Intro

For most solver classes the user only needs to select their parameters through a dedicated ISolverBuilder object from which the solver will be created. Afterwards the solver will be used by classes that perform the analysis.

Dense solver

Example usage of this solver

IModel model = ...; // created beforehand

// Set up the solver's parameters
var solverBuilder = new DenseMatrixSolver.Builder();
solverBuilder.IsMatrixPositiveDefinite = true; // true to use LDL factorization, false to use LUP factorization. LUP is applicable to all invertible matrices, but slower.

// Create the solver object
DenseMatrixSolver solver = solverBuilder.BuildSolver(model);

// Setup analysis
var problem = new ProblemStructural(model, solver);
var linearAnalyzer = new LinearAnalyzer(model, solver, problem);
var staticAnalyzer = new StaticAnalyzer(model, solver, problem, linearAnalyzer);

Skyline LDL solver

Example usage of this solver with AMD reordering

IModel model = ...; // created beforehand

// Set up the solver's parameters
var solverBuilder = new DenseMatrixSolver.Builder();
solverBuilder.DofOrderer = new DofOrderer(new NodeMajorDofOrderingStrategy(), AmdReordering.CreateWithCSparseAmd()); // AMD reordering
solverBuilder.FactorizationPivotTolerance = 1E-7; // if a diagonal (pivot) entry falls beneath this value during factorization, the matrix will not be considered positive definite and the algorithm will terminate

// Create the solver object
SkylineSolver solver = solverBuilder.BuildSolver(model);

// Setup analysis
var problem = new ProblemStructural(model, solver);
var linearAnalyzer = new LinearAnalyzer(model, solver, problem);
var staticAnalyzer = new StaticAnalyzer(model, solver, problem, linearAnalyzer);

If the user does not want to use any reordering algorithm, then the corresponding line should be replaced with

solverBuilder.DofOrderer = new DofOrderer(new NodeMajorDofOrderingStrategy(), new NullReordering());