Minutes_2018_01_18

Numba Meeting: 2018-01-18

Attendees: Siu, Todd, Stuart, Ehsan, Stefan

numba#2685: Radixsort.
- Dask sparse is looking for sort implementation that is faster than numpy.sort (quicksort).
numba#2688: Can't get numba work with CUDA
- Likely outdated driver
numba#2690: Array to Tuple conversion
- PR numba#2691. See below.
numba#2689: Numba 0.36.2 tests fail on npy114, py27, win64
- Sporadic error.
numba#2692: np.var and np.std impls do not handle complex types correctly

numba#2681: Allow Assigning to DeviceNDArrays
- User contribution based on some previous PRs fixing CUDA device arrays. @sklam reviewed and determined that the array-to-tuple conversion is better as an intrinsic than an implicit cast. Added dep on numba#2691.
numba#2691: Intrinsic to convert array to tuple
- Adding two intrinsic. 1) tuple_setitem for copying tuple & replacing element. 2) to_fixed_tuple for convert array to tuple.
numba#2674: ParallelAccelerator Refactor
- Reviewed.
numba#2672: Add ascontinguousarray
- @sklam. Still needs more work
numba#2668: Fix empty_like from readonly array
- Needs tests, not sure if submitter will add them, so @stuartarchibald will add commits to PR
numba#2660: Support bools from cffi in nopython
- Needs additional testing. @stuartarchibald will give submitter feedback for additional test scenarios.

CUDA 9.0/9.1 status:
- [stan] Figuring out best practice to get CUDA 9 on Ubuntu 14.04 environment for libgdf
Dask GDF: Apply transformations to groups
- Completed pygdf.apply* features for efficient row-wise and group-wise transformation.
Gufunc rewrite (aiming to start landing improvements in 0.38)
- Any feedback

Reach out to the community about the Gufunc rewrite.
Pipeline refactor question.
- How to handle https://github.com/numba/numba/pull/2398 ?
- Worried about opening up API that changes compiler behavior.
- Complicated by other issues with pipeline (i.e. rewrite pass mutate IR #2657)