mirror of https://gitlab.com/QEF/q-e.git
139 lines
4.9 KiB
Fortran
139 lines
4.9 KiB
Fortran
!
|
|
! Copyright (C) 2002-2011 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 .
|
|
!
|
|
|
|
#if defined(__CUDA)
|
|
#define PINMEM ,PINNED
|
|
#else
|
|
#define PINMEM
|
|
#endif
|
|
|
|
!=----------------------------------------------------------------------------=!
|
|
MODULE wavefunctions
|
|
!=----------------------------------------------------------------------------=!
|
|
!! Wavefunction arrays.
|
|
!
|
|
USE kinds, ONLY : DP
|
|
#if defined (__CUDA)
|
|
USE cudafor
|
|
#elif defined(__OPENMP_GPU)
|
|
USE omp_lib
|
|
#endif
|
|
|
|
IMPLICIT NONE
|
|
SAVE
|
|
|
|
!
|
|
!civn
|
|
! COMPLEX(DP), ALLOCATABLE, TARGET :: &
|
|
#if defined(__CUDA)
|
|
COMPLEX(DP), POINTER, CONTIGUOUS :: evc(:,:)
|
|
#else
|
|
COMPLEX(DP), ALLOCATABLE, TARGET :: evc(:,:)
|
|
#endif
|
|
!! wavefunctions in the PW basis set.
|
|
!! noncolinear case: first index is a combined PW + spin index
|
|
!
|
|
#if defined(__OPENMP_GPU)
|
|
! ... tools to pin evc with omp
|
|
INTEGER, PARAMETER :: ntraits = 1
|
|
INTEGER(omp_allocator_handle_kind) :: pinned_alloc
|
|
TYPE(omp_alloctrait) :: traits(ntraits) =[omp_alloctrait(omp_atk_pinned,1)]
|
|
#endif
|
|
!
|
|
COMPLEX(DP) , ALLOCATABLE, TARGET :: psic(:)
|
|
!! additional memory for FFT
|
|
COMPLEX(DP) , ALLOCATABLE, TARGET :: psic_nc(:,:)
|
|
!! additional memory for FFT for the noncolinear case
|
|
COMPLEX(DP) , ALLOCATABLE, TARGET :: psicg(:)
|
|
!! additional memory again for FFT
|
|
!
|
|
! electronic wave functions, CPV code
|
|
! distributed over gvector and bands
|
|
!
|
|
!dir$ attributes align: 4096 :: c0_bgrp, cm_bgrp, phi
|
|
COMPLEX(DP), ALLOCATABLE PINMEM :: c0_bgrp(:,:) ! wave functions at time t
|
|
COMPLEX(DP), ALLOCATABLE PINMEM :: cm_bgrp(:,:) ! wave functions at time t-delta t
|
|
COMPLEX(DP), ALLOCATABLE :: phi(:,:) ! |phi> = s'|c0> = |c0> + sum q_ij |i><j|c0>
|
|
COMPLEX(DP), ALLOCATABLE :: c0_d(:,:) ! wave functions at time t
|
|
COMPLEX(DP), ALLOCATABLE :: cm_d(:,:) ! wave functions at time t-delta t
|
|
#if defined (__CUDA)
|
|
ATTRIBUTES(DEVICE) :: c0_d, cm_d, phi
|
|
#endif
|
|
! for hybrid functionals in CP with Wannier functions
|
|
COMPLEX(DP), ALLOCATABLE :: cv0(:,:) ! Lingzhu Kong
|
|
|
|
CONTAINS
|
|
|
|
SUBROUTINE deallocate_wavefunctions
|
|
USE control_flags, ONLY : use_gpu
|
|
INTEGER :: istat
|
|
IF( ALLOCATED( cv0) ) DEALLOCATE( cv0) ! Lingzhu Kong
|
|
IF( ALLOCATED( c0_bgrp ) ) DEALLOCATE( c0_bgrp )
|
|
IF( ALLOCATED( cm_bgrp ) ) DEALLOCATE( cm_bgrp )
|
|
IF( ALLOCATED( phi ) ) DEALLOCATE( phi )
|
|
IF( ALLOCATED( psic_nc ) ) THEN
|
|
#if defined(__OPENMP_GPU)
|
|
!$omp target exit data map(delete:psic_nc)
|
|
#endif
|
|
DEALLOCATE( psic_nc )
|
|
ENDIF
|
|
IF( ALLOCATED( psic ) ) THEN
|
|
#if defined(__OPENMP_GPU)
|
|
!$omp target exit data map(delete:psic)
|
|
#endif
|
|
DEALLOCATE( psic )
|
|
ENDIF
|
|
IF( ALLOCATED( psicg ) ) DEALLOCATE( psicg )
|
|
#if defined(__CUDA)
|
|
!$acc exit data delete(evc)
|
|
IF(use_gpu) istat = cudaHostUnregister(C_LOC(evc(1,1)))
|
|
#endif
|
|
IF( ALLOCATED( evc ) ) DEALLOCATE( evc )
|
|
#if defined (__CUDA)
|
|
IF( ALLOCATED( c0_d ) ) DEALLOCATE( c0_d )
|
|
IF( ALLOCATED( cm_d ) ) DEALLOCATE( cm_d )
|
|
#endif
|
|
END SUBROUTINE deallocate_wavefunctions
|
|
|
|
SUBROUTINE allocate_cp_wavefunctions( ngw, nbspx, vnbsp, lwfpbe0nscf )
|
|
INTEGER, INTENT(IN) :: ngw, nbspx, vnbsp
|
|
LOGICAL, INTENT(IN) :: lwfpbe0nscf
|
|
INTEGER :: ierr
|
|
ALLOCATE( c0_bgrp( ngw, nbspx ), STAT=ierr )
|
|
IF( ierr /= 0 ) &
|
|
CALL errore( ' allocate_cp_wavefunctions ', ' allocating on CPU ', ABS( ierr ) )
|
|
c0_bgrp = (0_DP,0_DP)
|
|
ALLOCATE( cm_bgrp( ngw, nbspx ), STAT=ierr )
|
|
IF( ierr /= 0 ) &
|
|
CALL errore( ' allocate_cp_wavefunctions ', ' allocating on CPU ', ABS( ierr ) )
|
|
cm_bgrp = (0_DP,0_DP)
|
|
ALLOCATE( phi( ngw, nbspx ), STAT=ierr )
|
|
IF( ierr /= 0 ) &
|
|
CALL errore( ' allocate_cp_wavefunctions ', ' allocating on CPU ', ABS( ierr ) )
|
|
phi = (0_DP,0_DP)
|
|
IF(lwfpbe0nscf) THEN
|
|
ALLOCATE(cv0( ngw, vnbsp ), STAT=ierr ) ! Lingzhu Kong
|
|
IF( ierr /= 0 ) &
|
|
CALL errore( ' allocate_cp_wavefunctions ', ' allocating on CPU ', ABS( ierr ) )
|
|
cv0 = (0_DP,0_DP)
|
|
END IF
|
|
#if defined (__CUDA)
|
|
ALLOCATE( c0_d( ngw, nbspx ), STAT=ierr )
|
|
IF( ierr /= 0 ) &
|
|
CALL errore( ' allocate_cp_wavefunctions ', ' allocating on GPU ', ABS( ierr ) )
|
|
ALLOCATE( cm_d( ngw, nbspx ), STAT=ierr )
|
|
IF( ierr /= 0 ) &
|
|
CALL errore( ' allocate_cp_wavefunctions ', ' allocating on GPU ', ABS( ierr ) )
|
|
#endif
|
|
END SUBROUTINE
|
|
|
|
!=----------------------------------------------------------------------------=!
|
|
END MODULE wavefunctions
|
|
!=----------------------------------------------------------------------------=!
|
|
|