Skip to main content

Line-granularity, thread-aware deterministic pure-python profiler

Project description

Line-granularity, thread-aware deterministic pure-python profiler

Inspired from Robert Kern’s line_profiler .

Overview

Python’s standard profiling tools have a callable-level granularity, which means it is only possible to tell which function is a hot-spot, not which lines in that function.

Robert Kern’s line_profiler is a very nice alternative providing line-level profiling granularity, but in my opinion it has a few drawbacks which (in addition to the attractive technical chalenge) made me start pprofile:

  • It is not pure-python. This choice makes sense for performance but makes usage with pypy difficult and requires installation (I value execution straight from checkout).

  • It requires source code modification to select what should be profiled. I prefer to have the option to do an in-depth, non-intrusive profiling.

  • As an effect of previous point, it does not have a notion above individual callable, annotating functions but not whole files - preventing module import profiling.

  • Profiling recursive code provides unexpected results (recursion cost is accumulated on callable’s first line) because it doesn’t track call stack. This may be unintended, and may be fixed at some point in line_profiler.

Usage

As a command:

$ pprofile some_python_executable

Once some_python_executable returns, prints annotated code of each file involved in the execution (output can be directed to a file using -o/–out arguments).

As a command with conflicting argument names: use “–” before profiled executable name:

$ pprofile -- foo --out bla

As a module:

import pprofile

profiler = pprofile.Profile()
def someHotSpotCallable():
    with profiler:
        # Some hot-spot code

Alternative to with, allowing to end profiling in a different place:

def someHotSpotCallable():
    profiler.enable()
    # Some hot-spot code
    someOtherFunction()

def someOtherFunction():
    # Some more hot-spot code
    profiler.disable()

Then, to display anotated source on stdout:

profiler.print_stats()

(several similar methods are available).

Sample output (threading.py removed from output):

$ pprofile dummy.py
0.0
55
9.26535896605e-05
6765
Total duration: 0.245515s
dummy.py
Line #|      Hits|         Time| Time per hit|      %|Source code
------+----------+-------------+-------------+-------+-----------
     1|         0|            0|            0|  0.00%|#!/usr/bin/env python
     2|         1|  7.15256e-06|  7.15256e-06|  0.00%|import threading
     3|         1|  0.000106812|  0.000106812|  0.04%|from dummy_module.fibo import fibo, sin
     4|         0|            0|            0|  0.00%|
     5|         1|  5.96046e-06|  5.96046e-06|  0.00%|def sin_printer(n):
     6|         2|    0.0359957|    0.0179979| 14.66%|    print sin(n)
     7|         0|            0|            0|  0.00%|
     8|         1|  4.05312e-06|  4.05312e-06|  0.00%|def main():
     9|         1|  1.21593e-05|  1.21593e-05|  0.00%|    t1 = threading.Thread(target=sin_printer, args=(0, ))
    10|         1|  1.19209e-05|  1.19209e-05|  0.00%|    t2 = threading.Thread(target=sin_printer, args=(3.1415, ))
    11|         1|  4.29153e-05|  4.29153e-05|  0.02%|    t1.start()
    12|         1|  0.000106812|  0.000106812|  0.04%|    print fibo(10)
    13|         1|  4.22001e-05|  4.22001e-05|  0.02%|    t2.start()
    14|         1|  5.10216e-05|  5.10216e-05|  0.02%|    print fibo(20)
    15|         1|  1.78814e-05|  1.78814e-05|  0.01%|    t1.join()
    16|         1|  1.19209e-05|  1.19209e-05|  0.00%|    t2.join()
    17|         0|            0|            0|  0.00%|
    18|         1|  5.00679e-06|  5.00679e-06|  0.00%|if __name__ == '__main__':
    19|         1|  1.38283e-05|  1.38283e-05|  0.01%|    main()
dummy_module/__init__.py
Line #|      Hits|         Time| Time per hit|      %|Source code
------+----------+-------------+-------------+-------+-----------
     1|         1|  2.14577e-06|  2.14577e-06|  0.00%|
dummy_module/fibo.py
Line #|      Hits|         Time| Time per hit|      %|Source code
------+----------+-------------+-------------+-------+-----------
     1|         1|  3.38554e-05|  3.38554e-05|  0.01%|import math
     2|         0|            0|            0|  0.00%|
     3|         1|  7.15256e-06|  7.15256e-06|  0.00%|def fibo(n):
     4|     13638|    0.0266435|  1.95362e-06| 10.85%|    assert n > 0, n
     5|     13638|    0.0526528|  3.86074e-06| 21.45%|    if n < 3:
     6|      6820|    0.0255547|  3.74702e-06| 10.41%|        return 1
     7|      6818|     0.108189|  1.58681e-05| 44.07%|    return fibo(n - 1) + fibo(n - 2)
     8|         0|            0|            0|  0.00%|
     9|         1|  5.00679e-06|  5.00679e-06|  0.00%|def sin(n):
    10|         2|  8.91685e-05|  4.45843e-05|  0.04%|    return math.sin(n)

Thread-aware profiling

ThreadProfile class provides the same features are Profile, but uses threading.settrace to propagate tracing to threading.Thread threads started after profiling is enabled.

Limitations

The time spent in another thread is not discounted from interrupted line. On the long run, it should not be a problem if switches are evenly distributed among lines, but threads executing fewer lines will appear as eating more cpu time than they really do.

This is not specific to simultaneous multi-thread profiling: profiling a single thread of a multi-threaded application will also be polluted by time spent in other threads.

Example (lines are reported as taking longer to execute when profiled along with another thread - although the other thread is not profiled):

$ ./ppsinglethread.py
Total duration: 1.00009s
./ppsinglethread.py
Line #|      Hits|         Time| Time per hit|      %|Source code
------+----------+-------------+-------------+-------+-----------
     1|         0|            0|            0|  0.00%|#!/usr/bin/env python
     2|         0|            0|            0|  0.00%|import threading
     3|         0|            0|            0|  0.00%|import pprofile
     4|         0|            0|            0|  0.00%|import time
     5|         0|            0|            0|  0.00%|import sys
     6|         0|            0|            0|  0.00%|
     7|         0|            0|            0|  0.00%|def func():
     8|         0|            0|            0|  0.00%|  # Busy loop, so context switches happe, so context switches happenn
     9|         1|  5.96046e-06|  5.96046e-06|  0.00%|  end = time.time() + 1
    10|    141331|     0.513656|  3.63442e-06| 51.36%|  while time.time() < end:
    11|    141330|     0.486344|   3.4412e-06| 48.63%|    pass
    12|         0|            0|            0|  0.00%|
    13|         0|            0|            0|  0.00%|# Single-treaded run
    14|         0|            0|            0|  0.00%|prof = pprofile.Profile()
    15|         0|            0|            0|  0.00%|with prof:
    16|         0|            0|            0|  0.00%|  func()
    17|         0|            0|            0|  0.00%|prof.annotate(sys.stdout, __file__)
    18|         0|            0|            0|  0.00%|
    19|         0|            0|            0|  0.00%|# Dual-threaded run
    20|         0|            0|            0|  0.00%|t1 = threading.Thread(target=func)
    21|         0|            0|            0|  0.00%|prof = pprofile.Profile()
    22|         0|            0|            0|  0.00%|with prof:
    23|         0|            0|            0|  0.00%|  t1.start()
    24|         0|            0|            0|  0.00%|  func()
    25|         0|            0|            0|  0.00%|  t1.join()
    26|         0|            0|            0|  0.00%|prof.annotate(sys.stdout, __file__)
Total duration: 1.03361s
./ppsinglethread.py
Line #|      Hits|         Time| Time per hit|      %|Source code
------+----------+-------------+-------------+-------+-----------
[...]
     9|         1|   3.8147e-06|   3.8147e-06|  0.00%|  end = time.time() + 1
    10|     59771|     0.487474|   8.1557e-06| 47.16%|  while time.time() < end:
    11|     59770|     0.512529|  8.57502e-06| 49.59%|    pass
[...]

This also means that the sum of the percentage of all lines can exceed 100%. It can reach the number of concurrent threads (200% with 2 threads being busy for the whole profiled executiong time, etc).

Example with 3 threads:

$ ./pprofile.py ppthread.py
Total duration: 1.00541s
ppthread.py
Line #|      Hits|         Time| Time per hit|      %|Source code
------+----------+-------------+-------------+-------+-----------
     1|         1|  6.19888e-06|  6.19888e-06|  0.00%|import threading
     2|         1|  1.50204e-05|  1.50204e-05|  0.00%|import time
     3|         0|            0|            0|  0.00%|
     4|         1|   3.8147e-06|   3.8147e-06|  0.00%|def func():
     5|         3|      3.00359|       1.0012|298.74%|  time.sleep(1)
     6|         0|            0|            0|  0.00%|
     7|         1|  1.40667e-05|  1.40667e-05|  0.00%|t1 = threading.Thread(target=func)
     8|         1|  1.09673e-05|  1.09673e-05|  0.00%|t2 = threading.Thread(target=func)
     9|         1|  2.88486e-05|  2.88486e-05|  0.00%|t1.start()
    10|         1|  4.69685e-05|  4.69685e-05|  0.00%|t2.start()
    11|         1|  5.79357e-05|  5.79357e-05|  0.01%|func()
    12|         1|  5.67436e-05|  5.67436e-05|  0.01%|t1.join()
    13|         1|  3.88622e-05|  3.88622e-05|  0.00%|t2.join()

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pprofile-1.1.tar.gz (8.9 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page