Minutes_2019_09_17

Attendees: Pearu, Aaron, Stuart, Stan, Siu, Todd, James

0. Feature Discussion

Bump Python 3.8 to 0.47.1?
- Delay due to python3.8 packages are not ready
- Py3.8 to release on Oct 21
- Plan to do a Py3.8 patch at around that time

#4575 - NumPy allocator routines accept heterogeneous shape
#4574 - Default Compilation mode for CUDA Kernels
- Stan will respond
**** #4572 - Filter list comprehensions unsupported
- bug in list-comp
- may want breakup the list-comp passes
#4570 - Support arguments that are functions in (n)jit
**** #4569 - Error producing ndarray from parallel list comprehension
- likely prange issue with loop-comp (i.e. array-comp)
- Todd will take a look
#4566 - Define __cuda_array_interface__ equivalent for ROC GPUs?
#4563 - Unicode Array codegen triggers LLVM bug on PPC64LE
#4561 - c++ vs numba compiled code speed
#4558 - keep the parameter's reference(passed first time) and not free
**** #4555 - incorrect result with parfor and -=
- Todd will take a look
#4553 - How to deal with NumbaPendingDeprecationWarning?
- need to add set in 0.47
#4552 - datetimes don't implement __hash__
- stuart wrote a starter patch
#4550 - numpy array_equal implementaton
- See open PR
#4549 - Disabling on-disk cache
**** #4546 - Using Numba to generate LLVM/NVVM IR for different targets
- Pearu need hints to resolve this and the @overload issue
  - TODO: Siu
- may needs to refactor the backend to be more like the new style compiler passes
**** #4545 - Return a np.float32 and not a float
- we may need NumPyFloat32 types and friends
- probably low priority given other issues
#4543 - Request Numba to be able to accept keyword args with Print
#4542 - numba doesn't raise exception for broadcast error with +=

#4571 - Add axis argument for np.mean
#4568 - add np.array_equal support
#4567 - Add env var to disable performance warnings.
- maybe also have an option in the `parallel={...show_warnings=False...}``
#4565 - Fix typo in jit-compilation docs
#4564 - Skip unicode array tests on ppc64le that trigger an LLVM bug
#4560 - Add signal handler for SIGABRT.
- install_fatal_error_handler https://llvm.org/doxygen/ErrorHandling_8cpp_source.html
**** #4556 - 0.46 scheduled deprecations
**** #4554 - Refactor NRT C-API
#4551 - Add Basic CUDA Coordination Group Support
**** #4548 - Document numba.extending.intrinsic and inlining.
**** #4547 - Update to inliner cost model information.
- adding the ir.Call

Finish pending rewrite passes
Document best practices for constructing new compiler pipelines
Define autodiscovery system for Numba extensions (like numba_scipy or HPAT) that don't need direct user import
Allow opt-in dispatching of functions by literal value
Making caching aware of transitive dependencies
Define declarative typing system for @overload (to be used in future releases)
Numba Runtime C API for extensions to register reference-counted memory with the runtime.
Start using new CI system in parallel with existing one
Priority bug fixes:
- Low performance of JIT method calls (requested by Pandas devs)
- Others TBD