-
Notifications
You must be signed in to change notification settings - Fork 49
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
[cartesian] K-offset writes bound check is failing #1754
Comments
PR for temp fix: #1755 |
FlorianDeconinck
added a commit
that referenced
this issue
Nov 28, 2024
Following the issue logged as #1754 we are deactivating the K-offset write feature until we can figure out why it's failing. I will monitor any activity on the ticket if users are hit by this. --------- Co-authored-by: Hannes Vogt <hannes@havogt.de>
Probably related to the change in #1755: https://gitlab.com/cscs-ci/ci-testing/webhook-ci/mirrors/4455690602105886/4525297225819146/-/jobs/8498719906 |
Sorry @edopao that's a mistake on my end fixing that asap |
Fixing the Underlyign problem is that the |
FlorianDeconinck
changed the title
[cartesian] K-offset writes fail on GPU
[cartesian] K-offset writes bound check is failing
Jan 9, 2025
havogt
pushed a commit
that referenced
this issue
Jan 10, 2025
The `interval` analysis in the unit test `test_K_offset_write_conditional` fails to catch a mistake in the code that leads to a race condition. Work: - Fix the bad interval - Remove not needed restriction on CUDA version Further work to fix the underlying problem and the larger issue of bound check on variable indexing is covered [here](#1684) and [there](#1754)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Consider the test
column_physics_conditional
intests/cartesian_tests/integration_tests/multi_feature_tests/test_code_generation.py
. A fix inhttps://github.com/GridTools/gt4py/pull/1791
moved the interval tointerval(1,-1)
in line with the access and the domain defined.But a previous version had the interval at
inteval(1,None)
which leads to a race condition and an OOB.The stencil mix variable offset in K, where OOB can't be detected (see #1684), and scalar offset which should be detected.
Original ticket:
See failure on
gt:gpu
here; https://gitlab.com/cscs-ci/ci-testing/webhook-ci/mirrors/4455690602105886/4525297225819146/-/jobs/8381258008Translate tests is
test_K_offset_write_conditional
and should be modified when true fix comes into place.Temporary fix to deactivate the feature: #1755
The text was updated successfully, but these errors were encountered: