

In [1]:

    
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

from npgamma import calc_gamma



In [2]:

    
grid = 0.02
scale_factor = 1.3

x = np.arange(-1.8, 1.8 + 0.001, grid)
y = np.arange(-1.5, 1.5 + 0.001, grid)
coords = (y, x)

xx, yy = np.meshgrid(x, y)

dose_ref = np.exp(-(xx**10 + yy**10))
dose_eval = dose_ref * scale_factor

On a 4 core computer using more threads improves calculation speed



In [3]:

    
%%timeit

calc_gamma(
    coords, dose_ref,
    coords, dose_eval,
    0.1, 0.03, lower_dose_cutoff=0.2, 
    maximum_test_distance=0.2, num_threads=1)









    



1 loop, best of 3: 1.11 s per loop



In [4]:

    
%%timeit

calc_gamma(
    coords, dose_ref,
    coords, dose_eval,
    0.1, 0.03, lower_dose_cutoff=0.2, 
    maximum_test_distance=0.2, num_threads=2)









    



1 loop, best of 3: 699 ms per loop



In [5]:

    
%%timeit

calc_gamma(
    coords, dose_ref,
    coords, dose_eval,
    0.1, 0.03, lower_dose_cutoff=0.2, 
    maximum_test_distance=0.2, num_threads=3)









    



1 loop, best of 3: 526 ms per loop



In [6]:

    
%%timeit

calc_gamma(
    coords, dose_ref,
    coords, dose_eval,
    0.1, 0.03, lower_dose_cutoff=0.2, 
    maximum_test_distance=0.2, num_threads=4)









    



1 loop, best of 3: 500 ms per loop

When more threads than is available is used the calculation time is longer



In [7]:

    
%%timeit

calc_gamma(
    coords, dose_ref,
    coords, dose_eval,
    0.1, 0.03, lower_dose_cutoff=0.2, 
    maximum_test_distance=0.2, num_threads=5)









    



1 loop, best of 3: 575 ms per loop



In [8]:

    
%%timeit

calc_gamma(
    coords, dose_ref,
    coords, dose_eval,
    0.1, 0.03, lower_dose_cutoff=0.2, 
    maximum_test_distance=0.2, num_threads=10)









    



1 loop, best of 3: 688 ms per loop