While looking to write a pure ooc version of ftgl, I was reading the source of ftgl-gl3 and I stumbled upon this piece of code:

static inline GLuint NextPowerOf2(GLuint in)
{
     in -= 1;

     in |= in >> 16;
     in |= in >> 8;
     in |= in >> 4;
     in |= in >> 2;
     in |= in >> 1;

     return in + 1;
}

This is needed because dealing with power-of-two textures (32x32, 64x64, 128x128, etc.) is more efficient with OpenGL (some implementations don't even support non-power-of-two textures!).