auderson comments

Results 35 comments of


                                            auderson

poor performance with try except block in parallel

A workaround seems to be wrapping the block into another function: ```python from numba.extending import register_jitable @register_jitable def sum_te(s, x): try: s += np.sum(x, axis=0) except Exception: pass @njit("void(f8[:, :],...

poor performance with try except block in parallel

@guilhermeleobas Looks interesting, your result shows `test_1` and `test_4` have similar performance. But on my Linux server it has 2x difference (306 vs 688).

nan as dict key different behavior in py3.10

I'm using a dict to count values in an array. So in this case I prefer `nan == nan` when it's used as dict keys.

nan as dict key different behavior in py3.10

> This is a KeyError so I guess it's from nan != nan I took some time reading numba's source code, and found numba dict also uses hash. So this...

nan as dict key different behavior in py3.10

Yeah, I mean can we treat nan as a constant and also return _PyHASH_NAN in python 3.10?

nan as dict key different behavior in py3.10

Thanks, this is a valid option > For keeping the NumPy behaviour, it might be easier to convert the floats to an integer with the same bits: `np.float64(your_value).view(np.uint64)`

pycc compatibilty with versioneer

Not a big problem for me, since switching `get_cmdclass() & cc.distutils_extension()` can get around.

Deprecate `numba.pycc`

Will we have any substitutions for pycc? AOT functionality is quite essential if we want to use numba in production.

Texture healing makes some strings look out of place

Hi, may I ask how do you set the fonts in pycharm? I'm using win11 and pycharm can't see the fonts I installed.

Support time functions like `time.time()`

After some investigations, I found the following code may be a feasible solution: ```python from numba import njit import ctypes import time # Access the _PyTime_AsSecondsDouble and _PyTime_GetSystemClock functions from...