#include "VecFuncs.h"
#include <vector>
#include <iostream>
#include <cassert>
#include <cfloat>

void DebugVector(const std::vector<double>& xin, MPI_Comm icomm)
{
    int rank, size;
    MPI_Comm_rank(icomm, &rank);
    MPI_Comm_size(icomm, &size);
    int ierr;

    int active_rank = -1;

    for (int step = 0; step < size; ++step)
    {
        if (rank == 0)
        {
            std::cout << "Enter rank to display vector: " << std::endl;
            std::cin >> active_rank;
        }

        ierr = MPI_Bcast(&active_rank, 1, MPI_INT, 0, icomm);
        assert(ierr == 0);

        MPI_Barrier(icomm);

        if (rank == active_rank)
        {
            std::cout << "Output from process " << rank  << std::endl;

            for (size_t i = 0; i < xin.size(); ++i)
            {
                std::cout << "xin[" << i << "] = " << xin[i] << std::endl;
            }

            std::cout << std::endl;
        }

        MPI_Barrier(icomm);
    }
}

double par_scalar(const std::vector<double> &x, const std::vector<double> &y, const MPI_Comm &comm) {
    assert(x.size()==y.size());

    double local_sum = 0.0;
    for (int k = 0; k < x.size(); ++k) {
        local_sum += x[k] * y[k];
    }

    double global_sum = 0.0;
    int mpi_error = MPI_Allreduce(&local_sum, &global_sum, 1, MPI_DOUBLE, MPI_SUM, comm);
    assert(mpi_error == 0);

    return global_sum;
}

void par_minmax(std::vector<double> &x, double &min_value, double &max_value, const MPI_Comm &icomm)
{
	int myrank;
    MPI_Comm_rank(icomm, &myrank);
	
    int local_n = x.size();
	int global_offset = myrank * local_n;	
	
	struct {double value; int idx;} local_min, local_max, global_min, global_max;	// global index

	local_min.value = local_max.value = x[0];
    local_min.idx = local_max.idx = global_offset;

	// finding local min/max with the corresponding global index
	for (int i = 1; i < local_n; ++i) {
        int global_idx = global_offset + i;
	    if (x[i] < local_min.value){
	        local_min.value = x[i];
	        local_min.idx = global_idx;
	    }
	    if (x[i] > local_max.value){
	        local_max.value = x[i];
	        local_max.idx = global_idx;
	    }
	}
	
	// reduction to the global one including the global index (need it later for interchanging)
	MPI_Allreduce(&local_min, &global_min, 1, MPI_DOUBLE_INT, MPI_MINLOC, icomm);
	MPI_Allreduce(&local_max, &global_max, 1, MPI_DOUBLE_INT, MPI_MAXLOC, icomm);
	min_value = global_min.value;
	max_value = global_max.value;
	
	// calculating the process and the local index for interchanging the min and max value
	int rank_min = global_min.idx / local_n;
	int rank_max = global_max.idx / local_n;
	int local_min_idx = global_min.idx % local_n;
	int local_max_idx = global_max.idx % local_n;
	
	// interchanging
	if (rank_min == rank_max){
		std::swap(x[local_min_idx], x[local_max_idx]);
	}
	else {
	    double recv_value;
	    if (myrank == rank_min){
	        MPI_Sendrecv(&x[local_min_idx], 1, MPI_DOUBLE, rank_max, 0, &recv_value, 1, MPI_DOUBLE, rank_max, 0, icomm, MPI_STATUS_IGNORE);	
	        x[local_min_idx] = recv_value;
	    }
	    if (myrank == rank_max){
	        MPI_Sendrecv(&x[local_max_idx], 1, MPI_DOUBLE, rank_min, 0, &recv_value, 1, MPI_DOUBLE, rank_min, 0, icomm, MPI_STATUS_IGNORE);
	        x[local_max_idx] = recv_value;
	    }

	}
	

	return;
}