photutils performance

performance.py

import time
import numpy as np
import photutils
from collections import OrderedDict


c = OrderedDict()

name = "Small data, single small aperture"
c[name] = {}
c[name]['dims']     = (20, 20)
c[name]['pos']      = (10., 10.)
c[name]['circ']     = (5.,)
c[name]['circ_ann'] = (5., 6.)
c[name]['elli']     = (5., 2., 0.5)
c[name]['elli_ann'] = (2., 5., 4., 0.5)
c[name]['iter']     = 1000

name = "Big data, single small aperture"
c[name] = {}
c[name]['dims']     = (1000, 1000)
c[name]['pos']      = (500., 500.)
c[name]['circ']     = (5.,)
c[name]['circ_ann'] = (5., 6.)
c[name]['elli']     = (5., 2., 0.5)
c[name]['elli_ann'] = (2., 5., 4., 0.5)
c[name]['iter']     = 1000

name = "Big data, single big aperture"
c[name] = {}
c[name]['dims']     = (1000, 1000)
c[name]['pos']      = (500., 500.)
c[name]['circ']     = (50.,)
c[name]['circ_ann'] = (50., 60.)
c[name]['elli']     = (50., 20., 0.5)
c[name]['elli_ann'] = (20., 50., 40., 0.5)
c[name]['iter']     = 10

name = "Small data, multiple small apertures"
c[name] = {}
c[name]['dims']     = (20, 20)
c[name]['pos']      = (np.random.uniform(5., 15., 1000), np.random.uniform(5., 15., 1000))
c[name]['circ']     = (5.,)
c[name]['circ_ann'] = (5., 6.)
c[name]['elli']     = (5., 2., 0.5)
c[name]['elli_ann'] = (2., 5., 4., 0.5)
c[name]['iter']     = 1

name = "Big data, multiple small apertures"
c[name] = {}
c[name]['dims']     = (1000, 1000)
c[name]['pos']      = (np.random.uniform(250., 750., 1000), np.random.uniform(250., 750., 1000))
c[name]['circ']     = (5.,)
c[name]['circ_ann'] = (5., 6.)
c[name]['elli']     = (5., 2., 0.5)
c[name]['elli_ann'] = (2., 5., 4., 0.5)
c[name]['iter']     = 1

name = "Big data, multiple small apertures, multiple per object"
c[name] = {}
c[name]['dims']     = (1000, 1000)
c[name]['pos']      = (np.random.uniform(250., 750., 1000), np.random.uniform(250., 750., 1000))
c[name]['circ']     = (np.linspace(1., 10., 10).reshape((10, 1)),)
c[name]['iter']     = 1

name = "Big data, multiple big apertures"
c[name] = {}
c[name]['dims']     = (1000, 1000)
c[name]['pos']      = (np.random.uniform(250., 750., 1000), np.random.uniform(250., 750., 1000))
c[name]['circ']     = (50.,)
c[name]['circ_ann'] = (50., 60.)
c[name]['elli']     = (50., 20., 0.5)
c[name]['elli_ann'] = (20., 50., 40., 0.5)
c[name]['iter']     = 1


f = {}
f['circ'] = photutils.aperture_circular
f['circ_ann'] = photutils.annulus_circular
f['elli'] = photutils.aperture_elliptical
f['elli_ann'] = photutils.annulus_elliptical


# Select subset of defined tests and functions to run, to save time.
names_to_run = ["Small data, single small aperture",
                "Big data, single small aperture",
                "Big data, single big aperture",
                "Small data, multiple small apertures",
                "Big data, multiple small apertures, multiple per object",
                "Big data, multiple small apertures"]
functions_to_run = ['circ', 'circ_ann']

for name in names_to_run:

    print "=" * 70
    print name

    x, y = c[name]['pos']
    data = np.ones(c[name]['dims'])

    print "%15s " % ("subpixels ="),
    for subpixels in [1, 5, 10, 'exact']:
        print str(subpixels).center(10) + " ",
    print ""
    
    print "-" * 70

    t0 = time.time()

    for t in functions_to_run:

        if t not in c[name]: continue
        print "%15s " % t,

        for subpixels in [1, 5, 10, 'exact']:
            time1 = time.time()
            for i in range(c[name]['iter']):
                f[t](data, x, y, *c[name][t], subpixels=subpixels)
            time2 = time.time()
            print "%10.5f " % ((time2 - time1)/c[name]['iter']*1000.),
        print ""

    t1 = time.time()

    print "-" * 70
    print 'Real time:', t1 - t0

results_macbookair.txt

Note: Times are in ms for a single calculation (except real time which is in s)

======================================================================
Small data, single small aperture
    subpixels =      1           5           10        exact     
----------------------------------------------------------------------
           circ       0.442       0.496       0.791       0.485  
       circ_ann       0.398       0.538       1.058       0.806  
           elli       0.451       0.601       1.044       1.128  
       elli_ann       0.529       0.750       1.460       1.225  
----------------------------------------------------------------------
Real time: 12.2021780014
======================================================================
Big data, single small aperture
    subpixels =      1           5           10        exact     
----------------------------------------------------------------------
           circ       0.370       0.448       0.679       0.493  
       circ_ann       0.412       0.654       1.045       0.675  
           elli       0.452       0.618       1.064       1.131  
       elli_ann       0.531       0.739       1.417       1.228  
----------------------------------------------------------------------
Real time: 11.9569151402
======================================================================
Big data, single big aperture
    subpixels =      1           5           10        exact     
----------------------------------------------------------------------
           circ       0.830      16.548      65.521       8.165  
       circ_ann       1.329      28.172     104.839      12.488  
           elli       1.259      35.536     148.531      50.808  
       elli_ann       1.530      48.288     201.938      44.846  
----------------------------------------------------------------------
Real time: 7.70686602592
======================================================================
Small data, multiple small apertures
    subpixels =      1           5           10        exact     
----------------------------------------------------------------------
           circ     239.065     316.130     541.385     351.750  
       circ_ann     256.642     385.747     818.626     544.331  
           elli     324.661     481.354     898.504     981.842  
       elli_ann     352.118     582.204    1261.494    1089.414  
----------------------------------------------------------------------
Real time: 9.42676615715
======================================================================
Big data, multiple small apertures
    subpixels =      1           5           10        exact     
----------------------------------------------------------------------
           circ     250.987     319.424     553.005     355.616  
       circ_ann     256.599     391.445     818.673     550.130  
           elli     310.647     608.553     955.709     982.566  
       elli_ann     371.161     584.954    1259.741    1080.193  
----------------------------------------------------------------------
Real time: 9.64983105659
======================================================================
Big data, multiple big apertures
    subpixels =      1           5           10        exact     
----------------------------------------------------------------------
           circ     783.926   14467.406   57824.835    5358.977  
       circ_ann     960.781   24231.674   99426.035   12621.256  
           elli    1098.681   33718.230  149080.460   51604.902  
       elli_ann    1434.037   44646.395  196105.284   45405.847  
----------------------------------------------------------------------