Quote:Original post by silvermace
that sounds like _alot_ of overkill, i use a 2pass seperable blur, 8 samples per pass, so 16 blur samples per pixel, and i render to a P-Buffer bound as a texture so there is no readback of the framebuffer (a main killer)
all in all, i only ever render the scene 2 times, 4 times if i want excesive visual quality, i never readback the framebuffer and everything is done on the GPU.
read the article on Gamasutra, excelent as it provides some PS/VS code to help you along.
Ok, yeah that does sound better. I'm thinking of the following:
1) Render the scene once.
2) glCopyTexSub.
3) Draw this to the pbuffer bound as a texture 2 times with a 8 sample blur. (I would also slightly up the contrast, lower the brightness with each draw here)
4) Draw this over what was drawn in step one with additive blending.
I think this would be as fast as possible and get everything I wish to do done. My only question is: Is there anyway to have glCopyTexSub2d create a texture 1/2 the height and width of my screen? If not I imagine I would have to render twice instead (which I would rather not do as I have a complex scene).