I wanted to test how exponential moving average calculation would be faster with intel-numpy on intel python distribution. But my code runs slower on ipd when compared to native python.
- Intel Distribution for Python 3.6.3
- Intel(R) Xeon(R) CPU E5-2676 v3 @ 2.40GHz
- GCC 4.8.2 20140120 - Red Hat 4.8.2-15
def ema(): values = np.random.randint(2,9,(5000)) window = 20 start = time.time() weights = np.exp(np.linspace(-1., 0., window)) weights /= weights.sum() a = np.convolve(values, weights, mode='full')[:len(values)] a[:window] = a[window] end = time.time() print(end-start) return a
The numbers I got:
- Python - 0.00036263465881347656
- Intel Python Distribution - 0.005644321441650391
I don't think I've made a mistake in the installation as I get expected speed up with the sample code in below link:
What could be the reason? What's the perfect way to use intel distributions for optimizing ema calculations?
It is hard to conclude when execution times are on the order of milliseconds. Typically it is just a measurement noise. I advise increasing problem size or the number of repetitions to get conclusive results.
Along the same theme, I just installed Intel's Python Distribution on my i9 7980XE system running Windows 10. For my first performance test, I ran
import numpy as np
A = np.random.rand(30000,30000)
B = np.dot(A, A)
and was disappointed to see my CPU utilization nowhere near 100%, like I do when I run the same code with Python 3.7 and pip-installed numpy. Under full CPU load, the above code takes just over one minute to run under the latter set-up, but takes more than five minutes with Intel's Python (I lost patience and killed it).
I hope Intel can notice this issue. I am also using 7980XE and the fft2 performance is about 50% of the anaconda stock numpy.
Stock numpy takes 560ms to finish the test
import numpy as np a= np.random.random([8000,8000]) %timeit b=np.fft.fft2(a)
Intel python 2019 takes 980ms:
import numpy as np import mkl_fft as intel a=np.random.random([8000,8000]) %timeit b=intel.fft2(a)
7980XE cpu usage is only around 30% when using MKL, stock numpy can push 7980XE to 50%. Please fix this issue, giving up Intel python for now.