#include <mpi.h>
#include <hpc/mpi/matrix.h>
#include <hpc/matvec/gematrix.h>
#include <hpc/matvec/print.h>
#include <hpc/matvec/apply.h>

int main(int argc, char** argv) {
   MPI_Init(&argc, &argv);

   int nof_processes; MPI_Comm_size(MPI_COMM_WORLD, &nof_processes);
   int rank; MPI_Comm_rank(MPI_COMM_WORLD, &rank);

   using namespace hpc::matvec;
   using namespace hpc::mpi;

   using Matrix = GeMatrix<double>;
   using Index = typename Matrix::Index;
   int share = 3;
   int num_rows = nof_processes * share;
   int num_cols = 5;

   Matrix B(share, num_cols, StorageOrder::RowMajor); /* individual share */
   MPI_Datatype rowtype_B = get_row_type(B);

   if (rank == 0) {
      Matrix A(num_rows, num_cols); /* entire matrix */
      apply(A, [](double& val, Index i, Index j) -> void {
	 val = i * 100 + j;
      });
      MPI_Datatype rowtype_A = get_row_type(A);
      MPI_Scatter(&A(0, 0), share, rowtype_A,
	 &B(0, 0), share, rowtype_B,
	 0, MPI_COMM_WORLD);
      apply(B, [=](double& val, Index i, Index j) -> void {
	 val += 10000 * (rank + 1);
      });
      MPI_Gather(&B(0, 0), share, rowtype_B,
	 &A(0, 0), share, rowtype_A,
	 0, MPI_COMM_WORLD);
      print(A, "A");
   } else {
      MPI_Scatter(nullptr, 0, nullptr, /* ignored parameters */
	 &B(0, 0), share, rowtype_B,
	 0, MPI_COMM_WORLD);
      apply(B, [=](double& val, Index i, Index j) -> void {
	 val += 10000 * (rank + 1);
      });
      MPI_Gather(&B(0, 0), share, rowtype_B,
	 nullptr, 0, nullptr, /* ignored parameters */
	 0, MPI_COMM_WORLD);
   }
   MPI_Finalize();
}