Compiler Error

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi everyone,

I am getting a strange compiler error. The code is very simple inside my kernel function:

int MW_BLOCK_SIZE = 15;

__shared__ unsigned char Is[MW_BLOCK_SIZE * MW_BLOCK_SIZE * 3]; //Image in the shared memory

Compiler throws:

line 43: internal error: assertion failed:

          add_dyn_init_cleanup: missing destructible entity descr

          (lower_init.c, line 4799)

  __attribute__((__shared__)) unsigned char Is[MW_BLOCK_SIZE * MW_BLOCK_SIZE * 3]; //Image in the shared memory

                                             ^

1 catastrophic error detected in the compilation of "/tmp/tmp_00003033-3.ii".

Compilation aborted.

However if I write the value of MW_BLOCK_SIZE, 15, inside the brackets, it doesn’t complain.

One final question: Is the shared memory Is, big enough? I think 15153 = 675 should be fine. I just want to make sure. Thanks in advance.

2

shared array can’t have dynamic allocation you need to give a constant value for the size.

check you these post it will be very helpful

http://forums.nvidia.com/index.php?showtopic=28588

3

Thanks for the reply. I will make it constant.