-
Notifications
You must be signed in to change notification settings - Fork 234
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.
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
Newly-spawned tasks should re-set the device #851
Comments
API trace:
|
And under compute-sanitizer:
|
Well, this took me a bit. The problem is that your |
Wild! Thanks for the advice, pretty interesting. |
Describe the bug
We should be able to handle running multiple streams on multiple GPUs simultaneously using our old friends
@sync
and@async
. However, this currently leads to illegal address errors.To reproduce
The Minimal Working Example (MWE) for this bug:
Manifest.toml
master
for CUDA and dependencies.Expected behavior
The code should run without illegal address errors when Julia exits.
Version info
Details on Julia:
Details on CUDA:
Additional context
The script itself runs without issues, however when Julia exits with illegal address and finalizer errors.
Commenting out the assertion that fails in Julia's locking code (line 70 of
locks.h
), I get a segfault:The text was updated successfully, but these errors were encountered: