gc_stress.rb

# Adapted from a script by Evan Phoenix
require 'benchmark'

GC::Profiler.enable

class Simple
  attr_accessor :next
end

top = Simple.new

puts Benchmark.measure {
  outer = 10
  total = 100000
  per = 100

  outer.times do
    total.times do
      per.times { Simple.new }
      s = Simple.new
      top.next = s
      top = s
    end
  end
}

puts "GC time: #{GC::Profiler.total_time}s"

gistfile1.txt

system ~/projects/gc_tests $ rvm 1.9.3 do ruby -v gc_stress.rb 
ruby 1.9.3p0 (2011-10-30 revision 33570) [x86_64-darwin10.8.0]
 20.820000   0.240000  21.060000 ( 21.070793)
GC time: 2.29012300000049s

system ~/projects/gc_tests $ rvm ruby-head do ruby -v gc_stress.rb 
ruby 2.0.0dev (2012-02-19 trunk 34699) [x86_64-darwin10.8.0]
 21.740000   0.250000  21.990000 ( 22.002499)
GC time: 3.156919999999449s

system ~/projects/gc_tests $ (pickjdk 1 ; jruby -v gc_stress.rb)
New JDK: 1.6.0.jdk
jruby 1.7.0.dev (ruby-1.9.3-p28) (2012-02-19 348601d) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_29) [darwin-x86_64-java]
  8.655000   0.000000   8.655000 (  8.655000)
GC time: 0.223s

system ~/projects/gc_tests $ (pickjdk 2 ; jruby -v gc_stress.rb)
New JDK: 1.6.0_30-b12-409.jdk
jruby 1.7.0.dev (ruby-1.9.3-p28) (2012-02-19 348601d) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_30) [darwin-x86_64-java]
  9.354000   0.000000   9.354000 (  9.354000)
GC time: 0.217s

I don't know what tools exist for 1.8.7, but I would guess there might be something in REE?

So...if your results are to be believed, GC is taking more and more time in each major release of Ruby. I hope some of that is being shunted to GC threads (or something), but given that the overall times go up too...it seems like things are moving in the wrong direction.