Comments on: ‘inject’, ‘each’ and ‘times’ methods much slower in Ruby 1.9 on Mac OS X http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/ By Antonio Cangiano, Software Engineer & Technical Evangelist at IBM Sat, 17 May 2008 10:44:12 +0000 http://wordpress.org/?v=2.3.3 By: Arthur Lyman http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2986 Arthur Lyman Wed, 02 Apr 2008 14:10:36 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2986 The benchmarks seem very high ( slow) on Windows using ruby 1.8.6 (2007-03-13 patchlevel 0) [i386-mswin32] === 10^7 === inject: 35.735000 0.265000 36.000000 ( 36.297000) each: 23.171000 0.172000 23.343000 ( 23.550000) times: 22.422000 0.250000 22.672000 ( 22.804000) while: 30.407000 0.156000 30.563000 ( 30.882000) === 10^8 === inject: 362.296000 2.469000 364.765000 (368.944000) each: 235.922000 2.203000 238.125000 (240.274000) times: 228.188000 2.172000 230.360000 (231.533000) while: 249.031000 2.141000 251.172000 (252.498000) Is this typical for the Windows version of ruby.exe? The benchmarks seem very high ( slow) on Windows using ruby 1.8.6 (2007-03-13 patchlevel 0) [i386-mswin32]

=== 10^7 ===
inject: 35.735000 0.265000 36.000000 ( 36.297000)
each: 23.171000 0.172000 23.343000 ( 23.550000)
times: 22.422000 0.250000 22.672000 ( 22.804000)
while: 30.407000 0.156000 30.563000 ( 30.882000)
=== 10^8 ===
inject: 362.296000 2.469000 364.765000 (368.944000)
each: 235.922000 2.203000 238.125000 (240.274000)
times: 228.188000 2.172000 230.360000 (231.533000)
while: 249.031000 2.141000 251.172000 (252.498000)

Is this typical for the Windows version of ruby.exe?

]]> By: This Week in Ruby (March 31, 2008) | Zen and the Art of Programming http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2898 This Week in Ruby (March 31, 2008) | Zen and the Art of Programming Mon, 31 Mar 2008 05:39:41 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2898 [...] last week I spotted a bug that made Ruby 1.9 (built from trunk) significantly slower than Ruby 1.8. After a bit of [...] […] last week I spotted a bug that made Ruby 1.9 (built from trunk) significantly slower than Ruby 1.8. After a bit of […]

]]> By: Andrés Suárez http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2855 Andrés Suárez Wed, 26 Mar 2008 18:21:47 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2855 1) I'm running the script under Windows. 2) Sorry, but I don't know where to find the binaries of Ruby 1.9.0-1 3) You've reason. After installing the last jdk and using the "J-server" option the result are much better: http://pastie.caboo.se/171028 === 10^7 === inject: 3.266000 0.000000 3.266000 ( 3.270037) each: 2.406000 0.000000 2.406000 ( 2.402494) times: 2.344000 0.000000 2.344000 ( 2.343239) while: 1.031000 0.000000 1.031000 ( 1.045526) === 10^8 === inject: 33.094000 0.000000 33.094000 ( 33.102321) each: 24.828000 0.000000 24.828000 ( 24.827342) times: 23.922000 0.000000 23.922000 ( 23.919247) while: 10.297000 0.000000 10.297000 ( 10.299023) Thanks, and keep doing your good work Antonio. 1) I’m running the script under Windows.

2) Sorry, but I don’t know where to find the binaries of Ruby 1.9.0-1

3) You’ve reason. After installing the last jdk and using the “J-server” option the result are much better:

http://pastie.caboo.se/171028

=== 10^7 ===
inject: 3.266000 0.000000 3.266000 ( 3.270037)
each: 2.406000 0.000000 2.406000 ( 2.402494)
times: 2.344000 0.000000 2.344000 ( 2.343239)
while: 1.031000 0.000000 1.031000 ( 1.045526)
=== 10^8 ===
inject: 33.094000 0.000000 33.094000 ( 33.102321)
each: 24.828000 0.000000 24.828000 ( 24.827342)
times: 23.922000 0.000000 23.922000 ( 23.919247)
while: 10.297000 0.000000 10.297000 ( 10.299023)

Thanks, and keep doing your good work Antonio.

]]> By: Antonio Cangiano http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2853 Antonio Cangiano Wed, 26 Mar 2008 14:56:01 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2853 @Radarek: Done. :) @Radarek: Done. :)

]]> By: Radarek http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2852 Radarek Wed, 26 Mar 2008 14:41:40 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2852 Antonio Cangiano, I guess that you should change title to "‘inject’, ‘each’ and ‘times’ methods much slower in Ruby 1.9 on Mac OS X" :). Antonio Cangiano, I guess that you should change title to “‘inject’, ‘each’ and ‘times’ methods much slower in Ruby 1.9 on Mac OS X” :).

]]> By: Antonio Cangiano http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2851 Antonio Cangiano Wed, 26 Mar 2008 13:08:10 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2851 @Andrés Suárez: Your results are explained by three possible factors: <br/> 1) Are you running the script on Linux/Windows or Mac? The problem appears to be a Mac only issue;<br/> 2) The version of Ruby 1.9 that introduces the performance problem is a recent one, not the one that was released at Christmas that you tested;<br/> 3) Did you run JRuby with the -J-server option and with the latest version of the JDK? @Andrés Suárez: Your results are explained by three possible factors:

1) Are you running the script on Linux/Windows or Mac? The problem appears to be a Mac only issue;
2) The version of Ruby 1.9 that introduces the performance problem is a recent one, not the one that was released at Christmas that you tested;
3) Did you run JRuby with the -J-server option and with the latest version of the JDK?

]]> By: Andrés Suárez http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2849 Andrés Suárez Wed, 26 Mar 2008 09:24:30 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2849 http://pastie.caboo.se/170779 <pre> [x86 coreDuo@1800] ruby 1.9.0 (2007-12-25) === 10^7 === inject: 11.265000 0.000000 11.265000 ( 11.297000) each: 10.266000 0.000000 10.266000 ( 10.281000) times: 10.187000 0.000000 10.187000 ( 10.219000) while: 8.391000 0.000000 8.391000 ( 8.406000) === 10^8 === inject: 111.969000 0.000000 111.969000 (112.188000) each: 102.625000 0.000000 102.625000 (102.906000) times: 101.219000 0.000000 101.219000 (101.375000) while: 83.687000 0.000000 83.687000 ( 83.859000) jruby-1.1RC3 === 10^7 === inject: 48.422000 0.000000 48.422000 ( 48.427614) each: 9.234000 0.000000 9.234000 ( 9.241645) times: 9.156000 0.000000 9.156000 ( 9.165160) while: 15.109000 0.000000 15.109000 ( 15.083723) === 10^8 === inject: 490.485000 0.000000 490.485000 (490.388292) each: 92.656000 0.000000 92.656000 ( 92.660257) times: 91.906000 0.000000 91.906000 ( 91.896139) while: 148.156000 0.000000 148.156000 (148.153876) </pre> http://pastie.caboo.se/170779

[x86 coreDuo@1800]

ruby 1.9.0 (2007-12-25)  

=== 10^7 ===
inject:  11.265000   0.000000  11.265000 ( 11.297000)
each:    10.266000   0.000000  10.266000 ( 10.281000)
times:   10.187000   0.000000  10.187000 ( 10.219000)
while:    8.391000   0.000000   8.391000 (  8.406000)

=== 10^8 ===
inject: 111.969000   0.000000 111.969000 (112.188000)
each:   102.625000   0.000000 102.625000 (102.906000)
times:  101.219000   0.000000 101.219000 (101.375000)
while:   83.687000   0.000000  83.687000 ( 83.859000)

jruby-1.1RC3

=== 10^7 ===
inject:  48.422000   0.000000  48.422000 ( 48.427614)
each:     9.234000   0.000000   9.234000 (  9.241645)
times:    9.156000   0.000000   9.156000 (  9.165160)
while:   15.109000   0.000000  15.109000 ( 15.083723)
=== 10^8 ===
inject: 490.485000   0.000000 490.485000 (490.388292)
each:    92.656000   0.000000  92.656000 ( 92.660257)
times:   91.906000   0.000000  91.906000 ( 91.896139)
while:  148.156000   0.000000 148.156000 (148.153876)
]]> By: Antonio Cangiano http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2848 Antonio Cangiano Wed, 26 Mar 2008 07:01:56 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2848 That's excellent, Chris. :) That’s excellent, Chris. :)

]]> By: Chris http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2847 Chris Wed, 26 Mar 2008 06:34:33 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2847 I wrote a script to find when this happened. On my mac, using this inject benchmark I found on the Ruby talk mailing list (http://pastie.caboo.se/170752), the runtime jumps from 3 seconds to 35 seconds at revision 15124. The diff between 15123 and 15124 can be seen here: http://pastie.caboo.se/170751 I hope that helps figuring this out. I wrote a script to find when this happened. On my mac, using this inject benchmark I found on the Ruby talk mailing list (http://pastie.caboo.se/170752), the runtime jumps from 3 seconds to 35 seconds at revision 15124. The diff between 15123 and 15124 can be seen here: http://pastie.caboo.se/170751

I hope that helps figuring this out.

]]> By: Antonio Cangiano http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2842 Antonio Cangiano Wed, 26 Mar 2008 03:57:17 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2842 Hi James, it appears to be a Mac only issue, therefore I don't think it's something done intentionally. Hi James, it appears to be a Mac only issue, therefore I don’t think it’s something done intentionally.

]]> By: James Deville http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2840 James Deville Wed, 26 Mar 2008 03:45:54 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2840 I wonder if there is some sort of clean up going on inside of Enumerable. Anyone know if they are doing something that currently slows down these methods, but in the long term will regain the speed? JD I wonder if there is some sort of clean up going on inside of Enumerable. Anyone know if they are doing something that currently slows down these methods, but in the long term will regain the speed?

JD

]]> By: Antonio Cangiano http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2839 Antonio Cangiano Wed, 26 Mar 2008 03:01:45 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2839 Thanks Hans-Georg. There is definitely a huge performance regression for these methods in Ruby 1.9, at least on Mac. I'll try it out on Linux. Thanks Hans-Georg. There is definitely a huge performance regression for these methods in Ruby 1.9, at least on Mac. I’ll try it out on Linux.

]]> By: Hans-Georg http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2837 Hans-Georg Wed, 26 Mar 2008 02:47:27 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2837 <pre> === 10^7 === ruby 1.8.6 (2007-09-24 patchlevel 111) [universal-darwin9.0] inject: 11.130000 0.010000 11.140000 ( 11.156980) each: 7.080000 0.020000 7.100000 ( 7.095086) times: 7.080000 0.010000 7.090000 ( 7.095669) while: 8.820000 0.030000 8.850000 ( 8.847939) ruby 1.9.0 (2007-12-25 revision 14709) [i686-darwin9.2.0] inject: 3.730000 0.010000 3.740000 ( 3.740929) each: 3.510000 0.010000 3.520000 ( 3.513513) times: 3.510000 0.000000 3.510000 ( 3.531018) while: 2.660000 0.010000 2.670000 ( 2.654901) ruby 1.9.0 (2008-03-21 revision 0) [i686-darwin9.2.0] inject: 20.880000 7.410000 28.290000 ( 28.594721) each: 20.210000 7.330000 27.540000 ( 27.762472) times: 20.330000 7.320000 27.650000 ( 27.913742) while: 2.660000 0.020000 2.680000 ( 2.743546) MacRuby version 0.2 (ruby 1.9.0 2008-03-01) [universal-darwin9.0] inject: 31.220000 7.700000 38.920000 ( 39.308773) each: 30.450000 7.650000 38.100000 ( 38.361888) times: 30.680000 7.710000 38.390000 ( 39.138352) while: 12.220000 0.080000 12.300000 ( 12.460638) </pre> === 10^7 === ruby 1.8.6 (2007-09-24 patchlevel 111) [universal-darwin9.0] inject: 11.130000 0.010000 11.140000 ( 11.156980) each: 7.080000 0.020000 7.100000 ( 7.095086) times: 7.080000 0.010000 7.090000 ( 7.095669) while: 8.820000 0.030000 8.850000 ( 8.847939) ruby 1.9.0 (2007-12-25 revision 14709) [i686-darwin9.2.0] inject: 3.730000 0.010000 3.740000 ( 3.740929) each: 3.510000 0.010000 3.520000 ( 3.513513) times: 3.510000 0.000000 3.510000 ( 3.531018) while: 2.660000 0.010000 2.670000 ( 2.654901) ruby 1.9.0 (2008-03-21 revision 0) [i686-darwin9.2.0] inject: 20.880000 7.410000 28.290000 ( 28.594721) each: 20.210000 7.330000 27.540000 ( 27.762472) times: 20.330000 7.320000 27.650000 ( 27.913742) while: 2.660000 0.020000 2.680000 ( 2.743546) MacRuby version 0.2 (ruby 1.9.0 2008-03-01) [universal-darwin9.0] inject: 31.220000 7.700000 38.920000 ( 39.308773) each: 30.450000 7.650000 38.100000 ( 38.361888) times: 30.680000 7.710000 38.390000 ( 39.138352) while: 12.220000 0.080000 12.300000 ( 12.460638) ]]> By: Antonio Cangiano http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2836 Antonio Cangiano Wed, 26 Mar 2008 01:04:41 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2836 Radarek, that's odd. Charles' results are close to mine, but yours aren't. I wonder if this is a Linux vs Mac issue. Radarek, that’s odd. Charles’ results are close to mine, but yours aren’t. I wonder if this is a Linux vs Mac issue.

]]> By: Radarek http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2835 Radarek Wed, 26 Mar 2008 00:32:57 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2835 I can't reproduce that benchmarks. On my machine I get: <pre> === 10^8 === ruby1.8 inject: 70.060000 0.010000 70.070000 ( 70.133406) each: 46.510000 0.010000 46.520000 ( 46.551202) times: 46.070000 0.000000 46.070000 ( 46.121204) while: 61.160000 0.000000 61.160000 ( 61.208435) ruby1.9 inject: 30.810000 0.000000 30.810000 ( 30.835917) each: 27.990000 0.000000 27.990000 ( 28.014818) times: 27.300000 0.000000 27.300000 ( 27.338477) while: 20.150000 0.000000 20.150000 ( 20.162903) rubinius inject: 183.314781 0.000000 183.314781 (183.314766) each: 145.106427 0.000000 145.106427 (145.106411) times: 107.245163 0.000000 107.245163 (107.245142) while: 95.506226 0.000000 95.506226 ( 95.506214) jruby inject: 22.626000 0.000000 22.626000 ( 22.626000) each: 16.456000 0.000000 16.456000 ( 16.455000) times: 17.483000 0.000000 17.483000 ( 17.484000) while: 6.655000 0.000000 6.655000 ( 6.656000) </pre> All compiled from sourced and updated from svn today. For all results look at pastie: http://pastie.caboo.se/170627 <br/> So, ruby1.9 is more that 2x faster that MRI, but the fastest jruby :). I can’t reproduce that benchmarks. On my machine I get:

=== 10^8 ===
ruby1.8
inject:  70.060000   0.010000  70.070000 ( 70.133406)
each:    46.510000   0.010000  46.520000 ( 46.551202)
times:   46.070000   0.000000  46.070000 ( 46.121204)
while:   61.160000   0.000000  61.160000 ( 61.208435)

ruby1.9
inject:  30.810000   0.000000  30.810000 ( 30.835917)
each:    27.990000   0.000000  27.990000 ( 28.014818)
times:   27.300000   0.000000  27.300000 ( 27.338477)
while:   20.150000   0.000000  20.150000 ( 20.162903)

rubinius
inject: 183.314781   0.000000 183.314781 (183.314766)
each:   145.106427   0.000000 145.106427 (145.106411)
times:  107.245163   0.000000 107.245163 (107.245142)
while:   95.506226   0.000000  95.506226 ( 95.506214)

jruby
inject:  22.626000   0.000000  22.626000 ( 22.626000)
each:    16.456000   0.000000  16.456000 ( 16.455000)
times:   17.483000   0.000000  17.483000 ( 17.484000)
while:    6.655000   0.000000   6.655000 (  6.656000)

All compiled from sourced and updated from svn today. For all results look at pastie: http://pastie.caboo.se/170627

So, ruby1.9 is more that 2x faster that MRI, but the fastest jruby :).

]]> By: Antonio Cangiano http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2834 Antonio Cangiano Wed, 26 Mar 2008 00:10:36 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2834 Good idea Lawrence, I've added Rubinius' results. :) Good idea Lawrence, I’ve added Rubinius’ results. :)

]]> By: Charles Oliver Nutter http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2833 Charles Oliver Nutter Wed, 26 Mar 2008 00:00:10 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2833 Lawrence: I ran it against the five implementations I have access to: Ruby 1.8.6p111, Ruby 1.9 trunk, Rubinius trunk, MacRuby trunk, and JRuby trunk. The results on my system are here: http://pastie.org/170572 Lawrence: I ran it against the five implementations I have access to: Ruby 1.8.6p111, Ruby 1.9 trunk, Rubinius trunk, MacRuby trunk, and JRuby trunk. The results on my system are here:

http://pastie.org/170572

]]> By: Lawrence Oluyede http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2832 Lawrence Oluyede Tue, 25 Mar 2008 23:13:26 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2832 Have you tried running these benchmarks against Rubinius? I am curious Have you tried running these benchmarks against Rubinius? I am curious

]]> By: Antonio Cangiano http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2831 Antonio Cangiano Tue, 25 Mar 2008 23:05:13 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2831 Of course Charles, please go ahead. Of course Charles, please go ahead.

]]> By: Charles Oliver Nutter http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2830 Charles Oliver Nutter Tue, 25 Mar 2008 22:46:32 +0000 http://antoniocangiano.com/2008/03/25/inject-each-and-times-methods-much-slower-in-ruby-19/#comment-2830 Wow, great numbers for JRuby. May we include this benchmark in our suite? Wow, great numbers for JRuby. May we include this benchmark in our suite?

]]>