I was performing some code timing of cells in jupyter notebook where I was writing to segments of a numpy array and found that breaking up the code produces a different processing time, compared to original combined code.
The folowing is a minimal example code. First setting up some values:
import numpy as np
V=10000
K=1000
zrs=np.random.random(V)
Then timing full code:
%%timeit
a=np.zeros(V*K)
for n in range(0,K):
a[n*V:(n+1)*V]=zrs
this returns: 45.6 ms ± 2.82 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
However, if I run the code in separate cells, I get:
%%timeit
a=np.zeros(V*K)
307 µs ± 2.33 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)
and, first declaring 'a':
a=np.zeros(V*K)
then running timeit on the second part of the code:
%%timeit
for n in range(0,K):
a[n*V:(n+1)*V]=zrs
9.52 ms ± 214 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
To conclude running complete code i get ~46ms and running the code in separate segments I get ~10ms!
Why? I am confused! What can I expect when running the code?
Help appreciated :)
Python 3.6.3, jupyter notebook 5.0.0
