How to Scatter then Reduce using MPI and c++ - c++

Question

I'm trying to parallelize the computation of the norm2 of a vector (variable size). My approach was to first scatter the vector amoungs the processors, calculate the square and sum of each sub vectors and then reduce the reslut and apply the squareroot.
Here is my code :
#include <mpi.h>
#include <vector>
#include <iostream>
#include <cmath>
double SquareSum(std::vector<double> & v) {
double res;
for (std::vector<double>::iterator it = v.begin(); it != v.end(); it++){
if (*it){
res += (*it)*(*it);
}
else{
it++;
}
}
return res;
}
int main(int argc, char *argv[]){
std::vector<double> numbers;
double val;
while (std::cin >> val) numbers.push_back(val);
MPI_Init(&argc,&argv);
int rank, size;
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &size);
unsigned numbers_count = numbers.size();
MPI_Bcast(&numbers_count, 1, MPI_UNSIGNED, 0, MPI_COMM_WORLD);
unsigned local_share = numbers_count / size; // local_share is the floor function of numbers.size() / number of process
if (numbers_count % size > 0){
++local_share; // if size is not a multiple of numbers.size() add 1 to local_share in order to make it "fit"
}
if (rank == 0){
numbers.resize(local_share*size); //resize numbers by adding null empty spot if necessary
}
//std::cout << "I'm" << rank << std::endl;
std::vector<double> local(local_share);
MPI_Scatter(&numbers, local_share, MPI_DOUBLE, &local, local_share, MPI_DOUBLE, 0, MPI_COMM_WORLD);
double par_sum = SquareSum(local);
double sum = 0;
MPI_Reduce(&par_sum, &sum, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
if (rank == 0){
std::cout << "norm : " << std::sqrt(sum);
}
MPI_Finalize();
return 0;
}
When I try to execute the programm I have this error :
"mpiexec noticed that process rank 2 with PID 9823 on node laptop exited on signal 11 (Segmentation fault)."
I suppose there is a problem with the scatter but I cannot figure ou what.
Thanks for the help

Answer 1

In addition to the comment of #Hristo Iliev:
The size of numbers is likely not a multiple of size. Then, local_share becomes numbers_count % size+1. It is used as the size of the part each process retreives by MPI_Scatter(). As a result, the scattered vector must be of size (numbers_count % size+1)*size. Since the provided vector numbers is likely too small, the program tries to acceed to elements out of bound, which triggers undefined behaviors such as segmentation fault.
Two solutions:
Push back zeros until numbers.size()%size==0. It modifies the vector, which is not practical for future uses and rather ugly.
Use MPI_Scatterv() !