Repository navigation
Optimize pymalloc for non PGO build #81724
Description
| BPO | 37543 |
|---|---|
| Nosy | @tim-one, @methane |
| PRs |
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 2019-07-17.12:24:34.174>
created_at = <Date 2019-07-10.10:59:54.080>
labels = ['interpreter-core', '3.9', 'performance']
title = 'Optimize pymalloc for non PGO build'
updated_at = <Date 2019-08-18.10:11:20.109>
user = 'https://github.com/methane'bugs.python.org fields:
activity = <Date 2019-08-18.10:11:20.109>
actor = 'hansrajdas'
assignee = 'none'
closed = True
closed_date = <Date 2019-07-17.12:24:34.174>
closer = 'methane'
components = ['Interpreter Core']
creation = <Date 2019-07-10.10:59:54.080>
creator = 'methane'
dependencies = []
files = []
hgrepos = []
issue_num = 37543
keywords = ['patch']
message_count = 2.0
messages = ['347615', '348057']
nosy_count = 2.0
nosy_names = ['tim.peters', 'methane']
pr_nums = ['14674', '15309']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'performance'
url = 'https://bugs.python.org/issue37543'
versions = ['Python 3.9']
Activity
methane commentedon Jul 10, 2019
When PGO is not used, compilers don't know which part is hot.
So gcc failed to inline hot code in pymalloc_alloc and pymalloc_free
into _PyObject_Malloc and _PyObject_Free. For example, only this code is inlined into _PyObject_Malloc.
But the hottest part is taking memory block from freelist in the pool.
To optimize it,
In PR 14674, pymalloc is now as fast as mimalloc in spectral_norm benchmark.
Mean +- std dev: [python-master] 199 ms +- 1 ms -> [python] 176 ms +- 1 ms: 1.13x faster (-11%)
methane commentedon Jul 17, 2019
New changeset fb26504 by Inada Naoki in branch 'master':
bpo-37543: optimize pymalloc (bpo-14674)
fb26504