MPI Vector Addition

The root and other process codes can be clubbed together, but in this code root has been treated separately so as for better understanding.

	/***************************************
	AUTHOR: ANKIT MAHATO
	amahato@iitk.ac.in
	IIT KANPUR
	This code distributes data and adds
	two vectors a and b in parallel.
	The root and other process codes can be
	clubbed together but in this code root
	has been treated separately so as for
	better understanding.
	***************************************/
	#include "mpi.h"
	#include <stdio.h>
	#include <stdlib.h>
	#define MASTER 0
	int main (int argc, char *argv[])
	{
	int * a;
	int * b;
	int * c;
	// arrays a and b
	int total_proc;
	// total nuber of processes
	int rank;
	// rank of each process
	int T;
	// total number of test cases
	long long int n_per_proc;
	// elements per process
	long long int i, n;
	MPI_Status status;
	// Initialization of MPI environment
	MPI_Init (&argc, &argv);
	MPI_Comm_size (MPI_COMM_WORLD, &total_proc);
	//Now you know the total number of processes running in parallel
	MPI_Comm_rank (MPI_COMM_WORLD,&rank);
	//Now you know the rank of the current process
	int * ap;
	int * bp;
	int * cp;
	if (rank == MASTER) //you choose process rank 0 to be your root which will be used to perform input output.
	{
	scanf ("%d", &T);
	// you get the total number of test cases
	MPI_Bcast (&T, 1, MPI_INT, MASTER, MPI_COMM_WORLD);
	// Then you broadcast it to every process out there
	while (T--)
	{
	scanf ("%lld", &n);
	a = (int *) malloc(sizeof(int)*n);
	b = (int *) malloc(sizeof(int)*n);
	c = (int *) malloc(sizeof(int)*n);
	MPI_Bcast (&n, 1, MPI_LONG_LONG_INT, MASTER, MPI_COMM_WORLD);
	n_per_proc = n/total_proc;
	for(i=0;i<n;i++)
	scanf ("%d", &a[i]);
	for(i=0;i<n;i++)
	scanf ("%d", &b[i]);
	if(n%total_proc != 0)
	{
	n_per_proc+=1;
	for(i=0;i<(n_per_proc*total_proc - n);i++)
	{
	a[n+i] = 0;
	b[n+i] = 0;
	}
	} // to divide data evenly by the number of processors
	ap = (int *) malloc(sizeof(int)*n_per_proc);
	bp = (int *) malloc(sizeof(int)*n_per_proc);
	cp = (int *) malloc(sizeof(int)*n_per_proc);
	MPI_Bcast (&n_per_proc, 1, MPI_LONG_LONG_INT, MASTER, MPI_COMM_WORLD);
	//Broadcast element per process
	MPI_Scatter(a, n_per_proc, MPI_INT, ap, n_per_proc, MPI_INT, 0, MPI_COMM_WORLD);
	//scattering array a
	MPI_Scatter(b, n_per_proc, MPI_INT, bp, n_per_proc, MPI_INT, 0, MPI_COMM_WORLD);
	//scattering array b
	for(i=0;i<n_per_proc;i++)
	cp[i] = ap[i]+bp[i];
	MPI_Gather(cp, n_per_proc, MPI_INT, c, n_per_proc, MPI_INT, MASTER, MPI_COMM_WORLD);
	//gathering array c
	for(i=0;i<n;i++)
	printf ("%d ", c[i]);
	printf ("\n");
	}
	}
	else
	{ // Non-master tasks
	// Receive T
	MPI_Bcast (&T, 1, MPI_INT, MASTER, MPI_COMM_WORLD);
	//printf("%d from process %d\n",T,rank);
	while (T--)
	{
	MPI_Bcast (&n, 1, MPI_LONG_LONG_INT, MASTER, MPI_COMM_WORLD);
	MPI_Bcast (&n_per_proc, 1, MPI_LONG_LONG_INT, MASTER, MPI_COMM_WORLD);
	ap = (int *) malloc(sizeof(int)*n_per_proc);
	bp = (int *) malloc(sizeof(int)*n_per_proc);
	cp = (int *) malloc(sizeof(int)*n_per_proc);
	MPI_Scatter(a, n_per_proc, MPI_INT, ap, n_per_proc, MPI_INT, 0, MPI_COMM_WORLD);
	//Recieving Scattered a
	MPI_Scatter(b, n_per_proc, MPI_INT, bp, n_per_proc, MPI_INT, 0, MPI_COMM_WORLD);
	//Recieving Scattered b
	for(i=0;i<n_per_proc;i++)
	{
	cp[i] = ap[i]+bp[i];
	//printf("%d from %d\n",ap[i],rank);
	//printf("%d from %d\n",bp[i],rank);
	}
	MPI_Gather(cp, n_per_proc, MPI_INT, c, n_per_proc, MPI_INT, MASTER, MPI_COMM_WORLD);
	//To initiate the gathering of computed part of array C
	}
	}
	MPI_Finalize();
	//Terminate MPI Environment
	return 0;
	}

view raw vector_add.c hosted with ❤ by GitHub