enumerate() instantiation time reducing by using PEP 590 vectorcall

[corona10]

BPO	43706
Nosy	@vstinner, @JelleZijlstra, @corona10, @tirkarthi, @Fidget-Spinner
PRs	bpo-43706: Use PEP 590 vectorcall to speed up enumerate() #25154 bpo-46527: allow calling enumerate(iterable=...) again #30904
Files	bench_enumerate.py

^{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 2021-10-21.23:23:14.271>
created_at = <Date 2021-04-02.17:36:13.800>
labels = ['3.11', 'performance']
title = 'enumerate() instantiation time reducing by using PEP 590 vectorcall'
updated_at = <Date 2022-01-26.03:39:17.414>
user = 'https://github.com/corona10'

bugs.python.org fields:

activity = <Date 2022-01-26.03:39:17.414>
actor = 'JelleZijlstra'
assignee = 'none'
closed = True
closed_date = <Date 2021-10-21.23:23:14.271>
closer = 'corona10'
components = []
creation = <Date 2021-04-02.17:36:13.800>
creator = 'corona10'
dependencies = []
files = ['49929']
hgrepos = []
issue_num = 43706
keywords = ['patch']
message_count = 5.0
messages = ['390083', '404705', '404706', '404707', '404717']
nosy_count = 5.0
nosy_names = ['vstinner', 'JelleZijlstra', 'corona10', 'xtreak', 'kj']
pr_nums = ['25154', '30904']
priority = 'normal'
resolution = None
stage = 'resolved'
status = 'closed'
superseder = None
type = 'performance'
url = 'https://bugs.python.org/issue43706'
versions = ['Python 3.11']

[corona10]

Finally, I success to implement PEP-590 for enumerate type which is the well-used type in Python.

Amazingly enhanced!

+-----------------+------------------------+----------------------------+
| Benchmark | enumerate_bench_master | enumerate_bench_vectorcall |
+=================+========================+============================+
| bench enumerate | 527 ns | 380 ns: 1.39x faster |
+-----------------+------------------------+----------------------------+

[corona10]

See my last benchmark

#25154 (comment)

Same result in both Linux VM and Macbook pro.

[corona10]

New changeset 83f202a by Dong-hee Na in branch 'main':
bpo-43706: Use PEP-590 vectorcall to speed up enumerate() (GH-25154)
83f202a

[corona10]

Thanks Victor!

[corona10]

@vstinner

I noticed that the msg390083 was based on a non-optimized build.
#25154 (comment) ran with the same build option condition.

Here is the PGO + LTO based benchmark.
Still fast.

+-----------------+----------+----------------------+
| Benchmark | base_opt | vectorcall_opt |
+=================+==========+======================+
| bench enumerate | 384 ns | 277 ns: 1.39x faster |
+-----------------+----------+----------------------+