Continuing with the articles about profiling, another tool for profiling is called cProfile, a C extension introduced in Python 2.5. It is used for deterministic profiling (time is measured precisely rather than sampled).
The example I will develop is the same one as the inverse matrix generation; the difference is that the definition of the matrix's n value and its generation will be in a function called main. This function will then be called from cProfile.
Here is the code:
#!/usr/bin/env python
# -*- coding: utf-8 -*-
#Se importa numpy como np
import numpy as np
import cProfile
import sys
#funcion que genera matrices inversas.
def Inversa(n):
a = np.matrix(np.random.rand(n, n))
return a.I
def main():
#Se define una lista de tamaños de la matriz
tamagno = 2**np.arange(0,12)
#Se recorre la lista de tamaños y se invierte cada matriz con la
#funcion.
for n in tamagno:
Inversa(n)
#Se ejecuta la función main desde la llamada run de cProfile.
cProfile.run('main()')When you run the script, the result is the profiling of the inverse matrix generation:
ecrespo@jewel:~/proyectos/ernesto-ecrespo.blogspot/cprofile$ python matrizinversa.py
809 function calls in 5.764 seconds
Ordered by: standard name
ncalls tottime percall cumtime percall filename:lineno(function)
1 0.000 0.000 5.764 5.764 <string>:1(<module>)
24 0.000 0.000 0.058 0.002 defmatrix.py:233(__new__)
36 0.000 0.000 0.000 0.000 defmatrix.py:279(__array_finalize__)
12 0.000 0.000 0.000 0.000 defmatrix.py:55(asmatrix)
12 0.015 0.001 5.467 0.456 defmatrix.py:808(getI)
1 0.000 0.000 0.000 0.000 dual.py:12(<module>)
12 0.000 0.000 0.000 0.000 linalg.py:127(_to_native_byte_order)
12 0.000 0.000 0.257 0.021 linalg.py:139(_fastCopyAndTranspose)
12 0.000 0.000 0.000 0.000 linalg.py:151(_assertRank2)
12 0.000 0.000 0.000 0.000 linalg.py:157(_assertSquareness)
12 0.000 0.000 5.396 0.450 linalg.py:244(solve)
12 0.020 0.002 5.452 0.454 linalg.py:404(inv)
36 0.000 0.000 0.000 0.000 linalg.py:66(_makearray)
36 0.000 0.000 0.000 0.000 linalg.py:71(isComplexType)
24 0.000 0.000 0.000 0.000 linalg.py:84(_realType)
12 0.000 0.000 0.000 0.000 linalg.py:99(_commonType)
12 0.007 0.001 5.751 0.479 matrizinversa.py:12(Inversa)
1 0.013 0.013 5.764 5.764 matrizinversa.py:17(main)
36 0.000 0.000 0.000 0.000 numeric.py:167(asarray)
12 0.000 0.000 0.035 0.003 numeric.py:1830(identity)
36 0.000 0.000 0.000 0.000 {getattr}
72 0.000 0.000 0.000 0.000 {isinstance}
60 0.000 0.000 0.000 0.000 {issubclass}
60 0.000 0.000 0.000 0.000 {len}
12 0.000 0.000 0.000 0.000 {max}
24 0.000 0.000 0.000 0.000 {method '__array_prepare__' of 'numpy.ndarray' objects}
24 0.000 0.000 0.000 0.000 {method 'append' of 'list' objects}
12 0.059 0.005 0.059 0.005 {method 'astype' of 'numpy.ndarray' objects}
12 0.058 0.005 0.058 0.005 {method 'copy' of 'numpy.ndarray' objects}
1 0.000 0.000 0.000 0.000 {method 'disable' of '_lsprof.Profiler' objects}
24 0.000 0.000 0.000 0.000 {method 'get' of 'dict' objects}
12 0.220 0.018 0.220 0.018 {method 'rand' of 'mtrand.RandomState' objects}
12 0.000 0.000 0.000 0.000 {method 'transpose' of 'numpy.ndarray' objects}
12 0.000 0.000 0.000 0.000 {method 'view' of 'numpy.ndarray' objects}
12 0.000 0.000 0.000 0.000 {min}
24 0.257 0.011 0.257 0.011 {numpy.core.multiarray._fastCopyAndTranspose}
1 0.000 0.000 0.000 0.000 {numpy.core.multiarray.arange}
36 0.000 0.000 0.000 0.000 {numpy.core.multiarray.array}
24 0.035 0.001 0.035 0.001 {numpy.core.multiarray.zeros}
12 5.078 0.423 5.078 0.423 {numpy.linalg.lapack_lite.dgesv}The output is the same as the one shown in the article on timeit with ipython. In addition to the timeit article, the meaning of each column is shown below:
ncalls: Number of calls.tottime: Total time spent in a function.percall: Time per call, calculated as the total time divided by the number of calls.cumtime: Cumulative time spent in a function and in the functions it calls, including recursive calls.
You can download the script code from bitbucket (broken link).
Make a donation! If you liked the article you can make a donation in Bitcoin (BTC) using the digital wallet of your choice to the following address: 17MtNybhdkA9GV3UNS6BTwPcuhjXoPrSzV
Or by scanning the QR code from your wallet:
