/**
 * adi.c: This file is part of the PolyBench/C 3.2 test suite.
 *
 *
 * Contact: Louis-Noel Pouchet <pouchet@cse.ohio-state.edu>
 * Web address: http://polybench.sourceforge.net
 */
#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <math.h>

/* Include polybench common header. */
#include <polybench.h>

/* Include benchmark-specific header. */
/* Default data type is double, default size is 10x1024x1024. */
#include "adi.h"


/* Array initialization. */
static
void init_array (int n,
		 DATA_TYPE POLYBENCH_2D(X,N,N,n,n),
		 DATA_TYPE POLYBENCH_2D(A,N,N,n,n),
		 DATA_TYPE POLYBENCH_2D(B,N,N,n,n))
{
#pragma STDC FP_CONTRACT OFF
  int i, j;

  for (i = 0; i < n; i++)
    for (j = 0; j < n; j++)
      {
	X[i][j] = ((DATA_TYPE) i*(j+1) + 1) / n;
	A[i][j] = ((DATA_TYPE) i*(j+2) + 2) / n;
	B[i][j] = ((DATA_TYPE) i*(j+3) + 3) / n;
      }
}


/* DCE code. Must scan the entire live-out data.
   Can be used also to check the correctness of the output. */
static
void print_array(int n,
		 DATA_TYPE POLYBENCH_2D(X,N,N,n,n))

{
  int i, j;
  char *printmat = malloc(n*16 + 1); printmat[n*16] = 0;

  for (i = 0; i < n; i++) {
    for (j = 0; j < n; j++)
      print_element(X[i][j], j*16, printmat);
    fputs(printmat, stderr);
  }
  free(printmat);
}


/* Main computational kernel. The whole function will be timed,
   including the call and return. */
static
void kernel_adi(int tsteps,
		int n,
		DATA_TYPE POLYBENCH_2D(X,N,N,n,n),
		DATA_TYPE POLYBENCH_2D(A,N,N,n,n),
		DATA_TYPE POLYBENCH_2D(B,N,N,n,n))
{
  int t, i1, i2;

#pragma scop
  for (t = 0; t < _PB_TSTEPS; t++)
    {
      for (i1 = 0; i1 < _PB_N; i1++)
	for (i2 = 1; i2 < _PB_N; i2++)
	  {
	    X[i1][i2] = X[i1][i2] - X[i1][i2-1] * A[i1][i2] / B[i1][i2-1];
	    B[i1][i2] = B[i1][i2] - A[i1][i2] * A[i1][i2] / B[i1][i2-1];
	  }

      for (i1 = 0; i1 < _PB_N; i1++)
	X[i1][_PB_N-1] = X[i1][_PB_N-1] / B[i1][_PB_N-1];

      for (i1 = 0; i1 < _PB_N; i1++)
	for (i2 = 0; i2 < _PB_N-2; i2++)
	  X[i1][_PB_N-i2-2] = (X[i1][_PB_N-2-i2] - X[i1][_PB_N-2-i2-1] * A[i1][_PB_N-i2-3]) / B[i1][_PB_N-3-i2];

      for (i1 = 1; i1 < _PB_N; i1++)
	for (i2 = 0; i2 < _PB_N; i2++) {
	  X[i1][i2] = X[i1][i2] - X[i1-1][i2] * A[i1][i2] / B[i1-1][i2];
	  B[i1][i2] = B[i1][i2] - A[i1][i2] * A[i1][i2] / B[i1-1][i2];
	}

      for (i2 = 0; i2 < _PB_N; i2++)
	X[_PB_N-1][i2] = X[_PB_N-1][i2] / B[_PB_N-1][i2];

      for (i1 = 0; i1 < _PB_N-2; i1++)
	for (i2 = 0; i2 < _PB_N; i2++)
	  X[_PB_N-2-i1][i2] = (X[_PB_N-2-i1][i2] - X[_PB_N-i1-3][i2] * A[_PB_N-3-i1][i2]) / B[_PB_N-2-i1][i2];
    }
#pragma endscop

}

#if !FMA_DISABLED
// NOTE: FMA_DISABLED is true for targets where FMA contraction causes
// discrepancies which cause the accuracy checks to fail.
// In this case, the test runs with the option -ffp-contract=off
static void
kernel_adi_StrictFP(int tsteps,
                    int n,
                    DATA_TYPE POLYBENCH_2D(X,N,N,n,n),
                    DATA_TYPE POLYBENCH_2D(A,N,N,n,n),
                    DATA_TYPE POLYBENCH_2D(B,N,N,n,n))
{
#pragma STDC FP_CONTRACT OFF
  int t, i1, i2;

  for (t = 0; t < _PB_TSTEPS; t++)
    {
      for (i1 = 0; i1 < _PB_N; i1++)
	for (i2 = 1; i2 < _PB_N; i2++)
	  {
	    X[i1][i2] = X[i1][i2] - X[i1][i2-1] * A[i1][i2] / B[i1][i2-1];
	    B[i1][i2] = B[i1][i2] - A[i1][i2] * A[i1][i2] / B[i1][i2-1];
	  }

      for (i1 = 0; i1 < _PB_N; i1++)
	X[i1][_PB_N-1] = X[i1][_PB_N-1] / B[i1][_PB_N-1];

      for (i1 = 0; i1 < _PB_N; i1++)
	for (i2 = 0; i2 < _PB_N-2; i2++)
	  X[i1][_PB_N-i2-2] = (X[i1][_PB_N-2-i2] - X[i1][_PB_N-2-i2-1] * A[i1][_PB_N-i2-3]) / B[i1][_PB_N-3-i2];

      for (i1 = 1; i1 < _PB_N; i1++)
	for (i2 = 0; i2 < _PB_N; i2++) {
	  X[i1][i2] = X[i1][i2] - X[i1-1][i2] * A[i1][i2] / B[i1-1][i2];
	  B[i1][i2] = B[i1][i2] - A[i1][i2] * A[i1][i2] / B[i1-1][i2];
	}

      for (i2 = 0; i2 < _PB_N; i2++)
	X[_PB_N-1][i2] = X[_PB_N-1][i2] / B[_PB_N-1][i2];

      for (i1 = 0; i1 < _PB_N-2; i1++)
	for (i2 = 0; i2 < _PB_N; i2++)
	  X[_PB_N-2-i1][i2] = (X[_PB_N-2-i1][i2] - X[_PB_N-i1-3][i2] * A[_PB_N-3-i1][i2]) / B[_PB_N-2-i1][i2];
    }
}

/* Return 0 when one of the elements of arrays A and B do not match within the
   allowed FP_ABSTOLERANCE.  Return 1 when all elements match.  */
static inline int
check_FP(int n,
         DATA_TYPE POLYBENCH_2D(A,N,N,n,n),
         DATA_TYPE POLYBENCH_2D(B,N,N,n,n)) {
  int i, j;
  double AbsTolerance = FP_ABSTOLERANCE;
  for (i = 0; i < _PB_N; i++)
    for (j = 0; j < _PB_N; j++)
      {
        double V1 = A[i][j];
        double V2 = B[i][j];
        double Diff = fabs(V1 - V2);
        if (Diff > AbsTolerance) {
          fprintf(stderr, "A[%d][%d] = %lf and B[%d][%d] = %lf differ more than"
                  " FP_ABSTOLERANCE = %lf\n", i, j, V1, i, j, V2, AbsTolerance);
          return 0;
        }
      }

  /* All elements are within the allowed FP_ABSTOLERANCE error margin.  */
  return 1;
}
#endif

int main(int argc, char** argv)
{
  /* Retrieve problem size. */
  int n = N;
  int tsteps = TSTEPS;

  /* Variable declaration/allocation. */
  POLYBENCH_2D_ARRAY_DECL(X, DATA_TYPE, N, N, n, n);
#if !FMA_DISABLED
  POLYBENCH_2D_ARRAY_DECL(X_StrictFP, DATA_TYPE, N, N, n, n);
#endif
  POLYBENCH_2D_ARRAY_DECL(A, DATA_TYPE, N, N, n, n);
  POLYBENCH_2D_ARRAY_DECL(B, DATA_TYPE, N, N, n, n);


  /* Initialize array(s). */
  init_array (n, POLYBENCH_ARRAY(X), POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(B));

  /* Start timer. */
  polybench_start_instruments;

  /* Run kernel. */
  kernel_adi (tsteps, n, POLYBENCH_ARRAY(X),
	      POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(B));

  /* Stop and print timer. */
  polybench_stop_instruments;
  polybench_print_instruments;

#if FMA_DISABLED
  /* Prevent dead-code elimination. All live-out data must be printed
     by the function call in argument. */
  polybench_prevent_dce(print_array(n, POLYBENCH_ARRAY(X)));
#else
  init_array (n, POLYBENCH_ARRAY(X_StrictFP), POLYBENCH_ARRAY(A),
              POLYBENCH_ARRAY(B));
  kernel_adi (tsteps, n, POLYBENCH_ARRAY(X_StrictFP),
	      POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(B));
  if (!check_FP(n, POLYBENCH_ARRAY(X), POLYBENCH_ARRAY(X_StrictFP)))
    return 1;

  /* Prevent dead-code elimination. All live-out data must be printed
     by the function call in argument. */
  polybench_prevent_dce(print_array(n, POLYBENCH_ARRAY(X_StrictFP)));
#endif

  /* Be clean. */
  POLYBENCH_FREE_ARRAY(X);
#if !FMA_DISABLED
  POLYBENCH_FREE_ARRAY(X_StrictFP);
#endif
  POLYBENCH_FREE_ARRAY(A);
  POLYBENCH_FREE_ARRAY(B);

  return 0;
}