Performance Tests



In [1]:

    
import numpy as np
import timeit
from csxtools.image import rotate90
from csxtools.fastccd import correct_images

Rotations

Start by testing the speed of the rotations of the arrays using the fast stack. Lets create some big arrays!



In [2]:

    
big_array = np.ones((10000,960,960), dtype=np.float32)



In [3]:

    
t = timeit.timeit('rotate90(big_array, "cw")', globals=globals(), number=10) / 10
print('Rotation of {} matrix takes {} seconds'.format(big_array.shape, t))









    



Rotation of (10000, 960, 960) matrix takes 12.532802192610689 seconds

FastCCD Image Corrections

Test a stack of fast CCD images



In [4]:

    
big_array = np.ones((10000,960,960), dtype=np.uint16)
bgnd = np.zeros((960,960), dtype=np.float32)
bgnd = np.array((bgnd,bgnd,bgnd), dtype=np.float32)
ff = np.ones((960,960), dtype=np.float32)



In [5]:

    
def adjust_gain_and_subtract_bg(raw_data, bg, ff):
    G1 = int('0b1100000000000000',2)
    G2 = int('0b1000000000000000',2)
    G8 = int('0b0000000000000000',2)
    M  = int('0b0001111111111111',2)
 
    images = ( ((raw_data & M) - bg[0]) * ((raw_data & G1) == G1) * 8 +
               ((raw_data & M) - bg[1]) * ((raw_data & G1) == G2) * 4 +
               ((raw_data & M) - bg[2]) * ((raw_data & G1) == G8) * 1 )
    images = images * ff
    return images



In [6]:

    
t = timeit.timeit('adjust_gain_and_subtract_bg(big_array,bgnd,ff)', globals=globals(), number=1)/1
print('Correction of {} matrix takes {} seconds using NUMPY'.format(big_array.shape, t))









    



Correction of (10000, 960, 960) matrix takes 300.50499161100015 seconds using NUMPY



In [7]:

    
t = timeit.timeit('correct_images(big_array,bgnd,ff)', globals=globals(), number=10)/10
print('Correction of {} matrix takes {} seconds using CSXTOOLS'.format(big_array.shape, t))









    



Correction of (10000, 960, 960) matrix takes 9.618055118992924 seconds using CSXTOOLS

Mean of image stack

Calculate the mean of a stack of images



In [8]:

    
big_array = np.ones((10000,960,960), dtype=np.float32)



In [9]:

    
t = timeit.timeit('np.mean(big_array, axis=0)', globals=globals(), number=10)
print('Numpy Mean of {} matrix takes {} seconds'.format(big_array.shape, t/10))









    



Numpy Mean of (10000, 960, 960) matrix takes 5.237616855790838 seconds



In [ ]: