Minutes_2019_10_29
Valentin Haenel edited this page Mar 30, 2020
·
1 revision
Attendees: Siu, Aaron, Ehsan, Stuart, Todd, Val, James
- Numba 0.46.1 Status
- Python 3.8 bytecodes really mess with control flow assumptions
- WIP PR: https://github.com/numba/numba/pull/4755
- looplifting is completely broken
- tuple hash algo changed
- good news: 90+% of testsuite passes
- pytest
- platform specific issues
- pytests bugs
- non blockers
- IR legalizer catching a lot of issues (reported as warnings)
- test slicing/selection works
- platform specific issues
- Opportunity for Aaron, Pearu, and James to work on implementing functions as first-class types #3405 (IIUC currently each function is treated as it's own unique type)
- How large an task do we suspect this is?
- If a large undertaking, is there a related smaller task that would still be useful to Numba?
- related issue:
- tasks:
- Signature type
- Runtime value of a Function
- a pointer to the actual function
-
#4751 - Import numba crashes after import s3fs
- pyarrow problem
-
#4750 - jitting a jitted function raises AttributeError
- add warnings, "Don't @jit @jit"
- **** #4749 - Allow casting numpy scalars
- should we allow implicit cast from Array scalar to scalar?
- thought 1: No. because performance linter
- thought 2: Yes. too easy to have a array scalar i.e.
a_numpy_array[idx]
can yield array scalar - doesn't sound dangerous. should try
- should we allow implicit cast from Array scalar to scalar?
- **** #4748 - Parallel failing in no python mode
- slicing with a 0-d array as index
- #4747 - type unification problem
- #4746 - Support numpy.indices
- **** #4742 - problem with jitted function when integer indexing with zero element
- parfor
- #4740 - Structured array of jitclass instances (jitclass <-> dtype)
- #4739 - AOT always installed to site-package folder not the module folder
- **** #4738 - Lowering Error when using @njit(parallel = True)
- parfor
- should have better error message
-
#4735 - LoweringError when yielding multiple elements from an njit function
- tuple typing issue
-
#4733 - Python 3.7 warns on
from collections import Iterable
- #4732 - 'cannot import name '_typeconv' from 'numba.typeconv'
-
#4731 -
.strip()
not stripping\t
.
- #4745 - "Call to cuPointerGetAttribute" errors with multi-process + multi-GPU + external GPU arrays
- #4743 - question about the efficientcy of numba.typed.dict
- #4744 - Fix_Py27_Build_Azure
- #4741 - Update choosing the "CUDA_PATH" for windows
- #4734 - prefer to import from collections.abc
-
CPython 3.8
-
Requests for 0.47 (last release for the year) - jitclass performance issues - llvm 9 trial - CTK libcudadevrt.a - CI needs to take 50% of current time - Val & Stu already looking at this - also checking Azure CI config to avoid wasting compute time - Caching: - transitive dependency - other issues: i.e. function as argument,
with objmode
- distributing cache - Immutable list and deprecating reflected list - Switch to pytest (see above) - Using Numba to generate LLVM/NVVM IR for different targets https://github.com/numba/numba/issues/4546 -@overload
for gpu