Currently, ColorTypes.jl is disabling its precompilation (JuliaGraphics/ColorTypes.jl#269).
Also, the precompilation list of Colors.jl has not been verified for quite some time.
I have the impression that minimal precompilation tends to work well, but it would be desirable to obtain data-driven confirmation of how true that is and to what extent it varies depending on the version of Julia.
In addition, given the need to improve GPU suitability, it may be necessary to revisit the implementation in some cases.
Currently, the Julia ecosystem has the infrastructure in place to systematically benchmark TTFX.
https://discourse.julialang.org/t/a-visual-log-of-loading-and-precompilation-times-for-many-packages-over-a-variety-of-julia-versions/133887
Since Colors.jl is a "relatively upstream" package, I'm not sure if it's appropriate to register the benchmark in Julia-TTFX-Snippets, but I think it's good to have some kind of measure.
IMO, the benchmarks that include colordiff might perform well.
They involve String/Symbol (e.g. colorant"red") → RGB{N0f8} → XYZ{Float32} → Lab{Float32} conversions.
xref: JuliaGraphics/ColorTypes.jl#266
Currently, ColorTypes.jl is disabling its precompilation (JuliaGraphics/ColorTypes.jl#269).
Also, the precompilation list of Colors.jl has not been verified for quite some time.
I have the impression that minimal precompilation tends to work well, but it would be desirable to obtain data-driven confirmation of how true that is and to what extent it varies depending on the version of Julia.
In addition, given the need to improve GPU suitability, it may be necessary to revisit the implementation in some cases.
Currently, the Julia ecosystem has the infrastructure in place to systematically benchmark TTFX.
https://discourse.julialang.org/t/a-visual-log-of-loading-and-precompilation-times-for-many-packages-over-a-variety-of-julia-versions/133887
Since Colors.jl is a "relatively upstream" package, I'm not sure if it's appropriate to register the benchmark in Julia-TTFX-Snippets, but I think it's good to have some kind of measure.
IMO, the benchmarks that include
colordiffmight perform well.They involve
String/Symbol(e.g.colorant"red") →RGB{N0f8}→XYZ{Float32}→Lab{Float32}conversions.xref: JuliaGraphics/ColorTypes.jl#266