JRuby demos from recent conferences

Ruby JRuby
#1

I have moved the jruby-demos project from my github account to the
jruby github account. You can find them here:

Lots of fun stuff if you’re planning to do a presentation!

  • Charlie

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#2

I have moved the jruby-demos project from my github account to the
jruby github account. You can find them here:

http://github.com/jruby/jruby-demos

Lots of fun stuff if you’re planning to do a presentation!

  • Charlie

Sent you a pull with simple fixes for the ffi demos.

an error here:
$ jruby bench_threaded_reverse.rb
bench_threaded_reverse.rb:29:in do_test': divided by 0 (ZeroDivisionError) from bench_threaded_reverse.rb:70:inrun’
from bench_threaded_reverse.rb:77
from bench_threaded_reverse.rb:76:in `times’
from bench_threaded_reverse.rb:76

interesting timings for this benchmark:

bench_fractal.rb:

macruby 0.480132
maglev-ruby 1.405938
rbx 2.025917
jruby --fast 2.745000
jruby 2.957000
ruby19 4.044064
ruby 6.819144

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#3

Speaking of presentations, any word on when the videos from JRubyConf
will
show up?

Links are up here http://jrubyconf2009.confreaks.com/ but no videos.

//Peter

On Sat, Dec 12, 2009 at 4:54 AM, Charles Oliver N.

#4

On Sat, Dec 12, 2009 at 1:51 AM, Stephen B.
[email protected] wrote:

Sent you a pull with simple fixes for the ffi demos.

Thanks for the reminder; I just pushed your fixes.

an error here:
$ jruby bench_threaded_reverse.rb
bench_threaded_reverse.rb:29:in do_test': divided by 0 (ZeroDivisionError) Â Â Â Â from bench_threaded_reverse.rb:70:inrun’
    from bench_threaded_reverse.rb:77
    from bench_threaded_reverse.rb:76:in `times’
    from bench_threaded_reverse.rb:76

This one always requires a numeric parameter for the number of threads
to use. It’s an odd benchmark.

 ruby       6.819144
Yeah, it’s very math-heavy, which is where we’re going to slip behind
without JVM support for fixnums. But there’s also some heavy block use
in there, so if we improve block dispatch performance we’ll be able to
make up some of the gap.

Your numbers do seem a little slow, though…after warmup, my machine
runs the fractal benchmark on JRuby in 1.5s. Are these single runs?

  • Charlie

  • Charlie

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#5

On Wed, Dec 16, 2009 at 12:25 PM, Peter Sönnergren Lind [email protected]
wrote:

Speaking of presentations, any word on when the videos from JRubyConf will
show up?
Links are up here http://jrubyconf2009.confreaks.com/ but no videos.

Yeah, I’m checking every day as well… :slight_smile:

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#6

At 1:14 PM -0600 12/15/09, Charles Oliver N. wrote:

   from bench_threaded_reverse.rb:77
   from bench_threaded_reverse.rb:76:in `times'
   from bench_threaded_reverse.rb:76

This one always requires a numeric parameter for the number of threads
to use. It’s an odd benchmark.

OK, running 4 threads on my 2.5 GHz Intel Core Duo 2 MacBook Pro JRuby
and MagLev are much faster then the other VMs:

[I don’t know how to interpret the cases where the user time is more than the real time ??]

$ time jruby bench_threaded_reverse.rb 4

real 0m14.076s
user 0m21.300s
sys 0m1.139s

$ time maglev-ruby bench_threaded_reverse.rb 4

real 0m15.453s
user 0m10.599s
sys 0m0.365s

$ time ruby19 bench_threaded_reverse.rb 4

real 0m39.475s
user 0m38.497s
sys 0m0.414s

$ time ruby bench_threaded_reverse.rb 4

real 0m39.744s
user 0m38.867s
sys 0m0.387s

$ time rbx bench_threaded_reverse.rb 4

real 1m20.313s
user 1m14.747s
sys 0m0.722s

$ time macruby bench_threaded_reverse.rb 4

real 1m47.122s
user 2m46.045s
sys 0m19.389s

ruby 6.819144

Yeah, it’s very math-heavy, which is where we’re going to slip behind
without JVM support for fixnums. But there’s also some heavy block use
in there, so if we improve block dispatch performance we’ll be able to
make up some of the gap.

Your numbers do seem a little slow, though…after warmup, my machine
runs the fractal benchmark on JRuby in 1.5s. Are these single runs?

You’re right – I hadn’t realized that the benchmark app took an
argument.

This is better:

$ jruby --version
jruby 1.5.0.dev (ruby 1.8.7 patchlevel 174) (2009-12-12 a4a1bb7) (Java
HotSpot™ 64-Bit Server VM 1.6.0_17) [x86_64-java]

$ jruby bench_fractal.rb 4
Ruby Elapsed 2.540000
Ruby Elapsed 1.935000
Ruby Elapsed 1.891000
Ruby Elapsed 1.816000

$ jruby --fast bench_fractal.rb 4
Ruby Elapsed 2.534000
Ruby Elapsed 1.899000
Ruby Elapsed 1.657000
Ruby Elapsed 1.761000

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#7

On Tue, Dec 15, 2009 at 10:55 PM, Stephen B.
[email protected] wrote:

OK, running 4 threads on my 2.5 GHz Intel Core Duo 2 MacBook Pro JRuby and MagLev are much faster then the other VMs:

[I don’t know how to interpret the cases where the user time is more than the real time ??]

 $ time jruby bench_threaded_reverse.rb 4

 real  0m14.076s
 user  0m21.300s
 sys  0m1.139s
…
 $ time macruby bench_threaded_reverse.rb 4

 real  1m47.122s
 user  2m46.045s
 sys  0m19.389s

More user time than real time means that it’s using multiple cores to
get the same job done in less time. Notice that only MacRuby and JRuby
actually are utilizing multiple cores concurrently.

 Ruby Elapsed 1.891000
 Ruby Elapsed 1.816000

 $ jruby --fast bench_fractal.rb 4
 Ruby Elapsed 2.534000
 Ruby Elapsed 1.899000
 Ruby Elapsed 1.657000
 Ruby Elapsed 1.761000

Yes, much better. Still can be improved, certainly, but this again
shows that perf is still not a super-high priority compared to getting
things like Java integration 100%.

  • charlie

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#8

sys 0m1.139s
…
$ time macruby bench_threaded_reverse.rb 4

real 1m47.122s
user 2m46.045s
sys 0m19.389s

More user time than real time means that it’s using multiple cores to
get the same job done in less time. Notice that only MacRuby and JRuby
actually are utilizing multiple cores concurrently.

Make me wonder how maglev got so fast on the benchmark?

The maglev real time is 15s and the user time is 10s – I suspect they
are using multiple cores because they are almost as fastas JRuby but
don’t know for sure.

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#9

On Wed, Dec 16, 2009 at 5:25 AM, Peter Sönnergren Lind [email protected]
wrote:

Speaking of presentations, any word on when the videos from JRubyConf will
show up?
Links are up here http://jrubyconf2009.confreaks.com/ but no videos.
//Peter

No word, unfortunately. Email confreaks and report back? :slight_smile:

  • Cahrlie

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#10

We’ll have to wait a little bit longer:

On Thu, Dec 17, 2009 at 12:35 PM, Carl Y. wrote:

Both RubyConf and JRubyConf had some framerate sync issues
that are causing us problems. We are in the middle of redoing our
post-production in another editing program because our current system
wasn’t able to deal with the problem. We hope to have this issue
resolved by the end of the year.

On Wed, Dec 16, 2009 at 6:21 PM, Charles Oliver N.

#11

Well at least we know they’re on the way. Thanks for checking, Peter!

On Thu, Dec 17, 2009 at 6:23 AM, Peter Sönnergren Lind [email protected]
wrote:

wrote:

  • Cahrlie

To unsubscribe from this list, please visit:

  http://xircles.codehaus.org/manage_email

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#12

On Mon, Jan 25, 2010 at 6:32 PM, Peter Sönnergren Lind [email protected]
wrote:

Keynote is up now, thanks @grantmichaels for tweeting!
http://jrubyconf2009.confreaks.com/22-nov-2009-10-00-jruby-state-of-the-union-charles-nutter-and-tom-enebo.html
Now, when is that nice looking pragprog book going to be released?

Nice to meet you in Stockholm, Peter (I don’t think we’d met before,
right?).

The book is moving along; we’ve got chapters under review and
completed review, and are wrapping up some details for beta.

  • Charlie

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
#13

Keynote is up now, thanks @grantmichaels for tweeting!

http://jrubyconf2009.confreaks.com/22-nov-2009-10-00-jruby-state-of-the-union-charles-nutter-and-tom-enebo.html

Now, when is that nice looking pragprog book going to be released?

//Peter
http://twitter.com/peter_lind
http://peter.lind.be

On Thu, Dec 17, 2009 at 5:20 PM, Charles Oliver N.