Guilherme Leobas comments

Results 187 comments of


                                            Guilherme Leobas

Skip test if libm is not found.

I'm on BigSur and all tests pass without this patch: https://gist.github.com/guilhermeleobas/29863546c0e766bb438a7d4aa23c8f04

Skip test if libm is not found.

> > > > I'm on BigSur and all tests pass without this patch: https://gist.github.com/guilhermeleobas/29863546c0e766bb438a7d4aa23c8f04 > > OK. Thank you! Just to confirm: this patch is NOT needed for you...

call_conv holds a reference to a different CPUContext object

I have a fix for it in #8134, which just replace `@cached_property` by `@property`.

Support list/sets as dict value type

I'll close this one since reflected containers are unstable in jitted code.

Function accepting njitted functions as arguments is slow

Hi, if possible I'd like to work on this issue.

Function accepting njitted functions as arguments is slow

I couldn't reproduce this bug ```python In [1]: import numba In [2]: from numba import njit In [3]: @njit ...: def foo(x): ...: return x ...: In [4]: def bar(x,...

Function accepting njitted functions as arguments is slow

Sorry, I forgot to put the decorator on `bar`. With the `@njit` decorator I got the same results. I will start working on this issue.

Function accepting njitted functions as arguments is slow

I don't have that much knowledge of how Numba works under the hood but the LLVM IR generated for both functions is almost the same, except for the entry basic...

Function accepting njitted functions as arguments is slow

I believe that the decision tree is necessary. There is no need to keep more than one implementation of the same method because if an argument is not used in...

Refactor `numba.np.arraymath` methods from lower_builtins to overloads

`np.std` gives a different result in Numba when the input is an empty array. This is not exclusive to your refactoring and perhaps should be addressed in a future PR....