extern __shared__ does not allocate memory

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi All,

Can you please help me to understand the usage of shared ?

In my case “the block Dimension” and hence “the number of threads” in a blocks will be calculated at the run time.

So i cant pass a hard coded value in the shared memory, which i was doing by the below written line -

shared int s_data[4];

But if my array size has to be allocated dynamically, Can you please suggest me what is the best method to do so.

Additionally when i try to use -

extern shared int s_data;

It allocates values for first few elements only and rest of it displays me ZERO.

regards,
Nabarun.

2

ISSUE RESOLVED.

Shared Memory size is the third argument we pass in the kernel function call.

Thanks,

Nabarun.