JoeJ said:
What? I'm shocked.

The didnt even supported core contexts for a few years, and forced driver writers to cap the version number at 2.1 in the fermi era.

Geri said:
(I dont know if this function even exists in newer opengl contexts, so thats why i am suggesting not to waste your time on core profiles)

Core profiles are made to define a new standard for modern HW, deprecating the legacy stuff for very good reasons.
So your advise to stick at legacy may increase compatibility on old and outdated HW, running old and outdated OS, but it's the opposite from being future proof.

There surely are many ways to transition depth buffer to texture. Afaik, there is legacy OGL, the ‘old’ modern OGL, and a ‘new’ modern OGL building on direct state access and bindless.
Each model likely has at least one option to do this task, and a bit of googling should reveal it.

I kind of implemented depth of field.

grat, you seem to work very fast. was it copyteximage, or fbo?

I cheated. I use glReadPixels to copy to a cpu buffer, then I do the DOF on the cpu, then I render it to a full screen quad.

p.s. It’s super slow :( lol

p.p.s. still looking into how to read the Color attachment buffer.

check the times, might only your dof algo is slow. ofc vram to ram to vram copy will be very slow as well… ensure you dont use float, and use threads. my cpu dof algo took quite a few to reach the area of unnoticeable performance degradation.