Some for fun benchmarking of JRuby Trunk

[Zachary_Brown]

Hi all,

Did some benchmarking on MRI/JRuby to see if there’d be a noticeable
increase in
performance in some of the work I typically do. For background, most
of what I
use Ruby for these days is processing data output files that I create
from some
of my Java-coded bioinformatics tools. In other words, I have a whole
lot of data
that frequently needs to be processed.

The following are results of processing 30 files with 30000 lines each
in them,
all of which are manipulated in arrays (though not at the same time).
Here’s some
rough benchmarking just using require ‘benchmark’. To say the least, I
was
surprised at the difference between MRI and JRuby, nevermind JRuby –
server.

ruby
user system total real
341.500000 10.520000 352.020000 (371.315493)

jruby
user system total real
169.982000 0.000000 169.982000 (169.982000)

jruby --server
user system total real
125.332000 0.000000 125.332000 (125.332000)

I ran the results a few times and the numbers usually come out to
something like the above.

Its nice to see that we get this kind of speed up using JRuby :).

-Zac

---

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email

[Zachary_Brown]

Zachary Brown wrote:

The following are results of processing 30 files with 30000 lines each
in them,
all of which are manipulated in arrays (though not at the same time).
Here’s some
rough benchmarking just using require ‘benchmark’. To say the least, I was
surprised at the difference between MRI and JRuby, nevermind JRuby
–server.

Yeah, very nice numbers. And it sounds like a benchmark that’s a bit
more “real world” than many, pulling in real data from files, stuffing
into arrays, juggling them…not fib by any means.

Thanks for the numbers. If you see anything we can do to improve, or
feel like doing some profiling, let us know.

• Charlie

---

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email

[Zachary_Brown]

Charles Oliver N.:

And it sounds like a benchmark that’s a bit more “real world” than
many, pulling in real data from files, stuffing into arrays, juggling
them…not fib by any means.

Thanks for the numbers. If you see anything we can do to
improve, or feel like doing some profiling, let us know.

I’ll happily provide some benchmarks from my PhD (I’m rewriting it from
scratch at the moment, and I’m targeting Ruby 1.9 when I’m at it), but
have a simple question: how does one profile Ruby under JRuby? (I assume
ruby-prof does not work.)

— Shot