Cuda code generation using codegen command

Hi , I am generating cuda code for following equation :

function [out] = My_Fcn(A,B,C,D,X,U) %#codgen

n = size(U,2);

DU = D*U(:,1:n);

out = coder.nullcopy(zeros(size(U)));

coder.gpu.kernelfun();

for i=1:1:n

X(:,i+1) = A*X(:,i) + B*U(:,i);

end

out = C*X(:,1:n) + DU;

end

And I am generating code using following method :

A = double(zeros(3));

B = double(zeros(3));

C = double(zeros(3));

D = double(zeros(3));

X = double(zeros(3,4));

U = double(zeros(3));

cfg = coder.gpuConfig('lib');

codegen -args {A,B,C,D,X,U} -config cfg My_Fcn -report

When I am generating with 3*3 matrix, I am getting correct kernels as follows :

#include "My_Fcn.h"

#include "MWCudaDimUtility.hpp"

// Function Declarations

static __global__ void My_Fcn_kernel1(const double U[9], const double B[9],

const double X[12], const int i, const double A[9], double b_A[3], double b_B

[3]);

static __global__ void My_Fcn_kernel2(const double B[3], const double A[3],

const int i, double X[12]);

static __global__ void My_Fcn_kernel3(const double U[9], const double D[9],

const double X[12], const double C[9], double b_C[9], double b_D[9]);

static __global__ void My_Fcn_kernel4(const double D[9], const double C[9],

double out[9]);

But when I am increasing the size of matrices for example 100*100, The kernels created are as follows:

// Include Files

#include "My_Fcn.h"

#include "MWCudaDimUtility.hpp"

#include "My_Fcn_data.h"

#include "My_Fcn_initialize.h"

// Type Definitions

#include "cublas_v2.h"

// Function Declarations

static __global__ void My_Fcn_kernel1(double DU[10000]);

static __global__ void My_Fcn_kernel2(double C[100]);

static __global__ void My_Fcn_kernel3(double C[100]);

static __global__ void My_Fcn_kernel4(const int i, double C[100], double b_C[100],

double X[10100]);

static __global__ void My_Fcn_kernel5(double out[10000]);

static __global__ void My_Fcn_kernel6(double DU[10000], double out[10000]);

static cublasHandle_t getCublasGlobalHandle();

Here I cannot pass A,B,C,D,X,U matrices. I want to use this kernels to run ptx file on GPU, but as these kernels are not generating properly, I am not able to proceed.

Can you please help me with solution or workaround?

Thank You