Heavy invalidations from == overloads

[ChrisRackauckas]

The following:

# From: https://timholy.github.io/SnoopCompile.jl/stable/snoopr/
using SnoopCompile
invalidations = @snoopr begin
    using DifferentialEquations

    function lorenz(du,u,p,t)
     du[1] = 10.0(u[2]-u[1])
     du[2] = u[1]*(28.0-u[3]) - u[2]
     du[3] = u[1]*u[2] - (8/3)*u[3]
    end
    u0 = [1.0;0.0;0.0]
    tspan = (0.0,100.0)
    prob = ODEProblem(lorenz,u0,tspan)
    alg = Rodas5()
    tinf = solve(prob,alg)
end;

trees = SnoopCompile.invalidation_trees(invalidations);

@show length(SnoopCompile.uinvalidated(invalidations)) # show total invalidations

show(trees[end]) # show the most invalidated method

# Count number of children (number of invalidations per invalidated method)
n_invalidations = map(trees) do methinvs
    SnoopCompile.countchildren(methinvs)
end

found ChainRulesCore as one of the main culprits for invalidations:

julia> trees[end-2]
inserting ==(a, b::ChainRulesCore.AbstractThunk) in ChainRulesCore at C:\Users\accou\.julia\packages\ChainRulesCore\sHMAp\src\tangent_types\thunks.jl:27 invalidated:      
   backedges: 1: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Base.UUID, ::Any) (4 children)
              2: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Tokenize.Tokens.Kind, ::Any) (6 children)
              3: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Module, ::Any) (8 children)
              4: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Method, ::Any) (24 children)
              5: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Core.TypeName, ::Any) (70 children)
              6: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Symbol, ::Any) (213 children)


julia> trees[end-1]
inserting ==(a::ChainRulesCore.AbstractThunk, b) in ChainRulesCore at C:\Users\accou\.julia\packages\ChainRulesCore\sHMAp\src\tangent_types\thunks.jl:26 invalidated:      
   backedges:  1: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::LineNumberNode) (1 children)
               2: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Module) (1 children)
               3: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Core.MethodInstance) (1 children)
               4: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Base.CoreLogging.LogLevel) (2 children)
               5: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Core.Compiler.InferenceResult) (2 children)
               6: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Task) (4 children)
               7: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Base.UUID) (5 children)
               8: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Tokenize.Tokens.Kind) (5 children)
               9: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::typeof(show)) (6 children)
              10: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::REPL.LineEdit.Prompt) (7 children)
              11: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Atom.OutlineItem) (7 children)
              12: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Nothing) (22 children)
              13: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::Symbol) (152 children)
              14: superseding ==(x, y) in Base at Base.jl:119 with MethodInstance for ==(::Any, ::CSTParser.EXPR) (238 children)

But those overloads are completely unnecessary anyways as you can always just do a isa AbstractThunk etc. stuff, and it was only added for testing:

#367

Seems like @johnnychen94 noticed this too that these "only for test" overloads are now some of the main compile time culprits in the language... not a good trade-off 😅

[ChrisRackauckas]

This was solved.