Type definition for MUMPS solver.

Functions/Subroutines
subroutine, public	call_mumps (solver, job)
	Call MUMPS and look for error or warning (check INFOG(1)).

subroutine, public	mumps_setup (solver, n_nonzeros, rhs_size)
	Initialization of MUMPS (JOB = -1) and set-up the parameters for the analysis without filling the matrix.

subroutine, public	distribute_solution (solution, isol_loc, sol_loc, row_offsets)
	Redistribute the solution across the processes.

Function/Subroutine Documentation

◆ call_mumps()

subroutine, public type_solver_mumps::call_mumps	(	class(t_mumps_solver), intent(inout)	solver,
		integer, intent(in)	job )

Parameters

[in,out]	solver	MUMPS solver.
[in]	job	job number.

MUMPS interface

subroutine, public type_solver_mumps::distribute_solution	(	double precision, dimension(:), intent(out)	solution,
		integer, dimension(:), intent(in)	isol_loc,
		double precision, dimension(:), intent(in)	sol_loc,
		integer, dimension(:), intent(in)	row_offsets )

The solution returned by MUMPS is indexed differently from the global indexing. This routine redistribute the solution.

Parameters

[out]	solution	solution of the linear system.
[in]	isol_loc	indices of the solution returned by MUMPS.
[in]	sol_loc	values of the solution returned by MUMPS.
[in]	row_offsets	offsets of the global indices (size n_mpi_proc).

subroutine, public type_solver_mumps::mumps_setup	(	class(t_mumps_solver), intent(inout)	solver,
		integer(kind=8), intent(in)	n_nonzeros,
		integer, intent(in)	rhs_size )

Parameters

[in,out]	solver	MUMPS solver.
[in]	n_nonzeros	number of nonzero coefficients.
[in]	rhs_size	size of the global right-hand side.