hsolver::dngvd_op< T, Device > Struct Template Reference

#include <dngvd_op.h>

Collaboration diagram for hsolver::dngvd_op< T, Device >:

Public Types
using	Real = typename GetTypeReal< T >::type

Public Member Functions
void	operator() (const Device *d, const int nstart, const int ldh, const T *A, const T *B, Real *W, T *V)
	DNGVD computes all the eigenvalues and eigenvectors of a complex generalized Hermitian-definite eigenproblem. If eigenvectors are desired, it uses a divide and conquer algorithm.

Member Typedef Documentation

Real

template<typename T , typename Device >

using hsolver::dngvd_op< T, Device >::Real = typename GetTypeReal<T>::type

Member Function Documentation

operator()()

template<typename T , typename Device >

void hsolver::dngvd_op< T, Device >::operator()	(	const Device *	d,
		const int	nstart,
		const int	ldh,
		const T *	A,
		const T *	B,
		Real *	W,
		T *	V
	)

DNGVD computes all the eigenvalues and eigenvectors of a complex generalized Hermitian-definite eigenproblem. If eigenvectors are desired, it uses a divide and conquer algorithm.

In this op, the CPU version is implemented through the gvd interface, and the CUDA version is implemented through the gvd interface. API doc:

1. zhegvd: https://netlib.org/lapack/explore-html/df/d9a/group__complex16_h_eeigen_ga74fdf9b5a16c90d8b7a589dec5ca058a.html
2. cusolverDnZhegvd: https://docs.nvidia.com/cuda/cusolver/index.html#cusolverdn-t-sygvd

Input Parameters

Parameters

d	: the type of device
nstart	: the number of cols of the matrix
ldh	: the number of rows of the matrix
A	: the hermitian matrix A in A x=lambda B x (col major)
B	: the overlap matrix B in A x=lambda B x (col major) Output Parameter
W	: calculated eigenvalues
V	: calculated eigenvectors (col major)

---

The documentation for this struct was generated from the following file:

• /home/runner/work/abacus-develop/abacus-develop/source/source_hsolver/kernels/dngvd_op.h