CairoMakie allocations

[jkrumbiegel]

Noticed that the speed of CairoMakie svg generation was sometimes really really slow. Turns out this was mostly because of the slow sha512 hashing, I replaced that with crc32 (it doesn't have to be cryptographically secure, just random enough to be practical). And I improved the id replacement and added some manual union splitting that also seemed to allocate quite a bit when drawing scatters.

Net result is drastic, but of course usually you wouldn't want the svg format anyway if you have many data points:

## before

julia> @time begin
           f = Figure()
           ax = Axis(f[1, 1])
           scatter!(ax, 1:10000)
           save("test2.svg", f)
       end;
 12.317504 seconds (444.40 k allocations: 125.444 MiB, 0.60% gc time)

## after

julia> @time begin
           f = Figure()
           ax = Axis(f[1, 1])
           scatter!(ax, 1:10000)
           save("test2.svg", f)
       end;
  0.100446 seconds (59.69 k allocations: 15.657 MiB)

[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	3.45s (3.41, 3.55) 0.05+-	337.15ms (305.41, 413.44) 45.61+-	479.84ms (468.00, 518.56) 18.42+-	7.56ms (7.41, 7.71) 0.11+-	25.97ms (25.55, 26.36) 0.34+-
master	3.48s (3.40, 3.59) 0.08+-	399.71ms (387.43, 413.99) 8.51+-	481.47ms (464.08, 494.20) 11.27+-	7.58ms (7.44, 7.69) 0.09+-	25.72ms (25.37, 26.46) 0.49+-
evaluation	1.01x invariant, -0.03s (-0.38d, 0.49p, 0.07std)	1.19x faster✅, -62.56ms (-1.91d, 0.01p, 27.06std)	1.00x invariant, -1.62ms (-0.11d, 0.85p, 14.85std)	1.00x invariant, -0.02ms (-0.22d, 0.69p, 0.10std)	0.99x invariant, 0.25ms (0.58d, 0.30p, 0.42std)
CairoMakie	3.03s (3.01, 3.08) 0.02+-	324.33ms (322.16, 329.41) 2.37+-	144.62ms (143.17, 146.56) 1.14+-	7.69ms (7.62, 7.85) 0.08+-	611.20μs (606.15, 614.27) 3.23+-
master	3.03s (3.01, 3.08) 0.03+-	329.89ms (323.76, 339.45) 5.16+-	145.30ms (143.12, 149.23) 2.67+-	7.99ms (7.88, 8.23) 0.11+-	604.21μs (598.89, 610.39) 3.98+-
evaluation	1.00x invariant, -0.0s (-0.00d, 1.00p, 0.02std)	1.02x faster ✓, -5.56ms (-1.39d, 0.03p, 3.76std)	1.00x invariant, -0.67ms (-0.33d, 0.56p, 1.91std)	1.04x faster ✓, -0.3ms (-2.97d, 0.00p, 0.10std)	0.99x slower X, 6.98μs (1.93d, 0.00p, 3.61std)
WGLMakie	3.75s (3.73, 3.78) 0.02+-	331.34ms (321.11, 346.18) 7.75+-	8.82s (8.74, 8.89) 0.06+-	9.84ms (9.38, 11.87) 0.90+-	75.43ms (67.60, 81.58) 5.50+-
master	3.71s (3.66, 3.78) 0.04+-	327.58ms (318.29, 334.62) 6.33+-	8.89s (8.56, 9.16) 0.20+-	9.52ms (9.40, 9.69) 0.11+-	73.78ms (68.31, 79.78) 5.11+-
evaluation	0.99x invariant, 0.04s (1.09d, 0.08p, 0.03std)	0.99x invariant, 3.76ms (0.53d, 0.34p, 7.04std)	1.01x invariant, -0.07s (-0.45d, 0.43p, 0.13std)	0.97x invariant, 0.32ms (0.49d, 0.39p, 0.51std)	0.98x invariant, 1.65ms (0.31d, 0.57p, 5.31std)