Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

#

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Jump to bottom

Queue write only after processing all buffers #445

Merged
merged 6 commits into from
Sep 9, 2022
Merged

Queue write only after processing all buffers #445

merged 6 commits into from
Sep 9, 2022

Conversation

jakirkham
Copy link
Contributor

Partially addresses ( #404 )

Delays calling _queue_write until writelines has added all buffers with _write. Should allow libuv to leveraged vectorized IO (scatter) via calling sendmsg under-the-hood.

@jakirkham
Queue write only after processing all buffers
c8fc56d 
Allows `writelines` to leverage vectorized IO within uvloop to send
multiple buffers in one `sendmsg` call.
@jakirkham
Add copyright header
a12214a
jakirkham
jakirkham commented Oct 1, 2021
View reviewed changes
@jakirkham jakirkham mentioned this pull request Oct 1, 2021
Open
@jakirkham
Copy link
Contributor Author

Also given GitHub's new policy regarding running CI. Think someone needs to approve running CI here for that to happen

@jakirkham jakirkham requested a review from 1st1 October 5, 2021 18:40
@jakirkham
Copy link
Contributor Author

@1st1 any thoughts here?

Also need some help running the tests. GH doesn't run tests for new contributors on any commit without approval by default

@jakirkham jakirkham mentioned this pull request Oct 21, 2021
Merged
@jakirkham
Copy link
Contributor Author

@fantix, do you have any thoughts on this one? 🙂

@jakirkham
Copy link
Contributor Author

Friendly nudge 😉

@fantix
Copy link
Member

fantix commented Aug 13, 2022

Thanks for the PR and sorry for the long waiting!

Should allow libuv to leveraged vectorized IO (scatter) via calling sendmsg under-the-hood.

Hmm do you have some test results to show if gathered send is/not working before/after this change? I'm asking because I think _queue_write() doesn't call _exec_write() until the next loop iteration, so calling _queue_write() multiple times in a tight loop in writelines() is equivalent to calling it just once. I'll run a test later when I come back to this one.

1st1
1st1 approved these changes Aug 23, 2022
View reviewed changes
Copy link
Member

@1st1 1st1 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@fantix looks good. Please check that we call _queue_write whenever we call _write, but overall this is good.

@1st1
Copy link
Member

1st1 commented Aug 23, 2022

Hmm do you have some test results to show if gathered send is/not working before/after this change? I'm asking because I think _queue_write() doesn't call _exec_write() until the next loop iteration, so calling _queue_write() multiple times in a tight loop in writelines()

Yeah, this has to be benchmarked, ideally, but the change still looks good & logical. I think we should go with it.

Perhaps we should reshape it a little -- add a default arg "queue_write=True" to _write and call it with "False" from writelines() (to make this internal API a bit easier to use)

@jakirkham
Copy link
Contributor Author

Thanks all! 🙏

Yeah agree benchmarking makes sense. Is there a preferred set of benchmarks here we should try with initially? Also is there somewhere to add new benchmarks?

As to context about this change, did a fair bit of sleuthing in issue ( #404 ), which included a bit of digging into libuv itself. Could always be missing something. So would certainly appreciate feedback on that.

Sure happy to tweak the API to make things a bit more flexible.

@fantix fantix requested a review from 1st1 September 9, 2022 17:47
1st1
1st1 reviewed Sep 9, 2022
View reviewed changes
@fantix fantix merged commit 9c6ecb6 into MagicStack:master Sep 9, 2022
@fantix
Copy link
Member

fantix commented Sep 9, 2022

@jakirkham thank you for starting this anyways! Please feel free to comment on the PR or the original issue.

fantix added a commit that referenced this pull request Sep 9, 2022
@fantix
CRF: rename _start_write() to _initiate_write()
cdab9d3 
Tweaked the comments too.

Refs #445
fantix added a commit that referenced this pull request Sep 13, 2022
@fantix
uvloop v0.17.0
c4c3f9f 
This release adds Python 3.11 support, updates bundled libuv to 1.43.0
and fixes a handful of issues.

Changes
=======

* Expose uv_loop_t pointer for integration with other C-extensions (#310)
  (by @pranavtbhat in b332eb8 for #310)

* Support python 3.11+ (#473)
  (by @zeroday0619 in 8e42921 for #473)

* Expose libuv uv_fs_event functionality (#474)
  (by @jensbjorgensen @fantix in 74d381e for #474)

* Activate debug mode when `-X dev` is used
  (by @jack1142 in 637a77a)

* Expose uv_version() for libuv API compatibility (#491)
  (by @fantix in 089f6cb for #491)

* Fix loop.getaddrinfo() and tests (#495)
  (by @fantix in 598b16f for #495)

* Bump to libuv 1.43.0
  (by @fantix in 94e5e53)

Fixes
=====

* _TransProtPair is no longer defined in asyncio.events
  (by @jensbjorgensen in fae5f7f)

* use a TypeVar for asyncio.BaseProtocol (#478)
  (by @graingert in 3aacb35 for #478)

* Fix segfault in TimerHandle.when() after cleared
  (by @jensbjorgensen in c39afff for #469)

* Avoid self._errpipe_write double close (#466)
  (by @graingert in 72140d7 for #466)

* Fix typo in test (#456)
  (by @kianmeng in 033d52d for #456)

* Fix potential infinite loop (#446)
  (by @kfur in ada43c0 for #446)

* use a stack of self._fds_to_close to prevent double closes (#481)
  (by @graingert in 3214cf6 for #481)

* Fix incorrect main thread id value forking from a thread  (#453)
  (by @horpto @fantix in e7934c8 for #453)

* create_subprocess_exec should treat env={} as empty environment (#439) (#454)
  (by @byllyfish in e04637e for #439)

* Queue write only after processing all buffers (#445)
  (by @jakirkham @fantix in 9c6ecb6 for #445)

* Drop Python 3.6 support for thread ident
  (by @fantix in 9c37930)

* bugfix: write to another transport in resume_writing() fails (#498)
  (by @fantix in d2deffe for #498)

Build
=====

* Upgrade GitHub Actions (#477) (#480)
  (by @cclauss in fcbf422 for #477, 1008694 for #480)

* typo `same as same`
  (by @YoSTEALTH in fedba80)

* setup.py: allow to override extra_compile_args (#443)
  (by @giuliobenetti in a130375 for #443)

* Drop hack in setup.py in finalize_options (492)
  (by @fantix in 2f1bc83 for #492)

* Fix tests invocation on release CI worklow (#489)
  (by @ben9923 in d6a2b59 for #489)

Documentation
=============

* use asyncio.Runner loop_factory on 3.11+ (#472)
  (by @graingert in 31ba48c for #472)

* Fix CI badge in docs, remove remaining Travis CI references from docs
  (by @Nothing4You in c6901a7)

* Fix typo in README
  (by @monosans in 73d7253)
@fantix fantix mentioned this pull request Sep 13, 2022
Merged
fantix added a commit that referenced this pull request Sep 14, 2022
@fantix
uvloop v0.17.0
afb3268 
This release adds Python 3.11 support, updates bundled libuv to 1.43.0
and fixes a handful of issues.

Changes
=======

* Expose uv_loop_t pointer for integration with other C-extensions (#310)
  (by @pranavtbhat in b332eb8 for #310)

* Support python 3.11+ (#473)
  (by @zeroday0619 in 8e42921 for #473)

* Expose libuv uv_fs_event functionality (#474)
  (by @jensbjorgensen @fantix in 74d381e for #474)

* Activate debug mode when `-X dev` is used
  (by @jack1142 in 637a77a)

* Expose uv_version() for libuv API compatibility (#491)
  (by @fantix in 089f6cb for #491)

* Fix loop.getaddrinfo() and tests (#495)
  (by @fantix in 598b16f for #495)

* Bump to libuv 1.43.0
  (by @fantix in 94e5e53)

Fixes
=====

* _TransProtPair is no longer defined in asyncio.events
  (by @jensbjorgensen in fae5f7f)

* use a TypeVar for asyncio.BaseProtocol (#478)
  (by @graingert in 3aacb35 for #478)

* Fix segfault in TimerHandle.when() after cleared
  (by @jensbjorgensen in c39afff for #469)

* Avoid self._errpipe_write double close (#466)
  (by @graingert in 72140d7 for #466)

* Fix typo in test (#456)
  (by @kianmeng in 033d52d for #456)

* Fix potential infinite loop (#446)
  (by @kfur in ada43c0 for #446)

* use a stack of self._fds_to_close to prevent double closes (#481)
  (by @graingert in 3214cf6 for #481)

* Fix incorrect main thread id value forking from a thread  (#453)
  (by @horpto @fantix in e7934c8 for #453)

* create_subprocess_exec should treat env={} as empty environment (#439) (#454)
  (by @byllyfish in e04637e for #439)

* Queue write only after processing all buffers (#445)
  (by @jakirkham @fantix in 9c6ecb6 for #445)

* Drop Python 3.6 support for thread ident
  (by @fantix in 9c37930)

* bugfix: write to another transport in resume_writing() fails (#498)
  (by @fantix in d2deffe for #498)

Build
=====

* Upgrade GitHub Actions (#477) (#480)
  (by @cclauss in fcbf422 for #477, 1008694 for #480)

* typo `same as same`
  (by @YoSTEALTH in fedba80)

* setup.py: allow to override extra_compile_args (#443)
  (by @giuliobenetti in a130375 for #443)

* Drop hack in setup.py in finalize_options (492)
  (by @fantix in 2f1bc83 for #492)

* Fix tests invocation on release CI worklow (#489)
  (by @ben9923 in d6a2b59 for #489)

Documentation
=============

* use asyncio.Runner loop_factory on 3.11+ (#472)
  (by @graingert in 31ba48c for #472)

* Fix CI badge in docs, remove remaining Travis CI references from docs
  (by @Nothing4You in c6901a7)

* Fix typo in README
  (by @monosans in 73d7253)
@jakirkham
Copy link
Contributor Author

Thanks all for improving this and including it 🙏

Sorry for being unresponsive (was out at the time). Though am happy to see this conclusion while catching up 🙂

@jakirkham jakirkham deleted the queue_write_once branch September 26, 2022 08:53
# for free to join this conversation on GitHub. Already have an account? # to comment
Reviewers

@1st1 1st1 1st1 approved these changes

@fantix fantix fantix left review comments

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@jakirkham @fantix @1st1