reading a pdf from nvidia about the 8800
ofc it was full of catchphrases but it did give an overview so it may be worth finding it online

i think texture units are still on a special part of a gpu, but i might be wrong
(why would they still recommend n^2 texture sizes if it was using main memory, idk)

i found this also, intel 4k overview
http://www.ozone3d.net/public/jegx/201204/intel-ivy-bridge-hd4000-diagram-architecture.jpg
also some in depth http://people.freedesktop.org/~marcheu/linuxgraphicsdrivers.pdf (chapter 8 is about opengl)

so shaders are basically compute kernels for massively parallel execution
and you should look at them as just parallel data processors, not dedicated 3D processors
ofc opengl makes it easier to use them for 3D to 2D processing

bdw shaders dont know what you call the buffers you send them

http://pastebin.com/Qf4gfiuD
heres the small program i patched up
it uses the shaders from last post, named " fragmentshader.frag " and " vertexshader.vert "
and they have to be in the same directory as the program (had trouble embedding normal text in C)
fun thing is you can change the shader without recompiling the program itself

if you run windows you need to change usleep() to sleep(10) and maybe clock_gettime() (or just delete it, its to calculate fps)
if i remember GLFW can handle timings too, but i like clock_gettime since its a VDSO in linux glibc

compile with CC -lGL -lglfw -lSOIL
needs soil for loading pictures to RGB
and GLFW 3 for everything else
GLFW has excellent documentation and can do lots of things

Multitexturing in modern OpenGL:



Fragment shader:



I am using OpenGL everyday at work for more than 6 years now, so feel free to ask if you have any questions about this API.

BTW, very nice game!

randall,
thanks for the multitexturing sample. We can assume that GL_TEXTURE0 is active by default? I wanted to try some bump mapping in a hex game (different thread), but it looks like I will only need one texture since the actually look of the pieces is computed by the uv coordinates of the triangle.
I just uploaded a tuned version of this game that I think looks the best. It does a small radius gaussian blur followed by a large radius gassian blur on each of the two layers, and each layers is render at 4 times its display resolution.
I get about 1000fps (the top number in the lower left) on a gtx 850, However, I noticed that once the frame buffers hit a certain size, the performance on a i5-3320 with integrated graphics takes a huge hit.

Yes, GL_TEXTURE0 is active by default (this is required by the OpenGL specification).

getting out of memory error compiling it. looks cool, hope to learn something from it and perhaps become more valuable!

In the options>compiler setup you will have to increase the memory.
Let me know if it works for you. I didn't optimize either of the two Gaussian blurs, so if your graphics card can't handle the resolution, try changing as

Also, the particles and grid background require quite a bit a processing, so it is triple threaded.

You know, I just upgraded to latest FASM and compiled it with no problem. Not sure if replacing the old fasm version with this new one adjusted memory size. I forgot about that option though. Thanks for reminding me.

Wh.....a..t..? Seems I have to learn a lot. I'm way behind I guess.
This looks amazing! Better than GTA 5 . FASM is future. Assembly is future. This will tell C
Fanboys what asm can achieve.
I will try it when I'm home.
i3-2370M, 64 bit, 8 GB RAM,Intel HD 9000, nVIDIA GT400. Should that be enough?

I would like to clean up the code a bit, possibly optimize the graphics, and know where it crashed. Kevin, could you tell me where it crashed for you?

One exe was crashing when I tried to launch/run it with fasm. Not sure what was up. I deleted it and reassembled and now there is no problem like that.

tthsqe: One of the best game examples written in FASM. It works on my roomates PC, but it crashes on my PCs (laptop+itx) on Win7 32BIT. On startup, "C000001Dh: Illegal instruction" (just from this, I would suspect that esp became unbalanced or eip didn't return correctly or some jmp [table+ecx*4]).

Wow, this is a super mega example! I'd say it would be better categorized as a whole project rather than example.

m3ntal, maybe your computers don't support SSE4.1? Check with CPU-Z. Or run this program and make sure it terminates without crashing:

Loco: That was it! My CPU doesn't support SSE4.1. VC6 debugger displayed exactly the same error and address.

ok - I am in the process of cleaning up the code. It really does not need sse 4.1 dpps, so am I definitely replacing it. I'm also trying to speed up the graphics, as I am getting very poor results even on haswell GPU. Of course, gtx 580 just plows through it, but this is probably because they have a better driver.

[ Post removed by author. ]

... about SSE 4.1 you can find something useful here :
http://software.intel.com/sites/default/files/m/a/9/b/7/b/1000-SSE.pdf
and I have a question about your source code :
what is and where can be found :
call GLCleanup
and this
invoke GetClientRect,[hwnd],rc !!

call GLCleanup: misc.inc line 380
invoke GetClientRect,[hwnd],rc: I don't think I use this. Instead I have
CorrectMainWindowSize: misc.inc line 879

Now that I look back at the cleanup function, I think I am missing some frame buffers and the shader programs. I;ll have to add clean up code for those as well.

I take a look ... also why don't use glFlush and OpenGL.GL.ARB
Maybe one video tutorial about how you codding this into real time will help, the code seam clear , just is to big

@tthsqe : take a look at olly , can help you ...

@catafest: With what can ollydbg help me? Things seem to be working fine in this example, except that I might not be cleaning everything at the end.