[Python] Why numpy.array so slow when editing the data inside?

I wrote an algorithm that uses a long list. Since numpy.array should perform better in dealing with long data, I wrote another version using numpy.array.

list version:

import math
def PrimeTable(n:int) -> list[int]:
nb2m1 = n//2 - 1
l = [True]*nb2m1
for i in range(1,(math.isqrt(n)+1)//2):
if l[i-1]:
for j in range(i*3,nb2m1,i*2+1):
l[j] = False
return [2] + [i for i,v in zip(range(3,n,2), l) if v]


array version:

import math, numpy
def PrimeTable2(n:int) -> list[int]:
nb2m1 = n//2 - 1
l = numpy.full(nb2m1, True)
for i in range(1,(math.isqrt(n)+1)//2):
if l[i-1]:
for j in range(i*3,nb2m1,i*2+1):
l[j] = False
return [2] + [i for i,v in zip(range(3,n,2), l) if v]


It turns out that numpy.array version is 1x slower than list version. The graph:where y0 is PrimeTable and y1 is PrimeTable2

Why is numpy.array so slow, and how can I improve it?

Continue reading...