Siu Kwan Lam
Siu Kwan Lam
Current implementation of first-class function is going through the cfunc wrapper and it is intentionally ignoring exception in this part of the code: https://github.com/numba/numba/blob/88fef5b350f5b9b26b26ec963147c9dbaf3d0998/numba/core/cpu.py#L201-L211 This is probably not what user...
Interesting technique. Can you add a test for literal list? e.g. ```python @njit def foo(): return [1,"2"], [1,"2"] a, b = foo() ``` Would mutating `a` also mutate `b`?
@testhound, yes, having the function resolution tests will help isolate any problem in the future. I recall that it takes us awhile to get to the root cause when we...
there's some more conflict due to recent merges
We have to bump this to the next release.
Note-to-self: implementation may need compressing
Note: - seems like the NumPy version is properly SIMD-vectorized but not the Numba version - we should check the assembly or optimized LLVM IR for these two versions.
I found that reduction based implementation is superior in the worst case scenario and a Numba jitted version of that is faster than NumPy's. The downside is that the scalar-shortcircuiting...
Here's a better impl: ```python @njit def any_or_generic(arr): c = False for i in np.nditer(arr): c |= i.item() return c @njit def any_or_chunked(arr): c = False if arr.flags.c_contiguous: car =...
/azp run