#include "benchmarks.h"
#include <cassert>       // assert()
#include <cmath>
#include <iostream>
#include <vector>
#include <omp.h>

// (A) Inner product of two vectors (from skalar_stl)
double scalar_parallel(vector<double> const &x, vector<double> const &y)
{
    assert(x.size() == y.size());
    size_t const N = x.size();
    double sum = 0.0;
//#pragma omp parallel for default(none) shared(x, y, N) reduction(+:sum) schedule(runtime)
#pragma omp parallel for shared(x, y, N) reduction(+:sum)
    for (size_t i = 0; i < N; ++i)
    {
        sum += x[i] * y[i];
    }
    return sum;
}

// (A) Vector entry sum
double sum(vector<double> const &x)
{
    double sum = 0.0;
#pragma omp parallel for shared(x) reduction(+:sum)
    for (size_t i = 0; i < x.size(); ++i)
    {
        sum += x[i];
    }
    return sum;
}


// (B) Matrix-vector product (from intro_vector_densematrix)
vector<double> MatVec_parallel(vector<double> const &A, vector<double> const &x)
{
    size_t const nelem = A.size();
    size_t const N = x.size();
    assert(nelem % N == 0); // make sure multiplication is possible
    size_t const M = nelem/N;          

    vector<double> b(M);    

#pragma omp parallel for shared(A, x, N, M, b)
    for (size_t i = 0; i < M; ++i)
    {
        double tmp = 0.0;
        for (size_t j = 0; j < N; ++j)
            tmp += A[N*i + j] * x[j];
        b[i] = tmp;
    }

    return b;
}


// (C) Matrix-matrix product
vector<double> MatMat_parallel(vector<double> const &A, vector<double> const &B, size_t const &L)
{
    size_t const nelem_A = A.size();
    size_t const nelem_B = B.size();

    assert(nelem_A % L == 0 && nelem_B % L == 0);

    size_t const M = nelem_A/L;
    size_t const N = nelem_B/L;


    vector<double> C(M*N);

            
#pragma omp parallel for shared(A, B, M, N, L, C)
    for (size_t i = 0; i < M; ++i)
    {
        for (size_t k = 0; k < L; ++k)
        {
            for (size_t j = 0; j < N; ++j)
            {
                C[N*i + j] += A[L*i + k]*B[N*k + j];
            }
            
        }
    }

    return C;
}


// (D) Evaluation of a polynomial function
vector<double> poly_parallel(vector<double> const &a, vector<double> const &x)
{
    size_t const N = x.size();
    size_t const p = a.size() - 1;
    vector<double> y(N, 0);

#pragma omp parallel for shared(a, x, N, p, y)
    for (size_t i = 0; i < N; ++i)
    {
        double x_temp = x[i];
        double y_temp = 0;
        for (size_t k = 0; k < p + 1; ++k)
        {
            y_temp += x_temp*y_temp + a[p - k];
        }
        y[i] = y_temp;
    }

    return y;
}