Performance regression in preconditioner creation from v1.0.0 to v1.1.0

[ttsHub]

Hi,

There is a significant difference between versions 0.6.0 and 1.1.0 of AlgebraicMultigrid.jl for the following sparse matrix to create the preconditioner

using SparseArrays, LinearAlgebra, AlgebraicMultigrid

(Tv, n, p) = (Float64, 4000000, 1.0e-8)
A = sprand(Tv, n, n, p)
A1 = 0.5 * (A + A') + LinearAlgebra.I

Observed results

v1.0.0

julia> @time AlgebraicMultigrid.aspreconditioner(AlgebraicMultigrid.smoothed_aggregation(A1))
  1.355003 seconds (456 allocations: 717.090 MiB, 32.77% gc time)
AlgebraicMultigrid.Preconditioner{AlgebraicMultigrid.MultiLevel{AlgebraicMultigrid.Pinv{Float64}, GaussSeidel{SymmetricSweep}, GaussSeidel{SymmetricSweep}, SparseMatrixCSC{Float64, Int64}, SparseMatrixCSC{Float64, Int64}, Adjoint{Float64, SparseMatrixCSC{Float64, Int64}}, AlgebraicMultigrid.MultiLevelWorkspace{Vector{Float64}, 1}}, AlgebraicMultigrid.V}(Multilevel Solver
-----------------
Operator Complexity: 1.035
Grid Complexity: 1.037
No. of Levels: 4
Coarse Solver: Pinv
Level     Unknowns     NonZeros
-----     --------     --------
    1      4000000      4319678 [96.66%]
    2       148084       149020 [ 3.33%]
    3          462          462 [ 0.01%]
    4            0            0 [ 0.00%]
, :zero, AlgebraicMultigrid.V())

v1.1.0

julia> @time AlgebraicMultigrid.aspreconditioner(AlgebraicMultigrid.smoothed_aggregation(A1))
 45.310739 seconds (2.67 M allocations: 838.746 MiB, 1.28% gc time)
AlgebraicMultigrid.Preconditioner{AlgebraicMultigrid.MultiLevel{AlgebraicMultigrid.Pinv{Float64}, GaussSeidel{SymmetricSweep}, GaussSeidel{SymmetricSweep}, SparseMatrixCSC{Float64, Int64}, SparseMatrixCSC{Float64, Int64}, Adjoint{Float64, SparseMatrixCSC{Float64, Int64}}, AlgebraicMultigrid.MultiLevelWorkspace{Vector{Float64}, 1}}, AlgebraicMultigrid.V}(Multilevel Solver
-----------------
Operator Complexity: 1.034
Grid Complexity: 1.037
No. of Levels: 4
Coarse Solver: Pinv
Level     Unknowns     NonZeros
-----     --------     --------
    1      4000000      4319362 [96.67%]
    2       147639       148523 [ 3.32%]
    3          440          440 [ 0.01%]
    4            0            0 [ 0.00%]
, :zero, AlgebraicMultigrid.V())

Version Info

julia> versioninfo()
Julia Version 1.11.2
Commit 5e9a32e7af (2024-12-01 20:02 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Windows (x86_64-w64-mingw32)

Any help on this is much appreciated. Thank you.