Specialize UNPACK_SEQUENCE

[brandtbucher]

BPO	46702
Nosy	@markshannon, @brandtbucher
PRs	bpo-46702: Specialize UNPACK_SEQUENCE #31240

^{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/brandtbucher'
closed_at = <Date 2022-02-18.21:19:36.450>
created_at = <Date 2022-02-10.00:35:53.502>
labels = ['interpreter-core', '3.11', 'performance']
title = 'Specialize UNPACK_SEQUENCE'
updated_at = <Date 2022-02-18.21:19:36.449>
user = 'https://github.com/brandtbucher'

bugs.python.org fields:

activity = <Date 2022-02-18.21:19:36.449>
actor = 'brandtbucher'
assignee = 'brandtbucher'
closed = True
closed_date = <Date 2022-02-18.21:19:36.450>
closer = 'brandtbucher'
components = ['Interpreter Core']
creation = <Date 2022-02-10.00:35:53.502>
creator = 'brandtbucher'
dependencies = []
files = []
hgrepos = []
issue_num = 46702
keywords = ['patch']
message_count = 4.0
messages = ['412960', '412962', '413340', '413512']
nosy_count = 2.0
nosy_names = ['Mark.Shannon', 'brandtbucher']
pr_nums = ['31240']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'performance'
url = 'https://bugs.python.org/issue46702'
versions = ['Python 3.11']

[brandtbucher]

UNPACK_SEQUENCE already has fast paths for tuples and lists, which make up (literally) 99% of unpackings in the benchmark suite. What's more, two-element tuples make up about two-thirds of all unpackings (though I actually suspect it's even higher, since the unpack_sequence benchmark is definitely skewing the results towards 10-element lists and tuples).

These specializations are trivial to implement and result in a solid 1% improvement overall.

[brandtbucher]

(I also plan on looking into an adaptive super-duper-instruction for UNPACK_SEQUENCE_TWO_TUPLE__STORE_FAST__STORE_FAST after the current PR has landed).

[brandtbucher]

New changeset a9da085 by Brandt Bucher in branch 'main':
bpo-46702: Specialize UNPACK_SEQUENCE (GH-31240)
a9da085

[brandtbucher]

Closing as UNPACK_SEQUENCE_TWO_TUPLE__STORE_FAST__STORE_FAST results in lots of hits, but no performance improvement.