Added example on distributed array.

         VectorExample

)

set( MPI_COMMON_EXAMPLES

         DistributedArrayExample

)

SET( mpi_test_parameters -np 4 -H localhost:4 "${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/DistributedArrayTest${CMAKE_EXECUTABLE_SUFFIX}" )

if( BUILD_CUDA )

   foreach( target IN ITEMS ${COMMON_EXAMPLES} )

      cuda_add_executable( ${target}-cuda ${target}.cu OPTIONS )

      add_custom_command( COMMAND ${target}-cuda > ${TNL_DOCUMENTATION_OUTPUT_SNIPPETS_PATH}/${target}.out OUTPUT ${target}.out )

      set( CUDA_OUTPUTS ${CUDA_OUTPUTS} ${target}.out )

   endforeach()

   if( ${BUILD_MPI} )

      foreach( target IN ITEMS ${MPI_COMMON_EXAMPLES} )

         cuda_add_executable( ${target}-cuda ${target}.cu OPTIONS )

         add_custom_command( COMMAND "mpirun" ${mpi_test_parameters} ${target}-cuda > ${TNL_DOCUMENTATION_OUTPUT_SNIPPETS_PATH}/${target}.out OUTPUT ${target}.out )

         set( CUDA_OUTPUTS ${CUDA_OUTPUTS} ${target}.out )

      endforeach()

   endif( ${BUILD_MPI} )

else()

   foreach( target IN ITEMS ${HOST_EXAMPLES} )

      add_executable( ${target} ${target}.cpp )

      add_custom_command( COMMAND ${target} > ${TNL_DOCUMENTATION_OUTPUT_SNIPPETS_PATH}/${target}.out OUTPUT ${target}.out )

      set( HOST_OUTPUTS ${HOST_OUTPUTS} ${target}.out )

   endforeach()

   if( ${BUILD_MPI} )

      foreach( target IN ITEMS ${MPI_COMMON_EXAMPLES} )

         add_executable( ${target} ${target}.cpp )

         add_custom_command( COMMAND "mpirun" ${mpi_test_parameters} ${target}-cuda > ${TNL_DOCUMENTATION_OUTPUT_SNIPPETS_PATH}/${target}.out OUTPUT ${target}.out )

         set( HOST_OUTPUTS ${CUDA_OUTPUTS} ${target}.out )

      endforeach( ${BUILD_MPI} )

   endif()

endif()

IF( BUILD_CUDA )

#include <iostream>

#include <TNL/Containers/Partitioner.h>

#include <TNL/Containers/DistributedArray.h>

#include <TNL/MPI/ScopedInitializer.h>

using namespace TNL;

using namespace std;

/***

 * The following works for any device (CPU, GPU ...).

 */

template< typename Device >

void distributedArrayExample()

{

   using ArrayType = Containers::DistributedArray< int, Device >;

   using IndexType = typename ArrayType::IndexType;

   using LocalRangeType = typename ArrayType::LocalRangeType;

   const MPI_Comm group = TNL::MPI::AllGroup();

   const int rank = TNL::MPI::GetRank(group);

   const int nproc = TNL::MPI::GetSize(group);

   /***

    * We set size to prime number to force non-uniform distribution of the distributed array.

    */

   const int size = 97;

   const int ghosts = (nproc > 1) ? 4 : 0;

   const LocalRangeType localRange = Containers::Partitioner< IndexType >::splitRange( size, group );

   ArrayType a( localRange, ghosts, size, group );

}

int main( int argc, char* argv[] )

{

   TNL::MPI::ScopedInitializer mpi(argc, argv);

   std::cout << "The first test runs on CPU ..." << std::endl;

   distributedArrayExample< Devices::Host >();

#ifdef HAVE_CUDA

   std::cout << "The second test runs on GPU ..." << std::endl;

   distributedArrayExample< Devices::Cuda >();

#endif

}