Assorted ipaddress performance improvements

[pitrou]

BPO	21487
Nosy	@ncoghlan, @pitrou, @serhiy-storchaka
Files	ipaddr_perf.patch ipaddr_perf2.patch

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2014-05-15.19:05:59.847>
created_at = <Date 2014-05-12.20:04:23.234>
labels = ['library', 'performance']
title = 'Assorted ipaddress performance improvements'
updated_at = <Date 2014-05-15.19:05:59.847>
user = 'https://github.com/pitrou'

bugs.python.org fields:

activity = <Date 2014-05-15.19:05:59.847>
actor = 'pitrou'
assignee = 'none'
closed = True
closed_date = <Date 2014-05-15.19:05:59.847>
closer = 'pitrou'
components = ['Library (Lib)']
creation = <Date 2014-05-12.20:04:23.234>
creator = 'pitrou'
dependencies = []
files = ['35234', '35251']
hgrepos = []
issue_num = 21487
keywords = ['patch']
message_count = 6.0
messages = ['218357', '218515', '218525', '218568', '218607', '218622']
nosy_count = 5.0
nosy_names = ['ncoghlan', 'pitrou', 'pmoody', 'python-dev', 'serhiy.storchaka']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'performance'
url = 'https://bugs.python.org/issue21487'
versions = ['Python 3.5']

[pitrou]

Now that issue bpo-16531 has been committed, it becomes possible to make some operations faster. Attached patch makes summarize_address_range() ~2x faster and Network.subnets() ~4x faster.

[serhiy-storchaka]

Alternative implementations of _count_righthand_zero_bits():

def _count_righthand_zero_bits(number, bits):
    if not number:
        return bits
    return (~number & (number-1)).bit_length()

or

def _count_righthand_zero_bits(number, bits):
    if not number:
        return bits
    return (~(number | -number)).bit_length()

[pitrou]

Good point, this is a much faster implementation. Updated patch (and fixed the implementation to not return more than bits).

[serhiy-storchaka]

Looks as second alternative is few percents faster then first one.

[pitrou]

I find logical operations on negative numbers confusing in Python, so I'd rather stick with the first implementation.

[python-dev]

New changeset 2711677cf874 by Antoine Pitrou in branch 'default':
Issue bpo-21487: Optimize ipaddress.summarize_address_range() and ipaddress.{IPv4Network,IPv6Network}.subnets().
http://hg.python.org/cpython/rev/2711677cf874