#include <assert.h>
#include <cuda.h>
#include <cuda_runtime.h>
#include <stdio.h>
#include <stim/visualization/colormap.h>
#include <sstream>
#include <stim/math/vector.h>
#include <stim/cuda/cudatools/devices.h>
#include <stim/cuda/cudatools/threads.h>
#include <stim/cuda/cuda_texture.cuh>
	stim::cuda::cuda_texture tx;  //texture object.
	float* print;
	
	///Initialization function, allocates the memory and passes the necessary
	///handles from OpenGL and Cuda.
	///@param DIM_Y			--integer controlling how much memory to allocate.
	void initArray(int x, int y)
	{
			cudaMalloc( (void**) &print, x*y*sizeof(float));     ///temporary
	}

	///Deinit function that frees the memery used and releases the texture resource
	///back to OpenGL.
	void cleanUP()
	{
			cudaFree(print);         ///temporary
	}  

        __device__
        float templ(int x)                                                                      
        {
                if(x < 32/6 || x > 32*5/6 || (x > 32*2/6 && x < 32*4/6)){
                        return 1.0;
                }else{
                        return 0.0;
                }

        }
		
	///Find the difference of the given set of samples and the template
	///using cuda acceleration.
	///@param stim::cuda::cuda_texture t	--stim texture that holds all the references
	///					  to the data.
	///@param float* result			--a pointer to the memory that stores the result.
	__global__
	//void get_diff (float *result)
	void get_diff (cudaTextureObject_t texIn, float *print)
	{       
		int x   = threadIdx.x + blockIdx.x * blockDim.x;
		int y   = threadIdx.y + blockIdx.y * blockDim.y;
//		int idx = y*64+x;
		int idx = y*32+x;
	//	int idx = y*16+x;

		float valIn             = tex2D<unsigned char>(texIn, x, y);
		float templa		= templ(x);
		print[idx]              = valIn;             ///temporary
		//print[idx]              = abs(templa);             ///temporary

	}


	///External access-point to the cuda function
	///@param GLuint texbufferID 	--GLtexture (most be contained in a framebuffer object)
	///				  that holds the data that will be handed to cuda.
	///@param GLenum texType	--either GL_TEXTURE_1D, GL_TEXTURE_2D or GL_TEXTURE_3D
	///				  may work with other gl texture types, but untested.
	///@param DIM_Y, the number of samples in the template.
	void test(GLint texbufferID, GLenum texType, int x, int y)
	{

		//Bind the Texture in GL and allow access to cuda.
		tx.MapCudaTexture(texbufferID, texType);

		//initialize the return arrays.

		initArray(x,y);
		dim3 numBlocks(1, y);
		dim3 threadsPerBlock(x, 1);
		int max_threads = stim::maxThreadsPerBlock();
		//dim3 threads(max_threads, 1);
		//dim3 blocks(x / threads.x + 1, y);	
		//dim3 numBlocks(2, 2);
		//dim3 threadsPerBlock(8, 108);


//		get_diff <<< blocks, threads >>> (tx.getTexture(), print);
		get_diff <<< numBlocks, threadsPerBlock >>> (tx.getTexture(), print);

		cudaDeviceSynchronize();
		stringstream name;      //for debugging
		name << "FromTex.bmp";
		stim::gpu2image<float>(print, name.str(),x,y,0,255);
	  
		tx.UnmapCudaTexture();
		cleanUP();
	}

	void test(GLint texbufferID, GLenum texType, int x, int y, std::string nam)
	{

		//Bind the Texture in GL and allow access to cuda.
		tx.MapCudaTexture(texbufferID, texType);

		//initialize the return arrays.

		initArray(x,y);
		dim3 numBlocks(1, y);
		dim3 threadsPerBlock(x, 1);
		int max_threads = stim::maxThreadsPerBlock();
		//dim3 threads(max_threads, 1);
		//dim3 blocks(x / threads.x + 1, y);	
		//dim3 numBlocks(2, 2);
		//dim3 threadsPerBlock(8, 108);


//		get_diff <<< blocks, threads >>> (tx.getTexture(), print);
		get_diff <<< numBlocks, threadsPerBlock >>> (tx.getTexture(), print);

		cudaDeviceSynchronize();
		stringstream name;      //for debugging
		name << nam.c_str();
		stim::gpu2image<float>(print, name.str(),x,y,0,255);
	  
		tx.UnmapCudaTexture();
		cleanUP();
	}