Minutes_2021_02_02
Valentin Haenel edited this page Feb 2, 2021
·
1 revision
Attendees:
NOTE: All communication is subject to the Numba Code of Conduct.
- 0.53.0.RC1 update:
- Seems like we're going to be on time.
- chunk-size PR pushed to 0.53+
- Ivan raises issue about threads in TBB dynamically spinning up and in some cases this could lead to non-repeatable results due to reductions happening in different orders.
- F-array PR, stu to review
- ArrayLike
- #6673 - Requesting Implementation of Asynchronous Data Copies and Asynchronous Barrier since CUDA 11
-
#6669 - Unstable software
- closed
-
#6668 - Returning complex types from objmode fails in some cases
- Siu need to look at
- module-getattr
- CC
- better error handling at boxer
- #6665 - Nested dictionary construction erroneously accepting heterogeneous value types
- #6662 - Update C code with Python API calls parsing tuples/kwargs to use bools where possible
-
#6672 - support for standard library sum()
- due to duplication
- #6674 - DO NOT MERGE. Wip/get thread id schedules
-
#6671 - Update environment creation example to python 3.8
- RTM
- #6670 - Fast path for overload attribute matching
- #6667 - [WIP] Signature cache for overloads for faster compilation
- #6666 - CUDA: Add a function to query whether the runtime version is supported.
- #6664 - Adds tests for PR 5659
- #6663 - Fix error caused by missing "-static" libraries defined for some platforms
- #6661 - CUDA: Support NVVM70 / CUDA 11.2
- Relase branch: Feb 4th
- https://github.com/numba/llvmlite/milestone/8
- https://github.com/numba/numba/milestone/44