Sparse Solver

Intel® MKL provides a user-callable direct sparse solver subroutine to solve symmetric and symmetricly-structured matrices with real or complex coefficients. For sparse symmetric matrices, this solver can solve both positive definite and indefinite systems.

pardiso calculates the solution of a set of sparse linear equations with multiple right-hand sides.
Fortran:
call pardiso(pt, maxfct, mnum, mtype, phase, n, a, ia, ja, perm, nrhs, iparm, msglvl, b, x, error)
C:
pardiso(pt, &maxfct, &mnum, &mtype, &phase, &n, a, ia, ja, perm, &nrhs, iparm, &msglvl, b, x, &error);


Additionally to PARDISO interface, Intel MKL supports

• an alternative interface for the direct sparse solver referred to as Direct Sparse Solvers (DSS) interface
• an alternative interface for the interative sparse solvers based on Reverse Communication Interface referred to as RCI ISS interface.

DSS Interface

dss_create initializes the solver.
dss_create(handle, opt)


dss_define_structure communicates to the solver locations of nNonZeros number of non-zero elements in matrix of size nRows by nCols.
dss_define_structure(handle, opt, rowIndex, nRows, nCols, columns, nNonZeros)


dss_reorder computes permutation vector that minimizes the fill-in during the factorization phase.
dss_reorder(handle, opt, perm)


dss_factor_real / dss_factor_complex compute the factorization of the matrix with previously specified location.
dss_factor_real(handle, opt, rValues)
dss_factor_complex(handle, opt, cValues)

dss_solve_real / dss_solve_complex compute the corresponding solutions vector and place it in the output array.
dss_solve_real(handle, opt, rRhsValues, nRhs, rSolValues)
dss_solve_complex(handle, opt, cRhsValues, nRhs, cSolValues)


dss_delete deletes all of data structures created during the solutions process.
dss_delete(handle, opt)


dss_statistics returns statistics about various phases of the solving process.
dss_statistics(handle, opt, statArr, retValues)


mkl_cvt_to_null_terminated_str passes character strings from Fortran routines to C routines.
mkl_cvt_to_null_terminated_str(destStr, destLen, srcStr)


RCI ISS Interface

RCI Conjugate Gradient (CG) Solver

dcg_init initializes the solver.
dcg_init(n, x, b, RCI_request, ipar, dpar, tmp)


dcg_check checks the user defined data for consistency and correctness.
dcg_check(n, x, b, RCI_request, ipar, dpar, tmp)


dcg computes the approximate solution vector.
dcg(n, x, b, RCI_request, ipar, dpar, tmp)


dcg_get retrieves the number of the current iteration.
dcg_get(n, x, b, RCI_request, ipar, dpar, tmp, itercount)


RCI Flexible Generalized Minimal Residual (FGMRES) Solver

dfgmres_init initializes the solver.
dfgmres_init(n, x, b, RCI_request, ipar, dpar, tmp)


dfgmres_check checks the user defined data for consistency and correctness.
dfgmres_check(n, x, b, RCI_request, ipar, dpar, tmp)


dfgmres makes the FGMRES iterations.
dfgmres(n, x, b, RCI_request, ipar, dpar, tmp)


dfgmres_get retrieves the number of the current iteration and updates the solution.
dfgmres_get(n, x, b, RCI_request, ipar, dpar, tmp, itercount)


* Legal Information © 1999, 2002-2006, Intel Corporation