nonaligned_operators_m Module

Implementation of parallel operators with non-aligned methods, These methods are mostly for testing purpose or exploring numerical techniques, i.e. there is no focus on performance optimisation and no unit tests were created for these routines - Guenter's scheme second order: (see S. Guenter et al., Journal of Computational Physics 209 (2005) 354–370 - Guenter's scheme fourth order: (see S. Guenter et al., Journal of Computational Physics 226 (2007) 2306-2316

Contents

Subroutines

gradient_3D_staggered parallel_gradient_nonaligned parallel_divergence_nonaligned

Subroutines

public subroutine gradient_3D_staggered(comm_handler, equi, mesh, map, u, gradu, order)

Computes gradient (dudx, dudy, dudphi = dudz) at 3D staggered positions Gradient is computed according to symmetric formulation of Guenter's scheme Staggered point is at northeast-forward position w.r.t. its considered grid point on full grid Second and fourth order are implemented

Arguments

Type	Intent	Optional	Attributes		Name
type(comm_handler_t),	intent(in)			::	comm_handler	Communicator
class(equilibrium_t),	intent(inout)			::	equi	Equilibrium
type(mesh_cart_t),	intent(inout)			::	mesh	Mesh within poloidal plane
type(parallel_map_t),	intent(in)			::	map	Parallel map
type(variable_t)				::	u	Variable, must be defined on canonical grid
real(kind=GP),	intent(out),		dimension(mesh%get_n_points(), 3)	::	gradu	Values of gradient
integer,	intent(in),	optional		::	order	Order of scheme default: 2

public subroutine parallel_gradient_nonaligned(comm_handler, equi, mesh, map, u, pargradu, order)

Computes parallel gradient at 3D staggered positions Based on gradient_3D_staggered

Arguments

Type	Intent	Optional	Attributes		Name
type(comm_handler_t),	intent(in)			::	comm_handler	Communicator
class(equilibrium_t),	intent(inout)			::	equi	Equilibrium
type(mesh_cart_t),	intent(inout)			::	mesh	Mesh within poloidal plane
type(parallel_map_t),	intent(in)			::	map	Parallel map
type(variable_t),	intent(in)			::	u	Variable, must be defined on canonical grid
real(kind=GP),	intent(out),		dimension(mesh%get_n_points())	::	pargradu	Values of parallel gradient
integer,	intent(in),	optional		::	order	Order of scheme default: 2

public subroutine parallel_divergence_nonaligned(comm, equi, mesh, map, uflx, pardivu, order)

Computes parallel divergence from parallel flux defined at 3D staggered positions Result (pardivu) is defined on full grid Based on symmetric formulation of Guenter's scheme (2nd and 4th order available)

Arguments

Type	Intent	Optional	Attributes		Name
integer,	intent(in)			::	comm	MPI communicator
class(equilibrium_t),	intent(inout)			::	equi	Equilibrium
type(mesh_cart_t),	intent(inout)			::	mesh	Mesh within poloidal plane
type(parallel_map_t),	intent(in)			::	map	Parallel map
real(kind=GP),	intent(in),		dimension(mesh%get_n_points())	::	uflx	Parallel flux
real(kind=GP),	intent(out),		dimension(mesh%get_n_points_inner())	::	pardivu	Parallel divergence of uflx
integer,	intent(in),	optional		::	order	Order of scheme default: 2