Improve CairoMakie's 2D mesh performance

[asinghvi17]

Description

As seen in #4112, CairoMakie's 2D drawing routines are pretty slow. This PR attempts to speed them up a bit (currently has a ~30% improvement locally). Not done with the optimization or profiling yet though, and I have a sense that a lot of optimization is still doable.

Each change is usually a single commit so the commit log should show what's been done.

Type of change

• Bug fix (non-breaking change which fixes an issue)

[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	6.28s (6.14, 6.56) 0.12+-	162.02ms (157.57, 172.86) 4.43+-	470.59ms (462.70, 488.57) 6.82+-	6.78ms (6.69, 6.94) 0.07+-	25.74ms (25.42, 26.46) 0.29+-
master	6.30s (6.13, 6.69) 0.15+-	163.08ms (156.74, 177.50) 6.06+-	474.13ms (460.91, 504.48) 11.75+-	6.83ms (6.71, 7.28) 0.16+-	25.79ms (25.41, 26.56) 0.37+-
evaluation	1.00x invariant, -0.02s (-0.14d, 0.66p, 0.13std)	1.01x invariant, -1.05ms (-0.20d, 0.53p, 5.24std)	1.01x invariant, -3.54ms (-0.37d, 0.25p, 9.29std)	1.01x invariant, -0.05ms (-0.42d, 0.19p, 0.12std)	1.00x invariant, -0.05ms (-0.16d, 0.62p, 0.33std)
CairoMakie	5.92s (5.85, 6.01) 0.05+-	162.98ms (160.59, 169.14) 2.54+-	199.14ms (195.17, 206.62) 2.80+-	6.84ms (6.79, 6.95) 0.04+-	1.06ms (1.05, 1.07) 0.01+-
master	5.93s (5.85, 6.02) 0.05+-	163.95ms (159.78, 175.39) 3.65+-	199.26ms (196.00, 203.68) 1.88+-	6.78ms (6.68, 6.85) 0.05+-	1.06ms (1.05, 1.07) 0.01+-
evaluation	1.00x invariant, -0.01s (-0.17d, 0.60p, 0.05std)	1.01x invariant, -0.97ms (-0.31d, 0.34p, 3.09std)	1.00x invariant, -0.12ms (-0.05d, 0.88p, 2.34std)	0.99x slower X, 0.06ms (1.26d, 0.00p, 0.05std)	1.00x invariant, 0.0ms (0.09d, 0.79p, 0.01std)
WGLMakie	6.59s (6.47, 6.71) 0.05+-	164.97ms (160.05, 173.04) 3.70+-	5.61s (5.47, 5.93) 0.11+-	10.41ms (10.10, 10.80) 0.20+-	147.01ms (142.98, 156.44) 2.76+-
master	6.59s (6.49, 6.69) 0.05+-	166.61ms (160.57, 172.63) 3.87+-	5.87s (5.80, 5.96) 0.04+-	10.43ms (10.20, 10.89) 0.16+-	146.41ms (143.10, 148.94) 1.58+-
evaluation	1.00x invariant, 0.0s (0.02d, 0.96p, 0.05std)	1.01x invariant, -1.64ms (-0.43d, 0.18p, 3.78std)	1.05x faster ✓, -0.26s (-3.16d, 0.00p, 0.07std)	1.00x invariant, -0.02ms (-0.12d, 0.72p, 0.18std)	1.00x invariant, 0.61ms (0.27d, 0.40p, 2.17std)

[ffreyer]

Should we merge this since nothing more has happened here? 30% is already quite significant (and the GeometryBasics refactor may change some details anyway)

[asinghvi17]

Agreed, may as well merge. Will wait for the GeometryBasics refactor before pushing more on this! (And in the worst case, we can just switch all meshes to the 3d pipeline if it's significantly faster)

[asinghvi17]

WGLMakie failures are because the texture atlas is slightly larger than expected:

session_size = 6.095909
texture_atlas_size = 6.095909
memory leaks: Test Failed at /home/runner/work/Makie.jl/Makie.jl/WGLMakie/test/runtests.jl:57
  Expression: texture_atlas_size < 6
   Evaluated: 6.095909 < 6
Stacktrace:
 [1] macro expansion
   @ ~/work/Makie.jl/Makie.jl/WGLMakie/test/runtests.jl:57 [inlined]
 [2] macro expansion
   @ /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.6/Test/src/Test.jl:1151 [inlined]
 [3] top-level scope
   @ ~/work/Makie.jl/Makie.jl/WGLMakie/test/runtests.jl:43
Test Summary: | Pass  Fail  Total
memory leaks  |    3     1      4
ERROR: LoadError: Some tests did not pass: 3 passed, 1 failed, 0 errored, 0 broken.
in expression starting at /home/runner/work/Makie.jl/Makie.jl/WGLMakie/test/runtests.jl:42
ERROR: Package WGLMakie errored during testing
Stacktrace: