size_t global_work_size = size;
	size_t cb;

	cl_platform_id cpPlatform;
    clGetPlatformIDs(1, &cpPlatform, NULL);
    cl_device_id cdDevice;
    clGetDeviceIDs(cpPlatform, CL_DEVICE_TYPE_GPU, 1, &cdDevice, NULL);
    cl_context context = clCreateContext(0, 1, &cdDevice, NULL, NULL, NULL);
	cl_command_queue cmd_queue = clCreateCommandQueue(context,cdDevice,0,0);
	cl_program program = clCreateProgramWithSource(context,1,&ker,0,0);
	clBuildProgram(program,0,0,0,0,0);
	cl_kernel kernel = clCreateKernel(program,"vecadd",0);

	cl_mem DeviceMemA,DeviceMemB,DeviceMemC;
	DeviceMemA =   clCreateBuffer(context,CL_MEM_READ_ONLY|CL_MEM_COPY_HOST_PTR,size*sizeof(cl_float),A,0);
	DeviceMemB = clCreateBuffer(context,CL_MEM_READ_ONLY|CL_MEM_COPY_HOST_PTR,size*sizeof(cl_float),B,0);
	DeviceMemC = clCreateBuffer(context,CL_MEM_WRITE_ONLY,size*sizeof(cl_float),0,0);

	clSetKernelArg(kernel,0,sizeof(cl_mem),(void*)&DeviceMemA);
	clSetKernelArg(kernel,1,sizeof(cl_mem),(void*)&DeviceMemB);
	clSetKernelArg(kernel,2,sizeof(cl_mem),(void*)&DeviceMemC);

	clEnqueueNDRangeKernel(cmd_queue,kernel,1,0,&global_work_size,0,0,0,0);

	clEnqueueReadBuffer(cmd_queue,DeviceMemC,CL_TRUE,0,global_work_size*sizeof(cl_float),C,0,0,0);