Fortran test for real driver interface

Makefile.am

@@ -324,6 +324,8 @@ noinst_PROGRAMS = \
   elpa2_test_complex@SUFFIX@ \
   elpa2_test_complex_default@SUFFIX@ \
   elpa2_test_complex_api@SUFFIX@ \
+  elpa2_test_complex_api@SUFFIX@ \
+  elpa_driver_real@SUFFIX@ \
   elpa1_real_toeplitz@SUFFIX@ \
   elpa1_real_transpose_multiply@SUFFIX@ \
   elpa1_complex_transpose_multiply@SUFFIX@ \
@@ -348,7 +350,6 @@ noinst_PROGRAMS += \
   elpa1_complex_cholesky_single_precision@SUFFIX@ \
   elpa1_complex_invert_trm_single_precision@SUFFIX@ \
   elpa2_test_complex_api_single_precision@SUFFIX@ \
-  elpa_driver_real_c_version_single_precision@SUFFIX@ \
   elpa_driver_complex_c_version_single_precision@SUFFIX@
 endif
 
@@ -359,9 +360,11 @@ noinst_PROGRAMS += \
   elpa2_test_real_default_single_precision@SUFFIX@ \
   elpa2_test_real_qr_single_precision@SUFFIX@ \
   elpa2_test_real_api_single_precision@SUFFIX@ \
+  elpa_driver_real_single_precision@SUFFIX@ \
   elpa1_real_transpose_multiply_single_precision@SUFFIX@ \
   elpa1_real_cholesky_single_precision@SUFFIX@ \
   elpa1_real_invert_trm_single_precision@SUFFIX@ \
+  elpa_driver_real_c_version_single_precision@SUFFIX@ \
   elpa1_real_toeplitz_single_precision@SUFFIX@
 endif
 
@@ -519,6 +522,11 @@ elpa2_test_complex_api@SUFFIX@_LDADD = $(build_lib)
 elpa2_test_complex_api@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
 EXTRA_elpa2_test_complex_api@SUFFIX@_DEPENDENCIES = test/Fortran/elpa_print_headers.X90
 
+elpa_driver_real@SUFFIX@_SOURCES = test/Fortran/test_driver_real.F90
+elpa_driver_real@SUFFIX@_LDADD = $(build_lib)
+elpa_driver_real@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
+EXTRA_elpa_driver_real@SUFFIX@_DEPENDENCIES = test/Fortran/elpa_print_headers.X90
+
 elpa2_print_kernels@SUFFIX@_SOURCES = src/elpa2_print_kernels.F90
 elpa2_print_kernels@SUFFIX@_LDADD = $(build_lib)
 elpa2_print_kernels@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
@@ -535,6 +543,11 @@ elpa2_test_real_single_precision@SUFFIX@_LDADD = $(build_lib)
 elpa2_test_real_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
 EXTRA_elpa2_test_real_single_precision@SUFFIX@_DEPENDENCIES = test/Fortran/elpa_print_headers.X90
 
+elpa_driver_real_single_precision@SUFFIX@_SOURCES = test/Fortran/test_driver_real_single.F90
+elpa_driver_real_single_precision@SUFFIX@_LDADD = $(build_lib)
+elpa_driver_real_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
+EXTRA_elpa_driver_real_single_precision@SUFFIX@_DEPENDENCIES = test/Fortran/elpa_print_headers.X90
+
 elpa1_real_toeplitz_single_precision@SUFFIX@_SOURCES = test/Fortran/test_toeplitz_single.F90
 elpa1_real_toeplitz_single_precision@SUFFIX@_LDADD = $(build_lib)
 elpa1_real_toeplitz_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
@@ -545,6 +558,11 @@ elpa1_real_transpose_multiply_single_precision@SUFFIX@_LDADD = $(build_lib)
 elpa1_real_transpose_multiply_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
 EXTRA_elpa1_real_transpose_multiply_single_precision@SUFFIX@_DEPENDENCIES = test/Fortran/elpa_print_headers.X90
 
+elpa_driver_real_c_version_single_precision@SUFFIX@_SOURCES = test/C/elpa_driver_real_c_version_single.c
+elpa_driver_real_c_version_single_precision@SUFFIX@_LDADD = $(build_lib) $(FCLIBS)
+elpa_driver_real_c_version_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
+EXTRA_elpa_driver_real_c_version_single_precision@SUFFIX@_DEPENDENCIES = test/Fortran/elpa_print_headers.X90
+
 elpa1_real_cholesky_single_precision@SUFFIX@_SOURCES = test/Fortran/test_cholesky_real_single.F90
 elpa1_real_cholesky_single_precision@SUFFIX@_LDADD = $(build_lib)
 elpa1_real_cholesky_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
@@ -579,11 +597,6 @@ elpa1_test_complex_single_precision@SUFFIX@_LDADD = $(build_lib)
 elpa1_test_complex_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
 EXTRA_elpa1_test_complex_single_precision@SUFFIX@_DEPENDENCIES = test/Fortran/elpa_print_headers.X90
 
-elpa_driver_real_c_version_single_precision@SUFFIX@_SOURCES = test/C/elpa_driver_real_c_version_single.c
-elpa_driver_real_c_version_single_precision@SUFFIX@_LDADD = $(build_lib) $(FCLIBS)
-elpa_driver_real_c_version_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
-EXTRA_elpa_driver_real_c_version_single_precision@SUFFIX@_DEPENDENCIES = test/Fortran/elpa_print_headers.X90
-
 elpa_driver_complex_c_version_single_precision@SUFFIX@_SOURCES = test/C/elpa_driver_complex_c_version_single.c
 elpa_driver_complex_c_version_single_precision@SUFFIX@_LDADD = $(build_lib) $(FCLIBS)
 elpa_driver_complex_c_version_single_precision@SUFFIX@_FCFLAGS = $(AM_FCFLAGS) @FC_MODOUT@private_modules @FC_MODINC@private_modules
@@ -659,6 +672,7 @@ check_SCRIPTS = \
   elpa2_test_complex_default@SUFFIX@.sh \
   elpa2_test_real_api@SUFFIX@.sh \
   elpa2_test_complex_api@SUFFIX@.sh \
+  elpa_driver_real@SUFFIX@.sh \
   elpa1_real_toeplitz@SUFFIX@.sh \
   elpa1_test_real_with_c@SUFFIX@.sh \
   elpa1_real_cholesky@SUFFIX@.sh \
@@ -681,6 +695,7 @@ check_SCRIPTS += \
   elpa2_test_real_single_precision@SUFFIX@.sh \
   elpa2_test_real_default_single_precision@SUFFIX@.sh \
   elpa2_test_real_qr_single_precision@SUFFIX@.sh \
+  elpa_driver_real_single_precision@SUFFIX@.sh \
   elpa1_real_cholesky_single_precision@SUFFIX@.sh \
   elpa1_real_invert_trm_single_precision@SUFFIX@.sh \
   elpa1_real_transpose_multiply_single_precision@SUFFIX@.sh \

test/Fortran/test_driver_real.F90

+!    This file is part of ELPA.
+!
+!    The ELPA library was originally created by the ELPA consortium,
+!    consisting of the following organizations:
+!
+!    - Max Planck Computing and Data Facility (MPCDF), formerly known as
+!      Rechenzentrum Garching der Max-Planck-Gesellschaft (RZG),
+!    - Bergische Universität Wuppertal, Lehrstuhl für angewandte
+!      Informatik,
+!    - Technische Universität München, Lehrstuhl für Informatik mit
+!      Schwerpunkt Wissenschaftliches Rechnen ,
+!    - Fritz-Haber-Institut, Berlin, Abt. Theorie,
+!    - Max-Plack-Institut für Mathematik in den Naturwissenschaften,
+!      Leipzig, Abt. Komplexe Strukutren in Biologie und Kognition,
+!      and
+!    - IBM Deutschland GmbH
+!
+!
+!    More information can be found here:
+!    http://elpa.mpcdf.mpg.de/
+!
+!    ELPA is free software: you can redistribute it and/or modify
+!    it under the terms of the version 3 of the license of the
+!    GNU Lesser General Public License as published by the Free
+!    Software Foundation.
+!
+!    ELPA is distributed in the hope that it will be useful,
+!    but WITHOUT ANY WARRANTY; without even the implied warranty of
+!    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+!    GNU Lesser General Public License for more details.
+!
+!    You should have received a copy of the GNU Lesser General Public License
+!    along with ELPA.  If not, see <http://www.gnu.org/licenses/>
+!
+!    ELPA reflects a substantial effort on the part of the original
+!    ELPA consortium, and we ask you to respect the spirit of the
+!    license that we chose: i.e., please contribute any changes you
+!    may have back to the original ELPA library distribution, and keep
+!    any derivatives of ELPA under the same license that we chose for
+!    the original distribution, the GNU Lesser General Public License.
+!
+!
+#include "config-f90.h"
+!>
+!> Fortran test programm to demonstrates the use of
+!> ELPA 2 real case library.
+!> If "HAVE_REDIRECT" was defined at build time
+!> the stdout and stderr output of each MPI task
+!> can be redirected to files if the environment
+!> variable "REDIRECT_ELPA_TEST_OUTPUT" is set
+!> to "true".
+!>
+!> By calling executable [arg1] [arg2] [arg3] [arg4]
+!> one can define the size (arg1), the number of
+!> Eigenvectors to compute (arg2), and the blocking (arg3).
+!> If these values are not set default values (4000, 1500, 16)
+!> are choosen.
+!> If these values are set the 4th argument can be
+!> "output", which specifies that the EV's are written to
+!> an ascii file.
+!>
+!> The real ELPA 2 kernel is set as the default kernel.
+!> However, this can be overriden by setting
+!> the environment variable "REAL_ELPA_KERNEL" to an
+!> appropiate value.
+!>
+program test_real2
+
+!-------------------------------------------------------------------------------
+! Standard eigenvalue problem - REAL version
+!
+! This program demonstrates the use of the ELPA module
+! together with standard scalapack routines
+!
+! Copyright of the original code rests with the authors inside the ELPA
+! consortium. The copyright of any additional modifications shall rest
+! with their original authors, but shall adhere to the licensing terms
+! distributed along with the original code in the file "COPYING".
+!
+!-------------------------------------------------------------------------------
+   use precision
+   use ELPA
+   use elpa_utilities, only : error_unit
+#ifdef WITH_OPENMP
+   use test_util
+#endif
+   use mod_check_for_gpu, only : check_for_gpu
+   use mod_read_input_parameters
+   use mod_check_correctness
+   use mod_setup_mpi
+   use mod_blacs_infrastructure
+   use mod_prepare_matrix
+   use elpa_mpi
+#ifdef HAVE_REDIRECT
+  use redirect
+#endif
+#ifdef HAVE_DETAILED_TIMINGS
+ use timings
+#endif
+ use output_types
+   implicit none
+
+   !-------------------------------------------------------------------------------
+   ! Please set system size parameters below!
+   ! na:   System size
+   ! nev:  Number of eigenvectors to be calculated
+   ! nblk: Blocking factor in block cyclic distribution
+   !-------------------------------------------------------------------------------
+
+   integer(kind=ik)           :: nblk
+   integer(kind=ik)           :: na, nev
+
+   integer(kind=ik)           :: np_rows, np_cols, na_rows, na_cols
+
+   integer(kind=ik)           :: myid, nprocs, my_prow, my_pcol, mpi_comm_rows, mpi_comm_cols
+   integer(kind=ik)           :: i, mpierr, my_blacs_ctxt, sc_desc(9), info, nprow, npcol
+
+   integer, external          :: numroc
+
+   real(kind=rk8), allocatable :: a(:,:), z(:,:), tmp1(:,:), tmp2(:,:), as(:,:), ev(:)
+
+   integer(kind=ik)           :: iseed(4096) ! Random seed, size should be sufficient for every generator
+   integer(kind=ik)           :: STATUS
+#ifdef WITH_OPENMP
+   integer(kind=ik)           :: omp_get_max_threads,  required_mpi_thread_level, provided_mpi_thread_level
+#endif
+   type(output_t)             :: write_to_file
+   logical                    :: success
+   character(len=8)           :: task_suffix
+   integer(kind=ik)           :: j
+   logical                    :: gpuAvailable
+   integer(kind=ik)           :: numberOfDevices
+
+   success = .true.
+   gpuAvailable  = .false.
+
+   call read_input_parameters(na, nev, nblk, write_to_file)
+
+   !-------------------------------------------------------------------------------
+   !  MPI Initialization
+   call setup_mpi(myid, nprocs)
+
+   gpuAvailable = check_for_gpu(myid, numberOfDevices)
+
+   STATUS = 0
+
+#define DATATYPE REAL
+#include "elpa_print_headers.X90"
+
+#ifdef HAVE_DETAILED_TIMINGS
+
+   ! initialise the timing functionality
+
+#ifdef HAVE_LIBPAPI
+   call timer%measure_flops(.true.)
+#endif
+
+   call timer%measure_allocated_memory(.true.)
+   call timer%measure_virtual_memory(.true.)
+   call timer%measure_max_allocated_memory(.true.)
+
+   call timer%set_print_options(&
+#ifdef HAVE_LIBPAPI
+                print_flop_count=.true., &
+                print_flop_rate=.true., &
+#endif
+                print_allocated_memory = .true. , &
+                print_virtual_memory=.true., &
+                print_max_allocated_memory=.true.)
+
+
+  call timer%enable()
+
+  call timer%start("program")
+#endif
+
+   !-------------------------------------------------------------------------------
+   ! Selection of number of processor rows/columns
+   ! We try to set up the grid square-like, i.e. start the search for possible
+   ! divisors of nprocs with a number next to the square root of nprocs
+   ! and decrement it until a divisor is found.
+
+   do np_cols = NINT(SQRT(REAL(nprocs))),2,-1
+      if(mod(nprocs,np_cols) == 0 ) exit
+   enddo
+   ! at the end of the above loop, nprocs is always divisible by np_cols
+
+   np_rows = nprocs/np_cols
+
+   if(myid==0) then
+      print *
+      print '(a)','Standard eigenvalue problem - REAL version'
+      print *
+      print '(3(a,i0))','Matrix size=',na,', Number of eigenvectors=',nev,', Block size=',nblk
+      print '(3(a,i0))','Number of processor rows=',np_rows,', cols=',np_cols,', total=',nprocs
+      print *
+   endif
+
+   !-------------------------------------------------------------------------------
+   ! Set up BLACS context and MPI communicators
+   !
+   ! The BLACS context is only necessary for using Scalapack.
+   !
+   ! For ELPA, the MPI communicators along rows/cols are sufficient,
+   ! and the grid setup may be done in an arbitrary way as long as it is
+   ! consistent (i.e. 0<=my_prow<np_rows, 0<=my_pcol<np_cols and every
+   ! process has a unique (my_prow,my_pcol) pair).
+
+   call set_up_blacsgrid(mpi_comm_world, my_blacs_ctxt, np_rows, np_cols, &
+                         nprow, npcol, my_prow, my_pcol)
+
+   if (myid==0) then
+     print '(a)','| Past BLACS_Gridinfo.'
+   end if
+
+   ! All ELPA routines need MPI communicators for communicating within
+   ! rows or columns of processes, these are set in elpa_get_communicators.
+
+   mpierr = elpa_get_communicators(mpi_comm_world, my_prow, my_pcol, &
+                                   mpi_comm_rows, mpi_comm_cols)
+
+   if (myid==0) then
+     print '(a)','| Past split communicator setup for rows and columns.'
+   end if
+
+   call set_up_blacs_descriptor(na ,nblk, my_prow, my_pcol, np_rows, np_cols, &
+                                na_rows, na_cols, sc_desc, my_blacs_ctxt, info)
+
+   if (myid==0) then
+     print '(a)','| Past scalapack descriptor setup.'
+   end if
+
+   !-------------------------------------------------------------------------------
+   ! Allocate matrices and set up a test matrix for the eigenvalue problem
+#ifdef HAVE_DETAILED_TIMINGS
+   call timer%start("set up matrix")
+#endif
+   allocate(a (na_rows,na_cols))
+   allocate(z (na_rows,na_cols))
+   allocate(as(na_rows,na_cols))
+
+   allocate(ev(na))
+
+   call prepare_matrix_double(na, myid, sc_desc, iseed,  a, z, as)
+
+#ifdef HAVE_DETAILED_TIMINGS
+   call timer%stop("set up matrix")
+#endif
+   ! set print flag in elpa1
+   elpa_print_times = .true.
+
+   !-------------------------------------------------------------------------------
+   ! Calculate eigenvalues/eigenvectors
+
+   if (myid==0) then
+     print '(a)','| Entering one-stage ELPA solver ... '
+     print *
+   end if
+#ifdef WITH_MPI
+   call mpi_barrier(mpi_comm_world, mpierr) ! for correct timings only
+#endif
+   success = elpa_solve_evp_real_double(na, nev, a, na_rows, ev, z, na_rows,  nblk, na_cols, &
+                              mpi_comm_rows, mpi_comm_cols, mpi_comm_world, method="1stage")
+
+   if (.not.(success)) then
+      write(error_unit,*) "elpa_solve_evp_real produced an error! Aborting..."
+#ifdef WITH_MPI
+      call MPI_ABORT(mpi_comm_world, 1, mpierr)
+#endif
+   endif
+
+   if (myid==0) then
+     print '(a)','| One-step ELPA solver complete.'
+     print *
+   end if
+
+   a = as
+   z = as
+
+   if (myid==0) then
+     print '(a)','| Entering two-stage ELPA solver ... '
+     print *
+   end if
+#ifdef WITH_MPI
+   call mpi_barrier(mpi_comm_world, mpierr) ! for correct timings only
+#endif
+   success = elpa_solve_evp_real_double(na, nev, a, na_rows, ev, z, na_rows,  nblk, na_cols, &
+                              mpi_comm_rows, mpi_comm_cols, mpi_comm_world, method="2stage")
+
+   if (.not.(success)) then
+      write(error_unit,*) "elpa_solve_evp_real produced an error! Aborting..."
+#ifdef WITH_MPI
+      call MPI_ABORT(mpi_comm_world, 1, mpierr)
+#endif
+   endif
+
+   if (myid==0) then
+     print '(a)','| two-step ELPA solver complete.'
+     print *
+   end if
+
+   a = as
+   z = as
+
+   if (myid==0) then
+     print '(a)','| Entering auto-chosen ELPA solver ... '
+     print *
+   end if
+#ifdef WITH_MPI
+   call mpi_barrier(mpi_comm_world, mpierr) ! for correct timings only
+#endif
+   success = elpa_solve_evp_real_double(na, nev, a, na_rows, ev, z, na_rows,  nblk, na_cols, &
+                              mpi_comm_rows, mpi_comm_cols, mpi_comm_world, method="auto")
+
+   if (.not.(success)) then
+      write(error_unit,*) "elpa_solve_evp_real produced an error! Aborting..."
+#ifdef WITH_MPI
+      call MPI_ABORT(mpi_comm_world, 1, mpierr)
+#endif
+   endif
+
+   if (myid==0) then
+     print '(a)','| auto-chosen ELPA solver complete.'
+     print *
+   end if
+
+   if(myid == 0) print *,'Time transform to tridi :',time_evp_fwd
+   if(myid == 0) print *,'Time solve tridi        :',time_evp_solve
+   if(myid == 0) print *,'Time transform back EVs :',time_evp_back
+   if(myid == 0) print *,'Total time (sum above)  :',time_evp_back+time_evp_solve+time_evp_fwd
+
+
+!   if(write_to_file%eigenvectors) then
+!     write(unit = task_suffix, fmt = '(i8.8)') myid
+!     open(17,file="EVs_real2_out_task_"//task_suffix(1:8)//".txt",form='formatted',status='new')
+!     write(17,*) "Part of eigenvectors: na_rows=",na_rows,"of na=",na," na_cols=",na_cols," of na=",na
+!
+!     do i=1,na_rows
+!       do j=1,na_cols
+!         write(17,*) "row=",i," col=",j," element of eigenvector=",z(i,j)
+!       enddo
+!     enddo
+!     close(17)
+!   endif
+!
+!   if(write_to_file%eigenvalues) then
+!      if (myid == 0) then
+!         open(17,file="Eigenvalues_real2_out.txt",form='formatted',status='new')
+!         do i=1,na
+!            write(17,*) i,ev(i)
+!         enddo
+!         close(17)
+!      endif
+!   endif
+
+
+   !-------------------------------------------------------------------------------
+   ! Test correctness of result (using plain scalapack routines)
+   allocate(tmp1(na_rows,na_cols))
+   allocate(tmp2(na_rows,na_cols))
+
+   status = check_correctness_double(na, nev, as, z, ev, sc_desc, myid, tmp1, tmp2)
+
+   deallocate(a)
+   deallocate(as)
+
+   deallocate(z)
+   deallocate(tmp1)
+   deallocate(tmp2)
+   deallocate(ev)
+
+#ifdef HAVE_DETAILED_TIMINGS
+   call timer%stop("program")
+   print *," "
+   print *,"Timings program:"
+   call timer%print("program")
+   print *," "
+   print *,"End timings program"
+#endif
+#ifdef WITH_MPI
+   call blacs_gridexit(my_blacs_ctxt)
+   call mpi_finalize(mpierr)
+#endif
+   call EXIT(STATUS)
+end
+
+!-------------------------------------------------------------------------------

test/Fortran/test_driver_real_single.F90

+!    This file is part of ELPA.
+!
+!    The ELPA library was originally created by the ELPA consortium,
+!    consisting of the following organizations:
+!
+!    - Max Planck Computing and Data Facility (MPCDF), formerly known as
+!      Rechenzentrum Garching der Max-Planck-Gesellschaft (RZG),
+!    - Bergische Universität Wuppertal, Lehrstuhl für angewandte
+!      Informatik,
+!    - Technische Universität München, Lehrstuhl für Informatik mit
+!      Schwerpunkt Wissenschaftliches Rechnen ,
+!    - Fritz-Haber-Institut, Berlin, Abt. Theorie,
+!    - Max-Plack-Institut für Mathematik in den Naturwissenschaften,
+!      Leipzig, Abt. Komplexe Strukutren in Biologie und Kognition,
+!      and
+!    - IBM Deutschland GmbH
+!
+!
+!    More information can be found here:
+!    http://elpa.mpcdf.mpg.de/
+!
+!    ELPA is free software: you can redistribute it and/or modify
+!    it under the terms of the version 3 of the license of the
+!    GNU Lesser General Public License as published by the Free
+!    Software Foundation.
+!
+!    ELPA is distributed in the hope that it will be useful,
+!    but WITHOUT ANY WARRANTY; without even the implied warranty of
+!    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+!    GNU Lesser General Public License for more details.
+!
+!    You should have received a copy of the GNU Lesser General Public License
+!    along with ELPA.  If not, see <http://www.gnu.org/licenses/>
+!
+!    ELPA reflects a substantial effort on the part of the original
+!    ELPA consortium, and we ask you to respect the spirit of the
+!    license that we chose: i.e., please contribute any changes you
+!    may have back to the original ELPA library distribution, and keep
+!    any derivatives of ELPA under the same license that we chose for
+!    the original distribution, the GNU Lesser General Public License.
+!
+!
+#include "config-f90.h"
+!>
+!> Fortran test programm to demonstrates the use of
+!> ELPA 2 real case library.
+!> If "HAVE_REDIRECT" was defined at build time
+!> the stdout and stderr output of each MPI task
+!> can be redirected to files if the environment
+!> variable "REDIRECT_ELPA_TEST_OUTPUT" is set
+!> to "true".
+!>
+!> By calling executable [arg1] [arg2] [arg3] [arg4]
+!> one can define the size (arg1), the number of
+!> Eigenvectors to compute (arg2), and the blocking (arg3).
+!> If these values are not set default values (4000, 1500, 16)
+!> are choosen.
+!> If these values are set the 4th argument can be
+!> "output", which specifies that the EV's are written to
+!> an ascii file.
+!>
+!> The real ELPA 2 kernel is set as the default kernel.
+!> However, this can be overriden by setting
+!> the environment variable "REAL_ELPA_KERNEL" to an
+!> appropiate value.
+!>
+program test_real2
+
+!-------------------------------------------------------------------------------
+! Standard eigenvalue problem - REAL version
+!
+! This program demonstrates the use of the ELPA module
+! together with standard scalapack routines
+!
+! Copyright of the original code rests with the authors inside the ELPA
+! consortium. The copyright of any additional modifications shall rest
+! with their original authors, but shall adhere to the licensing terms
+! distributed along with the original code in the file "COPYING".
+!
+!-------------------------------------------------------------------------------
+   use precision
+   use ELPA
+   use elpa_utilities, only : error_unit
+#ifdef WITH_OPENMP
+   use test_util
+#endif
+
+   use mod_check_for_gpu, only : check_for_gpu
+   use mod_read_input_parameters
+   use mod_check_correctness
+   use mod_setup_mpi
+   use mod_blacs_infrastructure
+   use mod_prepare_matrix
+   use elpa_mpi
+#ifdef HAVE_REDIRECT
+  use redirect
+#endif
+#ifdef HAVE_DETAILED_TIMINGS
+ use timings
+#endif
+ use output_types
+   implicit none
+
+   !-------------------------------------------------------------------------------
+   ! Please set system size parameters below!
+   ! na:   System size
+   ! nev:  Number of eigenvectors to be calculated
+   ! nblk: Blocking factor in block cyclic distribution
+   !-------------------------------------------------------------------------------
+
+   integer(kind=ik)           :: nblk
+   integer(kind=ik)           :: na, nev
+
+   integer(kind=ik)           :: np_rows, np_cols, na_rows, na_cols
+
+   integer(kind=ik)           :: myid, nprocs, my_prow, my_pcol, mpi_comm_rows, mpi_comm_cols
+   integer(kind=ik)           :: i, mpierr, my_blacs_ctxt, sc_desc(9), info, nprow, npcol
+
+   integer, external          :: numroc
+
+   real(kind=rk4), allocatable :: a(:,:), z(:,:), tmp1(:,:), tmp2(:,:), as(:,:), ev(:)
+
+   integer(kind=ik)           :: iseed(4096) ! Random seed, size should be sufficient for every generator
+   integer(kind=ik)           :: STATUS
+#ifdef WITH_OPENMP
+   integer(kind=ik)           :: omp_get_max_threads,  required_mpi_thread_level, provided_mpi_thread_level
+#endif
+   type(output_t)             :: write_to_file
+   logical                    :: success
+   character(len=8)           :: task_suffix
+   integer(kind=ik)           :: j
+   logical                    :: gpuAvailable
+   integer(kind=ik)           :: numberOfDevices
+
+   success = .true.
+   gpuAvailable  = .false.
+   call read_input_parameters(na, nev, nblk, write_to_file)
+