IronRuby performance?

Question

While I know IronRuby isn't quite ready for the world to use it, I was wondering if anyone here tried it and tested how well it faired against the other Rubies out there in terms of raw performance?

If so, what are the results, and how did you go about measuring the performance (which benchmarks etc)?

Edit: The IronRuby team maintains a site on how they compare to Ruby MRI 1.8 at http://ironruby.info/. Below the spec pass rate table, they also have some information on how IronRuby performs on these specs. This table is not continuously updated, but I assume they update it often enough (you can see the last update at the top of the page).

Tipsy

Is this article up to date? Is this still the correct answer? – Lemar

No, it is not. I've seen updated numbers and I'll update the thread as soon as I remember where you can find them. – Tipsy

Answer 1

According to this article http://www.iunknown.com/2008/05/ironruby-and-rails.html. In may performance was nowhere near where they expected it to be. I heard in http://altnetpodcast.com/episodes/9-state-of-ironruby (3 days ago) that they're still working on performance. I guess they put compatability first and are now trying to get the performance up to par with other ruby implementations out there.

As far as I understand they're not nearly as performant as Iron Python that is developed by the same team. I don't know if this is because Iron Ruby is using the DLR a lot more and that still needs to be optimized or if they need to optimize the Iron Ruby implementation itself more. But I guess it is good news because they can get it a lot faster. So if you're already happy with performance you'll get a lot happier.

Tyus

Answer 2

Antonio Cangiano just published some new benchmarks (August 09), IronRuby seems to be faster across that board than MRI under Windows. John Lam and co have done a fantastic job and are not done optimising.

Lemar

Answer 3

The load time and the memory utilization are still the two weakest points in IronRuby. Once a particular piece of code has been loaded and is running in a sort of steady-state mode -- that is, little to no new source is being evaluated -- then the performance should be quite good.

To answer your specific question, consider this data.

Porpoise

Answer 4

I have used it and it has worked great for what I have done. However my measuring of performance isn't really scientific, because it was all visual. However I did notice that IronRuby seemed a little more snappier when I compared the two program on equal tasks. I really think this had to do more with the strong and tight binding with IIS that .NET has more than the speed of the framework.

But I could totally be wrong, because I didn't really stress my applications to the levels that Twitter might see. But from my .NET experience I know it would hold up just as well if not better than current production Ruby applications.

By the way I tested Ruby using FastCGI under IIS.

Salenasalene