Constant Memory Error Stack Smashing

CuPy Version          : 8.0.0b3
CUDA Root             : /usr/local/cuda
CUDA Build Version    : 11000
CUDA Driver Version   : 11000
CUDA Runtime Version  : 10020
cuBLAS Version        : 10202
cuFFT Version         : 10102
cuRAND Version        : 10102
cuSOLVER Version      : (10, 3, 0)
cuSPARSE Version      : 10301
NVRTC Version         : (10, 2)
Thrust Version        : 100906
cuDNN Build Version   : None
cuDNN Version         : None
NCCL Build Version    : None
NCCL Runtime Version  : None
CUB Version           : Enabled
cuTENSOR Version      : None

I’m trying to test constant memory usage with the master branch.

Here is my toy code

import cupy as cp

cuda_code = r"""
__constant__ float arr[10];
"""

module = cp.RawModule(code=cuda_code,)

memptr = module.get_global("arr")
print(memptr)

I’m getting the following error

*** stack smashing detected ***: terminated
Aborted (core dumped)

gdb output

(gdb) run constant_test.py 
Starting program: /home/belt/anaconda3/envs/lmc/bin/python3 constant_test.py
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
[New Thread 0x7ffff5384700 (LWP 172697)]
[New Thread 0x7ffff4b83700 (LWP 172853)]
[New Thread 0x7ffff2382700 (LWP 173023)]
[New Thread 0x7fffefb81700 (LWP 173864)]
[New Thread 0x7fffed380700 (LWP 173867)]
[New Thread 0x7fffeab7f700 (LWP 173873)]
[New Thread 0x7fffe637e700 (LWP 173884)]
[New Thread 0x7fffe5b7d700 (LWP 173894)]
[New Thread 0x7fffe337c700 (LWP 173904)]
[New Thread 0x7fffe0b7b700 (LWP 173905)]
[New Thread 0x7fffde37a700 (LWP 173909)]
[New Thread 0x7fffa6b73700 (LWP 173937)]
[New Thread 0x7fffa5ccf700 (LWP 173938)]
[New Thread 0x7fffa54ce700 (LWP 173939)]
here
*** stack smashing detected ***: terminated

Thread 1 "python3" received signal SIGABRT, Aborted.
__GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
50      ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.

[leofang]

I parsed the error output wrong 😅 It’s a new error for unclean builds that I haven’t seen before…

[kmaehashi]

Could you clean the build workspace (eg git clean -fdx), rebuild CuPy and try again? I guess Cython failed to detect module that needs to be recompiled for some reason…