#ifndef _TMP_KERNEL_H_
#define _TMP_KERNEL_H_

#include <math.h>

__device__ double calculate_error(double Ri, double Ra, double u, double f, double F, double amax, double OBS_actual[])
{
    return 0.01; //dummy double value
}

__global__ void compute_fit(int n_Ri, double a_Ri, double b_Ri, int n_Ra, double a_Ra, 
							  double b_Ra, int n_u, double a_u, double b_u, int n_f, double a_f,
							  double b_f, int n_F, double a_F, double b_F, double OBS_actual[], 
							  double amax, double* best_fit)
{
	int i_Ri = blockIdx.x-floor(blockIdx.x/n_Ri)*n_Ri;
	int i_Ra = blockIdx.y;
	int i_u = floor(1.0*blockIdx.x/n_Ri)+(blockIdx.x%n_Ri==0?0:1);
	int i_f = threadIdx.x;
	int i_F = threadIdx.y;
	
	double Ri = a_Ri + (b_Ri-a_Ri)/(n_Ri-1)*i_Ri;
	double Ra = a_Ra + (b_Ra-a_Ra)/(n_Ra-1)*i_Ra;
	double u = a_u + (b_u-a_u)/(n_u-1)*i_u;
	double f = a_f + (b_f-a_f)/(n_f-1)*i_f;
	double F = a_F + (b_F-a_F)/(n_F-1)*i_F;
	
	double val;

	
	__syncthreads();	
	if(true)
	{
		for(int i = 0; i < 18; i++)
		{
			val = calculate_error(Ri, Ra, u, f, F, amax, OBS_actual);
			best_fit[i] = val; // This line if uncommented causes nv4_disp error
		}		
	}
		
	return;
}

#endif // #ifndef _TMP_KERNEL_H_