Skip to content
Siu Kwan Lam edited this page Jun 11, 2019 · 1 revision

Attendees: Stan, Val, Todd, Siu, Stuart

0. Feature Discussion

  • Actions from last meeting:
    • stuart: ticket for rewriting internals of functions that could be @overload'd
      • need to do still
    • stuart: legalize IR just prior to lowering (run postproc and compare)
      • much collateral damage
      • may be able to switch on soon
      • need richer IR comparison (IR diffs that understand "logical equivalence", but only approximately)
    • stuart: fix array inliner bad ir.Del emission
      • moving forward
    • siu: #4092 pointer moves in CUDA context
      • See PR #4114
      • still failing build farm
    • stan: #4086 user question about shuffling strings
      • responded and issue is closed
  • icc_rt builds to main
    • Stan will relabel
  • Any blockers for Numba 0.44.0 release on Thursday?
    • Stan will update final release checklist
  • capstone
    • Linking Numba IR to LLVM IR to assembly mostly working
    • capstone is a reverse engineering tool
    • Can determine when SIMD is being used with specific loops
    • capstone could use a conda package
  • fixing numba-integration-testing on hpat (next time?)
    • HPAT repo moved
    • Val will ping both Todd and Ehsan in future
  • Numba 0.45 Planning
    • July 1 release candidate
    • Main feature focus:
      • Val: Typed list, typed set is bonus
      • Stuart: Compiler rewrite passes
      • Siu: Caching

1. New issues

  • #4123 - Use Apply + Numba is possible?
  • #4122 - Multi-threading example slow
  • #4121 - np.mean ans np.einsum not supported by numba
  • #4120 - cannot work for view_as_windows
  • #4119 - not helping for np.convolve
  • #4117 - Exponentially increasing compilation time
  • *** #4115 - CFFI void* in nopython mode, determine Numba type of _cffi_backend.CData
    • Common use case
    • How to deal with void pointer management and wrapping?
    • Siu will add comments on specific tasks needed
  • #4111 - NUMBA_ENABLE_CUDASIM broken with record arrays
  • #4109 - Add support for range as argument to contains
  • *** #4108 - incorrect handling of array.shape with nopython=True
    • related to #4113
    • should sort this out by removing static_getitem rewrite pass and do something with literals when we need to
    • "Poison types"
  • #4102 - Numba release 0.44 release checklist

Already Closed

  • #4116 - cannot reassign the value to the elements of vector
  • #4106 - Numba 0.44 RC1 checklist
  • #4103 - Depend on enum34 using PEP 508 environment markers
  • #4101 - Error: "cannot import vectorize from numba" after "python numba_test.py"

2. Open PRs

  • #4118 - Fix inline assembly support on CPU.
  • #4114 - Support 3rd party activated CUDA context
  • *** #4113 - Add rewrite for semantic constants.
  • #4112 - better error messages for np.transpose and tuples
  • #4110 - Add range attrs, start, stop, step

Already merged

  • #4107 - Updates to the change log for 0.44.0
  • #4105 - Remove info about CUDA ENVVAR potential replacement
  • #4104 - Use PEP 508 compliant env markers for install deps

Old Active

4. Next Release: Version 0.44, RC=done, Final=ASAP

  • LLVM 8
    • fixes twine bug
    • fixes common symbol bug
  • Type-inferred Dict
  • Caching improvements requirement gathering
    • transitive dependency
    • bug related to storing runtime-only info
    • may punt on jitclasses
    • may punt on issues of (g)ufunc
    • future: pre-filled/package-able cache
      • open questions:
        • numba version pinning(?)
        • regen cache for different numba versions
        • regen at post-install
  • Add deprecation warnings
    • objmode
    • list/set reflection
Clone this wiki locally