The Trilinos Solver Framework

More documents

Recommendations

Info

Adjoint, inverse, and adjoint-inverseoperators can be created at a high level•TSFAdjointOperator applies the adjoint of another operatorTSFLinearOperator adj = new TSFAdjointOperator(A);or more simply,TSFLinearOperator Aadj = A.adjoint();•TSFInverseOperator applies the inverse of another operatorusing a specified solver. The inverse is never formed.TSFLinearSolver solver = new BICGSTABSolver(...);TSFLinearOperator Ainv = A.inverse(solver);•TSFInverseAdjointOperator applies the inverse adjointTSFLinearSolver solver = new BICGSTABSolver(...);TSFLinearOperator AinvAdj = A.inverseAdjoint(solver);Computational Science and Mathematics Research DepartmentSandiaNationalLaboratories
Overloaded operators are used to formefficient compound operators; Operator composition with overloaded *, e.g. A*B*C; Operator addition with overloaded + and -, e.g. A-B; Scalar multiplication with overloaded *, e.g, a*A; All of these are implemented without forming the explicitmatrix product or sumComputational Science and Mathematics Research DepartmentSandiaNationalLaboratories
Page 1 and 2: The Trilinos Solver FrameworkKevin
Page 3 and 4: TSF• Interface for representation
Page 5: written with block and composedoper
Page 8 and 9: The overhead incurred by using TSF
Page 10: General code style; TSF is in C++;
Page 13: TSF provides an interface forconfig
Page 17 and 18: Preconditioning• TSFPreconditione
Page 19 and 20: Example: creating a Kay-Loghinpreco
Page 21 and 22: Where is TSF being used?; Sundance
Page 23 and 24: Acknowledgements; Helpful design cr
Page 25: Formulation and solution of a sourc

The Trilinos Solver Framework

Create successful ePaper yourself

Delete template?

Save as template?