Skip to content
Siu Kwan Lam edited this page Dec 13, 2018 · 1 revision

Numba Meeting: 2018-11-13

Attendees: Stuart, Stan, Siu, Todd

0. Feature Discussion

  • LLVM 7 status
    • problem with common linkage symbol allocation during caching
    • TODO:
      • fix SVML with new patch first
      • Look at LLVM master to see if anyone is fiddling with this code
  • String status
    • string PR is in but pending literal/constant refactor PR
    • next step optimizations:
      • Single character optimizations
      • faster substring search
  • Clean up old numba style
    • prefer @overload over manual template
    • tanh impl mismatch: cpython vs numpy
  • Cast is not handling refct.

1. New issues

  • #3490 - ARMv7l np.trace test failing

  • #3489 - kwarg support in extending API @overload_method broken

  • #3486 - numba.stencil does not have a “mode” parameter

  • #3485 - Implement mode “wrap” for numba.stencil

  • #3483 - cuda.to_device ignore current context

  • #3482 - Getting PTXAS info from cuda kernel

  • #3475 Flushing STDOUT in nopython mode

  • #3474 - LoweringError when passing kwargs to print()

  • #3472 - List comprehension in jit function will not trigger undefined variable errors

Already Closed

  • #3479 - name The name of the device (e.g. “GeForce GTX 970”) issue
  • #3477 - Random Shuffle error

2. Open PRs

New

  • #3491 - Prevent faulthandler installation on armv7l.
    • ready to merge
  • #3488 - Remove dead script.
    • ready to merge
  • #3487 - Raise unsupported for kwargs given to print()
    • ready to merge
  • #3484 - Size Numba logo in docs in px units. Fixes #3313
    • ready to merge
  • #3480 - np.tri, np.tril, np.triu - default optional args
    • Very close to ready
    • General observation: handling keyword arguments and optional arguments is too hard. Need streamlining and documentation.

Already Merged

  • #3481 - Permit dtype argument as sole kwarg in np.eye
  • #3478 - Fix np.random.shuffle sideeffect

Old

  • #3468 - Add support for np.clip and ndarray.clip.
  • 3437 - Changes to accommodate LLVM 7.0.x
  • 3450 - [WIP] generated_jit for CUDA kernels
  • 3449 - [WIP] Allow matching non-array objects in find_callname()
  • 3414 - [WIP] Refactor Const type
  • 3399 - Add max_registers Option to cuda.jit
  • 3397 - Fix with-objmode warning
    • ready to merge
  • 3392 - Launch and attach gdb directly from Numba.
  • 3390 - typeinfer: use unknown_loc object instead of string literal
  • 3385 - conda recipe: whitelist libiomp5.dylib
  • 3382 - CUDA_ERROR_MISALIGNED_ADDRESS Using Multiple Const Arrays
    • ready to merge
  • 3162 - Support constant dtype string in nopython mode in functions like numpy.empty.
    • Need to resolve #3195
  • 3160 - First attempt at parallel diagnostics
    • Stuart will implement Todd's suggestion
  • 3134 - [WIP] Cfunc x86 abi
    • Needs re-review
  • 3046 - Pairwise sum implementation.
  • #2999 - Support LowLevelCallable
  • #2950 - Fix dispatcher to only consider contiguous-ness.
    • ready to merge
  • #2942 - Fix linkage nature (declspec(dllexport)) of some test functions
  • #2894: [WIP] Implement jitclass default constructor arguments.
  • #2817: [WIP] Emit LLVM optimization remarks

===========================

4. Next Release: Version 0.41, RC=Nov 19, Final=Nov 26, 2018

  • Type refactoring
  • Parallel diagnostics
  • LLVM 7
  • Initial string support
  • Finishing off stalled PRs
  • Usual collection of bug fixes and small features
Clone this wiki locally