stringlib fastsearch could be improved on 64-bit builds

[pitrou]

BPO	7607
Nosy	@pitrou, @giampaolo, @florentx
Superseder	bpo-7622: [patch] improve unicode methods: split() rsplit() and replace()

^{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 2010-01-13.08:11:07.879>
created_at = <Date 2009-12-30.22:23:17.944>
labels = ['interpreter-core', 'performance']
title = 'stringlib fastsearch could be improved on 64-bit builds'
updated_at = <Date 2010-01-13.08:11:07.878>
user = 'https://github.com/pitrou'

bugs.python.org fields:

activity = <Date 2010-01-13.08:11:07.878>
actor = 'pitrou'
assignee = 'none'
closed = True
closed_date = <Date 2010-01-13.08:11:07.879>
closer = 'pitrou'
components = ['Interpreter Core']
creation = <Date 2009-12-30.22:23:17.944>
creator = 'pitrou'
dependencies = []
files = []
hgrepos = []
issue_num = 7607
keywords = []
message_count = 3.0
messages = ['97065', '97163', '97479']
nosy_count = 3.0
nosy_names = ['pitrou', 'giampaolo.rodola', 'flox']
pr_nums = []
priority = 'normal'
resolution = 'out of date'
stage = None
status = 'closed'
superseder = '7622'
type = 'performance'
url = 'https://bugs.python.org/issue7607'
versions = ['Python 2.7', 'Python 3.2']

[pitrou]

The fastsearch algorithm uses a 32-bit mask for the boyer-moore
compression table but stores it as a long.
Since longs can be wider than 32 bits on some platforms (especially,
most 64-bit Unixes), the actual mask width could be platform-dependant
so as to improve the efficiency of the algorithm. Using the SIZEOF_LONG
constant would probably do the trick.

[florentx]

Another place where this optimization will apply:
"Objects/unicodeobject.c" for the bloom filters:

#define BLOOM(mask, ch) ((mask & (1 << ((ch) & 0x1F))))

[florentx]

Proposed in the patch for issue bpo-7622.