Manual

parallel_make_matrix(
    fes,
    u,
    createsubd,
    matrixupdt!;
    ntasks = Threads.nthreads(),
    kind = :CSC,
)

Assemble a sparse matrix.

Either a :CSC matrix or a :CSR matrix is created. We shall refer to this matrix as a CSX matrix. The process is:

1. Construct the incidence relation node-to-elements.
2. Construct the incidence relation node-to-neighbors.
3. Make the sparse pattern and create a sparse CSX matrix with all values zero.
4. Construct the incidence relation element-to-neighbors.
5. Compute element coloring.
6. Set up domain decomposition.
7. Compute and assemble the matrix entries.

Here, createsubd could be

(fessubset) -> FEMMAcoust(IntegDomain(fessubset, GaussRule(3, 2)), material)

and matrixupdt! could be

(femm, assmblr) -> acousticstiffness(femm, assmblr, geom, P)

parallel_make_matrix(
    fes,
    dofnums,
    ndofs,
    FT,
    n2e,
    createsubd,
    matrixupdt!,
    ntasks,
    kind,
)

Assemble a sparse matrix.

1. Construct the incidence relation node-to-neighbors.
2. Make the sparse pattern and create a sparse CSX matrix with all values zero.
3. Construct the incidence relation element-to-neighbors.
4. Compute element coloring.
5. Set up domain decomposition.
6. Compute and assemble the matrix entries.