Reputation: 429
My objective is to use Intel Python to accelerate some NumPy
process, mainly the FFT function. But when I use Intel Python
inside a docker it has the same efficiency that "normal" Python. How do I solve it?
Some useful information, to test the efficiency of those python I used this simple script:
import numpy as np
A = np.random.rand(5000,5000)
B = np.dot(A, A)
The results:
Using a Python 2.7
outside the docker:
time python bench.py
real 0m5,207s
user 0m16,714s
sys 0m0,960s
Using the Intel Python
outside the docker (using environment as the installation shows at this link):
time python bench.py
real 0m3,744s
user 0m6,269s
sys 0m0,184s
The difference is clear, but when I do the same test but with dockers environment the same test shows same time.
Using a Python 2.7
inside the docker:
time python bench.py
real 0m5.686s
user 0m18.060s
sys 0m1.045s
time python bench.py
real 0m4.871s
user 0m16.641s
sys 0m1.134s
time python bench.py
real 0m4.681s
user 0m16.805s
sys 0m0.635s
time python bench.py
real 0m4.804s
user 0m17.058s
sys 0m0.706s
time python bench.py
real 0m4.710s
user 0m17.057s
sys 0m0.607s
A result really close to the result outside docker.
Using the Intel Python
inside the docker:
time python bench.py
real 0m7.394s
user 0m18.744s
sys 0m1.931s
time python bench.py
real 0m4.705s
user 0m16.449s
sys 0m0.786s
time python bench.py
real 0m4.721s
user 0m16.495s
sys 0m0.809s
time python bench.py
real 0m4.735s
user 0m16.929s
sys 0m0.694s
Why the both pythons has the same efficiency inside the docker? I'm doing something wrong? My test is invalid for some reason?
And I'm pretty sure that I'm using Intel Python
inside the docker, because when just use the command python
on bash inside the docker this is the output:
Python 2.7.14 |Intel Corporation| (default, Aug 3 2018, 14:28:40)
[GCC 4.8.2 20140120 (Red Hat 4.8.2-15)] on linux2
On the other docker the return is:
Python 2.7.9 (default, Sep 25 2018, 20:42:16)
[GCC 4.9.2] on linux2
Upvotes: 0
Views: 218