Minutes_2020_02_04

Numba Meeting: 2020-02-04

Attendees: Aaron, Pearu, Val, Siu, Todd

Apologies: Graham, Stuart

0. Feature Discussion

• Chainsawing

  • Hoping to release lock on Tues/Weds. Refactor of purely file moves almost done.
  • https://github.com/numba/numba/pull/5197

• Issue discussion:

  • https://github.com/numba/numba/pull/4615
    • CI failing probably because of refactoring on master
    • TODO: Siu check on TBB status & Siu'll check/review again

1. New issues

• #5190 - No way to convert strings to numbers not in object mode
  • duplicate
• #5188 - Convert the numba CUDA array to Pytorch CUDA tensor
• #5187 - Add a separate type to indicate the undefined start and stop of the slice
  • need more info.
• #5186 - Incorrect conversion by elements of the float array to string data
• #5179 - TupleTestHashing unit tests fails on 32-bit Windows in Python 3.8
• #5178 - Multiple tests involving NaT datetimes fail in Python 3.8 with NumPy 1.18.1
• #5175 - @vectorize doesn't work with Numpy arrays (NotImplementedError)
• #5171 - TypingError with "in" operator comparing range() items with list
• #5170 - Assigning to DeviceNDArray with Ellipsis fails
• #5168 - [DOC] Publishing docs on cuda_memory
• #5167 - numba.errors.LoweringError with @njit(parallel=True)
  • parallel tuple issue
• #5165 - Can't unbox heterogeneous list
• #5164 - [DOC] from_cuda_array_interface
• #5163 - Adding MemoryPointer to public API
• #5158 - DeviceNDArray.bind() does not seem to bind the stream to self
• #5157 - RewriteArrayExprs pass incorrectly rewrites all Array-returning operations
• ** #5156 - Cannot add edge as dest node on Arch Linux
  • trouble with CFG and loop in py3.8
• #5154 - Implementing __bytes__ on DeviceNDArray
• #5152 - typed list fails to refine with tuple in interpreter
• #5149 - access to np.array data / tobytes etc

Already Closed

• #5180 - Memory leak with raised error in cascaded overloads
• #5166 - Syntax error installing from pip3
• #5160 - pip>=19.3 does not install 0.48 wheel on py38/Windows
• #5148 - Process finished with exit code -1073741819 (0xC0000005)

2. New Open PRs

• #5189 - cuda_array_interface not requiring context
• #5185 - extract code-gen utilities from closures
• #5184 - implement hoistable getitem for typed-list
• #5183 - pass 'inline' explicitly to overload
• #5182 - Add getitem for Record type when index is StringLiteral
• related:
  • #5181 - np.sum(..., axis=...): remove 0 <= axis <= 3 constraint
  • #5177 - Implementation of np.moveaxis()
  • #5173 - Implement numpy.core.multiarray.normalize_axis_index
  • #5169 - canonize tuple creation intrinsics
• #5176 - Normalize kws going into fold_arguments.
  • ready to merge
• #5174 - Nested modules https://github.com/numba/numba/issues/4739
  • ready to merge
• ** #5172 - fix version spec for llvmlite
  • bizzare
• **** #5162 - [DISCUSSION] Specify synchronization and lifetime semantics of CUDA Array Interface
• #5161 - Fix README.rst formatting
• #5159 - Fix README.rst
• #5155 - Explain what inlining at Numba IR level will do
• #5153 - Test numpy 1.18
• #5150 - More explicit llvmlite requirement in requirements.txt

Already merged/closed

• #5151 - implement None-typed typed-list

4. Next Release: Version 0.49.0, RC=March ???

• Requests for 0.49
  • TBD