mirror of https://gitlab.com/QEF/q-e.git
723 lines
22 KiB
Fortran
723 lines
22 KiB
Fortran
!
|
|
! Copyright (C) 2001-2013 Quantum ESPRESSO group
|
|
! 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 .
|
|
!
|
|
!
|
|
#define ZERO ( 0.D0, 0.D0 )
|
|
#define ONE ( 1.D0, 0.D0 )
|
|
!
|
|
!----------------------------------------------------------------------------
|
|
SUBROUTINE laxlib_cdiaghg( n, m, h, s, ldh, e, v, me_bgrp, root_bgrp, intra_bgrp_comm )
|
|
!----------------------------------------------------------------------------
|
|
!
|
|
! ... calculates eigenvalues and eigenvectors of the generalized problem
|
|
! ... Hv=eSv, with H hermitean matrix, S overlap matrix.
|
|
! ... On output both matrix are unchanged
|
|
!
|
|
! ... LAPACK version - uses both ZHEGV and ZHEGVX
|
|
!
|
|
USE laxlib_parallel_include
|
|
IMPLICIT NONE
|
|
INCLUDE 'laxlib_kinds.fh'
|
|
!
|
|
INTEGER, INTENT(IN) :: n, m, ldh
|
|
! dimension of the matrix to be diagonalized
|
|
! number of eigenstates to be calculate
|
|
! leading dimension of h, as declared in the calling pgm unit
|
|
COMPLEX(DP), INTENT(INOUT) :: h(ldh,n), s(ldh,n)
|
|
! actually intent(in) but compilers don't know and complain
|
|
! matrix to be diagonalized
|
|
! overlap matrix
|
|
REAL(DP), INTENT(OUT) :: e(n)
|
|
! eigenvalues
|
|
COMPLEX(DP), INTENT(OUT) :: v(ldh,m)
|
|
! eigenvectors (column-wise)
|
|
INTEGER, INTENT(IN) :: me_bgrp, root_bgrp, intra_bgrp_comm
|
|
!
|
|
INTEGER :: lwork, nb, mm, info, i, j
|
|
! mm = number of calculated eigenvectors
|
|
REAL(DP) :: abstol
|
|
INTEGER, ALLOCATABLE :: iwork(:), ifail(:)
|
|
REAL(DP), ALLOCATABLE :: rwork(:), sdiag(:), hdiag(:)
|
|
COMPLEX(DP), ALLOCATABLE :: work(:)
|
|
! various work space
|
|
LOGICAL :: all_eigenvalues
|
|
! REAL(DP), EXTERNAL :: DLAMCH
|
|
INTEGER, EXTERNAL :: ILAENV
|
|
! ILAENV returns optimal block size "nb"
|
|
!
|
|
!
|
|
CALL start_clock( 'cdiaghg' )
|
|
!
|
|
! ... only the first processor diagonalizes the matrix
|
|
!
|
|
IF ( me_bgrp == root_bgrp ) THEN
|
|
!
|
|
! ... save the diagonal of input S (it will be overwritten)
|
|
!
|
|
ALLOCATE( sdiag( n ) )
|
|
DO i = 1, n
|
|
sdiag(i) = DBLE( s(i,i) )
|
|
END DO
|
|
!
|
|
all_eigenvalues = ( m == n )
|
|
!
|
|
! ... check for optimal block size
|
|
!
|
|
nb = ILAENV( 1, 'ZHETRD', 'U', n, -1, -1, -1 )
|
|
!
|
|
IF ( nb < 1 .OR. nb >= n) THEN
|
|
!
|
|
lwork = 2*n
|
|
!
|
|
ELSE
|
|
!
|
|
lwork = ( nb + 1 )*n
|
|
!
|
|
END IF
|
|
!
|
|
ALLOCATE( work( lwork ) )
|
|
!
|
|
IF ( all_eigenvalues ) THEN
|
|
!
|
|
ALLOCATE( rwork( 3*n - 2 ) )
|
|
!
|
|
! ... calculate all eigenvalues (overwritten to v)
|
|
!
|
|
v(:,:) = h(:,:)
|
|
!
|
|
CALL ZHEGV( 1, 'V', 'U', n, v, ldh, &
|
|
s, ldh, e, work, lwork, rwork, info )
|
|
!
|
|
ELSE
|
|
!
|
|
ALLOCATE( rwork( 7*n ) )
|
|
!
|
|
! ... save the diagonal of input H (it will be overwritten)
|
|
!
|
|
ALLOCATE( hdiag( n ) )
|
|
DO i = 1, n
|
|
hdiag(i) = DBLE( h(i,i) )
|
|
END DO
|
|
!
|
|
ALLOCATE( iwork( 5*n ) )
|
|
ALLOCATE( ifail( n ) )
|
|
!
|
|
! ... calculate only m lowest eigenvalues
|
|
!
|
|
abstol = 0.D0
|
|
! abstol = 2.D0*DLAMCH( 'S' )
|
|
!
|
|
! ... the following commented lines calculate optimal lwork
|
|
!
|
|
!lwork = -1
|
|
!
|
|
!CALL ZHEGVX( 1, 'V', 'I', 'U', n, h, ldh, s, ldh, &
|
|
! 0.D0, 0.D0, 1, m, abstol, mm, e, v, ldh, &
|
|
! work, lwork, rwork, iwork, ifail, info )
|
|
!
|
|
!lwork = INT( work(1) ) + 1
|
|
!
|
|
!IF( lwork > SIZE( work ) ) THEN
|
|
! DEALLOCATE( work )
|
|
! ALLOCATE( work( lwork ) )
|
|
!END IF
|
|
!
|
|
CALL ZHEGVX( 1, 'V', 'I', 'U', n, h, ldh, s, ldh, &
|
|
0.D0, 0.D0, 1, m, abstol, mm, e, v, ldh, &
|
|
work, lwork, rwork, iwork, ifail, info )
|
|
!
|
|
DEALLOCATE( ifail )
|
|
DEALLOCATE( iwork )
|
|
!
|
|
! ... restore input H matrix from saved diagonal and lower triangle
|
|
!
|
|
DO i = 1, n
|
|
h(i,i) = CMPLX( hdiag(i), 0.0_DP ,kind=DP)
|
|
DO j = i + 1, n
|
|
h(i,j) = CONJG( h(j,i) )
|
|
END DO
|
|
DO j = n + 1, ldh
|
|
h(j,i) = ( 0.0_DP, 0.0_DP )
|
|
END DO
|
|
END DO
|
|
!
|
|
DEALLOCATE( hdiag )
|
|
!
|
|
END IF
|
|
!
|
|
!
|
|
DEALLOCATE( rwork )
|
|
DEALLOCATE( work )
|
|
!
|
|
IF ( info > n ) THEN
|
|
CALL lax_error__( 'cdiaghg', 'S matrix not positive definite', ABS( info ) )
|
|
ELSE IF ( info > 0 ) THEN
|
|
CALL lax_error__( 'cdiaghg', 'eigenvectors failed to converge', ABS( info ) )
|
|
ELSE IF ( info < 0 ) THEN
|
|
CALL lax_error__( 'cdiaghg', 'incorrect call to ZHEGV*', ABS( info ) )
|
|
END IF
|
|
!
|
|
! ... restore input S matrix from saved diagonal and lower triangle
|
|
!
|
|
DO i = 1, n
|
|
s(i,i) = CMPLX( sdiag(i), 0.0_DP ,kind=DP)
|
|
DO j = i + 1, n
|
|
s(i,j) = CONJG( s(j,i) )
|
|
END DO
|
|
DO j = n + 1, ldh
|
|
s(j,i) = ( 0.0_DP, 0.0_DP )
|
|
END DO
|
|
END DO
|
|
!
|
|
DEALLOCATE( sdiag )
|
|
!
|
|
END IF
|
|
!
|
|
! ... broadcast eigenvectors and eigenvalues to all other processors
|
|
!
|
|
#if defined __MPI
|
|
CALL MPI_BCAST( e, SIZE(e), MPI_DOUBLE_PRECISION, root_bgrp, intra_bgrp_comm, info )
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'cdiaghg', 'error broadcasting array e', ABS( info ) )
|
|
CALL MPI_BCAST( v, SIZE(v), MPI_DOUBLE_COMPLEX, root_bgrp, intra_bgrp_comm, info )
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'cdiaghg', 'error broadcasting array v', ABS( info ) )
|
|
#endif
|
|
!
|
|
CALL stop_clock( 'cdiaghg' )
|
|
!
|
|
RETURN
|
|
!
|
|
END SUBROUTINE laxlib_cdiaghg
|
|
!
|
|
!----------------------------------------------------------------------------
|
|
SUBROUTINE laxlib_cdiaghg_gpu( n, m, h_d, s_d, ldh, e_d, v_d, me_bgrp, root_bgrp, intra_bgrp_comm)
|
|
!----------------------------------------------------------------------------
|
|
!
|
|
! ... calculates eigenvalues and eigenvectors of the generalized problem
|
|
! ... Hv=eSv, with H hermitean matrix, S overlap matrix.
|
|
! ... On output both matrix are unchanged
|
|
!
|
|
!
|
|
#if defined(__CUDA)
|
|
USE cudafor
|
|
!
|
|
#if defined(__USE_CUSOLVER)
|
|
USE cusolverdn
|
|
#else
|
|
USE zhegvdx_gpu
|
|
#endif
|
|
#endif
|
|
!
|
|
USE laxlib_parallel_include
|
|
!
|
|
!define __USE_GLOBAL_BUFFER
|
|
#if defined(__USE_GLOBAL_BUFFER)
|
|
USE gbuffers, ONLY : dev=>dev_buf, pin=>pin_buf
|
|
#define VARTYPE POINTER
|
|
#else
|
|
#define VARTYPE ALLOCATABLE
|
|
#endif
|
|
!
|
|
IMPLICIT NONE
|
|
INCLUDE 'laxlib_kinds.fh'
|
|
!
|
|
INTEGER, INTENT(IN) :: n, m, ldh
|
|
! dimension of the matrix to be diagonalized
|
|
! number of eigenstates to be calculate
|
|
! leading dimension of h, as declared in the calling pgm unit
|
|
COMPLEX(DP), INTENT(INOUT) :: h_d(ldh,n), s_d(ldh,n)
|
|
! actually intent(in) but compilers don't know and complain
|
|
! matrix to be diagonalized, allocated on the GPU
|
|
! overlap matrix, allocated on the GPU
|
|
REAL(DP), INTENT(OUT) :: e_d(n)
|
|
! eigenvalues, , allocated on the GPU
|
|
COMPLEX(DP), INTENT(OUT) :: v_d(ldh,n)
|
|
! eigenvectors (column-wise), , allocated on the GPU
|
|
! NB: the dimension of v_d this is different from cdiaghg !!
|
|
#if defined(__CUDA)
|
|
ATTRIBUTES(DEVICE) :: h_d, s_d, e_d, v_d
|
|
#endif
|
|
INTEGER, INTENT(IN) :: me_bgrp, root_bgrp, intra_bgrp_comm
|
|
!
|
|
INTEGER :: lwork, info
|
|
!
|
|
REAL(DP) :: abstol
|
|
INTEGER, ALLOCATABLE :: ifail(:)
|
|
INTEGER, VARTYPE :: iwork(:)
|
|
REAL(DP), VARTYPE :: rwork(:)
|
|
COMPLEX(DP), VARTYPE :: work(:)
|
|
!
|
|
COMPLEX(DP), VARTYPE :: v_h(:,:)
|
|
REAL(DP), VARTYPE :: e_h(:)
|
|
#if (! defined(__USE_GLOBAL_BUFFER)) && defined(__CUDA)
|
|
ATTRIBUTES( PINNED ) :: work, iwork, rwork, v_h, e_h
|
|
#endif
|
|
!
|
|
INTEGER :: lwork_d, lrwork_d, liwork, lrwork
|
|
REAL(DP), VARTYPE :: rwork_d(:)
|
|
COMPLEX(DP), VARTYPE :: work_d(:)
|
|
! various work space
|
|
!
|
|
! Temp arrays to save H and S.
|
|
REAL(DP), VARTYPE :: h_diag_d(:), s_diag_d(:)
|
|
#if defined(__CUDA)
|
|
ATTRIBUTES( DEVICE ) :: work_d, rwork_d, h_diag_d, s_diag_d
|
|
#endif
|
|
INTEGER :: i, j
|
|
#if defined( __USE_CUSOLVER )
|
|
INTEGER :: devInfo_d, h_meig
|
|
ATTRIBUTES( DEVICE ) :: devInfo_d
|
|
TYPE(cusolverDnHandle) :: cuSolverHandle
|
|
!
|
|
COMPLEX(DP), VARTYPE :: h_bkp_d(:,:), s_bkp_d(:,:)
|
|
ATTRIBUTES( DEVICE ) :: h_bkp_d, s_bkp_d
|
|
#endif
|
|
#undef VARTYPE
|
|
!
|
|
!
|
|
!
|
|
!
|
|
CALL start_clock_gpu( 'cdiaghg' )
|
|
!
|
|
! ... only the first processor diagonalizes the matrix
|
|
!
|
|
IF ( me_bgrp == root_bgrp ) THEN
|
|
!
|
|
! Keeping compatibility for both CUSolver and CustomEigensolver, CUSolver below
|
|
!
|
|
#if defined(__USE_CUSOLVER) && defined(__CUDA)
|
|
!
|
|
! vvv __USE_CUSOLVER
|
|
|
|
#if ! defined(__USE_GLOBAL_BUFFER)
|
|
ALLOCATE(h_bkp_d(n,n), s_bkp_d(n,n), STAT = info)
|
|
IF( info /= 0 ) CALL errore( ' cdiaghg_gpu ', ' cannot allocate h_bkp_d or s_bkp_d ', ABS( info ) )
|
|
#else
|
|
CALL dev%lock_buffer( h_bkp_d, (/ n, n /), info )
|
|
CALL dev%lock_buffer( s_bkp_d, (/ n, n /), info )
|
|
#endif
|
|
!
|
|
!$cuf kernel do(2)
|
|
DO j=1,n
|
|
DO i=1,n
|
|
h_bkp_d(i,j) = h_d(i,j)
|
|
s_bkp_d(i,j) = s_d(i,j)
|
|
ENDDO
|
|
ENDDO
|
|
!
|
|
info = cusolverDnCreate(cuSolverHandle)
|
|
IF ( info /= CUSOLVER_STATUS_SUCCESS ) CALL errore( ' cdiaghg_gpu ', 'cusolverDnCreate', ABS( info ) )
|
|
!
|
|
info = cusolverDnZhegvdx_bufferSize(cuSolverHandle, CUSOLVER_EIG_TYPE_1, CUSOLVER_EIG_MODE_VECTOR, CUSOLVER_EIG_RANGE_I, CUBLAS_FILL_MODE_UPPER, &
|
|
n, h_d, ldh, s_d, ldh, 0.D0, 0.D0, 1, m, h_meig, e_d, lwork_d)
|
|
!
|
|
#if ! defined(__USE_GLOBAL_BUFFER)
|
|
ALLOCATE(work_d(1*lwork_d), STAT = info)
|
|
IF( info /= 0 ) CALL errore( ' cdiaghg_gpu ', ' cannot allocate work_d ', ABS( info ) )
|
|
#else
|
|
CALL dev%lock_buffer( work_d, lwork_d, info )
|
|
#endif
|
|
!
|
|
info = cusolverDnZhegvdx(cuSolverHandle, CUSOLVER_EIG_TYPE_1, CUSOLVER_EIG_MODE_VECTOR, CUSOLVER_EIG_RANGE_I, CUBLAS_FILL_MODE_UPPER, &
|
|
n, h_d, ldh, s_d, ldh, 0.D0, 0.D0, 1, m, h_meig, e_d, work_d, lwork, devInfo_d)
|
|
!$cuf kernel do(2)
|
|
DO j=1,n
|
|
DO i=1,n
|
|
IF(j <= m) v_d(i,j) = h_d(i,j)
|
|
h_d(i,j) = h_bkp_d(i,j)
|
|
s_d(i,j) = s_bkp_d(i,j)
|
|
ENDDO
|
|
ENDDO
|
|
!
|
|
IF( info /= 0 ) CALL errore( ' cdiaghg_gpu ', ' cusolverDnZhegvdx failed ', ABS( info ) )
|
|
info = cusolverDnDestroy(cuSolverHandle)
|
|
IF( info /= 0 ) CALL errore( ' cdiaghg_gpu ', ' cusolverDnDestroy failed ', ABS( info ) )
|
|
!
|
|
#if ! defined(__USE_GLOBAL_BUFFER)
|
|
DEALLOCATE(work_d)
|
|
DEALLOCATE(h_bkp_d, s_bkp_d)
|
|
#else
|
|
CALL dev%release_buffer( work_d, info )
|
|
CALL dev%release_buffer( h_bkp_d, info )
|
|
CALL dev%release_buffer( s_bkp_d, info )
|
|
#endif
|
|
! ^^^ __USE_CUSOLVER
|
|
!
|
|
! Keeping compatibility for both CUSolver and CustomEigensolver, CustomEigensolver below
|
|
!
|
|
#elif defined(__CUDA)
|
|
! vvv not __USE_CUSOLVER
|
|
#if ! defined(__USE_GLOBAL_BUFFER)
|
|
! NB: dimension is different!
|
|
ALLOCATE(v_h(ldh,n), e_h(n))
|
|
ALLOCATE(h_diag_d(n) , s_diag_d(n))
|
|
#else
|
|
CALL pin%lock_buffer( v_h, (/ldh,n/), info )
|
|
CALL pin%lock_buffer( e_h, n, info )
|
|
!
|
|
CALL dev%lock_buffer( h_diag_d, n, info )
|
|
CALL dev%lock_buffer( s_diag_d, n, info )
|
|
#endif
|
|
!
|
|
lwork = n
|
|
lrwork = 1+5*n+2*n*n
|
|
liwork = 3+5*n
|
|
!
|
|
lwork_d = 2*64*64 + 65 * n
|
|
lrwork_d = n
|
|
!
|
|
#if ! defined(__USE_GLOBAL_BUFFER)
|
|
ALLOCATE(work(lwork), rwork(lrwork), iwork(liwork))
|
|
!
|
|
ALLOCATE(work_d(1*lwork_d), STAT = info)
|
|
IF( info /= 0 ) CALL errore( ' cdiaghg_gpu ', ' allocate work_d ', ABS( info ) )
|
|
!
|
|
ALLOCATE(rwork_d(1*lrwork_d), STAT = info)
|
|
IF( info /= 0 ) CALL errore( ' cdiaghg_gpu ', ' allocate rwork_d ', ABS( info ) )
|
|
#else
|
|
CALL pin%lock_buffer(work, lwork, info)
|
|
CALL pin%lock_buffer(rwork, lrwork, info)
|
|
CALL pin%lock_buffer(iwork, liwork, info)
|
|
CALL dev%lock_buffer( work_d, lwork_d, info)
|
|
CALL dev%lock_buffer( rwork_d, lrwork_d, info)
|
|
#endif
|
|
!
|
|
!$cuf kernel do(1) <<<*,*>>>
|
|
DO i = 1, n
|
|
h_diag_d(i) = DBLE( h_d(i,i) )
|
|
s_diag_d(i) = DBLE( s_d(i,i) )
|
|
END DO
|
|
CALL zhegvdx_gpu(n, h_d, ldh, s_d, ldh, v_d, ldh, 1, m, e_d, work_d,&
|
|
lwork_d, rwork_d, lrwork_d, &
|
|
work, lwork, rwork, lrwork, &
|
|
iwork, liwork, v_h, SIZE(v_h, 1), e_h, info, .TRUE.)
|
|
!
|
|
|
|
IF( info /= 0 ) CALL errore( ' cdiaghg_gpu ', ' zhegvdx_gpu failed ', ABS( info ) )
|
|
!
|
|
!$cuf kernel do(1) <<<*,*>>>
|
|
DO i = 1, n
|
|
h_d(i,i) = DCMPLX( h_diag_d(i), 0.0_DP)
|
|
s_d(i,i) = DCMPLX( s_diag_d(i), 0.0_DP)
|
|
DO j = i + 1, n
|
|
h_d(i,j) = DCONJG( h_d(j,i) )
|
|
s_d(i,j) = DCONJG( s_d(j,i) )
|
|
END DO
|
|
DO j = n + 1, ldh
|
|
h_d(j,i) = ( 0.0_DP, 0.0_DP )
|
|
s_d(j,i) = ( 0.0_DP, 0.0_DP )
|
|
END DO
|
|
END DO
|
|
#if ! defined(__USE_GLOBAL_BUFFER)
|
|
DEALLOCATE(h_diag_d, s_diag_d)
|
|
!
|
|
DEALLOCATE(work, rwork, iwork)
|
|
DEALLOCATE(work_d, rwork_d)
|
|
DEALLOCATE(v_h, e_h)
|
|
#else
|
|
CALL dev%release_buffer( h_diag_d, info )
|
|
CALL dev%release_buffer( s_diag_d, info)
|
|
!
|
|
CALL pin%release_buffer(work, info)
|
|
CALL pin%release_buffer(rwork, info)
|
|
CALL pin%release_buffer(iwork, info)
|
|
CALL dev%release_buffer( work_d, info)
|
|
CALL dev%release_buffer( rwork_d, info)
|
|
CALL pin%release_buffer(v_h, info)
|
|
CALL pin%release_buffer(e_h, info)
|
|
#endif
|
|
! ^^^ not __USE_CUSOLVER
|
|
#else
|
|
CALL lax_error__( 'cdiaghg', 'Called GPU eigensolver without GPU support', 1 )
|
|
#endif
|
|
!
|
|
END IF
|
|
!
|
|
! ... broadcast eigenvectors and eigenvalues to all other processors
|
|
!
|
|
#if defined __MPI
|
|
#if defined __GPU_MPI
|
|
info = cudaDeviceSynchronize()
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'cdiaghg', 'error synchronizing device (first)', ABS( info ) )
|
|
CALL MPI_BCAST( e_d, n, MPI_DOUBLE_PRECISION, root_bgrp, intra_bgrp_comm, info )
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'cdiaghg', 'error broadcasting array e_d', ABS( info ) )
|
|
CALL MPI_BCAST( v_d, ldh*m, MPI_DOUBLE_COMPLEX, root_bgrp, intra_bgrp_comm, info )
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'cdiaghg', 'error broadcasting array v_d', ABS( info ) )
|
|
info = cudaDeviceSynchronize() ! this is probably redundant...
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'cdiaghg', 'error synchronizing device (second)', ABS( info ) )
|
|
#else
|
|
ALLOCATE(e_h(n), v_h(ldh,m))
|
|
e_h(1:n) = e_d(1:n)
|
|
v_h(1:ldh, 1:m) = v_d(1:ldh, 1:m)
|
|
CALL MPI_BCAST( e_h, n, MPI_DOUBLE_PRECISION, root_bgrp, intra_bgrp_comm, info )
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'cdiaghg', 'error broadcasting array e_d', ABS( info ) )
|
|
CALL MPI_BCAST( v_h, ldh*m, MPI_DOUBLE_COMPLEX, root_bgrp, intra_bgrp_comm, info )
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'cdiaghg', 'error broadcasting array v_d', ABS( info ) )
|
|
e_d(1:n) = e_h(1:n)
|
|
v_d(1:ldh, 1:m) = v_h(1:ldh, 1:m)
|
|
DEALLOCATE(e_h, v_h)
|
|
#endif
|
|
#endif
|
|
!
|
|
CALL stop_clock_gpu( 'cdiaghg' )
|
|
!
|
|
RETURN
|
|
!
|
|
END SUBROUTINE laxlib_cdiaghg_gpu
|
|
!
|
|
!----------------------------------------------------------------------------
|
|
!----------------------------------------------------------------------------
|
|
SUBROUTINE laxlib_pcdiaghg( n, h, s, ldh, e, v, idesc )
|
|
!----------------------------------------------------------------------------
|
|
!
|
|
! ... calculates eigenvalues and eigenvectors of the generalized problem
|
|
! ... Hv=eSv, with H hermitean matrix, S overlap matrix.
|
|
! ... On output both matrix are unchanged
|
|
!
|
|
! ... Parallel version, with full data distribution
|
|
!
|
|
USE laxlib_parallel_include
|
|
USE laxlib_descriptor, ONLY : la_descriptor, laxlib_intarray_to_desc
|
|
USE laxlib_processors_grid, ONLY : ortho_parent_comm
|
|
#if defined __SCALAPACK
|
|
USE laxlib_processors_grid, ONLY : ortho_cntx, me_blacs, np_ortho, me_ortho, ortho_comm
|
|
USE zhpev_module, ONLY : pzheevd_drv
|
|
#endif
|
|
!
|
|
IMPLICIT NONE
|
|
!
|
|
INCLUDE 'laxlib_kinds.fh'
|
|
include 'laxlib_param.fh'
|
|
include 'laxlib_mid.fh'
|
|
include 'laxlib_low.fh'
|
|
!
|
|
INTEGER, INTENT(IN) :: n, ldh
|
|
! dimension of the matrix to be diagonalized
|
|
! leading dimension of h, as declared in the calling pgm unit
|
|
COMPLEX(DP), INTENT(INOUT) :: h(ldh,ldh), s(ldh,ldh)
|
|
! actually intent(in) but compilers don't know and complain
|
|
! matrix to be diagonalized
|
|
! overlap matrix
|
|
REAL(DP), INTENT(OUT) :: e(n)
|
|
! eigenvalues
|
|
COMPLEX(DP), INTENT(OUT) :: v(ldh,ldh)
|
|
! eigenvectors (column-wise)
|
|
INTEGER, INTENT(IN) :: idesc(LAX_DESC_SIZE)
|
|
!
|
|
TYPE(la_descriptor) :: desc
|
|
!
|
|
INTEGER, PARAMETER :: root = 0
|
|
INTEGER :: nx, info
|
|
#if defined __SCALAPACK
|
|
INTEGER :: descsca( 16 )
|
|
#endif
|
|
! local block size
|
|
COMPLEX(DP), ALLOCATABLE :: ss(:,:), hh(:,:), tt(:,:)
|
|
! work space used only in parallel diagonalization
|
|
!
|
|
! ... input s and h are copied so that they are not destroyed
|
|
!
|
|
CALL start_clock( 'cdiaghg' )
|
|
!
|
|
CALL laxlib_intarray_to_desc(desc,idesc)
|
|
!
|
|
IF( desc%active_node > 0 ) THEN
|
|
!
|
|
nx = desc%nrcx
|
|
!
|
|
IF( nx /= ldh ) &
|
|
CALL lax_error__(" pcdiaghg ", " inconsistent leading dimension ", ldh )
|
|
!
|
|
ALLOCATE( hh( nx, nx ) )
|
|
ALLOCATE( ss( nx, nx ) )
|
|
!
|
|
hh(1:nx,1:nx) = h(1:nx,1:nx)
|
|
ss(1:nx,1:nx) = s(1:nx,1:nx)
|
|
!
|
|
END IF
|
|
|
|
CALL start_clock( 'cdiaghg:choldc' )
|
|
!
|
|
! ... Cholesky decomposition of sl ( L is stored in sl )
|
|
!
|
|
IF( desc%active_node > 0 ) THEN
|
|
!
|
|
#if defined __SCALAPACK
|
|
CALL descinit( descsca, n, n, desc%nrcx, desc%nrcx, 0, 0, ortho_cntx, SIZE( ss, 1 ) , info )
|
|
!
|
|
IF( info /= 0 ) CALL lax_error__( ' cdiaghg ', ' desckinit ', ABS( info ) )
|
|
#endif
|
|
!
|
|
#if defined __SCALAPACK
|
|
|
|
CALL pzpotrf( 'L', n, ss, 1, 1, descsca, info )
|
|
|
|
IF( info /= 0 ) CALL lax_error__( ' cdiaghg ', ' problems computing cholesky ', ABS( info ) )
|
|
#else
|
|
CALL laxlib_pzpotrf( ss, nx, n, idesc )
|
|
#endif
|
|
!
|
|
END IF
|
|
!
|
|
CALL stop_clock( 'cdiaghg:choldc' )
|
|
!
|
|
! ... L is inverted ( sl = L^-1 )
|
|
!
|
|
CALL start_clock( 'cdiaghg:inversion' )
|
|
!
|
|
IF( desc%active_node > 0 ) THEN
|
|
!
|
|
#if defined __SCALAPACK
|
|
!CALL clear_upper_tr( ss )
|
|
! set to zero the upper triangle of ss
|
|
!
|
|
CALL sqr_setmat( 'U', n, ZERO, ss, size(ss,1), idesc )
|
|
!
|
|
CALL pztrtri( 'L', 'N', n, ss, 1, 1, descsca, info )
|
|
!
|
|
IF( info /= 0 ) CALL lax_error__( ' cdiaghg ', ' problems computing inverse ', ABS( info ) )
|
|
#else
|
|
CALL laxlib_pztrtri( ss, nx, n, idesc )
|
|
#endif
|
|
!
|
|
END IF
|
|
!
|
|
CALL stop_clock( 'cdiaghg:inversion' )
|
|
!
|
|
! ... vl = L^-1*H
|
|
!
|
|
CALL start_clock( 'cdiaghg:paragemm' )
|
|
!
|
|
IF( desc%active_node > 0 ) THEN
|
|
!
|
|
CALL sqr_mm_cannon( 'N', 'N', n, ONE, ss, nx, hh, nx, ZERO, v, nx, idesc )
|
|
!
|
|
END IF
|
|
!
|
|
! ... hl = ( L^-1*H )*(L^-1)^T
|
|
!
|
|
IF( desc%active_node > 0 ) THEN
|
|
!
|
|
CALL sqr_mm_cannon( 'N', 'C', n, ONE, v, nx, ss, nx, ZERO, hh, nx, idesc )
|
|
!
|
|
! ensure that "hh" is really Hermitian, it is sufficient to set the diagonal
|
|
! properly, because only the lower triangle of hh will be used
|
|
!
|
|
CALL sqr_setmat( 'H', n, ZERO, hh, size(hh,1), idesc )
|
|
!
|
|
END IF
|
|
!
|
|
CALL stop_clock( 'cdiaghg:paragemm' )
|
|
!
|
|
!
|
|
IF ( desc%active_node > 0 ) THEN
|
|
!
|
|
#ifdef TEST_DIAG
|
|
CALL test_drv_begin()
|
|
#endif
|
|
|
|
#if defined(__SCALAPACK)
|
|
!
|
|
CALL pzheevd_drv( .true., n, desc%nrcx, hh, e, ortho_cntx, ortho_comm )
|
|
!
|
|
#else
|
|
!
|
|
CALL laxlib_pzheevd( .true., n, idesc, hh, SIZE( hh, 1 ), e )
|
|
!
|
|
#endif
|
|
!
|
|
#ifdef TEST_DIAG
|
|
CALL test_drv_end()
|
|
#endif
|
|
!
|
|
END IF
|
|
!
|
|
! ... v = (L^T)^-1 v
|
|
!
|
|
CALL start_clock( 'cdiaghg:paragemm' )
|
|
!
|
|
IF ( desc%active_node > 0 ) THEN
|
|
!
|
|
CALL sqr_mm_cannon( 'C', 'N', n, ONE, ss, nx, hh, nx, ZERO, v, nx, idesc )
|
|
!
|
|
END IF
|
|
!
|
|
#if defined __MPI
|
|
CALL MPI_BCAST( e, SIZE(e), MPI_DOUBLE_PRECISION, root, ortho_parent_comm, info )
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'pcdiaghg', 'error broadcasting array e', ABS( info ) )
|
|
#endif
|
|
!
|
|
CALL stop_clock( 'cdiaghg:paragemm' )
|
|
!
|
|
IF ( desc%active_node > 0 ) THEN
|
|
DEALLOCATE( ss, hh )
|
|
END IF
|
|
!
|
|
CALL stop_clock( 'cdiaghg' )
|
|
!
|
|
RETURN
|
|
!
|
|
CONTAINS
|
|
!
|
|
SUBROUTINE test_drv_begin()
|
|
ALLOCATE( tt( n, n ) )
|
|
CALL laxlib_zsqmcll_x( n, hh, nx, tt, n, desc, desc%comm )
|
|
RETURN
|
|
END SUBROUTINE test_drv_begin
|
|
!
|
|
SUBROUTINE test_drv_end()
|
|
!
|
|
INTEGER :: i, j, k
|
|
COMPLEX(DP), ALLOCATABLE :: diag(:,:)
|
|
!
|
|
IF( desc%myc == 0 .AND. desc%myr == 0 ) THEN
|
|
|
|
write( 100, fmt="(A20,2D18.10)" ) ' e code = ', e( 1 ), e( n )
|
|
ALLOCATE( diag( n*(n+1)/2, 1 ) )
|
|
k = 1
|
|
!write( 100, fmt="(I5)" ) n
|
|
DO j = 1, n
|
|
DO i = j, n
|
|
diag( k, 1 ) = tt( i, j )
|
|
!write( 100, fmt="(2I5,2D18.10)" ) i, j, tt( i, j )
|
|
k = k + 1
|
|
END DO
|
|
END DO
|
|
call zhpev_drv( 'V', 'L', N, diag(:,1), e, tt, n )
|
|
write( 100, fmt="(A20,2D18.10)" ) ' e test = ', e( 1 ), e( n )
|
|
!write( 100, * ) 'eigenvalues and eigenvectors'
|
|
DO j = 1, n
|
|
!write( 100, fmt="(1I5,1D18.10,A)" ) j, e( j )
|
|
DO i = 1, n
|
|
!write( 100, fmt="(2I5,2D18.10)" ) i, j, tt( i, j )
|
|
END DO
|
|
END DO
|
|
close(100)
|
|
DEALLOCATE( diag )
|
|
END IF
|
|
#if defined __MPI
|
|
CALL MPI_BCAST( tt, SIZE(tt), MPI_DOUBLE_COMPLEX, 0, desc%comm, info )
|
|
IF ( info /= 0 ) &
|
|
CALL lax_error__( 'test_drv_end', 'error broadcasting array e', ABS( info ) )
|
|
#endif
|
|
CALL laxlib_zsqmdst_x( n, tt, n, hh, nx, desc )
|
|
DEALLOCATE( tt )
|
|
CALL lax_error__('cdiaghg','stop serial',1)
|
|
RETURN
|
|
END SUBROUTINE test_drv_end
|
|
!
|
|
END SUBROUTINE laxlib_pcdiaghg
|
|
!
|