gistfile1.txt

karol@brenta:~/projects/personal/rubinius (master) $ ./bin/benchmark -t ./bin/rbx -t ^maglev-head -t ^jruby-head -t ^ruby-2.0.0-preview1 -t ^ruby-1.8.7-p248 benchmark/octane/bench_raytrace.rb
Found rvm-managed ruby: maglev-head
Found rvm-managed ruby: jruby-head
Found rvm-managed ruby: ruby-2.0.0-preview1
Found rvm-managed ruby: ruby-1.8.7-p248
=== ./bin/rbx ===
       #render_scene      527.3 (±3.0%) i/s -       2640 in   5.011487s (cycle=33)
=== /Users/karol/.rvm/rubies/maglev-head/bin/ruby ===
       #render_scene      191.0 (±1.6%) i/s -        969 in   5.074275s (cycle=19)
=== /Users/karol/.rvm/rubies/jruby-head/bin/ruby ===
       #render_scene      359.3 (±25.3%) i/s -       1470 in   4.976999s (cycle=2)
=== /Users/karol/.rvm/rubies/ruby-2.0.0-preview1/bin/ruby ===
       #render_scene      257.5 (±1.2%) i/s -       1300 in   5.048274s (cycle=25)
=== /Users/karol/.rvm/rubies/ruby-1.8.7-p248/bin/ruby ===
       #render_scene       70.0 (±1.4%) i/s -        357 in   5.097288s (cycle=7)

Comparing benchmark/octane/bench_raytrace.rb:#render_scene:
                                            ./bin/rbx:        527 i/s
         /Users/karol/.rvm/rubies/jruby-head/bin/ruby:        359 i/s - 1.47x slower
/Users/karol/.rvm/rubies/ruby-2.0.0-preview1/bin/ruby:        257 i/s - 2.05x slower
        /Users/karol/.rvm/rubies/maglev-head/bin/ruby:        191 i/s - 2.76x slower
    /Users/karol/.rvm/rubies/ruby-1.8.7-p248/bin/ruby:         70 i/s - 7.53x slower

FYI, running JRuby via that benchmark script will leave the benchmark's blocks interpreted, leading to significantly lower performance. I am also guessing this is not using invokedynamic. JRuby should be much faster than this...at least 2-3x faster than Ruby 2.0.0.