Confusion about data type alignment

I need to transfer 128 structures from host to FPGA. In the host code,I define the structure as following:

*********************************************************************# define AOCL_ALIGNMENT 64

typedef struct

{

double S;

double X;

double u;

double e_rt_p;

double e_rt_q;

int CP;

}struct_A;

struct_A* input_A;

input_A=(struct_A *)_aligned_malloc(128*sizeof(struct_A),AOCL_ALIGNMENT);

***********************************************************************

in the kernel code,I define the structure as following:

**********************************************************************

typedef struct __attribute__((packed)) __attribute__((aligned(64)))

{

double S;

double X;

double u;

double e_rt_p;

double e_rt_q;

int CP;

}struct_A;

__kernel void kernels(__global double* restrict output_data, __global struct_A* restrict input_data)

for (int i=0;i<128;i++){

S = input_data.S;

X = input_data.X;

u = input_data.u;

e_rt_p = input_data.e_rt_p;

e_rt_q = input_data.e_rt_q;

CP = input_data.CP;

}

*************************************************************************************

BUT, I can't get the values of S~cp in right order, when I remove "__attribute__((packed)) __attribute__((aligned(64)))", I can get the right values of S~cp.

Can anyone figure out what's the problem in my code?Thanks very much.