ruby-core@ruby-lang.org archive (unofficial mirror)
 help / color / mirror / Atom feed
From: "alanwu (Alan Wu) via ruby-core" <ruby-core@ml.ruby-lang.org>
To: ruby-core@ml.ruby-lang.org
Cc: "alanwu (Alan Wu)" <noreply@ruby-lang.org>
Subject: [ruby-core:117283] [Ruby master Bug#20338] certain **kwsplat calls have regressed allocations over past few ruby releases
Date: Thu, 21 Mar 2024 17:33:00 +0000 (UTC)	[thread overview]
Message-ID: <redmine.journal-107409.20240321173259.53225@ruby-lang.org> (raw)
In-Reply-To: redmine.issue-20338.20240314140511.53225@ruby-lang.org

Issue #20338 has been updated by alanwu (Alan Wu).

Subject changed from calling new has regressed allocations over past few ruby releases to certain **kwsplat calls have regressed allocations over past few ruby releases
Status changed from Open to Closed

After commit:aceee71c35e0b387691836e756b4e008efd84cf1 and commit:15dc3aaa311b32203d8ffb414bcf9b8e55ce5691, it's now back down to 3.2 levels on the master branch:

```
$ ./miniruby -v test.rb
ruby 3.4.0dev (2024-03-21T17:20:32Z master 15dc3aaa31) [arm64-darwin23]
3006
2005
```

The extra allocations were general issues with kwsplat calls which `Class#new` uses internally.

----------------------------------------
Bug #20338: certain **kwsplat calls have regressed allocations over past few ruby releases
https://bugs.ruby-lang.org/issues/20338#change-107409

* Author: fablestales (Fable Tales)
* Status: Closed
* Backport: 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN, 3.3: UNKNOWN
----------------------------------------
Given this script:

```
class Foo
  def initialize(**kwargs)
  end
end

class Bar
  def initialize(y:)
  end
end

b = GC.stat[:total_allocated_objects]
1000.times {
  Foo.new(y:1)
}
p GC.stat[:total_allocated_objects]-b


b = GC.stat[:total_allocated_objects]
1000.times {
  Bar.new(y:1)
}
p GC.stat[:total_allocated_objects]-b
```

On Ruby 3.2 it prints
```
 (fable-3_3_dead_with_patches) $ ruby --version
ruby 3.2.3 (2024-01-18 revision 52bb2ac0a6) [arm64-darwin23]
 (fable-3_3_dead_with_patches) $ ruby hi.rb
3006
2005
```

On Ruby 3.3 it prints:
```
 ((HEAD detached at v3_3_0)) $ ./ruby --version
ruby 3.3.0 (2023-12-25 revision 5124f9ac75) [arm64-darwin23]
 ((HEAD detached at v3_3_0)) $ ./ruby hi.rb
3008
3005
```

On master:
```
 (master) $ git rev-parse HEAD
83618f2cfa004accdd1514de7dcbba291aa7e831
 (master) $ ./ruby --version
ruby 3.4.0dev (2024-03-14T12:33:30Z master 83618f2cfa) [arm64-darwin23]
 (master) $ ./ruby hi.rb
3006
4005
```

so that is: since Ruby 3.2 `.new`ing an object with static keyword args has grown from 2 allocations per object to 4 allocations per object.

I have a pull request which introduces a new instruction to the VM `opt_new` which basically inlines calling allocate and initialize directly, avoiding a lot of what is causing these extra objects to be allocated. My PR is here: https://github.com/ruby/ruby/pull/10254. With my patch the script prints 2006 and 1005 respectively. This is the smallest number of objects that _can_ be allocated for such calls.




-- 
https://bugs.ruby-lang.org/
 ______________________________________________
 ruby-core mailing list -- ruby-core@ml.ruby-lang.org
 To unsubscribe send an email to ruby-core-leave@ml.ruby-lang.org
 ruby-core info -- https://ml.ruby-lang.org/mailman3/postorius/lists/ruby-core.ml.ruby-lang.org/

      parent reply	other threads:[~2024-03-21 17:33 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-03-14 14:05 [ruby-core:117169] [Ruby master Bug#20338] calling new has regressed allocations over past few ruby releases fablestales (Fable Tales) via ruby-core
2024-03-14 14:12 ` [ruby-core:117170] " kjtsanaktsidis (KJ Tsanaktsidis) via ruby-core
2024-03-14 14:15 ` [ruby-core:117171] " kjtsanaktsidis (KJ Tsanaktsidis) via ruby-core
2024-03-21 17:33 ` alanwu (Alan Wu) via ruby-core [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-list from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://www.ruby-lang.org/en/community/mailing-lists/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=redmine.journal-107409.20240321173259.53225@ruby-lang.org \
    --to=ruby-core@ruby-lang.org \
    --cc=noreply@ruby-lang.org \
    --cc=ruby-core@ml.ruby-lang.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).