Postprocessing refactor

[ffreyer]

Here's a draft for MakieOrg/Makie.jl#604.

I moved most of the postprocessing stuff into a new file and wrapped it up in a PostProcessor object. Beyond ssao, fxaa and copying I also added an "empty" postprocessor, which doesn't do anything when executed. Depending on some global variables enable_SSAO and enable_FXAA either the empty or the ssao/fxaa postprocessor ends up in a list in Screen. The renderloop then executes postprocessors from that list, in order.

So if for example screen.postprocessors[1] is an empty postprocessor, the renderloop will do nothing instead of the SSAO stuff. This is the case when enable_SSAO[] = false. I also added a bit to the GLVisualizeShader so that the position and normal buffers are removed from fragment_ouput when enable_SSAO = false.

[SimonDanisch]

Nice :) I will need a bit of time to review this ;)

[ffreyer]

I might make some more changes this weekend. I noticed, for example, that render_frame still attaches all 4 draw buffers even if only two are active/exist.

[SimonDanisch]

Cool, with all the issue, it seems to be a good idea to only turn it on when actually used!

[SimonDanisch]

Let me know once this is ready :)

[ffreyer]

I think I'm done with this.

It would be cool to turn on/off ssao (and maybe fxaa) on the fly in the future, i.e. to cause a shader recompilation when the scene demands it. That would also be nice to control shading and I think it would be needed for depth peeling. But I didn't want to do this here.

[ffreyer]

AbstractPlotting tests pass locally for me, on current AbstractPlotting#master (JuliaPlots/AbstractPlotting.jl@95d3375)

[ffreyer]

]test GLMakie fails locally on idx = N - 1 though it passes here and if I run it manually in a terminal. So I don't think it's an issue.

Should be ready to merge

[pbouffard]

Just gave it a try, works great. Thank you!!

[SimonDanisch]

Thanks!