Sparse matrices

In IncompressibleNavierStokes, all operators are implemented as matrix-free kernels. However, access to the underlying matrices can sometimes be useful, for example to precompute matrix factorizations. We therefore provide sparse matrix versions of some of the linear operators (see full list below).

Example

Consider a simple setup

using IncompressibleNavierStokes
ax = range(0, 1, 17);
setup = Setup(; x = (ax, ax), Re = 1e3);

(grid = (xlims = ((0.0, 1.0), (0.0, 1.0)), dimension = IncompressibleNavierStokes.Dimension{2}(), N = (18, 18), Nu = ((16, 16), (16, 16)), Np = (16, 16), Iu = (CartesianIndices((2:17, 2:17)), CartesianIndices((2:17, 2:17))), Ip = CartesianIndices((2:17, 2:17)), x = ([-0.0625, 0.0, 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.5625, 0.625, 0.6875, 0.75, 0.8125, 0.875, 0.9375, 1.0, 1.0625], [-0.0625, 0.0, 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.5625, 0.625, 0.6875, 0.75, 0.8125, 0.875, 0.9375, 1.0, 1.0625]), xu = (([0.0, 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.5625, 0.625, 0.6875, 0.75, 0.8125, 0.875, 0.9375, 1.0, 1.0625], [-0.03125, 0.03125, 0.09375, 0.15625, 0.21875, 0.28125, 0.34375, 0.40625, 0.46875, 0.53125, 0.59375, 0.65625, 0.71875, 0.78125, 0.84375, 0.90625, 0.96875, 1.03125]), ([-0.03125, 0.03125, 0.09375, 0.15625, 0.21875, 0.28125, 0.34375, 0.40625, 0.46875, 0.53125, 0.59375, 0.65625, 0.71875, 0.78125, 0.84375, 0.90625, 0.96875, 1.03125], [0.0, 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.5625, 0.625, 0.6875, 0.75, 0.8125, 0.875, 0.9375, 1.0, 1.0625])), xp = ([-0.03125, 0.03125, 0.09375, 0.15625, 0.21875, 0.28125, 0.34375, 0.40625, 0.46875, 0.53125, 0.59375, 0.65625, 0.71875, 0.78125, 0.84375, 0.90625, 0.96875, 1.03125], [-0.03125, 0.03125, 0.09375, 0.15625, 0.21875, 0.28125, 0.34375, 0.40625, 0.46875, 0.53125, 0.59375, 0.65625, 0.71875, 0.78125, 0.84375, 0.90625, 0.96875, 1.03125]), Δ = ([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625], [0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625]), Δu = ([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.03125], [0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.03125]), A = ((([1.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5], [0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 1.0]), ([1.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5], [0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 1.0])), (([1.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5], [0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 1.0]), ([1.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5], [0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 1.0])))), boundary_conditions = ((PeriodicBC(), PeriodicBC()), (PeriodicBC(), PeriodicBC())), Re = 1000.0, bodyforce = nothing, issteadybodyforce = false, closure_model = nothing, backend = CPU(false), workgroupsize = 64, temperature = nothing)

The matrices for the linear operators are named by appending _mat to the function name, for example: divergence, pressuregradient, and diffusion become divergence_mat, pressuregradient_mat, diffusion_mat etc.

Let's assemble some matrices:

divergence_mat(setup)

324×648 SparseArrays.SparseMatrixCSC{Float64, Int64} with 1024 stored entries:
⎡⠀⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣄⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎤
⎢⠀⠈⠹⢄⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣷⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠙⢦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣷⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠙⢦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣷⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠙⢤⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣷⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠙⢦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣗⣄⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠙⢦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣷⣄⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠙⢦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣷⣄⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠙⢦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣕⣄⠀⠀⎥
⎣⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠙⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠃⠑⠀⎦

pressuregradient_mat(setup)

648×324 SparseArrays.SparseMatrixCSC{Float64, Int64} with 1024 stored entries:
⎡⠀⢤⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎤
⎢⠀⠀⠓⢆⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠈⠳⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠈⠳⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠈⠓⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠳⣄⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠳⣄⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠳⣄⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠳⣄⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠓⠀⎥
⎢⠀⢤⣤⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠙⢿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠙⢿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠙⢿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠙⢿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠙⢽⣦⡀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠙⢿⣦⡀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠙⢿⣦⡀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠑⢽⢦⡀⎥
⎣⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠑⠙⎦

diffusion_mat(setup)

648×648 SparseArrays.SparseMatrixCSC{Float64, Int64} with 2560 stored entries:
⎡⣄⣤⣠⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎤
⎢⠈⠻⢷⢗⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠈⠻⣿⣿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠈⠻⣿⣿⢦⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠈⠛⣟⣽⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣵⣿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠛⠛⠙⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣄⣤⣤⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⢟⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣿⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣟⣽⣤⡀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠳⣿⣿⣦⡀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣿⣦⡀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⢵⢷⣦⡀⎥
⎣⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠋⠛⠙⎦

Note the sparsity pattern with matrix concatenation of two scalar operators for operators acting on or producing vector fields. The pressuregradient_mat converts a scalar field to a vector field, and is thus the vertical concatenation of the matrices for $\partial /\partial x$ and $\partial /\partial y$, while the divergence_mat is a horizontal concatenation of two similar matrices. The periodic boundary conditions are not included in the operators above, and are implemented via their own matrix. The periodic extension is visible:

bc_u_mat(setup)

648×648 SparseArrays.SparseMatrixCSC{Float64, Int64} with 648 stored entries:
⎡⠀⢄⠄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠨⠻⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎤
⎢⠀⠁⠱⢅⢀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠐⢑⢔⢀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠐⢑⢔⢀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠓⢄⠄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠁⠕⢅⠄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠁⠕⢅⠄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠁⢑⢔⢀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠐⢑⢔⢀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⣦⡂⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠐⠑⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢄⠄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠨⠻⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠁⠕⢅⠄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠁⠕⢅⢀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠐⢑⢔⢀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠐⢑⢔⢀⠀⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠐⠑⢤⠀⠀⠀⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠁⠕⢅⠄⠀⠀⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠁⠕⢅⠄⠀⠀⠀⎥
⎢⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠁⢑⢆⢀⠀⎥
⎣⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣦⡂⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠐⠑⠀⎦

We can verify that the diffusion matrix gives the same results as the diffusion kernel (without viscosity):

using Random
u = randn!(vectorfield(setup))
B = bc_u_mat(setup)
D = diffusion_mat(setup)
d_kernel = diffusion(apply_bc_u(u, 0.0, setup), setup; use_viscosity = false)
d_matrix = reshape(D * B * u[:], size(u))
maximum(abs, d_matrix - d_kernel)

4.547473508864641e-13

Matrices only work on flattened fields u[:], while the kernels work on $(D+1)$-array-shaped fields for a dimension $D\in \{2,3\}$.

Boundary conditions and matrices

Matrices can only be used to represent boundary conditions that depend linearly on the input, such as periodic or Neumann boundary conditions. Non-zero Dirichlet boundary conditions need to be accounted for separately. Consider the following inflow-setup:

setup = Setup(;
    x = (ax, ax),
    boundary_conditions = (
        (DirichletBC((10.0, 0.0)), PressureBC()),
        (DirichletBC(), DirichletBC()),
    ),
)

(grid = (xlims = ((0.0, 1.0), (0.0, 1.0)), dimension = IncompressibleNavierStokes.Dimension{2}(), N = (18, 18), Nu = ((16, 16), (16, 15)), Np = (16, 16), Iu = (CartesianIndices((2:17, 2:17)), CartesianIndices((2:17, 2:16))), Ip = CartesianIndices((2:17, 2:17)), x = ([0.0, 0.0, 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.5625, 0.625, 0.6875, 0.75, 0.8125, 0.875, 0.9375, 1.0, 1.0], [0.0, 0.0, 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.5625, 0.625, 0.6875, 0.75, 0.8125, 0.875, 0.9375, 1.0, 1.0]), xu = (([0.0, 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.5625, 0.625, 0.6875, 0.75, 0.8125, 0.875, 0.9375, 1.0, 1.0], [0.0, 0.03125, 0.09375, 0.15625, 0.21875, 0.28125, 0.34375, 0.40625, 0.46875, 0.53125, 0.59375, 0.65625, 0.71875, 0.78125, 0.84375, 0.90625, 0.96875, 1.0]), ([0.0, 0.03125, 0.09375, 0.15625, 0.21875, 0.28125, 0.34375, 0.40625, 0.46875, 0.53125, 0.59375, 0.65625, 0.71875, 0.78125, 0.84375, 0.90625, 0.96875, 1.0], [0.0, 0.0625, 0.125, 0.1875, 0.25, 0.3125, 0.375, 0.4375, 0.5, 0.5625, 0.625, 0.6875, 0.75, 0.8125, 0.875, 0.9375, 1.0, 1.0])), xp = ([0.0, 0.03125, 0.09375, 0.15625, 0.21875, 0.28125, 0.34375, 0.40625, 0.46875, 0.53125, 0.59375, 0.65625, 0.71875, 0.78125, 0.84375, 0.90625, 0.96875, 1.0], [0.0, 0.03125, 0.09375, 0.15625, 0.21875, 0.28125, 0.34375, 0.40625, 0.46875, 0.53125, 0.59375, 0.65625, 0.71875, 0.78125, 0.84375, 0.90625, 0.96875, 1.0]), Δ = ([2.220446049250313e-16, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 2.220446049250313e-16], [2.220446049250313e-16, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 2.220446049250313e-16]), Δu = ([0.03125, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.03125, 2.220446049250313e-16], [0.03125, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.03125, 2.220446049250313e-16]), A = ((([1.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5], [0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 1.0]), ([1.0, 1.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.0], [0.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 1.0, 1.0])), (([1.0, 1.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.0], [0.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 1.0, 1.0]), ([1.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5], [0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 1.0])))), boundary_conditions = ((DirichletBC{Tuple{Float64, Float64}}((10.0, 0.0)), PressureBC()), (DirichletBC{Nothing}(nothing), DirichletBC{Nothing}(nothing))), Re = 1000.0, bodyforce = nothing, issteadybodyforce = false, closure_model = nothing, backend = CPU(false), workgroupsize = 64, temperature = nothing)

We can assert that the kernel and matrix versions of the divergence give different results:

using Random
u = randn!(vectorfield(setup))
B = bc_u_mat(setup)
M = divergence_mat(setup)
div_kernel = divergence(apply_bc_u(u, 0.0, setup), setup)
div_matrix = reshape(M * B * u[:], size(div_kernel))
maximum(abs, div_matrix - div_kernel)

160.00000000000003

The solution is to create a vector containing the boundary conditions. This is done by applying the BC kernel on an empty field:

uzero = zero(u)
yu = apply_bc_u(uzero, 0.0, setup)
yM = divergence(yu, setup)

18×18 Matrix{Float64}:
 0.0     0.0     0.0     0.0     0.0  …     0.0     0.0     0.0     0.0  0.0
 0.0  -160.0  -160.0  -160.0  -160.0     -160.0  -160.0  -160.0  -160.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0  …     0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0  …     0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0  …     0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0
 0.0     0.0     0.0     0.0     0.0        0.0     0.0     0.0     0.0  0.0

By adding yM, we get the equality:

maximum(abs, (div_matrix + yM) - div_kernel)

2.842170943040401e-14

API

IncompressibleNavierStokes.bc_p_mat Function

Matrix for applying boundary conditions to pressure fields p.

source

IncompressibleNavierStokes.bc_temp_mat Function

Matrix for applying boundary conditions to temperature fields temp.

source

IncompressibleNavierStokes.bc_u_mat Function

Create matrix for applying boundary conditions to velocity fields u. This matrix only applies the boundary conditions depending on u itself (e.g. PeriodicBC). It does not apply constant boundary conditions (e.g. non-zero DirichletBC).

source

IncompressibleNavierStokes.diffusion_mat Method

diffusion_mat(
    setup
) -> SparseArrays.SparseMatrixCSC{Tv, Int64} where Tv

Diffusion matrix.

source

IncompressibleNavierStokes.divergence_mat Method

divergence_mat(
    setup
) -> SparseArrays.SparseMatrixCSC{Tv, Int64} where Tv

Divergence matrix.

source

IncompressibleNavierStokes.laplacian_mat Method

laplacian_mat(setup) -> Any

Get matrix for the Laplacian operator (for the pressure-Poisson equation). This matrix takes scalar field inputs restricted to the actual degrees of freedom.

source

IncompressibleNavierStokes.pad_scalarfield_mat Method

pad_scalarfield_mat(setup) -> Any

Create matrix for padding inner scalar field with boundary volumes. This can be useful for algorithms that require vectors with degrees of freedom only, and not the ghost volumes. To go back, simply transpose the matrix.

See also: pad_vectorfield_mat.

source

IncompressibleNavierStokes.pad_vectorfield_mat Method

pad_vectorfield_mat(
    setup
) -> SparseArrays.SparseMatrixCSC{Tv, Int64} where Tv

Create matrix for padding inner vector field with boundary volumes, similar to pad_scalarfield_mat.

source

IncompressibleNavierStokes.pressuregradient_mat Method

pressuregradient_mat(
    setup
) -> SparseArrays.SparseMatrixCSC{Tv, Int64} where Tv

Pressure gradient matrix.

source

IncompressibleNavierStokes.volume_mat Method

volume_mat(setup) -> Any

Volume-size matrix.

source