10000 Numpy 1.24.1 crash/reboot Intel i9-10900KF · Issue #23082 · numpy/numpy · GitHub
[go: up one dir, main page]
Skip to content

Numpy 1.24.1 crash/reboot Intel i9-10900KF #23082

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
jacopomoi opened this issue Jan 24, 2023 · 5 comments
Closed

Numpy 1.24.1 crash/reboot Intel i9-10900KF #23082

jacopomoi opened this issue Jan 24, 2023 · 5 comments

Comments

@jacopomoi
Copy link

I have a machine with the intel I9 and ubuntu 22.04.

I have set a conda environment for scientific production, but when i try in python 3.11:

import numpy as np

the machine crash and reboot like every time. I've tried to find something in journalctl but it doesn't seem to be a kernel related problem. I have also tried without conda environment and the same behavior is reproducible every time i call numpy.

So i have downgraded numpy several times and i found that 1.23.4 seems more stable, but anyway sometimes it crash.

The strange thing is that if i was successful to import numpy i can re-run the same script but if the time between the two calls is too tiny the system crash and reboot.
i have the sensation that is something related to openBLAS library, just to make a bet.
@seberg
Copy link
Member
seberg commented Jan 24, 2023

gh-22516 was experiencing a hard computer crash, it also includes a few settings to attempt to mitigate. However, that issue was with load. Nothing should just crash the machine, but if you suspect openblas is involved, the issue should give some things to try (i.e. reduce threads/pick a different kernel type).
Similar changes (picking other instructions) can be done within NumPy by setting an environment variable like NPY_DISABLE_CPU_FEATURES="SSE3,SSSE3,SSE41,POPCNT,SSE42 (this has nothing to do with openblas) see also here.

I am not sure where to point at exactly, since even if buggy, NumPy shouldn't be able to crash your machine...

@mattip
Copy link
Member
mattip commented Jan 24, 2023

Is this running inside a VM or other emulation software?

@jacopomoi
Copy link
Author

Is this running inside a VM or other emulation software?

Nope, it is running on Ubuntu 22.04. Conda is just a virtual environment for various python packages.
I can provide more details on motherboard ecc. if is needed.

@seberg thank you! I will try some workaround and i'll communicate the results

@jacopomoi
Copy link
Author

Ok, so:
i've made several attempts disabling SIMDs. i think it's completely related to microcode since disabling AVX2 it works (but sometimes fail). Anyway I've compiled and installed numpy 1.25.0 from source and seems fine for now.

@jacopomoi
Copy link
Author

i updated the BIOS microcode and everything is REALLY FINE. I close this issue

@seberg seberg mentioned this issue Feb 15, 2023
Closed
julianpryde added a commit to julianpryde/numpy that referenced this issue Mar 28, 2023
@julianpryde
BUG: fix numpygh-23082, allow Sequence[Union[type, np.dtype]] in ufuncs
3173755
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
3A42
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@seberg @mattip @jacopomoi
0