gh-121423: Improve import time of `socket` by writing `socket.errorTab` as a constant and lazy import modules#121424

Wulian233 · 2024-07-06T04:39:55Z

hyperfine:

$ hyperfine -w 16 './python -c "import socket"' Time (mean ± σ): 14.6 ms ± 0.7 ms [User: 11.9 ms, System: 2.6 ms] Range (min … max): 13.4 ms … 16.5 ms 192 runs $ git switch pr/121424 $ hyperfine -w 16 './python -c "import socket"' Time (mean ± σ): 13.9 ms ± 0.6 ms [User: 11.4 ms, System: 2.4 ms] Range (min … max): 12.6 ms … 15.5 ms 197 runs

≈ 30% faster

Issue: Speed up socket.errorTab and lazy import selectors #121423

Issue: Improve import time of various stdlib modules #118761

sobolevn

I don't think that your benchmark is correct. Because you put import socket into your setup part. So, the import time itself is not counted.

Lib/socket.py

Wulian233 · 2024-07-06T09:39:00Z

You are right, ↓ benchmark remove -s, slower 0.02us than -s because import

>python -m timeit -n 1000 "import socket" "print(socket.__all__)" 1000 loops, best of 5: 424 usec per loop >python -m timeit -n 1000 "import socket" "print(socket.__all__)" 1000 loops, best of 5: 371 usec per loop

1.14x faster

Misc/NEWS.d/next/Library/2024-07-06-12-37-10.gh-issue-121423.vnxrl4.rst

…nxrl4.rst Co-authored-by: Pieter Eendebak <[email protected]>

Misc/NEWS.d/next/Library/2024-07-06-12-37-10.gh-issue-121423.vnxrl4.rst

…nxrl4.rst Co-authored-by: Bénédikt Tran <[email protected]>

barry-scott · 2024-08-31T12:09:58Z

The benchmark only imports socket once. the other 999 imports are no-op as socket in in sys.modules.

Wulian233 · 2024-08-31T13:19:23Z

~~I carried out two parameter tests, one is with -s, is in the pr description, 1.15x. One is to do without -s in the comments 1.14x~~

Am I right? I'm not familiar with benchmark, sorry

I'm wrong

effigies · 2024-08-31T13:34:59Z

You can use hyperfine to measure the entire Python process, in which case using a python -c pass test is useful as a baseline:

❯ hyperfine -w 16 -u microsecond 'python -c pass'Benchmark 1: python -c pass Time (mean ± σ): 10921.2 µs ± 844.5 µs [User: 8745.1 µs, System: 2091.4 µs] Range (min … max): 9720.3 µs … 15591.1 µs 242 runs ❯ hyperfine -w 16 -u microsecond 'python -c "import socket; socket.__all__"'Benchmark 1: python -c "import socket; socket.__all__" Time (mean ± σ): 14554.3 µs ± 1154.4 µs [User: 12061.3 µs, System: 2368.5 µs] Range (min … max): 12826.1 µs … 20450.0 µs 206 runs

hugovk · 2024-08-31T15:34:17Z

Using a PGO+LTO build on macOS, so this will just be measuring the import selectors change, and saves about 1.1 ms:

❯ hyperfine --warmup 16 \ --prepare "git checkout socket"'./python.exe -c "import socket; socket.__all__"' \ --prepare "git checkout 6239d41527d5977aa5d44e4b894d719bc045860e"'./python.exe -c "import socket; socket.__all__"' Benchmark 1: ./python.exe -c "import socket; socket.__all__" Time (mean ± σ): 16.2 ms ± 0.8 ms [User: 13.1 ms, System: 2.5 ms] Range (min … max): 15.5 ms … 21.7 ms 91 runs Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs. Benchmark 2: ./python.exe -c "import socket; socket.__all__" Time (mean ± σ): 17.3 ms ± 1.0 ms [User: 13.9 ms, System: 2.7 ms] Range (min … max): 16.6 ms … 25.1 ms 88 runs Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs. Summary ./python.exe -c "import socket; socket.__all__" ran 1.06 ± 0.08 times faster than ./python.exe -c "import socket; socket.__all__"

hugovk

By the way, we have a general purpose issue for import time improvements -- #118761 -- shall we use that as an umbrella issue for this PR as well?

Using tuna to visualise import times (again, on macOS):

./python.exe -X importtime -c "import socket"2> import.log && tuna import.log

Before: 3ms

Current PR: 2ms

Also: 1ms

Moving the couple of import arrays into two functions that call it:

Lib/socket.py

Co-authored-by: Hugo van Kemenade <[email protected]>

vstinner · 2024-09-02T11:12:25Z

Lib/socket.py


 def _sendfile_use_sendfile(self, file, offset=0, count=None):
+ # Lazy import to improve module import time
+ import selectors


I suggest to use a global variable to avoid the import at each call:
globalselectorsifselectorsisNone: importselectors
You can define the selectors variable to None at the top of the file with a comment:
# module imported lazilyselectors=None

I checked several recent PRs making module loading lazy, but most of them do not use the pattern with a global variable (I suspect for readability reasons).
@vstinner Are there any other reasons besides the small performance improvement for using the global variable?

@serhiy-storchaka: Do you think that it's still useful in 2024 to use a global variable to avoid import selectors at each function call?

I'd just do the import. It's a mere dict lookup without a conditional when the import has already happened.

Just measure. It is more than a mere dict lookup (we also need to check that the module is not partially initialized, this adds 2 more dict lookups or like).
In this case, I think that the difference may be small even in comparison with a single os.fstat() call. The idiom proposed by @vstinner may be used when the whole function is very fast.

vstinner · 2024-09-02T11:15:53Z

Also: 1ms Moving the couple of import arrays into two functions that call it:

Can you also make the array module import lazy in this PR?

vstinner · 2024-09-02T11:44:41Z

I measured that the change saves 0.7 ms on import socket: 1.72 ms => 1.24 ms.

timeit:

$ ./python -m timeit -s 'import sys; state=dict(sys.modules)' 'import socket; del socket; sys.modules.clear(); sys.modules.update(state)' 200 loops, best of 5: 1.72 msec per loop $ git switch pr/121424 $ ./python -m timeit -s 'import sys; state=dict(sys.modules)' 'import socket; del socket; sys.modules.clear(); sys.modules.update(state)' 200 loops, best of 5: 1.24 msec per loop

hyperfine:

$ hyperfine -w 16 './python -c "import socket"' Time (mean ± σ): 14.6 ms ± 0.7 ms [User: 11.9 ms, System: 2.6 ms] Range (min … max): 13.4 ms … 16.5 ms 192 runs $ git switch pr/121424 $ hyperfine -w 16 './python -c "import socket"' Time (mean ± σ): 13.9 ms ± 0.6 ms [User: 11.4 ms, System: 2.4 ms] Range (min … max): 12.6 ms … 15.5 ms 197 runs

Wulian233 · 2024-09-02T13:02:27Z

Thank you! I just replaced the incorrect benchmark results in the PR with the correct ones

hugovk · 2024-09-02T14:50:34Z

Also: 1ms Moving the couple of import arrays into two functions that call it:
Can you also make the array module import lazy in this PR?

@Wulian233 Please could you do this as well?

Wulian233 · 2024-09-03T12:13:08Z

Of course! I just lazy imported array and haven't done full benchmarking yet. I also mentioned this optimization in 3.14.rst, do you think it should be included? @hugovk

vstinner

LGTM.

What's New entry:

which results in a 30% speed up in standard pyperformance benchmarks.

This sentence can be misunderstood as "everything is 30% faster".

Which pyperformance benchmark is now faster?

Lib/socket.py

serhiy-storchaka · 2024-09-02T18:07:28Z

Lib/socket.py


 def _sendfile_use_sendfile(self, file, offset=0, count=None):
+ # Lazy import to improve module import time
+ import selectors


Just measure. It is more than a mere dict lookup (we also need to check that the module is not partially initialized, this adds 2 more dict lookups or like).
In this case, I think that the difference may be small even in comparison with a single os.fstat() call. The idiom proposed by @vstinner may be used when the whole function is very fast.

Wulian233 · 2024-09-03T12:34:44Z

Improve import time of :mod:socket by lazy importing modules and
writing :data:!socket.errorTab as a constant, which results in
a 30% speed-up in the import time pyperformance benchmarks.

What do you think of this👀

hugovk · 2024-09-03T12:57:44Z

For What's New, we can follow the example @AlexWaygood wrote for 3.13, which grouped a few import improvements together:

Several standard library modules have had their import times significantly improved. For example, the import time of the typing module has been reduced by around a third by removing dependencies on re and contextlib. Other modules to enjoy import-time speedups include email.utils, enum, functools, importlib.metadata, and threading. (Contributed by Alex Waygood, Shantanu Jain, Adam Turner, Daniel Hollas, and others in gh-109653.)

Let's do the same with #118761 in 3.14. We have two under that issue so far.

I recommend we also group this PR under #118761 as well -> rename this PR title gh-118761: ....

So we can follow something like that now, or leave it out for now and add a grouped summary later.

serhiy-storchaka · 2024-09-03T13:02:40Z

Omit details that are not interested to the end user.

I would also remove any mention from Ehat's New -- this is an insignificant change. I am sure that if you measure import time of different modules, you will find larger changes between versions (maybe even between bugfix releases) without anybody noticing.

Wulian233 · 2024-09-03T13:17:34Z

I recommend we also group this PR under #118761 as well -> rename this PR title gh-118761: ....

Okay, now this pr belongs to 118761. I revert the changes to 3.14 so we can write them together future when there are more modules optimizations :)

hugovk · 2024-09-03T13:32:31Z

Misc/NEWS.d/next/Library/2024-07-06-12-37-10.gh-issue-121423.vnxrl4.rst

Please rename to Misc/NEWS.d/next/Library/2024-07-06-12-37-10.gh-issue-118761.vnxrl4.rst (containing gh-issue-118761) and we're ready to merge :)

#121423 is the right issue.

serhiy-storchaka

LGTM.

vstinner · 2024-09-04T10:01:32Z

Merged. Thanks for your nice enhancement @Wulian233.

Speed up socket.errotTab and lazy import selectors
1724a3c

bedevere-appbot mentioned this pull request Jul 6, 2024
Speed up socket.errorTab and lazy import selectors#121423
Closed

bedevere-appbot added the awaiting review label Jul 6, 2024

lint
6559ebb

Wulian233 changed the title ~~gh-121423: Speed up socket.errotTab and lazy import selectors~~gh-121423: Speed up socket.errorTab and lazy import selectorsJul 6, 2024

typo
984a3b2

sobolevn reviewed Jul 6, 2024
View reviewed changes

Lib/socket.py Outdated Show resolvedHide resolved

lower().startswith("win")
1823c9a

eendebakpt reviewed Jul 7, 2024
View reviewed changes

Misc/NEWS.d/next/Library/2024-07-06-12-37-10.gh-issue-121423.vnxrl4.rst Outdated Show resolvedHide resolved

sobolevn requested a review from gpshead July 7, 2024 07:24

Update Misc/NEWS.d/next/Library/2024-07-06-12-37-10.gh-issue-121423.v…
dce6cd2
…nxrl4.rst Co-authored-by: Pieter Eendebak <[email protected]>

Wulian233 changed the title ~~gh-121423: Speed up socket.errorTab and lazy import selectors~~gh-121423: Improve import time of socket by writing socket.errorTab as a constant and lazy import of selectorsJul 7, 2024

lint
276d0bb

picnixz reviewed Jul 7, 2024
View reviewed changes

Misc/NEWS.d/next/Library/2024-07-06-12-37-10.gh-issue-121423.vnxrl4.rst Outdated Show resolvedHide resolved

Update Misc/NEWS.d/next/Library/2024-07-06-12-37-10.gh-issue-121423.v…
feae2fa
…nxrl4.rst Co-authored-by: Bénédikt Tran <[email protected]>

hugovk reviewed Aug 31, 2024
View reviewed changes

Lib/socket.pyShow resolvedHide resolved

comment describing the intentional lazy import
bd771c8
Co-authored-by: Hugo van Kemenade <[email protected]>

gpshead approved these changes Aug 31, 2024
View reviewed changes

bedevere-appbot added awaiting merge and removed awaiting review labels Aug 31, 2024

Merge branch 'main' into socket
fa169f8

vstinner reviewed Sep 2, 2024
View reviewed changes

lazy array
1773d04

vstinner approved these changes Sep 3, 2024
View reviewed changes

serhiy-storchaka reviewed Sep 3, 2024
View reviewed changes

separate imports
412a3d2

Wulian233 changed the title ~~gh-121423: Improve import time of socket by writing socket.errorTab as a constant and lazy import of selectors~~gh-118761: Improve import time of socket by writing socket.errorTab as a constant and lazy import of selectorsSep 3, 2024

bedevere-appbot mentioned this pull request Sep 3, 2024
Improve import time of various stdlib modules #118761
Closed

undo 3.14.rst
fa49395

Wulian233 changed the title ~~gh-118761: Improve import time of socket by writing socket.errorTab as a constant and lazy import of selectors~~gh-118761: Improve import time of socket by writing socket.errorTab as a constant and lazy import modulesSep 3, 2024

hugovk reviewed Sep 3, 2024
View reviewed changes

serhiy-storchaka approved these changes Sep 4, 2024
View reviewed changes

vstinner changed the title ~~gh-118761: Improve import time of socket by writing socket.errorTab as a constant and lazy import modules~~gh-121423: Improve import time of socket by writing socket.errorTab as a constant and lazy import modulesSep 4, 2024

vstinner merged commit 7bd964d into python:mainSep 4, 2024

bedevere-appbot removed the awaiting merge label Sep 4, 2024

Wulian233 deleted the socket branch September 4, 2024 10:14

Uh oh!

gh-121423: Improve import time of socket by writing socket.errorTab as a constant and lazy import modules#121424

gh-121423: Improve import time of socket by writing socket.errorTab as a constant and lazy import modules #121424

Uh oh!

Conversation

Wulian233 commented Jul 6, 2024• edited by bedevere-app botLoading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

sobolevn left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Wulian233 commented Jul 6, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

barry-scott commented Aug 31, 2024

Uh oh!

Wulian233 commented Aug 31, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

effigies commented Aug 31, 2024

Uh oh!

hugovk commented Aug 31, 2024

Uh oh!

hugovk left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

vstinnerSep 2, 2024• edited by hugovk Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

eendebakptSep 2, 2024

Choose a reason for hiding this comment

Uh oh!

vstinnerSep 2, 2024

Choose a reason for hiding this comment

Uh oh!

gpsheadSep 2, 2024

Choose a reason for hiding this comment

Uh oh!

serhiy-storchakaSep 2, 2024

Choose a reason for hiding this comment

Uh oh!

vstinner commented Sep 2, 2024

Uh oh!

vstinner commented Sep 2, 2024

Uh oh!

Wulian233 commented Sep 2, 2024

Uh oh!

hugovk commented Sep 2, 2024

Uh oh!

Wulian233 commented Sep 3, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vstinner left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

serhiy-storchakaSep 2, 2024

Choose a reason for hiding this comment

Uh oh!

Wulian233 commented Sep 3, 2024

Uh oh!

hugovk commented Sep 3, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

serhiy-storchaka commented Sep 3, 2024

Uh oh!

Wulian233 commented Sep 3, 2024• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hugovkSep 3, 2024

Choose a reason for hiding this comment

Uh oh!

vstinnerSep 4, 2024

Choose a reason for hiding this comment

Uh oh!

gh-121423: Improve import time of `socket` by writing `socket.errorTab` as a constant and lazy import modules#121424

gh-121423: Improve import time of `socket` by writing `socket.errorTab` as a constant and lazy import modules #121424

Wulian233 commented Jul 6, 2024•
edited by bedevere-app bot
Loading

Wulian233 commented Jul 6, 2024•
edited
Loading

Wulian233 commented Aug 31, 2024•
edited
Loading

vstinnerSep 2, 2024•
edited by hugovk
Loading

Wulian233 commented Sep 3, 2024•
edited
Loading

hugovk commented Sep 3, 2024•
edited
Loading

Wulian233 commented Sep 3, 2024•
edited
Loading