Julia: using SIMD hangs on master

[KristofferC]

using SIMD (https://github.com/eschnett/SIMD.jl) hangs Julia (runs at 100% CPU). This is a regression compared to v0.6-rc2 where it takes a second to load.

[fredrikekre]

git bisect suggests it's 8bacb65bbab9cbb7da2a266dc384cdd62df372aa

[KristofferC]

cc @vtjnash

[vtjnash]

I suppose this is the ultimate sort of punishment for defining a method ambiguity.

Tuple{typeof(Base.:(!=)), SIMD.Vec{N, T}, Union{Bool, Int8, Int16, Int32, Int64, Int128, UInt8, UInt16, UInt32, UInt64, UInt128, Float16, Float32, Float64}} where T<:Union{Float16, Float32, Float64} where N
Tuple{typeof(Base.:(!=)), SIMD.Vec{N, T}, Union{Bool, Int8, Int16, Int32, Int64, Int128, UInt8, UInt16, UInt32, UInt64, UInt128}} where T<:Union{Bool, Int8, Int16, Int32, Int64, Int128, UInt8, UInt16, UInt32, UInt64, UInt128} where N

(the author probably intended to write Union{(SIMD.Vec{N, T} for T in outer_union)...}, but instead has an ambiguity when using objects of type SIMD.Vec{0, Union{}})