#include<mpi.h>
#include<stdio.h>
#include<stdlib.h>
#include<math.h>
#include<string.h>
#define N 10

const int data[] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};

int main(int argc, char * argv[]) {
	int rank;
	int nProcesses;

	int *v = (int*)malloc(sizeof(int)*N);
	int *aux = (int*)calloc(N, sizeof(int));

	v = (int*)memcpy(v, data, sizeof(data));

	MPI_Init(&argc, &argv);

	MPI_Comm_rank(MPI_COMM_WORLD, &rank);
	MPI_Comm_size(MPI_COMM_WORLD, &nProcesses);
	printf("Hello from %i/%i\n", rank, nProcesses);

	int i;
	int *local = (int*)malloc(sizeof(int) * N / nProcesses);

	MPI_Scatter(aux, N / nProcesses, MPI_INT, local, N / nProcesses, MPI_INT, 0, MPI_COMM_WORLD);

	int local_startIndex = rank * N / nProcesses;
	int local_stopIndex = (rank + 1) * N / nProcesses;

	printf("rank = %i startIndex = %i stopIndex = %i\n", rank, local_startIndex, local_stopIndex);
	int j, k;
	for( j = local_startIndex; j < local_stopIndex; j++ ) {
		for( k = 0; k < N; k++ ) {
			if (v[j] > v[k]) {
				local[j]++;
				// printf("local[%d] = %d\n", j, local[j]);
			}
		}
	}

	MPI_Gather(local, N / nProcesses, MPI_INT, aux, N / nProcesses, MPI_INT, 0, MPI_COMM_WORLD);

	if(rank == 0) {
		for(i = 0; i < N; i++) {
			printf("%d ", v[local[i]]);
		}
	}
	printf("\n");

	free(aux);
	free(v);

	MPI_Finalize();
	return 0;
}