I have an algorithm for prime numbers, coded in python.
There I need to calculate exponentiations modulo large numbers.
To be more precise: For h and k let n = h* 2^k + 1 and I want to calculate 17^((n-1)/2) mod n.
It works, for example I can calculate this with k = 10^4 and h = 123456 in 2 second.
With k = 10^5 and h = 123.456, the calculation needs 20 minutes.
But I want even larger exponents, k = 10^9 for example. But I would like to know whether this is possible in a day or a week.
Can I estimate the time needed?
Or can python even show me the progress?
Here is my code:
import time
start = time.time()
k = 10**4
h = 123456
n = h*2**k+1
print(pow(17, (n-1)//2,n))
print("Time needed:", time.time() - start, "seconds.")
Edit:
I took the exponents k from 10^3 to 10^4 and measured the times it needed for the modular calculation. It’s an exponential growth. How do I estimate now how long it will take for k = 10^9?
How do I estimate now how long it will take for k = 10^9?
You can use numpy.polyfit
to produce an equation that estimates the time required for any k
.
First, we generate some data, which you’ve already done:
import time
import matplotlib.pyplot as plt
import numpy as np
x = []
y = []
i = 4 # Average each time sample across this many iterations
step = 100 # Increase k by this much between samples
end = 8001 # End of our range for k
h = 123456
# Measure the execution time for lots of k values
for k in range(1, end, step):
total_time = 0
# Find the average time across `i` runs to reduce some noise
for _ in range(i):
start = time.time()
n = h*2**k+1
result = pow(17, (n-1)//2,n)
stop = time.time()
total_time += stop-start
# Collect results
x.append(k)
y.append(total_time/i) # Average time across i runs
The execution time look quadratic in your plot, so let’s fit a 2nd degree polynomial to the data using numpy.polyfit and numpy.poly1d
# Find polynomial fit for
poly_degree = 2 # We'll plot this against the actual data to compare
xarr = np.array(x)
yarr = np.array(y)
# Determine polynomial coefficients
z = np.polyfit(xarr, yarr, poly_degree)
print(f"Coefficients: {z}")
# Coefficients: [ 3.52708730e-08 -1.24620082e-04 8.82253129e-02]
# Make a calculator for our polynomial
p = np.poly1d(z)
xp = np.linspace(1,x[-1],step)
# Plot the results
fig, ax = plt.subplots()
ax.plot(x, y) # Empiricle data
ax.plot(xp,p(xp),'--') # Our polynomial curve
ax.set(xlabel="k", ylabel="time (s)")
ax.grid()
plt.show()
How does the quadratic fit against the measured data?
Not bad!
Now we plug in 10**9 for k in the polynomial
k = 10**9
seconds = z[0]*k**2 + z[1]*k + z[2]
hours = seconds/3600
days = hours/24
years = days/365
print(f"years required when k == 10**9: {years}")
Output: about 1118 years on my computer with these specific parameters. What will really happen is that your computer will fall flat on its face when trying to calculate n for k = 10**9 and the polynomial won’t actually hold.
Edit: Another thing to note is that I sampled k on the order of 10^3 so any small jitter or inaccuracy can produce an error of many thousands of years once you project all the way out to 10^9
That’s an exponential complexity. I’d think that for
k=10^9
you’d need around 4-5 million years.