Path: csiph.com!fu-berlin.de!uni-berlin.de!not-for-mail
From: Terry Reedy <tjreedy@udel.edu>
Newsgroups: comp.lang.python
Subject: Re: Pyhon 2.x or 3.x, which is faster?
Date: Mon, 7 Mar 2016 22:39:49 -0500
Lines: 60
Message-ID: <mailman.32.1457408419.15725.python-list@python.org>
References: <mailman.238.1457265255.20602.python-list@python.org> <nbjmv7$ad5$1@dont-email.me> <87d1r6iltx.fsf@elektro.pacujo.net> <nbjp1e$jhv$1@dont-email.me> <nbjrjm$m16$1@gioia.aioe.org> <nbjvas$h22$1@dont-email.me> <mailman.17.1457364684.10335.python-list@python.org> <nbkhei$dg6$1@dont-email.me> <mailman.43.1457377845.10335.python-list@python.org> <nbknir$avu$1@dont-email.me> <mailman.49.1457383632.10335.python-list@python.org> <nbkvq2$j5s$1@dont-email.me> <mailman.7.1457395587.15725.python-list@python.org> <nbl81v$hii$1@dont-email.me> <mailman.18.1457400239.15725.python-list@python.org> <nbl9vp$mhm$1@dont-email.me>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:38.0) Gecko/20100101 Thunderbird/38.6.0
In-Reply-To: <nbl9vp$mhm$1@dont-email.me>
Precedence: list
Xref: csiph.com comp.lang.python:104319

On 3/7/2016 8:33 PM, BartC wrote:
>> On Tue, Mar 8, 2016 at 12:00 PM, BartC <bc@freeuk.com> wrote:

>>> def whiletest():
>>> |   i=0
>>> |   while i<=100000000:
>>> |   |   i+=1
>>>
>>> whiletest()
>>>
>>> Python 2.7:  8.4 seconds
>>> Python 3.1: 12.5 seconds
>>> Python 3.4: 18.0 seconds
>>>
>>> Even if you don't care about speed, you must admit that there appears
>>> to be
>>> something peculiar going on here: why would 3.4 take more than twice
>>> as long
>>> as 2.7? What do they keep doing to 3.x to cripple it on each new
>>> version?

> Let me ask you a follow-on question first: how slow does a new Python
> version have to be before even you would take notice?

We now run a suite of benchmarks daily on latest versions of 2.7 and 
default (3.6) to compare not to each other but to previous days to 
detect changes within either branch.

I verified your result with installed 64 bit 2.7.11 versus 3.5.1

import time
def test():
     i=0
     while i<=100000000:
         i+=1
start = time.time()
test()
print(time.time() - start)

4.4 (2.7) versus 10.7 (3.5)

Running loop at top level instead of inside the function doubled the 
time.  Replacing globals dict lookup with function locals array lookup 
really helps.

Next, I replaced the body of the function with
     for i in range(100000000): pass

This time, 3.5 wins: 3.8 versus 2.7.  Whoops, unfair. Change range to 
xrange in 2.7 and the time is 1.5.

Neither version optimizes the do-nothing loop away.  A further version 
of CPython might.  There are people working on improving the speed of 
CPython.  Integer operations are not their focus, though, because that 
can be done in numpy.  Text operations have been getting work, and at 
least one person is actively working on an ast optimizer.

-- 
Terry Jan Reedy