"__eh_curr_region" is undefined

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi all,

I’m getting the following error during compilation of cuda. Does anyone have any clue what is this referring to?

Thanks.

error: identifier “__eh_curr_region” is undefined

1 error detected in the compilation of “/tmp/tmpxft_00002a5e_00000000-5_mycode.cpp2.i”.

2

I met such error. The reason is i did the following:

int b;
shared float a[b];

the size b can only be a constant.

3

Hi, I also have the same error, but I did not use the shared memory at all. I use global memory, and my code is :

matrix[ty*N+tx] -= row[tx] * col[ty];

__syncthreads();

Is the index of my matrix have error?

Thanks!