-
Notifications
You must be signed in to change notification settings - Fork 1.1k
Minutes_2021_06_01
Valentin Haenel edited this page Jun 2, 2021
·
2 revisions
Attendees: Graham, Val, Stuart, ToddA, Siu, Brandon
NOTE: All communication is subject to the Numba Code of Conduct.
- #7069 - Support accessing static values of classes in JIT functions
- #7065 - Cudasim tests will fail when NumPy 1.20 is added to the test matrix
- #7063 - Numba installation scripts exhibit a race condition during parallel builds
-
#7062 - Compile time constant overflow
- Stuff can overflow and the compiler folds constants
- Perhaps leave some notes on the issue that this may be potentially solvable with overflow checks
- Pow function is due for a review anyway
- #7060 - prange with randomised / shuffled iteration
-
#7059 - CUDA typing / lowering can't find
contains_impl
for char sequences -
#7055 - Make repr on
TypeRef
more verbose
- #7071 - Native installation of Numba using pip on M1 macOS
- #7054 - TypingError: movemen is not defined
-
#7070 - Improve the issue templates and pull request template.
- Recently made the template much shorter
- Maybe we don't need this overhead?
- Most of the user-base are not compiler engineers
- It is sufficient to hand in a reproducer, just hand in a small code snippet
- We may want to keep the barrier of entry nice and low
- Perhaps most of this information can be queried
- Dumping environments destroys the search facility of issue trackers --> e.g. the Pandas issue tracker when you search for `numba``
- A few things may be good though:
- A refernecable issue when a PR is opened (!!!) <-- THIS MUST GO IN.
- Maybe a short sentence "You did X, you expected Y and Z happened instead".
- We may need to reject this PR, but since it is well intentioned, we need to be deliberate in our communication
- The effort should go into the
reproducer.py
rather than filling out long templates
- #7068 - Enhancing docs about PRNG seeding
- #7067 - Implement np.isclose
- #7066 - Fix #7065: Add expected exception messages for NumPy 1.20 to tests
- #7064 - Testhound/host array performance warning
- #7061 - Add ndarray allocator classmethod
- #7058 - Convert operator module functions to overloads.
- #7057 - Fix #7041: Add charseq registry to CUDA target
- #7056 - Change prange conversion approach to reuse header block.