sharing of local_memory between Work Items on SoC FPGA (Cyclone V)

According to OpenCL ,

1. __local address space inside a __kernel function are allocated for each work-group executing the kernel.

2. variables that need to be allocated in local memory and are shared by all work-items of a work-group.

for the following kernel code for image of 512 *512 pixels

# define W 512# define H 512# define global_size_x 512# define global_size_y 512# define local_size_x 512# define local_size_y 1

__attribute__((reqd_work_group_size(local_size_x,local_size_y,1))) //dimensions

__kernel void sobel_kernel (__global unsigned char * restrict image_in,

__global unsigned char * restrict image_out)

{

__local int n;

int sum;

//Index of the pixel

__private short int row_id = get_global_id(1);

__private short int col_id = get_local_id(0);

sum = image_in[(row_id )*W + (col_id )] ; //read global to local

n=n+1; //update local value

barrier (CLK_GLOBAL_MEM_FENCE);

barrier(CLK_LOCAL_MEM_FENCE);

if((row_id <10) && (col_id <10))

printf("\n%d",n); //observe local value

//global mem write transaction

image_out[(row_id)*W + (col_id) ] = sum;

}

the above kernel when compiled on

1. emulator in linux (default s5_ref board) the output printed for n was n = 1 2 3 4 5.....

2. but when I compiled the kernel with Intel FPGA SDK and deployed .aocx file on the Cyclone V SoC FPGA the values for n were always 1 1 1 1 1...../

Can someone explain why the local variable declared doesn't have scope for all the work items in the work group.

and why emulator and FPGA may show different results.

Thanks