mirror of https://gitlab.com/QEF/q-e.git
66 lines
1.8 KiB
Fortran
66 lines
1.8 KiB
Fortran
!
|
|
! Copyright (C) 2003 A. Smogunov
|
|
! This file is distributed under the terms of the
|
|
! GNU General Public License. See the file `License'
|
|
! in the root directory of the present distribution,
|
|
! or http://www.gnu.org/copyleft/gpl.txt .
|
|
!
|
|
subroutine hev_ab (n, amt, lda, eigen, veigen, el, eh, m)
|
|
!
|
|
! It solves EP: A x = lambda X with hermitean matrix A.
|
|
!
|
|
#include "f_defs.h"
|
|
USE kinds, only : DP
|
|
implicit none
|
|
character :: lrange
|
|
integer :: &
|
|
n, & ! dimension of the matrix to be diagonalized
|
|
m, & ! input: flag
|
|
! output: number of eigenvalues in the [el,eh]
|
|
il, ih, & ! number of lowest and highest eigenvalues
|
|
lda, & ! leading dimension of A
|
|
lwork, & ! aux. var.
|
|
info ! -1 --> all eigenvalues are computed
|
|
|
|
integer, allocatable :: iwork(:), ifail(:)
|
|
real(kind=DP) :: &
|
|
eigen(n), & ! eigenvalues
|
|
el, eh, & ! interval for eigenvalue searching
|
|
abstol ! accuracy for eigenvalues
|
|
real(kind=DP), allocatable :: rwork(:)
|
|
complex(kind=DP), allocatable :: work(:)
|
|
complex(kind=DP) :: &
|
|
amt(lda, n), & ! A
|
|
veigen(lda, n) ! X
|
|
|
|
lwork = 16*n
|
|
|
|
allocate( work( lwork ) )
|
|
allocate( rwork( 7*n ) )
|
|
allocate( iwork( 5*n ) )
|
|
allocate( ifail( n ) )
|
|
|
|
abstol=0.d0
|
|
|
|
!
|
|
! If m=-1 find all the solutions, otherwise only el<lambda<eh
|
|
!
|
|
if (m.eq.-1) then
|
|
lrange='A'
|
|
else
|
|
lrange='V'
|
|
endif
|
|
|
|
call ZHEEVX('V', lrange, 'U', n, amt, lda, el, eh, il, ih, abstol, &
|
|
m, eigen, veigen, n, work, lwork, rwork, iwork, ifail, &
|
|
info)
|
|
call errore ('hev_ab','info =/= 0',abs(info))
|
|
|
|
deallocate(rwork)
|
|
deallocate(work)
|
|
deallocate(iwork)
|
|
deallocate(ifail)
|
|
|
|
return
|
|
end subroutine hev_ab
|