From: Brian Mitchell <binary42@...>
Date: 2007-08-07T23:28:02+09:00
Subject: Re: continuations in Ruby 1.9?

On 8/7/07, Koichi Sasada <ko1@atdot.net> wrote:
> Hi,
>
> Brian Mitchell wrote:
> > Right now callcc seems to have a _large_ negative performance gap in
> > regard to the 1.8 implementation (tested on OS X, I'll try another OS
> > when I get a chance) so I am not sure I would use them in their
> > current state. Are there plans to address performance (or am I the
> > only one getting this problem)?
>
> Do you have any concrete data?

I did some benchmarking on some machines I have access to. Here is what I get:

$ cat callcc.rb
require 'benchmark'

N = 10_000

result = Benchmark.measure do
  N.times {callcc {|cc| cc}.call(lambda {|*| 42})}
end

puts RUBY_VERSION
puts result

On FreeBSD 6.2 amd64:
$ ruby -v
ruby 1.8.6 (2007-03-13 patchlevel 0) [amd64-freebsd6]
$ ruby callcc.rb
1.8.6
  0.164062   0.007812   0.171875 (  0.178217)
$ ruby19/bin/ruby -v
ruby 1.9.0 (2007-08-07 patchlevel 0) [x86_64-freebsd6.2]
$ ruby19/bin/ruby callcc.rb
1.9.0
 15.226562  12.570312  27.796875 ( 55.368435)

On Mac OS X (10.4.8):
$ ruby -v
ruby 1.8.5 (2006-08-25) [i686-darwin8.6.2]
$ ruby callcc.rb
1.8.5
  0.180000 0.010000 0.190000 ( 0.193483)
$ bin/ruby -v
ruby 1.9.0 (2007-08-07 patchlevel 0) [i686-darwin8.10.1]
$ bin/ruby callcc.rb
1.9.0
 11.230000   6.340000  17.570000 ( 21.121573)

> BTW I think Ruby's continuation has lack of important features.

I would be interested in hearing more about this. Is this related to
the fibers underlying the current trunk implementation of callcc?

Brian.