Minutes_2024_04_02
esc edited this page Apr 8, 2024
·
1 revision
Attendees: Siu Kwan Lam, Todd Anderson, Da Li, Guilherme, Graham markall, Tesla42 FPOC (last week): Guilherme FPOC (incoming): Graham
NOTE: All communication is subject to the Numba Code of Conduct.
Please refer to this calendar for the next meeting date.
- NP2 status
- NP2 binary compatibility PR is merged
- Buildfarm doing build on NP2 and test on NP1 + NP2
- also build on NP1 and test on NP1
- Refactoring builtin ufuncs to use libm
-
ufunc.resolve_dtypes
for NEP50 - Currently, splitting type system for NP2
@njit
def foo(a):
x = math.log2(a) # python float
y = np.log2(a) # np.float64
return x + y # np.float64
@njit
def bar(a):
x = foo(a) # np.float64
return math.log2(x)
- Plan to release Numba 0.60.0 ASAP when NP2 is official released
- IMPORTANT upcoming release do not have Typecast Semantic Compatibility against NP2.
- Re-enable Coverage on CI?
- related PR: https://github.com/numba/numba/pull/9508
- Siu will try
codecov
- State of parallel dictionary implementation:
- object based implementation: https://github.com/TheTesla/py-par-dict/blob/master/pardictimpl.py
- numba#9512 - Add support for more random functions.
- numba#9514 - Type inference leads to libNVVM bugs on 0.59.1
- numba#9515 - jitclass apraisal and performance revisit request.
- numba#9518 - numba's bug with prange without warning
-
numba#9519 - [Feature Request]
key_equal
,copy_key
,zero_key
in dict is slower than direct assignment if key type is primitive
- numba#9517 - Numba doing worse than raw Python on default example
- numba#9513 - Type system implementation #1: Added initial implementation for a new type system using redundancies.
-
numba#9516 - Replace
context.compile_internal
by@overload
in cmathimpl.py - numba#9520 - give primitive types a faster key&value compare, copy, and zero methods in TypedDict
- llvmlite#1036 - LLVM 15 conda recipe
2024-gantt: TBD 2023-gantt: https://github.com/numba/numba/issues/8971