#include <stdio.h>
#include "mpi.h"

/* 

Exercise 6.10.3, see Chap 6, p. 110 in PPMPI.
by John Weatherwax

Here we implement a dense matrix transpose from processor 0 to processor 1 
and print.

*/ 

#define MAX_ORDER 100

typedef float LOCAL_MATRIX_T[MAX_ORDER][MAX_ORDER];

main( int argc, char* argv[] ){
    int             my_rank;
    int             p;
    LOCAL_MATRIX_T  local_A; 
    int             n;
    int             i,j; 

    /* declarations: */ 
    void Send_Transpose(LOCAL_MATRIX_T local_A, int n,int my_rank); 

    MPI_Init(&argc,&argv); 
    MPI_Comm_size(MPI_COMM_WORLD,&p);
    MPI_Comm_rank(MPI_COMM_WORLD,&my_rank); 

    if( my_rank==0 ){
      if( p != 2 ){
	printf("Number of processors must be 2.\n" ); 
	MPI_Finalize(); 
	exit(1); 
      }
      printf("Enter the size of the matrix (n x n):\n"); 
      scanf("%d",&n);
      if( n > MAX_ORDER ){
	printf("n is too large...exiting\n"); 
	MPI_Finalize(); 
	exit(1); 
      }
    }
    MPI_Bcast(&n,1,MPI_INT,0,MPI_COMM_WORLD); 

    if( my_rank==0 ){ /* generate matrix A */ 
      for( i=0; i < n; i++ ){
	for( j=0; j < n; j++ ){
	  local_A[i][j] = i+1;
	}
      }
      /* for debugging we print ... */ 
      for( i=0; i < n; i++ ){
	for( j=0; j < n; j++ ){
	  printf("%7.2f ",local_A[i][j]);
	}
	printf("\n"); 
      }
      printf("\n"); 
    }

    Send_Transpose(local_A,n,my_rank);

    if( my_rank==1 ){ /* print matrix A^T */ 
      for( i=0; i < n; i++ ){
	for( j=0; j < n; j++ ){
	  printf("%7.2f ",local_A[i][j]);
	}
	printf("\n"); 
      }
    }

    MPI_Finalize(); 
}

/**********************************************************************/


/**********************************************************************/
void Send_Transpose(
         LOCAL_MATRIX_T  local_A  /* in/out */, 
         int                   n  /* in  */, 
         int             my_rank  /* in  */) {

    int             i, j, tag;
    MPI_Datatype    column_mpi_t; 
    MPI_Status      status; 

    MPI_Type_vector(n,1,MAX_ORDER,MPI_FLOAT,&column_mpi_t); 
    MPI_Type_commit(&column_mpi_t);

    if( my_rank==0 ){ /* I am sending...send each column */ 
      for( j=0; j < n; j++ ){
	/*tag=j;*/
	tag=0; 
	MPI_Send(&(local_A[0][j]),1,column_mpi_t,1,tag,MPI_COMM_WORLD); 
      }
    }else{ /* I am recieving...assumed I am process==1 */ 
      for( i=0; i < n; i++ ){
	/*tag=i;*/
	tag=0;
	MPI_Recv(&(local_A[i][0]),n,MPI_FLOAT,0,tag,MPI_COMM_WORLD,&status); 
      }
    }
}
/**********************************************************************/