Remove stencil buffer

[ffreyer]

Description

I wanted to give fixing the scene id problem (#665) an attempt and tried removing all the stencil code bits to see what actually breaks... which seems to be nothing? Locally all tests still passed. I also tested all combinations of ssao, fxaa and transparancy with

fig = Figure()
ax = Axis(fig[1, 1])
scatter!(ax, rand(10000), color = rand(RGBf, 10000))
top = campixel(ax.blockscene)
text!(
    top, Point2f(400), text = "Testing", strokewidth=2f0, strokecolor = :white, 
    ssao = true, fxaa = !true, transparency = true
)

and get the same ordering as before.

From my understanding the stencil code restricts rendering to scene areas which is also what glViewport and glScissor do... So I guess it's redundant? Translating Axis and LScene content is still correctly cut off at least.

Fixes #665. (Well not yet but we can remove/freely type screenid if we remove stencil tests)

[MakieBot]

Compile Times benchmark

Note, that these numbers may fluctuate on the CI servers, so take them with a grain of salt. All benchmark results are based on the mean time and negative percent mean faster than the base branch. Note, that GLMakie + WGLMakie run on an emulated GPU, so the runtime benchmark is much slower. Results are from running:

using_time = @ctime using Backend
# Compile time
create_time = @ctime fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @ctime Makie.colorbuffer(display(fig))
# Runtime
create_time = @benchmark fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @benchmark Makie.colorbuffer(display(fig))

	using	create	display	create	display
GLMakie	39.19s (38.89, 39.33) 0.17+-	25.44s (25.05, 25.71) 0.23+-	22.98s (22.66, 23.25) 0.20+-	25.68ms (24.96, 26.80) 0.59+-	62.97ms (62.00, 63.58) 0.65+-
master	38.98s (38.81, 39.33) 0.19+-	25.47s (25.21, 25.88) 0.27+-	23.03s (22.62, 23.76) 0.42+-	25.72ms (24.84, 26.24) 0.54+-	63.40ms (62.39, 65.24) 1.01+-
evaluation	+0.55%, 0.22s slower X (1.18d, 0.05p, 0.18std)	-0.10%, -0.03s invariant (-0.11d, 0.85p, 0.25std)	-0.22%, -0.05s invariant (-0.15d, 0.78p, 0.31std)	-0.17%, -0.04ms invariant (-0.08d, 0.88p, 0.57std)	-0.68%, -0.43ms invariant (-0.51d, 0.36p, 0.83std)
CairoMakie	33.62s (32.37, 35.45) 0.96+-	23.72s (22.96, 24.88) 0.71+-	3.32s (3.20, 3.42) 0.07+-	24.76ms (24.31, 25.85) 0.53+-	28.80ms (27.90, 30.61) 0.86+-
master	33.44s (32.60, 34.89) 0.77+-	23.69s (22.97, 24.49) 0.59+-	3.49s (3.27, 3.74) 0.18+-	25.23ms (24.57, 26.61) 0.69+-	29.03ms (28.04, 29.86) 0.74+-
evaluation	+0.53%, 0.18s invariant (0.21d, 0.71p, 0.86std)	+0.13%, 0.03s invariant (0.05d, 0.93p, 0.65std)	-5.32%, -0.18s faster✅ (-1.29d, 0.04p, 0.13std)	-1.89%, -0.47ms invariant (-0.76d, 0.18p, 0.61std)	-0.82%, -0.24ms invariant (-0.29d, 0.59p, 0.80std)
WGLMakie	34.73s (34.25, 35.18) 0.29+-	29.72s (29.40, 30.40) 0.35+-	44.63s (43.28, 49.46) 2.18+-	24.59ms (23.46, 25.52) 0.67+-	121.92ms (117.44, 135.21) 6.30+-
master	34.66s (34.37, 35.08) 0.24+-	29.33s (28.93, 29.55) 0.21+-	44.62s (43.65, 47.46) 1.42+-	23.44ms (23.10, 24.04) 0.30+-	81.60ms (71.85, 91.57) 7.29+-
evaluation	+0.20%, 0.07s invariant (0.25d, 0.65p, 0.27std)	+1.30%, 0.39s slower X (1.34d, 0.03p, 0.28std)	+0.01%, 0.01s invariant (0.00d, 1.00p, 1.80std)	+4.67%, 1.15ms slower X (2.22d, 0.00p, 0.48std)	+33.07%, 40.31ms slower❌ (5.92d, 0.00p, 6.80std)

[SimonDanisch]

I remember, I had to resort to use the stencil buffer for a pretty obscure reason... There is a chance, that the reason went away, but there's also a chance, that it's only needed for a very annoying corner case.
Would be really cool to just get rid of it for good :D

[ffreyer]

Got introduced here: JuliaPlots/GLMakie.jl#35

I'll check if I find any issues with picking. Overlapping scenes seem fine so far...

[SimonDanisch]

Hm, I also tried the most common suspects and they do seem to work fine, as far as I can tell...
Jeez, really a reminder to comment odd decisions much better -.-

[ffreyer]

This also looks the same as before and DataInspector works fine.

fig = Figure()
for i in (1:2, 2:3), j in (1:2, 2:3)
    ax = Axis(fig[i, j])
    scatter!(ax, rand(Point2f, 1000), color = rand(RGBf, 1000))
    s1 = campixel(ax.scene)
    mesh!(s1, Rect2f(100, 100, 200, 200), color = RGBAf(0.3, 0.1, 0.5, 0.3), transparency = true)
    translate!(s1, 0, 0, 100)
    s2 = campixel(ax.scene)
    mesh!(s2, Rect2f(200, 200, 200, 200), color = RGBAf(0.7, 0.3, 0.1, 0.3), transparency = !true)
    xlims!(ax, 0, 1)
    ylims!(ax, 0, 1)
end
DataInspector(fig)
fig

Didn't see any problems rendering 192 Menus generating 385 scenes either. None with Toggles or sliders either (though those don't hit > 256 scenes)

[jkrumbiegel]

Love that I kept repeating "it's not trivial to remove the 256 limit" and now this 😂