[ruby-core:102141] [Ruby master Bug#17497] Ractor performance issue

Issue #17497 has been updated by keithrbennett (Keith Bennett).


I too have seen strange results testing ractors. I used the code at https://github.com/keithrbennett/keithrbennett-ractor-test/blob/master/my_ractor.rb to do some arbitrary but predictable work. I have a 24-core Ryzen 9 CPU, and I compared using 1 ractor with using 24. With 24, htop reported that all the CPU's were at 100% most of the time, yet the elapsed time using 24 CPU's was only about a third less than when using 1 CPU. Also, the CPU's seemed to be working collectively about ten times harder with 24 CPU's. Here is the program output:

```
1 CPU:
Many HTOP readings are < 100% for all CPU's
 time ractor/my_ractor.rb ruby '*.rb'
Running the following command to find all filespecs to process: find -L ruby -type f -name '*.rb' -print
Processing 8218 files in 1 slices, whose sizes are:
[8218]
ractor/my_ractor.rb ruby '*.rb'  2513.90s user 6.75s system 99% cpu 42:03.01 total


24 CPU's:
% time ractor/my_ractor.rb ruby '*.rb' ; espeak finished
Running the following command to find all filespecs to process: find -L ruby -type f -name '*.rb' -print
Processing 8218 files in 24 slices, whose sizes are:
[343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 343, 329]
ractor/my_ractor.rb ruby '*.rb'  22986.42s user 14.98s system 1134% cpu 33:47.96 total
```

(In the command, `ruby` refers to the directory in which I've cloned the Github Ruby repo.)

Here is the current content of the test program:

```
#!/usr/bin/env ruby

require 'amazing_print'
require 'etc'
require 'set'
require 'shellwords'
require 'yaml'

raise "This script requires Ruby version 3 or later." unless RUBY_VERSION.split('.').first.to_i >= 3


# An instance of this parser class is created for each ractor.
class RactorParser

  attr_reader :dictionary_words

  def initialize(dictionary_words)
    @dictionary_words = dictionary_words
  end

  def parse(filespecs)
    filespecs.inject(Set.new) do |found_words, filespec|
      found_words | process_one_file(filespec)
    end
  end

  private def word?(string)
    dictionary_words.include?(string)
  end

  private def strip_punctuation(string)
    punctuation_regex = /[[:punct:]]/
    string.gsub(punctuation_regex, ' ')
  end

  private def file_lines(filespec)
    command = "strings #{Shellwords.escape(filespec)}"
    text = `#{command}`
    strip_punctuation(text).split("\n")
  end

  private def line_words(line)
    line.split.map(&:downcase).select { |text| word?(text) }
  end

  private def process_one_file(filespec)
    file_words = Set.new
    file_lines(filespec).each do |line|
      line_words(line).each { |word| file_words << word }
    end

    # puts "Found #{file_words.count} words in #{filespec}."
    file_words
  end
end


class Main

  BASEDIR =  ARGV[0] || '.'
  FILEMASK = ARGV[1]
  CPU_COUNT = Etc.nprocessors

  def call
    check_arg_count
    slices = get_filespec_slices
    ractors = create_and_populate_ractors(slices)
    all_words = collate_ractor_results(ractors)
    yaml = all_words.to_a.sort.to_yaml
    File.write('ractor-words.yaml', yaml)
    puts "Words are in ractor-words.yaml."
  end

  private def check_arg_count
    if ARGV.length > 2
      puts "Syntax is ractor [base_directory] [filemask], and filemask must be quoted so that the shell does not expand it."
      exit -1
    end
  end

  private def collate_ractor_results(ractors)
    ractors.inject(Set.new) do |all_words, ractor|
      all_words | ractor.take
    end
  end

  private def get_filespec_slices
    all_filespecs = find_all_filespecs
    slice_size = (all_filespecs.size / CPU_COUNT) + 1
    # slice_size = all_filespecs.size # use this line instead of previous to test with 1 ractor
    slices = all_filespecs.each_slice(slice_size).to_a
    puts "Processing #{all_filespecs.size} files in #{slices.size} slices, whose sizes are:\n#{slices.map(&:size).inspect}"
    slices
  end

  private def create_and_populate_ractors(slices)
    words = File.readlines('/usr/share/dict/words').map(&:chomp).map(&:downcase).sort

    slices.map do |slice|
      ractor = Ractor.new do
        filespecs = Ractor.receive
        dictionary_words = Ractor.receive
        RactorParser.new(dictionary_words).parse(filespecs)
      end
      ractor.send(slice)
      ractor.send(words)
      ractor
    end
  end

  private def find_all_filespecs
    filemask = FILEMASK ? %Q{-name '#{FILEMASK}'} : ''
    command = "find -L #{BASEDIR} -type f #{filemask} -print"
    puts "Running the following command to find all filespecs to process: #{command}"
    `#{command}`.split("\n")
  end
end

Main.new.call
```

----------------------------------------
Bug #17497: Ractor performance issue
https://bugs.ruby-lang.org/issues/17497#change-89993

* Author: marcandre (Marc-Andre Lafortune)
* Status: Open
* Priority: Normal
* Assignee: ko1 (Koichi Sasada)
* ruby -v: ruby 3.0.0p0 (2020-12-25 revision 95aff21468) [x86_64-darwin18]
* Backport: 2.5: UNKNOWN, 2.6: UNKNOWN, 2.7: UNKNOWN
----------------------------------------
There's a strange performance issue with Ractor (at least on MacOS, didn't run on other OS).

I ran a benchmark doing 3 different types of work:
* "fib": method calls (naive fibonacci calculation)
* "cpu": `(0...1000).inject(:+)`
* "sleep": call `sleep`

I get the kind of results I was excepting for the `fib` and for sleeping, but the results for the "cpu" workload show a problem.

It is so slow that my pure Ruby backport (using Threads) is 65x faster  on my Mac Pro (despite having 6 cores). Expected results would be 6x slower, so in that case Ractor is 400x slower than it should 仭

On my MacBook (2 cores) the results are not as bad, the `cpu` workload is 3x faster with my pure-Ruby backport (only) instead of ~2x slower, so the factor is 6x too slow.


```
$ gem install backports
Successfully installed backports-3.20.0
1 gem installed
$ ruby ractor_test.rb
<internal:ractor>:267: warning: Ractor is experimental, and the behavior may change in future versions of Ruby! Also there are many implementation issues.
fib: 110 ms  |  cpu: 22900 ms  |  sleep: 206 ms
$ B=t ruby ractor_test.rb
Using pure Ruby implementation
fib: 652 ms  |  cpu: 337 ms  |  sleep: 209 ms
```

Notice the `sleep` run takes similar time, which is good, and `fib` is ~6x faster on my 6-core CPU (and ~2x faster on my 2-core MacBook), again that's good as the pure ruby version uses Threads and thus runs with a single GVL.

The `cpu` version is the problem.

Script is here: https://gist.github.com/marcandre/bfed626e538a3d0fc7cad38dc026cf0e



-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>