The field_storage struct contains minimal information for using the field in a loop. It is communicated to CUDA kernels and other accelerator functions. More...

#include <field_storage.h>

Public Member Functions
auto	get_element (const unsigned i, const lattice_struct &lattice) const

void	gather_elements (T __restrict__ buffer, const unsigned __restrict__ index_list, int n, const lattice_struct &lattice) const
	CUDA implementation of gather_elements without CUDA aware MPI.

void	gather_elements_negated (T __restrict__ buffer, const unsigned __restrict__ index_list, int n, const lattice_struct &lattice) const

void	place_comm_elements (Direction d, Parity par, T *__restrict__ buffer, const lattice_struct::comm_node_struct &from_node, const lattice_struct &lattice)
	Place boundary elements from neighbour.

void	place_elements (T __restrict__ buffer, const unsigned __restrict__ index_list, int n, const lattice_struct &lattice)
	CUDA implementation of place_elements without CUDA aware MPI.

void	set_local_boundary_elements (Direction dir, Parity par, const lattice_struct &lattice, bool antiperiodic)
	Place boundary elements from local lattice (used in vectorized version)

template<typename T >
void	set_element (const T &value, const unsigned idx)

Detailed Description

template<typename T>
class field_storage< T >

The field_storage struct contains minimal information for using the field in a loop. It is communicated to CUDA kernels and other accelerator functions.

Definition at line 29 of file field_storage.h.

Member Function Documentation

◆ gather_elements()

template<typename T >

void field_storage< T >::gather_elements	(	T *__restrict__	buffer,
		const unsigned *__restrict__	index_list,
		int	n,
		const lattice_struct &	lattice
	)		const

CUDA implementation of gather_elements without CUDA aware MPI.

Fetch elements from the field to buffer using sites in index_list.

Definition at line 41 of file backend_cpu/field_storage_backend.h.

◆ gather_elements_negated()

template<typename T >

void field_storage< T >::gather_elements_negated	(	T *__restrict__	buffer,
		const unsigned *__restrict__	index_list,
		int	n,
		const lattice_struct &	lattice
	)		const

requires unary - !!

Definition at line 54 of file backend_cpu/field_storage_backend.h.

◆ get_element()

template<typename T >

T field_storage< T >::get_element	(	const unsigned	i,
		const lattice_struct &	lattice
	)		const

inline

get_element gathers one T-element from vectorized store again, idx is the "site" index

Definition at line 125 of file backend_cpu/field_storage_backend.h.

◆ place_elements()

template<typename T >

void field_storage< T >::place_elements	(	T *__restrict__	buffer,
		const unsigned *__restrict__	index_list,
		int	n,
		const lattice_struct &	lattice
	)

CUDA implementation of place_elements without CUDA aware MPI.

Vectorized implementation of setting elements.

Definition at line 72 of file backend_cpu/field_storage_backend.h.

◆ set_element()

template<typename T >

void field_storage< T >::set_element	(	const T &	value,
		const unsigned	idx
	)

inline

set_element scatters one individual T-element to vectorized store, using the "site" index idx.

Definition at line 130 of file backend_vector/field_storage_backend.h.

◆ set_local_boundary_elements()

template<typename T >

void field_storage< T >::set_local_boundary_elements	(	Direction	dir,
		Parity	par,
		const lattice_struct &	lattice,
		bool	antiperiodic
	)

Place boundary elements from local lattice (used in vectorized version)

Loop over the boundary sites - i is the vector index location where the vectors are copied from are in halo_index

get ptrs to target and source vec elements

Definition at line 81 of file backend_cpu/field_storage_backend.h.

The documentation for this class was generated from the following files:

Public Member Functions

Detailed Description

Member Function Documentation

◆ gather_elements()

◆ gather_elements_negated()

◆ get_element()

◆ place_elements()

◆ set_element()

◆ set_local_boundary_elements()