eigs fails when called on some matrix with which=:SM

[laurentbartholdi]

I have a largeish sparse, symmetric matrix on which eigs(mat,which=:SM) fails; but it succeeds on small submatrices.

TL;dr: it seems bunchkaufman! is called with a sparse matrix, but it can only accept strided matrices.

julia> mat = sparse([1, 2, 34, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 2, 33, 34], [1, 2, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 33, 34, 34, 34], Float32[12.0, 8.0, 1.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 8.0, 1.0, 1.0, 1.0, 4.0])
34×34 SparseMatrixCSC{Float32, Int64} with 38 stored entries:
⠑⢄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠐
⠀⠀⠑⢄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠑⢄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠑⢄⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠑⢄⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠑⢄⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠑⢄⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠑⢄⠀
⠐⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠛

julia> eigs(mat) # this is fine
(Float32[12.000001, 8.449489, 8.000001, 8.0, 8.0, 8.0], Float32[1.0000001 -7.030933f-8 … 0.0 0.0; 5.9604645f-8 0.6738876 … 0.027224585 -0.00036482513; … ; 0.0 0.67388695 … -0.0272246 0.0003648214; 1.8559231f-7 0.30290544 … 0.0 0.0], 6, 1, 20, Float32[-1.2639064f-15, 6.4955343f-9, 2.1896764f-9, -2.1639128f-8, -3.943517f-9, -2.0025592f-8, -4.8012203f-9, -3.2126188f-8, 1.7540275f-8, 8.445834f-9  …  7.8725595f-9, 1.7422437f-8, 3.789941f-8, -2.3730575f-8, 5.321133f-8, 1.9782162f-8, 1.8557815f-8, -1.6891136f-8, -6.495541f-9, -1.921318f-15])

julia> eigs(mat,which=:SM) # ka-boom!
ERROR: MethodError: no method matching bunchkaufman!(::Hermitian{Float32, SparseMatrixCSC{Float32, Int64}}, ::Bool; check=true)
Closest candidates are:
  bunchkaufman!(::Hermitian{T, S} where {T<:Union{ComplexF32, ComplexF64}, S<:StridedMatrix{T}}, ::Bool; check) at /usr/local/src/julia/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/bunchkaufman.jl:102
  bunchkaufman!(::Union{Hermitian{T, S}, Symmetric{T, S}, Symmetric{Complex{T}, S}} where {T<:Union{Float32, Float64}, S<:(StridedMatrix{T} where T)}, ::Bool; check) at /usr/local/src/julia/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/bunchkaufman.jl:96
  bunchkaufman!(::StridedMatrix{var"#s815"} where var"#s815"<:Union{Float32, Float64, ComplexF32, ComplexF64}, ::Bool; check) at /usr/local/src/julia/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/bunchkaufman.jl:108
Stacktrace:
 [1] bunchkaufman(A::Hermitian{Float32, SparseMatrixCSC{Float32, Int64}}, rook::Bool; check::Bool)
   @ LinearAlgebra /usr/local/src/julia/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/bunchkaufman.jl:196
 [2] _factorize(A::Hermitian{Float32, SparseMatrixCSC{Float32, Int64}}; check::Bool)
   @ LinearAlgebra /usr/local/src/julia/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/symmetric.jl:645
 [3] _factorize(A::Hermitian{Float32, SparseMatrixCSC{Float32, Int64}})
   @ LinearAlgebra /usr/local/src/julia/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/symmetric.jl:643
 [4] factorize(A::Hermitian{Float32, SparseMatrixCSC{Float32, Int64}})
   @ LinearAlgebra /usr/local/src/julia/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/symmetric.jl:641
 [5] factorize(A::SparseMatrixCSC{Float32, Int64})
   @ SparseArrays /usr/local/src/julia/usr/share/julia/stdlib/v1.7/SparseArrays/src/linalg.jl:1612
 [6] _eigs(A::SparseMatrixCSC{Float32, Int64}, B::UniformScaling{Bool}; nev::Int64, ncv::Int64, which::Symbol, tol::Float64, maxiter::Int64, sigma::Nothing, v0::Vector{Float32}, ritzvec::Bool, explicittransform::Symbol, check::Int64)
   @ Arpack ~/.julia/packages/Arpack/S16ni/src/Arpack.jl:211
 [7] #eigs#10
   @ ~/.julia/packages/Arpack/S16ni/src/Arpack.jl:51 [inlined]
 [8] #eigs#9
   @ ~/.julia/packages/Arpack/S16ni/src/Arpack.jl:50 [inlined]
 [9] top-level scope
   @ REPL[153]:1

julia> eigs(mat[1:33,1:33],which=:SM) # on that top block, all is fine too
(Float32[7.999999, 8.0, 8.0, 8.0, 8.0, 8.0], Float32[-1.4790526f-8 1.7125398f-17 … -1.5129569f-9 2.8769254f-10; 0.29630464 0.0011543621 … 0.018754192 0.09862298; … ; 0.16326842 0.32161772 … 0.04730363 0.013291076; -0.06902978 0.38974732 … -0.41718462 -0.14601785], 6, 1, 20, Float32[-2.1406748f-18, -2.7734992f-10, 1.06227f-10, 3.3392237f-11, -6.1807434f-11, 5.5903016f-10, 6.26343f-11, 3.1210298f-10, -1.538405f-10, -3.69939f-10  …  -1.0335148f-10, 3.7906647f-11, 1.1831452f-10, 7.2083906f-10, -9.563203f-11, -9.36515f-11, -6.767832f-10, 8.40918f-10, 5.816487f-10, 3.533424f-10])

[PetrKryslUCSD]

This is probably related to #130