So with the previous optimizations I discovered, I've been rewriting my engine. But I got sort of stuck when I came to the shaders, because there are so many different ways of doing it. At first I used different shaders for different effects. One for texture mapping, one for texture mapping with normal mapping, one for texture mapping with normal mapping with some other effect, and so on. It was very tedious to have more than one, mainly because of all the extra code in each shader that is necessary to make it work. After that I swapped to a more unified system, where instead of using different shaders, I just put everything into one shader, but I could enable and disable features as I wanted. This is of course very nice for demonstration purposes, because I can change stuff at runtime with a simple function call. But the problem with this is that even though I turned things off, it would still send all the information, so turning everything off would change the output image, but not the performance.
Now I've sort of gone back to my first system, but I've changed it around a bit. Instead of adding in all that extra code, I created an include preprocessor directive. Currently it only works with 3 different things, GL_UNIFORMS, GL_LIB, and GL_MAIN, but I could of course make it work with filenames and such as well. Uniforms is the one with the uniforms (duh), sent from the lua code, having the projection information and such. Lib has a library of handy functions to make life easier, including the 'uv' function you see. And Main does all the main things, reverse projection, raycasting, uv mapping, all the good stuff to get the proper information you need. With all of this, I can create easy shaders:
That shader effectively takes in the 3 vertex positions, 3 vertex colors, interpolates the color, and scales it by the depth to the pixel, giving this result:
Here's a different shader, which also does texture mapping along with the per vertex color:
And its result: (I pushed the camera further ahead, and reduced the intensity on the depth for a better image)
Here is one last shader, where I took out the per vertex color, and added per vertex texture. =O
And the result: (I left only one cube alive, and I rearranged the vertices for a better image)
What do you guys think?
