// #define TEST
#define X	100
#define Y	100

//#define USE_DBL
#ifdef USE_DBL
#define TYPE_FLT  double
#define TYPE_INT  long
#define MASK      0xFFFF
#define SHIFT     16
#define EPSILSON  4.94065645841247E-324
#else
#define TYPE_FLT  float
#define TYPE_INT  int
#define MASK      0x00FF
#define SHIFT     8
#define EPSILSON  1.401298E-45
#endif

constant sampler_t sampler = CLK_NORMALIZED_COORDS_FALSE |
                                CLK_ADDRESS_CLAMP_TO_EDGE |
                                CLK_FILTER_LINEAR;


struct input {
  int gain;
  int dyn_cont_selector;
  float gamma;
};

TYPE_FLT read_data(image2d_t input_frame, int x, int z)
{
  int2 gid = (int2)(x, z);
  uint4 pixel = read_imageui(input_frame, sampler, gid);
  TYPE_INT temp =  (TYPE_INT)((TYPE_INT)pixel.x & MASK) | 
                    (TYPE_INT)(((TYPE_INT)pixel.y & MASK) << SHIFT) |
                    (TYPE_INT)(((TYPE_INT)pixel.z & MASK) << (SHIFT * 2)) | 
                    (TYPE_INT)(((TYPE_INT)pixel.w & MASK) << (SHIFT * 3));
  TYPE_FLT raw_data = *((TYPE_FLT*)(&temp));
  return raw_data;
}


kernel void DynCont(read_only image2d_t input_frame, read_write image2d_t output_frame, global TYPE_FLT* scratch_pad, struct input params) 
{
  int2 gid = (int2)(get_global_id(0), get_global_id(1));
  TYPE_FLT input = read_data(input_frame, gid.x, gid.y);
  TYPE_FLT output_data  = 0;

  if (params.dyn_cont_selector >= 12)
  {
      output_data = log2((params.gamma - 1) * input + 1) / log2(params.gamma);
  }
  else
  {
      output_data = pow(input, params.gamma);
  }

  output_data = output_data * pow(10.0, params.gain / 20.0);

  if(output_data < 0)
    output_data = 0;

  if(output_data > 1)
    output_data = 1;
  
    TYPE_INT out = *((TYPE_INT*)(&output_data));
    uint4 pixel;
    pixel.x = (TYPE_INT)(out & MASK);
    pixel.y = (TYPE_INT)((out >> SHIFT) & MASK);
    pixel.z = (TYPE_INT)((out >> (SHIFT *2)) & MASK);
    pixel.w = (TYPE_INT)((out >> (SHIFT * 3)) & MASK);

    write_imageui(output_frame, gid, pixel);
}