Invalidations and problems with stack traces

[ranocha]

I don't really know where to report this, so please move it somewhere else.

I had a brief look at load times using Julia v1.9.0-rc1:

julia> @time_imports using OrdinaryDiffEq
[...]
    416.1 ms  TruncatedStacktraces 93.53% compilation time (63% recompilation)
[...]
    237.7 ms  SciMLBase 4.98% compilation time (100% recompilation)
[...]

This made me suspicious. Thus, I wanted to check whether invalidations are the reason but ran into some problems:

julia> using Pkg; Pkg.activate(temp=true); Pkg.add("SciMLBase")
[...]
  [0bca4576] + SciMLBase v1.89.0
[...]
  [781d530d] + TruncatedStacktraces v1.1.0
[...]

julia> using SnoopCompileCore; invalidations = @snoopr(using SciMLBase); using SnoopCompile

julia> length(uinvalidated(invalidations))
2113

julia> trees = invalidation_trees(invalidations)
5-element Vector{SnoopCompile.MethodInvalidations}:
Error showing value of type Vector{SnoopCompile.MethodInvalidations}:
ERROR: UndefVarError: `isinplace` not defined

Some of the types have been truncated in the stacktrace for improved reading. To emit complete information
in the stack trace, evaluate `TruncatedStacktraces.VERBOSE[] = true` and re-run the code.

Stacktrace:
  [1] show(io::IOContext{IOBuffer}, t::
SYSTEM (REPL): showing an error caused an error
ERROR: UndefVarError: `isinplace` not defined

Some of the types have been truncated in the stacktrace for improved reading. To emit complete information
in the stack trace, evaluate `TruncatedStacktraces.VERBOSE[] = true` and re-run the code.

Stacktrace:
  [1] show(io::IOContext{IOBuffer}, t::
SYSTEM (REPL): caught exception of type UndefVarError while trying to handle a nested exception; giving up

So there are indeed quite a lot of invalidations. However, it's hard to see the full report since something fails. Verbose printing doesn't help:

julia> Pkg.add("TruncatedStacktraces"); using TruncatedStacktraces; TruncatedStacktraces.VERBOSE[] = true

julia> trees[5]
inserting Error showing value of type SnoopCompile.MethodInvalidations:
ERROR: UndefVarError: `uType` not defined
Stacktrace:
  [1] show(io::IOContext{IOBuffer}, t::
SYSTEM (REPL): showing an error caused an error
ERROR: UndefVarError: `uType` not defined
Stacktrace:
  [1] show(io::IOContext{IOBuffer}, t::
SYSTEM (REPL): caught exception of type UndefVarError while trying to handle a nested exception; giving up

I suspect TruncatedStacktraces.jl to play a role here, though:

julia> trees[5].method.name
:show

julia> trees[5].backedges
2-element Vector{SnoopCompile.InstanceNode}:
 MethodInstance for show(::IOContext{IOBuffer}, ::Type) at depth 0 with 107 children
 MethodInstance for show(::IOBuffer, ::Type) at depth 0 with 3064 children

The first four methods do not have many invalidations and can be displayed without problems:

julia> trees[1]
inserting unique(x::FillArrays.AbstractFill{T}) where T @ FillArrays ~/.julia/packages/FillArrays/xqile/src/FillArrays.jl:555 invalidated:
   backedges: 1: superseding unique(A::AbstractArray; dims) @ Base multidimensional.jl:1640 with MethodInstance for unique(::AbstractArray) (1 children)


julia> trees[2]
inserting Base.IteratorSize(::Type{<:FillArrays.AbstractFill{T, N, Axes}}) where {T, N, Axes} @ FillArrays ~/.julia/packages/FillArrays/xqile/src/FillArrays.jl:64 invalidated:
   backedges: 1: superseding Base.IteratorSize(::Type) @ Base generator.jl:94 with MethodInstance for Base.IteratorSize(::Type{<:AbstractString}) (5 children)
   87 mt_cache


julia> trees[3]
inserting print_matrix_row(io::IO, X::Union{FillArrays.AbstractFill{<:Any, 1}, FillArrays.AbstractFill{<:Any, 2}, FillArrays.RectDiagonal, LinearAlgebra.AbstractTriangular{<:Any, <:FillArrays.AbstractFill{<:Any, 2}}, LinearAlgebra.Diagonal{<:Any, <:FillArrays.AbstractFill{<:Any, 1}}}, A::Vector, i::Integer, cols::AbstractVector, sep::AbstractString, idxlast::Integer) @ FillArrays ~/.julia/packages/FillArrays/xqile/src/FillArrays.jl:688 invalidated:
   backedges: 1: superseding print_matrix_row(io::IO, X::AbstractVecOrMat, A::Vector, i::Integer, cols::AbstractVector, sep::AbstractString, idxlast::Integer) @ Base arrayshow.jl:97 with MethodInstance for Base.print_matrix_row(::IOContext{Base.TTY}, ::AbstractVecOrMat, ::Vector{Tuple{Int64, Int64}}, ::Int64, ::Vector{Int64}, ::String, ::Int64) (9 children)


julia> trees[4]
inserting broadcasted(::Base.Broadcast.DefaultArrayStyle{N}, op, r::FillArrays.AbstractFill{T, N}) where {T, N} @ FillArrays ~/.julia/packages/FillArrays/xqile/src/fillbroadcast.jl:78 invalidated:
   mt_backedges: 1: signature broadcasted(::S, f, args...) where S<:BroadcastStyle @ Base.Broadcast broadcast.jl:1319 (formerly broadcasted(::S, f, args...) where S<:BroadcastStyle @ Base.Broadcast broadcast.jl:1319) triggered MethodInstance for Base.Broadcast.broadcasted(::typeof(esc), ::Any) (20 children)
   66 mt_cache

[devmotion]

Could it be that these invalidations are caused by the definitions of show(::IO, ::Type) that are required when using TruncatedStacktraces? At least such definitions tend to cause invalidations in other places, e.g., FluxML/Zygote.jl#1356 PainterQubits/Unitful.jl#321 JuliaGeometry/GeometryBasics.jl#67 SciML/RuntimeGeneratedFunctions.jl#16

[ranocha]

Yes, I suspect so.

[KristofferC]

julia> struct Foo end

julia> Base.show(io::IO, t::Type{Foo}) = @invoke show(io::IO, t::Type)

julia> @time using Random
  0.402905 seconds (576.12 k allocations: 39.977 MiB, 1.94% gc time, 99.02% compilation time: 100% of which was recompilation)

[ChrisRackauckas]

Fixed by making stacktraces huge.