OpenGL Tiles/Blitting/Clipping

In OpenGL, how can I select an area from an image-file that was loaded using IMG_Load()? (I am working on a tilemap for a simple 2D game)

I'm using the following principle to load an image-file into a texture:

GLuint loadTexture( const std::string &fileName ) {

    SDL_Surface *image = IMG_Load(fileName.c_str());

    unsigned object(0);                        
    glGenTextures(1, &object);                 
    glBindTexture(GL_TEXTURE_2D, object);   

    glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, image->w, im开发者_如何学编程age->h, 0, GL_RGBA, GL_UNSIGNED_BYTE, image->pixels);

    SDL_FreeSurface(image);
    return object;
}

I then use the following to actually draw the texture in my rendering-part:

glColor4ub(255,255,255,255);                
glBindTexture(GL_TEXTURE_2D, texture);      
glBegin(GL_QUADS);                          
glTexCoord2d(0,0);  glVertex2f(x,y);
glTexCoord2d(1,0);  glVertex2f(x+w,y);
glTexCoord2d(1,1);  glVertex2f(x+w,y+h);
glTexCoord2d(0,1);  glVertex2f(x,y+h);
glEnd(); 

Now what I need is a function that allows me to select certain rectangular parts from the GLuint that I get from calling loadTexture( const std::string &fileName ), such that I can then use the above code to bind these parts to rectangles and then draw them to the screen. Something like:

GLuint getTileTexture( GLuint spritesheet, int x, int y, int w, int h )

Go ahead and load the entire collage into a texture. Then select a subset of it using glTexCoord when you render your geometry.

glTexSubImage2D will not help in any way. It allows you to add more than one file to a single texture, not create multiple textures from a single file.

Example code:

void RenderSprite( GLuint spritesheet, unsigned spritex, unsigned spritey, unsigned texturew, unsigned textureh, int x, int y, int w, int h )
{
    glColor4ub(255,255,255,255);                
    glBindTexture(GL_TEXTURE_2D, spritesheet);
    glBegin(GL_QUADS);                          
    glTexCoord2d(spritex/(double)texturew,spritey/(double)textureh);
    glVertex2f(x,y);
    glTexCoord2d((spritex+w)/(double)texturew,spritey/(double)textureh);
    glVertex2f(x+w,y);
    glTexCoord2d((spritex+w)/(double)texturew,(spritey+h)/(double)textureh);
    glVertex2f(x+w,y+h);
    glTexCoord2d(spritex/(double)texturew,(spritey+h)/(double)textureh);
    glVertex2f(x,y+h);
    glEnd(); 
}

Although Ben Voigt's answer is the usual way to go, if you really want an extra texture for the tiles (which may help with filtering at the edges) you can use glGetTexImage and play a bit with the glPixelStore parameters:

GLuint getTileTexture(GLuint spritesheet, int x, int y, int w, int h)
{
    glBindTexture(GL_TEXTURE_2D, spritesheet);

    // first we fetch the complete texture
    GLint width, height;
    glGetTexLevelParameteriv(GL_TEXTURE_2D, 0, GL_TEXTURE_WIDTH, &width);
    glGetTexLevelParameteriv(GL_TEXTURE_2D, 0, GL_TEXTURE_HEIGHT, &height);
    GLubyte *data = new GLubyte[width*height*4];
    glGetTexImage(GL_TEXTURE_2D, 0, GL_RGBA, GL_UNSIGNED_BYTE, data);

    // now we take only a sub-rectangle from this data
    GLuint texture;
    glGenTextures(1, &texture);
    glBindTexture(GL_TEXTURE_2D, texture);
    glTexParameteri(GL_TEXTURE_2D, /*filter+wrapping*/, /*whatever*/);
    glPixelStorei(GL_UNPACK_ROW_LENGTH, width);
    glTexImage2D(GL_TEXTURE_2D, 0, RGBA, w, h, 0, 
        GL_RGBA, GL_UNSIGNED_BYTE, data+4*(y*width+x));

    // clean up
    glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
    delete[] data;
    return texture;
}

But keep in mind, that this function always reads the whole texture atlas into CPU memory and then copies a sub-part into the new smaller texture. So it would be a good idea to create all needed sprite textures in one go and only read the data in once. In this case you can also just drop the atlas texture completely and only read the image into system memory with IMG_Load to distribute it into the individual sprite textures. Or, if you really need the large texture, then at least use a PBO to copy its data into (with GL_DYNAMIC_COPY usage or something the like), so it need not leave the GPU memory.