LCOV - code coverage report
Current view: top level - test/Fortran/elpa1/legacy_interface - legacy_complex.F90 (source / functions) Hit Total Coverage
Test: coverage_50ab7a7628bba174fc62cee3ab72b26e81f87fe5.info Lines: 63 79 79.7 %
Date: 2018-01-10 09:29:53 Functions: 2 2 100.0 %

          Line data    Source code
       1             : !    This file is part of ELPA.
       2             : !
       3             : !    The ELPA library was originally created by the ELPA consortium,
       4             : !    consisting of the following organizations:
       5             : !
       6             : !    - Max Planck Computing and Data Facility (MPCDF), formerly known as
       7             : !      Rechenzentrum Garching der Max-Planck-Gesellschaft (RZG),
       8             : !    - Bergische Universität Wuppertal, Lehrstuhl für angewandte
       9             : !      Informatik,
      10             : !    - Technische Universität München, Lehrstuhl für Informatik mit
      11             : !      Schwerpunkt Wissenschaftliches Rechnen ,
      12             : !    - Fritz-Haber-Institut, Berlin, Abt. Theorie,
      13             : !    - Max-Plack-Institut für Mathematik in den Naturwissenschaften,
      14             : !      Leipzig, Abt. Komplexe Strukutren in Biologie und Kognition,
      15             : !      and
      16             : !    - IBM Deutschland GmbH
      17             : !
      18             : !
      19             : !    More information can be found here:
      20             : !    http://elpa.mpcdf.mpg.de/
      21             : !
      22             : !    ELPA is free software: you can redistribute it and/or modify
      23             : !    it under the terms of the version 3 of the license of the
      24             : !    GNU Lesser General Public License as published by the Free
      25             : !    Software Foundation.
      26             : !
      27             : !    ELPA is distributed in the hope that it will be useful,
      28             : !    but WITHOUT ANY WARRANTY; without even the implied warranty of
      29             : !    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
      30             : !    GNU Lesser General Public License for more details.
      31             : !
      32             : !    You should have received a copy of the GNU Lesser General Public License
      33             : !    along with ELPA.  If not, see <http://www.gnu.org/licenses/>
      34             : !
      35             : !    ELPA reflects a substantial effort on the part of the original
      36             : !    ELPA consortium, and we ask you to respect the spirit of the
      37             : !    license that we chose: i.e., please contribute any changes you
      38             : !    may have back to the original ELPA library distribution, and keep
      39             : !    any derivatives of ELPA under the same license that we chose for
      40             : !    the original distribution, the GNU Lesser General Public License.
      41             : !
      42             : !
      43             : #include "config-f90.h"
      44             : !>
      45             : !> Fortran test programm to demonstrates the use of
      46             : !> ELPA 1 complex case library.
      47             : !> If "HAVE_REDIRECT" was defined at build time
      48             : !> the stdout and stderr output of each MPI task
      49             : !> can be redirected to files if the environment
      50             : !> variable "REDIRECT_ELPA_TEST_OUTPUT" is set
      51             : !> to "true".
      52             : !>
      53             : !> By calling executable [arg1] [arg2] [arg3] [arg4]
      54             : !> one can define the size (arg1), the number of
      55             : !> Eigenvectors to compute (arg2), and the blocking (arg3).
      56             : !> If these values are not set default values (500, 150, 16)
      57             : !> are choosen.
      58             : !> If these values are set the 4th argument can be
      59             : !> "output", which specifies that the EV's are written to
      60             : !> an ascii file.
      61             : !>
      62         192 : program test_complex_double_precision
      63             : 
      64             : !-------------------------------------------------------------------------------
      65             : ! Standard eigenvalue problem - COMPLEX version
      66             : !
      67             : ! This program demonstrates the use of the ELPA module
      68             : ! together with standard scalapack routines
      69             : !
      70             : ! Copyright of the original code rests with the authors inside the ELPA
      71             : ! consortium. The copyright of any additional modifications shall rest
      72             : ! with their original authors, but shall adhere to the licensing terms
      73             : ! distributed along with the original code in the file "COPYING".
      74             : !-------------------------------------------------------------------------------
      75         192 :    use elpa1
      76             :    use elpa_utilities, only : error_unit
      77             : 
      78             :    use test_util
      79             :    use test_read_input_parameters
      80             :    use test_check_correctness
      81             :    use test_setup_mpi
      82             :    use test_blacs_infrastructure
      83             :    use test_prepare_matrix
      84             : #ifdef HAVE_REDIRECT
      85             :    use test_redirect
      86             : #endif
      87             : 
      88             :    use test_output_type
      89             :    implicit none
      90             : 
      91             :    !-------------------------------------------------------------------------------
      92             :    ! Please set system size parameters below!
      93             :    ! na:   System size
      94             :    ! nev:  Number of eigenvectors to be calculated
      95             :    ! nblk: Blocking factor in block cyclic distribution
      96             :    !-------------------------------------------------------------------------------
      97             : 
      98             :    integer(kind=ik)              :: nblk
      99             :    integer(kind=ik)              :: na, nev
     100             : 
     101             :    integer(kind=ik)              :: np_rows, np_cols, na_rows, na_cols
     102             : 
     103             :    integer(kind=ik)              :: myid, nprocs, my_prow, my_pcol, mpi_comm_rows, mpi_comm_cols
     104             :    integer(kind=ik)              :: i, mpierr, my_blacs_ctxt, sc_desc(9), info, nprow, npcol
     105             : 
     106         192 :    real(kind=rk8), allocatable    :: ev(:)
     107             : 
     108         384 :    complex(kind=ck8), allocatable :: a(:,:), z(:,:), as(:,:)
     109             : 
     110             :    complex(kind=ck8), parameter   :: CZERO = (0._rk8,0.0_rk8), CONE = (1._rk8,0._rk8)
     111             : 
     112             :    integer(kind=ik)              :: STATUS
     113             : #ifdef WITH_OPENMP
     114             :    integer(kind=ik)              :: omp_get_max_threads,  required_mpi_thread_level, provided_mpi_thread_level
     115             : #endif
     116             :    type(output_t)                :: write_to_file
     117             :    logical                       :: success
     118             :    character(len=8)              :: task_suffix
     119             :    integer(kind=ik)              :: j
     120             : 
     121             : #define DOUBLE_PRECISION_COMPLEX 1
     122             : 
     123         192 :    success = .true.
     124             :    ! read input parameters if they are provided
     125         192 :    call read_input_parameters(na, nev, nblk, write_to_file)
     126             : 
     127             :    !-------------------------------------------------------------------------------
     128             :    !  MPI Initialization
     129         192 :    call setup_mpi(myid, nprocs)
     130             : 
     131         192 :    STATUS = 0
     132             : 
     133             : #define COMPLEXCASE
     134             : #define ELPA1
     135             : #include "../../elpa_print_headers.F90"
     136             : 
     137             :    !-------------------------------------------------------------------------------
     138             :    ! Selection of number of processor rows/columns
     139             :    ! We try to set up the grid square-like, i.e. start the search for possible
     140             :    ! divisors of nprocs with a number next to the square root of nprocs
     141             :    ! and decrement it until a divisor is found.
     142             : 
     143         192 :    do np_cols = NINT(SQRT(REAL(nprocs))),2,-1
     144           0 :       if(mod(nprocs,np_cols) == 0 ) exit
     145             :    enddo
     146             :    ! at the end of the above loop, nprocs is always divisible by np_cols
     147             : 
     148         192 :    np_rows = nprocs/np_cols
     149             : 
     150         192 :    if(myid==0) then
     151         128 :       print *
     152         128 :       print '(a)','Standard eigenvalue problem - ELPA1, COMPLEX version'
     153         128 :       print *
     154         128 :       print '((a,i0))', 'Matrix size: ', na 
     155         128 :       print '((a,i0))', 'Num eigenvectors: ', nev
     156         128 :       print '((a,i0))', 'Blocksize: ', nblk 
     157         128 :       print '((a,i0))', 'Num MPI proc: ', nprocs 
     158         128 :       print '((a))', 'Using gpu: NO'
     159         128 :       print '((a,i0))', 'Num gpu devices: ', 0
     160         128 :       print '((a))', 'Number type: complex'
     161         128 :       print '((a))', 'Number precision: double'
     162         128 :       print *
     163         128 :       print '(3(a,i0))','Number of processor rows=',np_rows,', cols=',np_cols,', total=',nprocs
     164         128 :       print *
     165             :    endif
     166             : 
     167             :   !-------------------------------------------------------------------------------
     168             :    ! Set up BLACS context and MPI communicators
     169             :    !
     170             :    ! The BLACS context is only necessary for using Scalapack.
     171             :    !
     172             :    ! For ELPA, the MPI communicators along rows/cols are sufficient,
     173             :    ! and the grid setup may be done in an arbitrary way as long as it is
     174             :    ! consistent (i.e. 0<=my_prow<np_rows, 0<=my_pcol<np_cols and every
     175             :    ! process has a unique (my_prow,my_pcol) pair).
     176             : 
     177             :    call set_up_blacsgrid(mpi_comm_world, np_rows, np_cols, 'C', &
     178         192 :                          my_blacs_ctxt, my_prow, my_pcol)
     179             : 
     180         192 :    if (myid==0) then
     181         128 :      print '(a)','| Past BLACS_Gridinfo.'
     182             :    end if
     183             : 
     184             :    ! All ELPA routines need MPI communicators for communicating within
     185             :    ! rows or columns of processes, these are set in elpa_get_communicators.
     186             : 
     187             :    mpierr = elpa_get_communicators(mpi_comm_world, my_prow, my_pcol, &
     188         192 :                                    mpi_comm_rows, mpi_comm_cols)
     189             : 
     190         192 :    if (myid==0) then
     191         128 :      print '(a)','| Past split communicator setup for rows and columns.'
     192             :    end if
     193             : 
     194             :    ! Determine the necessary size of the distributed matrices,
     195             :    ! we use the Scalapack tools routine NUMROC for that.
     196             : 
     197             :    call set_up_blacs_descriptor(na ,nblk, my_prow, my_pcol, np_rows, np_cols, &
     198         192 :                                 na_rows, na_cols, sc_desc, my_blacs_ctxt, info)
     199             : 
     200         192 :    if (myid==0) then
     201         128 :      print '(a)','| Past scalapack descriptor setup.'
     202             :    end if
     203             : 
     204             :    !-------------------------------------------------------------------------------
     205             :    ! Allocate matrices and set up a test matrix for the eigenvalue problem
     206             : 
     207         192 :    allocate(a (na_rows,na_cols))
     208         192 :    allocate(z (na_rows,na_cols))
     209         192 :    allocate(as(na_rows,na_cols))
     210             : 
     211         192 :    allocate(ev(na))
     212             : 
     213         192 :    call prepare_matrix_random(na, myid, sc_desc, a, z, as)
     214         192 :    elpa_print_times = .true.
     215             :    !-------------------------------------------------------------------------------
     216             :    ! Calculate eigenvalues/eigenvectors
     217             : 
     218         192 :    if (myid==0) then
     219         128 :      print '(a)','| Entering one-step ELPA solver ... '
     220         128 :      print *
     221             :    end if
     222             : #ifdef WITH_MPI
     223         128 :    call mpi_barrier(mpi_comm_world, mpierr) ! for correct timings only
     224             : #endif
     225             :    success = elpa_solve_evp_complex_1stage_double(na, nev, a, na_rows, ev, z, na_rows, nblk, &
     226         192 :                                na_cols, mpi_comm_rows, mpi_comm_cols, mpi_comm_world)
     227             : 
     228         192 :    if (.not.(success)) then
     229           0 :       write(error_unit,*) "solve_evp_complex produced an error! Aborting..."
     230             : #ifdef WITH_MPI
     231           0 :       call MPI_ABORT(mpi_comm_world, 1, mpierr)
     232             : #else
     233           0 :       call exit(1)
     234             : #endif
     235             :    endif
     236             : 
     237         192 :    if (myid==0) then
     238         128 :      print '(a)','| One-step ELPA solver complete.'
     239         128 :      print *
     240             :    end if
     241             : 
     242         192 :    if(myid == 0) print *,'Time tridiag_complex  :',time_evp_fwd
     243         192 :    if(myid == 0) print *,'Time solve_tridi      :',time_evp_solve
     244         192 :    if(myid == 0) print *,'Time trans_ev_complex :',time_evp_back
     245         192 :    if(myid == 0) print *,'Total time (sum above):',time_evp_back+time_evp_solve+time_evp_fwd
     246             : 
     247         192 :    if(write_to_file%eigenvectors) then
     248           0 :      write(unit = task_suffix, fmt = '(i8.8)') myid
     249           0 :      open(17,file="EVs_complex_out_task_"//task_suffix(1:8)//".txt",form='formatted',status='new')
     250           0 :      write(17,*) "Part of eigenvectors: na_rows=",na_rows,"of na=",na," na_cols=",na_cols," of na=",na
     251             : 
     252           0 :      do i=1,na_rows
     253           0 :        do j=1,na_cols
     254           0 :          write(17,*) "row=",i," col=",j," element of eigenvector=",z(i,j)
     255             :        enddo
     256             :      enddo
     257           0 :      close(17)
     258             :    endif
     259             : 
     260         192 :    if(write_to_file%eigenvalues) then
     261           0 :       if (myid == 0) then
     262           0 :          open(17,file="Eigenvalues_complex_out.txt",form='formatted',status='new')
     263           0 :          do i=1,na
     264           0 :             write(17,*) i,ev(i)
     265             :          enddo
     266           0 :          close(17)
     267             :       endif
     268             :    endif
     269             : 
     270             : 
     271             :    !-------------------------------------------------------------------------------
     272             :    ! Test correctness of result (using plain scalapack routines)
     273             : 
     274         192 :    status = check_correctness_evp_numeric_residuals(na, nev, as, z, ev, sc_desc, nblk, myid, np_rows, np_cols, my_prow, my_pcol)
     275             : 
     276         192 :    deallocate(a)
     277         192 :    deallocate(as)
     278             : 
     279         192 :    deallocate(z)
     280         192 :    deallocate(ev)
     281             : 
     282             : #ifdef WITH_MPI
     283         128 :    call blacs_gridexit(my_blacs_ctxt)
     284         128 :    call mpi_finalize(mpierr)
     285             : #endif
     286         192 :    call EXIT(STATUS)
     287             : end
     288             : 
     289             : !-------------------------------------------------------------------------------

Generated by: LCOV version 1.12