Just a thought: what is the preferable solution (for glow/bloom effect):
- Use a FBO and several render textures, each texture is generated from the previous one, using a 4x4 downsampling shader? (each step has the same size, the original one, hence the use of a single FBO)
- Use multiple FBO’s (one per downsampled texture), each FBO rendered into a smaller viewport each frame? (must use several FBO’s because texture sizes are different)
> The first solution implies less FBO switching, but larger textures processing into the fragment shader,
> the second one implies more FBO switching, but less and less pixels to process as the texture becomes smaller
Your ideas? 
HT
