GPUHamiltonian< T > Class Template Reference

#include <hamgpu.h>

Inheritance diagram for GPUHamiltonian< T >:

Collaboration diagram for GPUHamiltonian< T >:

Public Member Functions
	GPUHamiltonian (int L, int Nu, int Nd, double J, double U)
	GPUHamiltonian (int L, int D, int Nu, int Nd, double J, double U)
virtual	~GPUHamiltonian ()
virtual void	mvprod (double *x, double *y, double alpha) const
double	LanczosDiagonalize (int m=0)
template<>
	GPUHamiltonian (int Ns, int Nu, int Nd, double J, double U)
template<>
	GPUHamiltonian (int L, int D, int Nu, int Nd, double J, double U)

Protected Attributes
double *	Umat_gpu
	The diagonal of the full hamiltonian (onsite interaction) More...
double *	Up_data_gpu
	The array to hold the data (ELL format) for the up hamiltonian. More...
double *	Down_data_gpu
	The array to hold the data (ELL format) for the down hamiltonian. More...
unsigned int *	Up_ind_gpu
	The array to hold the indices for the up hamiltonian. More...
unsigned int *	Down_ind_gpu
	The array to hold the indices for the down hamiltonian. More...

Detailed Description

template<class T>
class GPUHamiltonian< T >

This template class is used to do the Lanczos calculation on the GPU.

Author

Ward Poelmans wpoel.nosp@m.y86@.nosp@m.gmail.nosp@m..com

Definition at line 28 of file hamgpu.h.

Constructor & Destructor Documentation

template<class T >

GPUHamiltonian< T >::GPUHamiltonian	(	int	L,
		int	Nu,
		int	Nd,
		double	J,
		double	U
	)

template<class T >

GPUHamiltonian< T >::GPUHamiltonian	(	int	L,
		int	D,
		int	Nu,
		int	Nd,
		double	J,
		double	U
	)

template<class T >

GPUHamiltonian< T >::~GPUHamiltonian ( )

virtual

Definition at line 62 of file hamgpu.cu.

template<>

GPUHamiltonian< SparseHamiltonian >::GPUHamiltonian	(	int	Ns,
		int	Nu,
		int	Nd,
		double	J,
		double	U
	)

The constructor for the SparseHamiltonian Class

Definition at line 47 of file hamgpu.cu.

template<>

GPUHamiltonian< SparseHamiltonian2D >::GPUHamiltonian	(	int	L,
		int	D,
		int	Nu,
		int	Nd,
		double	J,
		double	U
	)

The constructor for the SparseHamiltonian2D Class

Definition at line 56 of file hamgpu.cu.

Member Function Documentation

template<class T >

double GPUHamiltonian< T >::LanczosDiagonalize

(

int

m = 0

)

Calculates the lowest eigenvalue of the hamiltonian matrix using the lanczos algorithm. Needs lapack.

Parameters

m	an optional estimate for the lanczos space size

Returns

the lowest eigenvalue

Definition at line 141 of file hamgpu.cu.

Here is the call graph for this function:

template<class T >

void GPUHamiltonian< T >::mvprod ( double *  x, double *  y, double  alpha  ) const

virtual

The matrix vector product. The method should calculate y = A*x + alpha * y

Parameters

x	the input vector
y	the output vector
alpha	the multiplicative constant

Definition at line 114 of file hamgpu.cu.

Member Data Documentation

template<class T >

double* GPUHamiltonian< T >::Down_data_gpu

protected

The array to hold the data (ELL format) for the down hamiltonian.

Definition at line 47 of file hamgpu.h.

template<class T >

unsigned int* GPUHamiltonian< T >::Down_ind_gpu

protected

The array to hold the indices for the down hamiltonian.

Definition at line 52 of file hamgpu.h.

template<class T >

double* GPUHamiltonian< T >::Umat_gpu

protected

The diagonal of the full hamiltonian (onsite interaction)

Definition at line 42 of file hamgpu.h.

template<class T >

double* GPUHamiltonian< T >::Up_data_gpu

protected

The array to hold the data (ELL format) for the up hamiltonian.

Definition at line 45 of file hamgpu.h.

template<class T >

unsigned int* GPUHamiltonian< T >::Up_ind_gpu

protected

The array to hold the indices for the up hamiltonian.

Definition at line 50 of file hamgpu.h.

---

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

• hamgpu.h
• hamgpu.cu