Python and STL efficiency

[Tim N. van der Leeuw]

Mc Osten wrote:
> Ray <ray_usenet at yahoo.com> wrote:
>
> > Yeah, my guess would be either he used the Debug configuration or he
> > actually created a Managed executable instead of a pure Win32
> > application. Sigh, now I can't wait to get home and try it out :)
>
> Can be. But I suppose a Managed should not get *that* slow.
> IronPython on Tim's machine is still faster than C++ (even though not as
> fast as CPython).
>
> --
> blog:  http://www.akropolix.net/rik0/blogs | Uccidete i filosofi,
> site:  http://www.akropolix.net/rik0/      | tenetevi riso e
> forum: http://www.akropolix.net/forum/     | bacchette per voi.

I have to admit to a stupid mistake, for which I feel quite ashamed - I
got the loop-size wrong in the Python code. So all Python results
posted by me were off by a factor of 10 :-(
I feel quite bad about that!

With the nr of loops corrected, Python on my laptop performs worse than
C++ under all circumstances, by a factor of about 2:

============ Python 2.4 =============
LeeuwT at nlshl-leeuwt ~/My Documents/Python
$ /cygdrive/c/Python24/python.exe SpeedTest.py
Begin Test
Number of unique string objects: 4
so long...
What do you know
fool
chicken crosses road
Number of unique string objects: 4000000
so long...
What do you know
fool
chicken crosses road
Fast - Elapsed: 4.239721 seconds
Slow - Elapsed: 11.883234 seconds

============ Python 2.5 =============
LeeuwT at nlshl-leeuwt ~/My Documents/Python
$ /cygdrive/c/Python25/python.exe SpeedTest.py
Begin Test
Number of unique string objects: 4
so long...
What do you know
fool
chicken crosses road
Number of unique string objects: 4000000
so long...
What do you know
fool
chicken crosses road
Fast - Elapsed: 4.031873 seconds
Slow - Elapsed: 11.314742 seconds


============ GCC 3.4.5, MinGW, -O6 =============
LeeuwT at nlshl-leeuwt ~/My Documents/Python
$ ./SpeedTest.exe
Begin Test
What do you know?
chicken crosses road
fool
so long...
What do you know?
chicken crosses road
fool
so long...
Fast - Elapsed: 2.088 seconds
Slow - Elapsed: 7.033 seconds

============ VC++ 6, 'release' build =============
LeeuwT at nlshl-leeuwt ~/My Documents/Python
$ ./SpeedTest_VC.exe
Begin Test
What do you know?
chicken crosses road
fool
so long...
What do you know?
chicken crosses road
fool
so long...
Fast - Elapsed: 4.585 seconds
Slow - Elapsed: 5.024 seconds

========== GCC 3.4.5, MinGW, -O6, with most optimized C++ code
==========
LeeuwT at nlshl-leeuwt ~/My Documents/Python
$ ./testcpp.exe
print_occurence_of_strings
What do you know?
chicken crosses road
fool
so long...
print_occurence_of_unique_strings
What do you know?
chicken crosses road
fool
so long...
print_occurence_of_unique_strings_compared_by_address
What do you know?
chicken crosses road
fool
so long...
strings : 2.338
unique strings : 1.109
compared by address : 0.23

LeeuwT at nlshl-leeuwt ~/My Documents/Python

============ IronPython 1.0rc2 =============

IronPython had a hard time coping with it; creating 4 million string
objects is a bit too much and the CLR was eating nearly a gigabyte of
memory near the end.
Here are the numbers:

IronPython 1.0.60816 on .NET 2.0.50727.42
Copyright (c) Microsoft Corporation. All rights reserved.
>>> import sys
>>> sys.path.append('C:/Documents and Settings/LeeuwT/My Documents/Python')
>>> import SpeedTest
>>> SpeedTest.run_test()
Begin Test
Number of unique string objects: 4
What do you know
so long...
chicken crosses road
fool
Number of unique string objects: 4000000
What do you know
so long...
chicken crosses road
fool
Fast - Elapsed: 10.501273 seconds
Slow - Elapsed: 371.047343 seconds
>>>

============ Java 1.6.0 b2 =============
Set size: 4
chicken crosses road
What do you know
fool
so long...
Set size: 4
chicken crosses road
What do you know
fool
so long...
Fast - Elapsed 1.003 seconds
Slow - Elapsed 3.96 seconds

============ Java 1.5.0  =============
Set size: 4
fool
What do you know
so long...
chicken crosses road
Set size: 4
fool
What do you know
so long...
chicken crosses road
Fast - Elapsed 1.754 seconds
Slow - Elapsed 5.044 seconds
=========================

Note that the Python code creates a set of all unique id's of all
objects in list a, and prints the length of this set, to verify that
all strings are really unique instances or duplicate instances. The C++
versions don't do that (at least not for 4 million strings); so Python
is at a slight disadvantage here. Printing the number of strings still
didn't help me catch the off-by-ten errors though.

I included a Java version of the program, and it looks like it performs
quite well compared to C++ both with jdk1.5 and jdk1.6.


I humbly apologize for my misinformation yesterday.

Regards,

--Tim