/****************************************************************************
* FILE: datatypes2.c
* DESCRIPTION:
*  Simple comparison between using a derived datatype and individual 
*  send/receive operations for different data types.
* AUTHOR:  01/09/99 Blaise Barney
* LAST REVISED:
******************************************************************************/

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

#define TRIALS    10
#define REPS      10000

int main(argc,argv)
int argc;
char *argv[];  {
int numtasks, rank, tag=1, i, k;
double bandwidth, tbytes, indiv_best, derived_best, t1, t2, ttime;

typedef struct {
  float  f1,f2,f3,f4;
  int    i1,i2;
  }      f4i2;
f4i2     rbuff, sbuff;
MPI_Datatype newtype, oldtypes[2]; 
int          blockcounts[2];
MPI_Aint     offsets[2], extent;
MPI_Status   stat;

MPI_Init(&argc,&argv);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &numtasks);
 
if (numtasks != 2) {
  printf("Please run this test with 2 tasks. Terminating\n");
  MPI_Finalize();
  }

/* Setup MPI structured type for the 4 floats and 2 ints */
offsets[0] = 0;
oldtypes[0] = MPI_FLOAT;
blockcounts[0] = 4;
MPI_Type_extent(MPI_FLOAT, &extent);
offsets[1] = 4 * extent;
oldtypes[1] = MPI_INT;
blockcounts[1] = 2;
MPI_Type_struct(2, blockcounts, offsets, oldtypes, &newtype);
MPI_Type_commit(&newtype);

/* Initialize the derived type send data */
sbuff.f1 = 1.0;
sbuff.f2 = 2.0;
sbuff.f3 = 3.0;
sbuff.f4 = 4.0;
sbuff.i1 = 1;
sbuff.i2 = 2;

/**************************** task 0 ***********************************/
if (rank == 0) {

  /* Greetings */
  printf("\n****** Derived Datatype vs. Individual Send/Receive *****\n");
  printf("Trials=       %8d\n",TRIALS);
  printf("Reps/trial=   %8d\n\n",REPS);
  printf("D/I  #    ttime      tbytes   bytes/sec\n");
  printf("--- --  ---------  ---------  ----------\n");

  /* Begin timings  - save best result over all trials */
  derived_best = 0.0;
  indiv_best = 0.0;
  tbytes = sizeof(f4i2) * 2 * (float)REPS;
  for (k=0; k<TRIALS; k++) {

    /* First do derived datatype send/receive */
    t1 = MPI_Wtime();
    for (i=1; i<=REPS; i++){
      MPI_Send(&sbuff, 1, newtype, 1, tag, MPI_COMM_WORLD);
      MPI_Recv(&rbuff, 1, newtype, 1, tag, MPI_COMM_WORLD, &stat);
      }
    t2 = MPI_Wtime();
    ttime = t2 - t1;
    bandwidth = tbytes/ttime;
    if (bandwidth > derived_best)
      derived_best = bandwidth;
    printf(" D %3d   %3.6f %10.0f  %10.0f\n",k,ttime,tbytes,bandwidth);

    /* Now do individual send/receives for structure components */
    t1 = MPI_Wtime();
    for (i=1; i<=REPS; i++){
      MPI_Send(&sbuff.f1, 4, MPI_FLOAT, 1, tag, MPI_COMM_WORLD);
      MPI_Send(&sbuff.i1, 2, MPI_INT, 1, tag, MPI_COMM_WORLD);
      MPI_Recv(&rbuff.f1, 4, MPI_FLOAT, 1, tag, MPI_COMM_WORLD, &stat);
      MPI_Recv(&rbuff.i1, 2, MPI_INT, 1, tag, MPI_COMM_WORLD, &stat);
      }
    t2 = MPI_Wtime();
    ttime = t2 - t1;
    bandwidth = tbytes/ttime;
    if (bandwidth > indiv_best)
       indiv_best = bandwidth;
    printf(" I %3d   %3.6f %10.0f  %10.0f\n",k,ttime,tbytes,bandwidth);

    }  /* k loop */

    printf("\nDerived bandwidth= \t%10.0f bytes/sec.\n",derived_best);
    printf("Individual bandwidth= \t%10.0f bytes/sec.\n",indiv_best);
  }    /* task 0 */


/**************************** task 1 ***********************************/
if (rank == 1) {
  for (k=0; k<TRIALS; k++) {

    for (i=1; i<=REPS; i++) { 
      MPI_Recv(&rbuff, 1, newtype, 0, tag, MPI_COMM_WORLD, &stat);
      MPI_Send(&sbuff, 1, newtype, 0, tag, MPI_COMM_WORLD);
      }

    for (i=1; i<=REPS; i++) { 
      MPI_Recv(&rbuff.f1, 4, MPI_FLOAT, 0, tag, MPI_COMM_WORLD, &stat);
      MPI_Recv(&rbuff.i1, 2, MPI_INT, 0, tag, MPI_COMM_WORLD, &stat);
      MPI_Send(&sbuff.f1, 4, MPI_FLOAT, 0, tag, MPI_COMM_WORLD);
      MPI_Send(&sbuff.i1, 2, MPI_INT, 0, tag, MPI_COMM_WORLD);
      } 

    }   /* k loop */
  }     /* task 1 */

   
MPI_Finalize();
}