Minutes_2018_03_01

Numba Meeting: 2018-03-01

Attendees: Siu, Stan, Stuart, Ehsan, Todd

New:

#2781 Lowering error with global default value for a parameter
- solution related to immutability of globals
#2776 Support mergesort in `np.sort()` / `np.argsort()`
- Just need to implement the algorithm, already have string keyword arg support
#2775 `side` kwarg for numpy.searchsorted
- Minor tweak to existing implement
#2774 reversed is not supported?
#2773 IntFlag converted to raw int inside njit function
#2772 mask_var not defined in _find_mask of numba/parfor.py
- Probably broken by refactoring
#2771 The tuple built-in is not supported in nopython mode
- Could handle some use cases with tuples
- Need some documentation to describe limitations of type system
#2770 Function calls with an unpacked list to *args are not handled by numba
- Segfault due to object mode handling
- Could support constant list *args in nopython?
- *args as tuple already works
#2769 compilation issue with infinite loop
- Not clear what is going on here.
- Needs more investigation

#2780 PowerPC reference counting fences & minor fixes
- Related: llvmlite#330 Add support for LLVM fence instruction
#2779 Implement np.random.permutation
- Random impl in Numba follows CPython rather than NumPy, which makes it hard to verify new NumPy random functions
- Can switch between Python and NumPy implementations
- Need some compatibility option if we switch RNG sequence on people.
#2778 Add More Device Array API Functions to CUDA Simulator
- Can review next week
#2777 Add support for np.correlate and np.convolve
- Ready for review
#2768 Don't Compile Code on Every Assignment
- Avoid regenerating code over and over
- Siu can review next week
#2767 Add len(..) Support to DeviceNDArray
- Siu review next week
#2765 Fix Bit-ness
- Review ASAP and merge
#2762 Support transpose with axes arguments.
- Reviewed and requested changes
#2748 Added Intel SVML optimizations as opt-out choice working by default
- Stuart has built LLVM and is testing
#2727 Changes to enable vectorization in ParallelAccelerator.
- Stuart and Todd will sort out
- Very close to merge, but likely future PR work related to SVML patch.
- Revealed new issues that will be patched in another patches.
#2660 Support bools from cffi in nopython. [now merged]