scnc
/
quantum-espresso
mirror of https://gitlab.com/QEF/q-e.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
quantum-espresso/Modules/environment.f90

399 lines
13 KiB
Fortran
Raw Blame History

!
! Copyright (C) 2002-2011 Quantum ESPRESSO groups
! 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 .
!
! Uncomment next line to print compilation info. BEWARE: may occasionally
! give compilation errors due to lines too long if paths are very long
!#define __HAVE_CONFIG_INFO
#if defined(HAVE_GITREV)
#include "git-rev.h"
#endif
!
!==-----------------------------------------------------------------------==!
MODULE environment
!==-----------------------------------------------------------------------==!
!! Environment management.
USE kinds, ONLY: DP
USE io_files, ONLY: crash_file, nd_nmbr
USE io_global, ONLY: stdout, meta_ionode
USE mp_world, ONLY: nproc, nnode
USE mp_images, ONLY: me_image, my_image_id, root_image, nimage, &
nproc_image
USE mp_pools, ONLY: npool
USE mp_bands, ONLY: ntask_groups, nproc_bgrp, nbgrp, nyfft
USE global_version, ONLY: version_number
USE fox_init_module, ONLY: fox_init
USE command_line_options, ONLY : nmany_
USE clib_wrappers, ONLY : get_mem_avail
#if defined(__HDF5)
USE qeh5_base_module, ONLY: initialize_hdf5, finalize_hdf5
#endif
IMPLICIT NONE
! ... title of the simulation
CHARACTER(LEN=75) :: title
SAVE
PRIVATE
PUBLIC :: environment_start
PUBLIC :: environment_end
PUBLIC :: opening_message
PUBLIC :: compilation_info
PUBLIC :: parallel_info
PUBLIC :: print_cuda_info
!==-----------------------------------------------------------------------==!
CONTAINS
!==-----------------------------------------------------------------------==!
SUBROUTINE environment_start( code )
CHARACTER(LEN=*), INTENT(IN) :: code
LOGICAL :: exst, debug = .false.
CHARACTER(LEN=80) :: code_version, uname
CHARACTER(LEN=6), EXTERNAL :: int_to_char
CHARACTER(LEN=3) :: env_maxdepth
INTEGER :: ios, crashunit, max_depth
! ... The Intel compiler allocates a lot of stack space
! ... Stack limit is often small, thus causing SIGSEGV and crash
! ... One may use "ulimit -s unlimited" but it doesn't always work
! ... The following call does the same and always works
!
#if defined(__INTEL_COMPILER)
CALL remove_stack_limit ( )
#endif
! ... use ".FALSE." to disable all clocks except the total cpu time clock
! ... use ".TRUE." to enable clocks
#if defined(__TRACE)
CALL get_environment_variable('ESPRESSO_MAX_DEPTH', env_maxdepth)
IF (env_maxdepth .NE. ' ') THEN
READ(env_maxdepth,'(I3)',iostat=ios) max_depth
IF (ios == 0 ) CALL set_trace_max_depth( max_depth )
END IF
#endif
CALL init_clocks(.TRUE.)
CALL start_clock( TRIM(code) )
code_version = TRIM (code) // " v." // TRIM (version_number)
! ... for compatibility with PWSCF
#if defined(__MPI)
nd_nmbr = TRIM ( int_to_char( me_image+1 ))
#else
nd_nmbr = ' '
#endif
IF( meta_ionode ) THEN
! ... search for file CRASH and delete it
INQUIRE( FILE=TRIM(crash_file), EXIST=exst )
IF( exst ) THEN
OPEN( NEWUNIT=crashunit, FILE=TRIM(crash_file), STATUS='OLD',IOSTAT=ios )
IF (ios==0) THEN
CLOSE( UNIT=crashunit, STATUS='DELETE', IOSTAT=ios )
ELSE
WRITE(stdout,'(5x,"Remark: CRASH file could not be deleted")')
END IF
END IF
ELSE
! ... one processor per image (other than meta_ionode)
! ... or, for debugging purposes, all processors,
! ... open their own standard output file
!#define DEBUG
#if defined(DEBUG)
debug = .true.
#endif
IF (me_image == root_image .OR. debug ) THEN
uname = 'out.' // trim(int_to_char( my_image_id )) // '_' // &
trim(int_to_char( me_image))
OPEN ( unit = stdout, file = TRIM(uname),status='unknown')
ELSE
#if defined(_WIN32)
OPEN ( unit = stdout, file='NUL:', status='unknown' )
#else
OPEN ( unit = stdout, file='/dev/null', status='unknown' )
#endif
END IF
END IF
!
CALL opening_message( code_version )
CALL compilation_info ( )
#if defined(__MPI)
CALL parallel_info ( code )
#else
CALL serial_info()
#endif
CALL fox_init()
#if defined(__HDF5)
CALL initialize_hdf5()
#endif
!
WRITE(stdout,'(5x, I0, A, A)') get_mem_avail()/1024, &
&" MiB available memory on the printing compute node ", &
&"when the environment starts"
WRITE(stdout, *)
END SUBROUTINE environment_start
!==-----------------------------------------------------------------------==!
SUBROUTINE environment_end( code )
CHARACTER(LEN=*), INTENT(IN) :: code
#if defined(_HDF5)
CALL finalize_hdf5()
#endif
IF ( meta_ionode ) WRITE( stdout, * )
CALL stop_clock( TRIM(code) )
CALL print_clock( TRIM(code) )
CALL closing_message( )
IF( meta_ionode ) THEN
WRITE( stdout,'(A)') ' JOB DONE.'
WRITE( stdout,3335)
END IF
3335 FORMAT('=',78('-'),'=')
FLUSH(stdout)
RETURN
END SUBROUTINE environment_end
!==-----------------------------------------------------------------------==!
SUBROUTINE opening_message( code_version )
CHARACTER(LEN=*), INTENT(IN) :: code_version
CHARACTER(LEN=9) :: cdate, ctime
CALL date_and_tim( cdate, ctime )
!
WRITE( stdout, '(/5X,"Program ",A," starts on ",A9," at ",A9)' ) &
TRIM(code_version), cdate, ctime
#if defined(HAVE_GITREV)
WRITE( stdout, '(8X, "Git branch: ", A)' ) &
GIT_BRANCH_RAW
WRITE( stdout, '(8X, "Last git commit: ", A)' ) &
GIT_HASH_RAW
WRITE( stdout, '(8X, "Last git commit date: ", A)' ) &
GIT_COMMIT_LAST_CHANGED_RAW
WRITE( stdout, '(8X, "Last git commit subject: ", A)' ) &
GIT_COMMIT_SUBJECT_RAW
#endif
!
WRITE( stdout, '(/5X,"This program is part of the open-source Quantum ",&
& "ESPRESSO suite", &
&/5X,"for quantum simulation of materials; please cite", &
&/9X,"""P. Giannozzi et al., J. Phys.:Condens. Matter 21 ",&
& "395502 (2009);", &
&/9X,"""P. Giannozzi et al., J. Phys.:Condens. Matter 29 ",&
& "465901 (2017);", &
&/9X,"""P. Giannozzi et al., J. Chem. Phys. 152 ",&
& "154105 (2020);", &
&/9X," URL http://www.quantum-espresso.org"", ", &
&/5X,"in publications or presentations arising from this work. More details at",&
&/5x,"http://www.quantum-espresso.org/quote")' )
RETURN
END SUBROUTINE opening_message
!==-----------------------------------------------------------------------==!
SUBROUTINE closing_message( )
CHARACTER(LEN=9) :: cdate, ctime
CHARACTER(LEN=80) :: time_str
CALL date_and_tim( cdate, ctime )
time_str = 'This run was terminated on: ' // ctime // ' ' // cdate
IF( meta_ionode ) THEN
WRITE( stdout,*)
WRITE( stdout,3334) time_str
WRITE( stdout,3335)
END IF
3334 FORMAT(3X,A60,/)
3335 FORMAT('=',78('-'),'=')
RETURN
END SUBROUTINE closing_message
!==-----------------------------------------------------------------------==!
SUBROUTINE parallel_info ( code )
!
CHARACTER(LEN=*), INTENT(IN) :: code
#if defined(_OPENMP)
INTEGER, EXTERNAL :: omp_get_max_threads
!
WRITE( stdout, '(/5X,"Parallel version (MPI & OpenMP), running on ",&
&I7," processor cores")' ) nproc * omp_get_max_threads()
!
WRITE( stdout, '(5X,"Number of MPI processes: ",I7)' ) nproc
!
WRITE( stdout, '(5X,"Threads/MPI process: ",I7)' ) &
omp_get_max_threads()
#else
WRITE( stdout, '(/5X,"Parallel version (MPI), running on ",&
&I5," processors")' ) nproc
#endif
!
WRITE( stdout, '(/5X,"MPI processes distributed on ",&
&I5," nodes")' ) nnode
IF ( nimage > 1 ) WRITE( stdout, &
'(5X,"path-images division: nimage = ",I7)' ) nimage
IF ( npool > 1 ) WRITE( stdout, &
'(5X,"K-points division: npool = ",I7)' ) npool
IF ( nbgrp > 1 ) WRITE( stdout, &
'(5X,"band groups division: nbgrp = ",I7)' ) nbgrp
IF ( nproc_bgrp > 1 ) WRITE( stdout, &
'(5X,"R & G space division: proc/nbgrp/npool/nimage = ",I7)' ) nproc_bgrp
IF ( nyfft > 1 ) WRITE( stdout, &
'(5X,"wavefunctions fft division: Y-proc x Z-proc = ",2I7)' ) &
nyfft, nproc_bgrp / nyfft
IF ( ntask_groups > 1 ) WRITE( stdout, &
'(5X,"wavefunctions fft division: task group distribution",/,34X,"#TG x Z-proc = ",2I7)' ) &
ntask_groups, nproc_bgrp / ntask_groups
IF ( nmany_ > 1) WRITE( stdout, '(5X,"FFT bands division: nmany = ",I7)' ) nmany_
!
END SUBROUTINE parallel_info
!==-----------------------------------------------------------------------==!
SUBROUTINE serial_info ( )
!
#if defined(_OPENMP)
INTEGER, EXTERNAL :: omp_get_max_threads
#endif
!
#if defined(_OPENMP)
WRITE( stdout, '(/5X,"Serial multi-threaded version, running on ",&
&I4," processor cores")' ) omp_get_max_threads()
!
#else
WRITE( stdout, '(/5X,"Serial version")' )
#endif
!
END SUBROUTINE serial_info
!==-----------------------------------------------------------------------==!
SUBROUTINE compilation_info ( )
!
! code borrowed by WanT - prints architecture / compilation details
!
#if defined(__HAVE_CONFIG_INFO)
#include "configure.h"
! #include "build_date.h"
!
!WRITE( stdout, "(2x,' BUILT :',4x,a)" ) TRIM( ADJUSTL( &
!__CONF_BUILD_DATE ))
WRITE( stdout, * )
! note: if any preprocessed variables __CONF_* exceeds 128 characters,
! the compilation may give error because the line exceeds 132 characters
WRITE( stdout, "(2x,' ARCH :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_ARCH))
WRITE( stdout, "(2x,' CC :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_CC))
WRITE( stdout, "(2x,' CPP :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_CPP))
WRITE( stdout, "(2x,' F90 :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_MPIF90))
WRITE( stdout, "(2x,' F77 :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_F77))
WRITE( stdout, "(2x,' DFLAGS :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_DFLAGS))
WRITE( stdout, "(2x,' BLAS LIBS :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_BLAS_LIBS))
WRITE( stdout, "(2x,' LAPACK LIBS :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_LAPACK_LIBS))
WRITE( stdout, "(2x,' FFT LIBS :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_FFT_LIBS))
WRITE( stdout, "(2x,' MASS LIBS :',4x,a)" ) TRIM( ADJUSTL( &
__CONF_MASS_LIBS))
!
#endif
END SUBROUTINE compilation_info
!
!-----------------------------------------------------------------------
SUBROUTINE print_cuda_info(check_use_gpu)
!-----------------------------------------------------------------------
!
USE io_global, ONLY : stdout
USE control_flags, ONLY : use_gpu_=> use_gpu, iverbosity
USE mp_world, ONLY : nnode, nproc
USE mp, ONLY : mp_sum, mp_max
#if defined(__CUDA)
USE cudafor
#endif
!
IMPLICIT NONE
!
LOGICAL, OPTIONAL,INTENT(IN) :: check_use_gpu
!! if present and trues the internal variable use_gpu is checked
#if defined (__CUDA)
INTEGER :: idev, ndev, ierr
TYPE (cudaDeviceProp) :: prop
LOGICAL :: use_gpu = .TRUE.
!
IF ( PRESENT(check_use_gpu) ) THEN
IF (check_use_gpu) use_gpu = use_gpu_
END IF
IF (use_gpu) THEN
WRITE( stdout, '(/,5X,"GPU acceleration is ACTIVE.")' )
#if defined(__GPU_MPI)
WRITE( stdout, '(5x, "GPU-aware MPI enabled")')
#endif
WRITE( stdout, '()' )
ELSE
WRITE( stdout, '(/,5X,"GPU acceleration is NOT ACTIVE.",/)' )
END IF
!
ierr = cudaGetDevice( idev )
IF (ierr /= 0) CALL errore('summary', 'cannot get device id', ierr)
ierr = cudaGetDeviceCount( ndev )
IF (ierr /= 0) CALL errore('summary', 'cannot get device count', ierr)
!
! User friendly, approximated warning.
! In order to get this done right, one needs an intra_node communicator
!
IF (nproc > ndev * nnode * 2) &
CALL infomsg('print_cuda_info', &
'High GPU oversubscription detected. Are you sure this is what you want?')
!
! Verbose information for advanced users
IF (iverbosity > 0) THEN
WRITE( stdout, '(/,5X,"GPU used by master process:",/)' )
! Device info taken from
! https://devblogs.nvidia.com/how-query-device-properties-and-handle-errors-cuda-fortran/
ierr = cudaGetDeviceProperties(prop, idev)
WRITE(stdout,"(5X,' Device Number: ',i0)") idev
WRITE(stdout,"(5X,' Device name: ',a)") trim(prop%name)
WRITE(stdout,"(5X,' Compute capability : ',i0, i0)") prop%major, prop%minor
WRITE(stdout,"(5X,' Ratio of single to double precision performance : ',i0)") prop%singleToDoublePrecisionPerfRatio
WRITE(stdout,"(5X,' Memory Clock Rate (KHz): ', i0)") &
prop%memoryClockRate
WRITE(stdout,"(5X,' Memory Bus Width (bits): ', i0)") &
prop%memoryBusWidth
WRITE(stdout,"(5X,' Peak Memory Bandwidth (GB/s): ', f6.2)") &
2.0*prop%memoryClockRate*(prop%memoryBusWidth/8)/10.0**6
END IF
!
#endif
!
END SUBROUTINE print_cuda_info
!==-----------------------------------------------------------------------==!
END MODULE environment
!==-----------------------------------------------------------------------==!
Reference in New Issue View Git Blame Copy Permalink