Minutes_2022_05_24
esc edited this page May 25, 2022
·
1 revision
Attendees: Siu Kwan Lam, Andre Masella, Benjamin Graham, Graham Markall, LI Da, stuart, Todd A. Anderson, Val, Guilherme Leobas, Shannon Quinn, Luk
NOTE: All communication is subject to the Numba Code of Conduct.
Please refer to this calendar for the next meeting date.
- Numba 0.55.2 & llvmlite 0.38.1 Release status
- llvmlite 0.38.1 is out
- Numba 0.55.2
- backported 1.22.3
- Numba 0.56.0
- #8062 - Jit functions with optional shape arguments for creating numpy arrays
- #8066 - Numba 0.55.2 Checklist
- #8076 - Please document required versions in installation documentation
-
#8085 -
cache=True
slow forparallel
target - #8087 - ModuleNotFoundError: No module named 'numba.misc'
-
#8090 - BUG: Memory leak when
PyModule_AddObject
fails (static analyzer report) -
#8091 - BUG: Memory leak when the module created by
MOD_DEF
is not decreased on failure (static analyzer report) - #8092 - BUG: Memory leak due to new reference is not decreased on failure (static analyzer reports)
- #8095 - cuda stubs raising mypy SyntaxError
- #8101 - Bad multithreaded performance of allocations
- #8105 - Missing check for failed allocation in nrt.c
- #8107 - scipy-1.8.1 breaks numba. Downgrade fixes issues.
llvmlite
- #838 - Segfault calling llvm function from ctypes under Python 3.10
- #839 - Building from source: should llvm_11_consecutive_registers.patch be used?
- #840 - llvmlite 0.38.1 checklist
- #843 - Cross-building from M1 to x86 may not work
- #8063 - call insert method on reflect list(NamedTuple) reports segfault
- #8086 - Occasional finalizer exceptions in CUDA target
-
#8088 - BUG: Double free in function
MOD_INIT(_devicearray)
due to unnecessary refcnt decrement (static analyzer report) - #8096 - Potential regression in libopenblas 0.3.20 for M1
-
#8061 - Migrate
glue_lowering
tooverload_method
where easy - #8064 - Rename "nb:run_pass" to "numba:run_pass" and document it.
- #8065 - Fix PyLowering mishandling starargs
- #8068 - update changelog for 0.55.2
- #8077 - change return type of np.broadcast_shapes to a tuple
- #8081 - Fix erroneous array count in parallel gufunc kernel generation.
- #8083 - Misc/backport0.55 combine
- #8084 - Pin meta.yaml to llvmlite 0.38 series
- #8089 - [WIP] CUDA caching
- #8098 - Backport #8097: Exclude libopenblas 0.3.20 on osx-arm64
- #8102 - Combine numpy run_constrained in meta.yaml to the run requirements
- #8103 - Backport #8102 to fix numpy requirements
- #8104 - WIP: support numpy array comparison with string
- #8106 - Remove injection of atomic JIT functions into NRT memsys.
llvmlite:
-
#846 - Cherry-Pick: #842 -->
main
:Changelog for 0.38.1
- #8067 - Backport #8027: Support for NumPy 1.22
- #8069 - Install llvmlite 0.38 for Numba 0.55.*
- #8070 - Release0.55
- #8071 - Combine #8067 #8069
- #8072 - Combine patches needed for 0.55.2
- #8073 - minimal rename from master to main for CI only
- #8074 - 0.55.2/update max numpy
- #8075 - update max NumPy for 0.55.2
- #8078 - Backport #7804: update local references from master -> main
- #8079 - [WIP] on-disk caching of CUDA kernels
- #8080 - Fix windows test failure due to timeout when the machine is slow poss…
- #8082 - Backport #8080: fix windows failure due to timeout
- #8093 - Backport #7904: Support m1
- #8094 - Backport #8052 Ensure pthread is linked in when building for ppc64le.
- #8097 - Exclude libopenblas 0.3.20 on osx-arm64
- #8099 - Fix Py_DECREF use in case of error state (for devicearray).
- #8100 - Backport #7786 for 0.55.2: Remove dependency on intel-openmp for OSX
llvmlite