actually update texture atlas on change

[SimonDanisch]

No description provided.

[ffreyer]

closes #3733
fixes #3232

[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(fig)

	using	create	display	create	display
GLMakie	3.57s (3.47, 3.64) 0.06+-	181.34ms (169.71, 194.45) 7.93+-	499.07ms (471.60, 519.88) 16.45+-	8.07ms (7.59, 8.25) 0.26+-	26.45ms (26.20, 26.70) 0.22+-
master	3.46s (3.41, 3.57) 0.05+-	170.27ms (164.72, 182.41) 6.31+-	478.20ms (469.72, 495.95) 9.21+-	7.65ms (7.49, 7.89) 0.17+-	26.14ms (25.86, 26.46) 0.22+-
evaluation	0.97x slower X, 0.11s (1.86d, 0.00p, 0.06std)	0.94x slower❌, 11.08ms (1.55d, 0.01p, 7.12std)	0.96x slower X, 20.88ms (1.57d, 0.02p, 12.83std)	0.95x slower❌, 0.42ms (1.92d, 0.00p, 0.21std)	0.99x slower X, 0.31ms (1.41d, 0.02p, 0.22std)
CairoMakie	2.98s (2.92, 3.03) 0.05+-	165.93ms (161.55, 170.90) 3.82+-	100.95ms (97.93, 104.40) 2.30+-	7.31ms (7.17, 7.63) 0.16+-	874.88μs (866.39, 884.71) 6.87+-
master	3.02s (2.94, 3.11) 0.07+-	168.26ms (163.38, 174.46) 4.58+-	103.23ms (99.79, 107.71) 2.61+-	7.19ms (7.02, 7.42) 0.14+-	875.24μs (863.93, 901.10) 13.47+-
evaluation	1.01x invariant, -0.04s (-0.68d, 0.23p, 0.06std)	1.01x invariant, -2.34ms (-0.55d, 0.32p, 4.20std)	1.02x invariant, -2.28ms (-0.93d, 0.11p, 2.45std)	0.98x invariant, 0.12ms (0.82d, 0.15p, 0.15std)	1.00x invariant, -0.35μs (-0.03d, 0.95p, 10.17std)
WGLMakie	3.63s (3.57, 3.71) 0.05+-	375.12ms (362.19, 394.08) 11.10+-	8.63s (8.46, 8.81) 0.13+-	9.39ms (9.14, 9.64) 0.16+-	68.75ms (67.64, 70.04) 0.83+-
master	3.61s (3.56, 3.76) 0.07+-	258.21ms (168.68, 390.75) 111.16+-	8.95s (8.41, 9.31) 0.32+-	9.67ms (9.11, 10.46) 0.41+-	69.02ms (67.35, 72.37) 1.66+-
evaluation	0.99x invariant, 0.02s (0.36d, 0.51p, 0.06std)	0.69x slower❌, 116.91ms (1.48d, 0.03p, 61.13std)	1.04x faster ✓, -0.32s (-1.33d, 0.04p, 0.23std)	1.03x invariant, -0.28ms (-0.89d, 0.13p, 0.28std)	1.00x invariant, -0.27ms (-0.21d, 0.71p, 1.25std)

[ffreyer]

Refimg test I added: (WGLMakie)

Step	master	pr
1
2

[SimonDanisch]

The 0.69x slower❌ turned up two times now and seems a bit big to be noise, but, create should be a purely Makie.jl based benchmark, and we haven't change a single Makie related function and the Julia changes are minimal as well, so I think we should be able to ignore this