jcrudy / cython-argsort Goto Github PK

import numpy as np

v = np.random.rand(1000)

%timeit v.argsort()
>> 19.6 µs ± 107 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

o = np.empty(v.shape[0], dtype="int")
%timeit cython_implementation.argsort(v, o)
>> 71 µs ± 275 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

This holds true for bigger arrays as well. Is there something I'm missing?

Hello,

when i tried the test_argsort.py (with n = 10000000), I get the following result:
Cython time: 2.513885974884033
Numpy time: 2.006075620651245

Shouldn't cython be faster? Do you also get the same results?

PS: I was new to cython and spent a lot of time making this work. it'd be nice if you could update readme with more details. Also in test_argsort.py,

import argsort
should be
from cyargsort import argsort
otherwise I get this error : SystemError: Parent module '' not loaded, cannot perform relative import

Thanks