Minutes_2018_05_10

Numba Meeting: 2018-05-10

Attendees: Ehsan, Stuart, Stan

#2954: prange generates internal numba exception when using ndarray.flat
- Todd will take a look at
#2952: Function accepting njitted functions as arguments is slow bug
- Might be dispatcher speed issue
#2951: Partial type inference
- Will ask Siu to comment on this
#2949: Value-based Specialization and Dispatch
- Will ask Siu to take a look
- Code explosion?
- Does return type of function depend on value?
#2948: Branching on sets not supported in nopython mode
- Should be easy to fix
#2947: Attribute annotations in jitclass-wrapped classes cause errors
- Also not hard to fix
- Could we use annotations as a more readable substitute for locals in JIT functions?
#2943: @vectorize'd functions can't be pickled
- Known issue
- Will be fixed with refactoring of vectorize/guvectorize to use same callable object as JIT

#2953 [WIP] percentile and nanpercentile
#2950 Fix dispatcher to only consider contiguous-ness.
- Raises new issues related to read-only arrays
- Need to solve the issue of read-only
#2942 Fix linkage nature (declspec(dllexport)) of some test functions
- Need to run this PR through the build farm to check Windows

#2939 comprehensive SVML test
- Review in progress
#2930 Add configuration file and color schemes.
- Review in progress
#2929 Add Flag To Prevent Unneccessary D->H Copies
- Switched to PyCUDA-style wrappers, but still has global convenience flag
#2928 [WIP] Making error about untyped list more informative.
- Check with Siu as to whether this is ready to review
#2910: [WIP] More CUDA intrinsics
- Check if it works on GPU
- Make sure it goes thru the buildfarm before merging
- Blocked on how to simulate in CUDA simulator. See recent njwhite packages?
#2905: Fix for #2862
- need to move test to right location
#2897: [WIP] Fix line position of delete statement in numba ir
- Ready for review
#2894: [WIP] Implement jitclass default constructor arguments.
- Need to give some feedback to contributor on approach
#2869: WIP: use pairwise summation in sum
- Producing wrong results in some cases
- Open new PR with Stuart's non-recursive implementation
#2860: __cuda_array_interface__
- ready for review
#2840: Support list of refcounted types
- Review started
- Problems found in review that need to be fixed
#2817: [WIP] Emit LLVM optimization remarks
- Want to refactor to not require temp file on disk
#2793: Simplify and remove javascript from html_annotate templates.
- Need to push some refactoring onto the PR

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

Numba in Anaconda defaults channel status
- Stuart to add to changelog, .1 release is no functional change, just patch to llvmdev recipe.
Preparing for Python 3.7
- Stuart to add 3.7 to build system, open PR with patch for bytecode.
Stuart to add an issue full of things about Parfors that seem to be common questions with view of turning this into a FAQ.