#define _USE_MATH_DEFINES

#include "cuda_runtime.h"
#include "device_launch_parameters.h"

#include <math.h>

__global__ void radialAverage(double** image, int x_center, int y_center)
{
    int i, x, y;

    const int x_size = 400;
    const int y_size = 400;

    int thread = blockIdx.x*blockDim.x+threadIdx.x;

    double angle     = 0;
    double dAngle    = M_PI/360;

    double radImgMat[x_size][y_size];
    double angleMatPi[x_size][y_size];


    //radImMap erstellen
    for( x = 0; x < x_size; x++) {
        for( y = 0; y < y_size; y++) {
        radImgMat[x][y]  = sqrt((double)(x+1-x_center)*(x+1-x_center) + (y+1-y_center)*(y+1-y_center));
        }
    }

    //Angle-Matrix (pi-pi) erstellen
    for ( x = 0; x < x_size; x++) {
        for( y = 0; y < y_size; y++) {
            float xD = x+1-x_center;
            float yD = y+1-y_center;
            if(yD>0) {
                angleMatPi[x][y] = -1*(atan(xD/yD)+M_PI/2)+M_PI;
            } else if(yD==0 && xD<0) {
                angleMatPi[x][y] = M_PI;
            } else if(yD==0 && xD>0) {
                angleMatPi[x][y] = 0;
            } else {
                angleMatPi[x][y] = -1*(atan(xD/yD)+M_PI/2);
            }
        }
    }

    //Cut ImgMat
    for(x=0; x < x_size; x++) {
        for(y = 0; y < y_size; y++) {
            if((angleMatPi[x][y] < (angle-dAngle)) || (angleMatPi[x][y] > (angle+dAngle))) {
                image[x][y] = 0;
            }
        }
    }
}

int *d_image;
cudaMalloc((void**)&d_image,x_size*y_size*sizeof(int)); //allocate memory on the GPU for the image

cudaMemcpy(d_image,image,x_size*y_size*sizeof(int),cudaMemcpyHostToDevice); //copy the image to GPU global memory

radialAverage<<<numBlocks,numThreads>>>(d_image,x_size,y_size,x_center,y_center); //Call CUDA kernel with device pointer to image data

image[y*x_size + x];