I am new to cuda programming. In this code, c matric return by GPU is Zero matrix. I tried different...

/*#ifndef CUDACC
#define CUDACC
#endif
*/
#include <time.h>
#include <stdio.h>
#include “cuda_runtime.h”
#include “device_launch_parameters.h”
//#include <cuda.h>
//#include <cuda_runtime_api.h>
//#include <device_functions.h>
#define N 32
#define TILE_WIDTH 2
double r2()
{
return (double)rand() / (double)RAND_MAX ;
}

 __global__ void test(double *a, double *b, double *c,int num)

{
/* shared int A_tile[ TILE_WIDTH ][ TILE_WIDTH ];
shared int B_tile[ TILE_WIDTH ][ TILE_WIDTH ];

	double accu=0;
	

	// target element coordinates
	int row =  blockIdx.y * TILE_WIDTH + threadIdx.y;
	int column = blockIdx.x * TILE_WIDTH + threadIdx.x;
// compute target element value
for(int i=0;i<(N/(int)TILE_WIDTH);i++){
	// move the tiles and update shared memory value for new tile positions
	if(row < N && (i*TILE_WIDTH + threadIdx.x)<N)
		A_tile[threadIdx.y][threadIdx.x] = a[row*N + i*TILE_WIDTH + threadIdx.x];

// else
// A_tile[threadIdx.y][threadIdx.x] = 0;
if(column < N && (iTILE_WIDTH + threadIdx.y)<N)
B_tile[threadIdx.x][threadIdx.y] = b[(i
TILE_WIDTH + threadIdx.y)*N + column];
//else
//B_tile[threadIdx.y][threadIdx.x] = 0;

	// after the entire tile's values are available, proceed
	__syncthreads();

	for(int j=0;j<TILE_WIDTH;j++)
		accu += A_tile[threadIdx.y][j] * B_tile[j][threadIdx.x];
	// after the entire tile's values have been used, proceed
	__syncthreads();
}
if(row < N && column < N)
	c[row*N+column] = accu;
*/


__shared__ int A_tile[ TILE_WIDTH ][  TILE_WIDTH ];
__shared__ int B_tile[ TILE_WIDTH ][  TILE_WIDTH ];
double accu=0;
for(int tileIdx = 0; tileIdx<(num/ blockDim.x-1) ; tileIdx++)
{
	int   i = blockIdx.y *  blockDim.y + threadIdx.y ;     
	int	  j = tileIdx *  blockDim.x + threadIdx.x ;
	A_tile[threadIdx.y][threadIdx.x] = a[i*(num/ blockDim.x-1) +tileIdx]; 
	B_tile[threadIdx.x][threadIdx.y] = b[tileIdx*(num/ blockDim.x-1)+j]; 
	__syncthreads();

	for(int k=0 ; k<2 ; k++)
	accu = accu + A_tile[threadIdx.y][k] * B_tile[k][threadIdx.x];


	__syncthreads();

}




int i = blockIdx.y *  blockDim.y + threadIdx.y  ;

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

c[i*num+j]=accu;
__syncthreads();
	


/*

// correct used code
int i = blockIdx.y * blockDim.y + threadIdx.y;
int j = blockIdx.x * blockDim.x + threadIdx.x;
double accu=0;
if(i<num && j<num)

    {
		accu = accu + a[i*num+n] * b[n*num+j];
	}
c[i*num+j]=accu;

// correct used code
*/

}

int main()
{

float ms;
cudaEvent_t start_g,stop_g;
clock_t start, end;
double cpu_time_used;

//declare host and device arrays;
double a[N][N], *a_device, b[N][N], *b_device, c[N][N], *c_device;

//declaring size of device array
const size_t a_size = sizeof(double) * size_t(N*N);

cudaMalloc((void **)&a_device, a_size); 
cudaMalloc((void **)&b_device, a_size); 
cudaMalloc((void **)&c_device, a_size); 

/*
double d_A[4][4];
double d_B[4][4];
double d_C[4][4];
double C[4][4];
double check[4][4]; */
// initializing 2d arrays d_A and d_B between 0 and 1 as well as d_C with 0;
for(int i=0;i<N;i++)
for(int j=0;j<N;j++)
{
a[i][j]=r2();
b[i][j]=r2();
// c[i][j] = 0;
}

  cudaMemcpy(a_device, a, a_size, cudaMemcpyHostToDevice); 
  cudaMemcpy(b_device, b, a_size, cudaMemcpyHostToDevice); 






dim3 dimBlock(16,16); //32 threads in one block in x and y directions;

// dim3 dimGrid(1, 1); // 2 blocks in one grid in both directions;
dim3 dimGrid(N/dimBlock.x, N/dimBlock.y);// 16 blocks in one grid in both directions;

// cudaDeviceSynchronize() ;

cudaEventCreate(&start_g);
cudaEventCreate(&stop_g);
cudaEventRecord(start_g, 0); 
cudaEventRecord(stop_g, 0);
//GPU timer starts;

test<<<dimGrid, dimBlock>>>(a_device,b_device,c_device,N); 

// cudaDeviceSynchronize();
cudaMemcpy(&c, c_device, a_size, cudaMemcpyDeviceToHost);
cudaEventSynchronize(stop_g);//GPU timer ends;
cudaEventElapsedTime(&ms, start_g, stop_g);
// cudaMemcpy(&C,d_C,4*4 , cudaMemcpyDeviceToHost);

cudaEventDestroy(start_g);
cudaEventDestroy(stop_g);
printf("GPU: %f ms\n",ms);

// cudaDeviceSynchronize() ;

 start = clock();

//CPU matrix multiplication
double check[N][N];
int i, j, k;
for (i = 0; i < N; i++)
{
for (j = 0; j < N; j++)
{
check[i][j] = 0;
for (k = 0; k < N; k++)
check[i][j] += a[i][k]*b[k][j];
}
}
end = clock();
//printf("start:=%f/n"start);
//printf("end:=%f/n"end);
cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC;

cpu_time_used*=1000;
printf (" cpu_time_used=%fms\n",cpu_time_used);

//CPU-GPU matrix matching

for(int i =0;i<N;i++)
{
	for(int j=0;j<N;j++)
	{
		if(check[i][j]==c[i][j])
		{	//	printf("correct\n");
				//printf("check = %f, c= %f\n",check[i][j],c[i][j]);
				continue;
		}
		else 
			{
				printf("not equal at i= %d and at j= %d\n",i,j);
				printf("check = %f, c= %f",check[i][j],c[i][j]);

		//		return 0;
			}
	}
}
cudaFree(a_device); cudaFree(b_device); cudaFree(c_device);

return 0;
}

/*

#include “cuda_runtime.h”
#include “device_launch_parameters.h”
cudaError_t addWithCuda(int *c, const int *a, const int *b, unsigned int size);

global void addKernel(int *c, const int *a, const int *b)
{
int i = threadIdx.x;
c[i] = a[i] + b[i];
}

int main()
{
const int arraySize = 5;
const int a[arraySize] = { 1, 2, 3, 4, 5 };
const int b[arraySize] = { 10, 20, 30, 40, 50 };
int c[arraySize] = { 0 };

// Add vectors in parallel.
cudaError_t cudaStatus = addWithCuda(c, a, b, arraySize);
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "addWithCuda failed!");
    return 1;
}

printf("{1,2,3,4,5} + {10,20,30,40,50} = {%d,%d,%d,%d,%d}\n",
    c[0], c[1], c[2], c[3], c[4]);

// cudaDeviceReset must be called before exiting in order for profiling and
// tracing tools such as Nsight and Visual Profiler to show complete traces.
cudaStatus = cudaDeviceReset();
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaDeviceReset failed!");
    return 1;
}

return 0;

}

// Helper function for using CUDA to add vectors in parallel.
cudaError_t addWithCuda(int *c, const int *a, const int *b, unsigned int size)
{
int *dev_a = 0;
int *dev_b = 0;
int *dev_c = 0;
cudaError_t cudaStatus;

// Choose which GPU to run on, change this on a multi-GPU system.
cudaStatus = cudaSetDevice(0);
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaSetDevice failed!  Do you have a CUDA-capable GPU installed?");
    goto Error;
}

// Allocate GPU buffers for three vectors (two input, one output)    .
cudaStatus = cudaMalloc((void**)&dev_c, size * sizeof(int));
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaMalloc failed!");
    goto Error;
}

cudaStatus = cudaMalloc((void**)&dev_a, size * sizeof(int));
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaMalloc failed!");
    goto Error;
}

cudaStatus = cudaMalloc((void**)&dev_b, size * sizeof(int));
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaMalloc failed!");
    goto Error;
}

// Copy input vectors from host memory to GPU buffers.
cudaStatus = cudaMemcpy(dev_a, a, size * sizeof(int), cudaMemcpyHostToDevice);
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaMemcpy failed!");
    goto Error;
}

cudaStatus = cudaMemcpy(dev_b, b, size * sizeof(int), cudaMemcpyHostToDevice);
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaMemcpy failed!");
    goto Error;
}

// Launch a kernel on the GPU with one thread for each element.
addKernel<<<1, size>>>(dev_c, dev_a, dev_b);

// Check for any errors launching the kernel
cudaStatus = cudaGetLastError();
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus));
    goto Error;
}

// cudaDeviceSynchronize waits for the kernel to finish, and returns
// any errors encountered during the launch.
cudaStatus = cudaDeviceSynchronize();
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching addKernel!\n", cudaStatus);
    goto Error;
}

// Copy output vector from GPU buffer to host memory.
cudaStatus = cudaMemcpy(c, dev_c, size * sizeof(int), cudaMemcpyDeviceToHost);
if (cudaStatus != cudaSuccess) {
    fprintf(stderr, "cudaMemcpy failed!");
    goto Error;
}

Error:
cudaFree(dev_c);
cudaFree(dev_a);
cudaFree(dev_b);

return cudaStatus;

}
*/