Minutes_2019_07_23

Attendees: Siu, Stuart, Val, Aaron, Todd, Stan, Eshan, James

0. Feature Discussion

• 0.45.1 patch release: https://github.com/numba/numba/milestone/29
  • numpy 1.17 changes (TBD)
  • fix semaphore leak (#4348)
  • fix typing of 0-d arrays (#4325)
  • cuda context management bug due to externally attached GPU context (i.e. pytorch, rapids cudf) (#4352)
• Priorities for 0.46
  • Target late Sept
  • 2. Rewrite passes
  • 3. Fix exception handling
  • (as time allows) Start work on new Numba IR implementation
    • Historical IR implementation is an object graph that do not make creating rewrite passes easy, and easy to create an inconsistent state. No "rollback" option.
    • Goal is create a new data structure that can solve these limitations.
    • Need to be able to round-trip between new IR container and old-style container so that existing rewrite passes will continue to work.
    • Need to prepare by refactoring existing rewrite passes to make them more amenable to a new IR representation.
    • Also need to document "best practices" for creating IR so that contributors will avoid known bad practices. (i.e Inline-closure-pass)
    • experimental repo for the new container: https://github.com/sklam/etude-okvmap
  • 1. Continuing on caching:
    • Catching transitive dependencies
    • More unusual cases (mixed mode, function parameters, etc)
  • New CI system!!!!!!!!
  • (proposal) Declarative typing option for @overload/@overload_method
• scumba - Numba extension adding overloads for scipy functions so they can be used from nopython mode
• llvm - patches causing behavioural deviations
  • need to make llvmlite pull down llvm and build it if not present

1. New issues

• **** #4348 - Global locks cause semaphore leaks.
  • needs to ping cpython dev
• #4347 - Can't use typed list in pandas constructor
  • needs to ping pandas dev
• **** #4344 - Meta issue: Error message improvements
• **** #4342 - SciPy 2019 feedback meta-issue
• #4340 - Dynamic container to fixed tuple helpers
  • good idea, we should do this
• #4335 - LoweringError: No definition for lowering <built-in function radians>(float64,) -> float64
  • quick fix, "good second issue"
• #4333 - Support numpy.isnan and numpy.isnat in cuda.jit kernels
  • need to write a generic implementation that can be used on CPU and GPU and add isnat
• #4332 - Definitions of a variable share the same reference to the Var object
  • discussion on implementation details

Already Closed

• #4339 - Memory leak when looping over slices
• #4336 - CUDA memory copy slower than C++
• #4329 - Numba 0.45.0 release checklist

2. Open PRs

• **** #4350 - Use process level locks for fork() only.
• #4349 - test using jitclass in typed-list
  • some jitclass limitations
• #4346 - Fix incorrect alg in isupper for ascii strings.
  • ready for merge
• #4345 - Replace "import *" with explicit imports in numba/types
  • seems reasonable, need to investigate bug it uncovered
• #4343 - [WIP]istitle() method for unicode strings implementation
• #4341 - Fix some coding lines at the top of some files (utf8 -> utf-8)
  • ready to merge
• #4338 - Fix #4299. Parfors reduction vars not deleted.
  • ready to merge
• #4337 - [WIP]title() method implementation for unicode strings
• **** #4331 - Automatic JIT of called functions
• #4330 - Loosen up typed container casting checks

Already merged

• #4334 - Update 0.45.0 changelog

4. Next Release: Version 0.45.1, Final=August 2

• bug fixes