/* F08AH_A1W_F C++ Header Example Program.
 *
 * Copyright 2021 Numerical Algorithms Group.
 * Mark 27.2, 2021.
 */

#include <dco.hpp>
#include <iostream>
#include <math.h>
#include <nag.h>
#include <nagad.h>
#include <stdio.h>
#include <string>
using namespace std;

int main(void)
{
  /* Scalars */
  Integer exit_status = 0;

  cout << "F08AH_A1W_F C++ Header Example Program Results\n\n";

  // Create AD tape
  dco::ga1s<double>::global_tape = dco::ga1s<double>::tape_t::create();

  void *  ad_handle = 0;
  Integer ifail     = 0;
  nag::ad::x10aa(ad_handle, ifail);

  // Skip heading in data file
  string mystr;
  getline(cin, mystr);

  Integer m, n;
  cin >> m;
  cin >> n;

  Integer pda   = m;
  Integer lwork = 64 * n;

  Integer tau_len;
  if (m < n)
    {
      tau_len = m;
    }
  else
    {
      tau_len = n;
    }

  // Allocate arrays
  nagad_a1w_w_rtype *a = 0, *tau = 0, *work = 0, *a_in = 0;
  double *           l = 0, *dlda = 0;
  a    = new nagad_a1w_w_rtype[m * n];
  work = new nagad_a1w_w_rtype[lwork];
  a_in = new nagad_a1w_w_rtype[m * n];
  tau  = new nagad_a1w_w_rtype[tau_len];
  l    = new double[m * m];
  dlda = new double[m * n];

  // Read A from data file
  double tmp;
  for (int i = 0; i < m; ++i)
    {
      for (int j = 0; j < n; ++j)
        {
          cin >> tmp;
          a_in[i + j * m] = tmp;
        }
    }

  for (int i = 0; i < n; i++)
    {
      dco::ga1s<double>::global_tape->register_variable(a_in[i * m]);
    }
  for (int i = 0; i < m * n; i++)
    {
      a[i] = a_in[i];
    }

  nag::ad::f08ah(ad_handle, m, n, a, pda, tau, work, lwork, ifail);
  if (ifail != 0)
    {
      printf("Error from F08AH_A1W_F .\n%" NAG_IFMT " ", ifail);
      exit_status = 1;
      goto END;
    }

  for (int i = 0; i < m * m; ++i)
    {
      l[i] = dco::value(a[i]);
    }

  for (int i = 0; i < m; ++i)
    {
      double d_one = 1.0;
      int    ld    = i * (m + 1);
      dco::ga1s<double>::global_tape->zero_adjoints();
      dco::derivative(a[ld]) += d_one;

      dco::ga1s<double>::global_tape->sparse_interpret() = true;
      dco::ga1s<double>::global_tape->interpret_adjoint();

      for (int j = 0; j < n; ++j)
        {
          int k   = j * m + i;
          dlda[k] = dco::derivative(a_in[j * m]);
        }
    }
  NagError fail;
  INIT_FAIL(fail);
  x04cac(Nag_ColMajor, Nag_LowerMatrix, Nag_NonUnitDiag, m, m, l, m,
         "L from Q factorization of A", 0, &fail);
  printf("\nDerivatives of diagonal of L w.r.t. first row of A\n");
  x04cac(Nag_ColMajor, Nag_GeneralMatrix, Nag_NonUnitDiag, m, n, dlda, m,
         "dL_ii/dA_j1", 0, &fail);

END:

  // Remove computational data object and tape
  nag::ad::x10ab(ad_handle, ifail);
  dco::ga1s<double>::tape_t::remove(dco::ga1s<double>::global_tape);
  delete[] a;
  delete[] a_in;
  delete[] tau;
  delete[] l;
  delete[] dlda;
  return exit_status;
}