>100k alloc wasted on startup

[mikesfpy]

BPO	875689
Nosy	@tim-one, @rhettinger
Files	intobject.c.diff: Patch vs. Objects/intobject.c

^{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 = 'https://github.com/rhettinger'
closed_at = <Date 2004-02-08.18:59:53.000>
created_at = <Date 2004-01-12.21:30:50.000>
labels = ['interpreter-core']
title = '>100k alloc wasted on startup'
updated_at = <Date 2004-02-08.18:59:53.000>
user = 'https://bugs.python.org/mikesfpy'

bugs.python.org fields:

activity = <Date 2004-02-08.18:59:53.000>
actor = 'rhettinger'
assignee = 'rhettinger'
closed = True
closed_date = None
closer = None
components = ['Interpreter Core']
creation = <Date 2004-01-12.21:30:50.000>
creator = 'mikesfpy'
dependencies = []
files = ['5781']
hgrepos = []
issue_num = 875689
keywords = ['patch']
message_count = 3.0
messages = ['45236', '45237', '45238']
nosy_count = 3.0
nosy_names = ['tim.peters', 'rhettinger', 'mikesfpy']
pr_nums = []
priority = 'normal'
resolution = 'accepted'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue875689'
versions = []

[mikesfpy]

A glaring error in Objects/intobject.c function _PyInt_Init():

It allocates a whole BLOCK of ints each time through the loop
(i.e. 106 blocks of 1K size) and not just 106 int objects.

Allocating 106 blocks means a preallocation of more than 8600
int objects which is VERY unlikely to be needed by most programs.
So this bug wastes more than 100K on every startup.

The appended patch is vs. the current CVS. The bug is present
in 2.3.* and also in 2.2.? (not sure if I interpret the CVS logs right).

If you don't believe that this bug exists, just add a putc('#',stderr);
right before the code in fill_free_list() and see what happens when
you startup python. :-)

[tim-one]

Logged In: YES
user_id=31435

Good eye! Your analysis is correct, and the patch looks good
too. It's probably bad that the current code is spraying
the "frequently used" little integers all over memory too. The
patch also fixes that.

[rhettinger]

Logged In: YES
user_id=80475

Accepted and applied to Objects/intobject.c 2.108 and 2.105.8.1
.