From: rise <rise@...>
Date: 2001-05-01T02:26:39+09:00
Subject: [ruby-talk:14445] Re: Rosetta Code

On Tue, 1 May 2001, Amos wrote:

> bff> "The Great Computer Language Shootout":
> bff> http://www.bagley.org/~doug/shootout/
>
> Any chance Ruby performance might improve as the language evolves,
> or does the dynamic/OO overhead hold us back from gaining too much
> ground?

I've taken a look at some of the implementations and as Doug himself
points out they're often noticeably naive. This is true for some varying
set of tests in all of the languages I'm competent to judge, not just
Ruby.  Rewritten versions of a few of the test showed noticeable
improvements[1] even without using optimizations of dubious generality[2].
The Shootout is a nice page, but suffers from the same problem most such
attempts I've seen suffer from, in that the coding skill and effort put
into the programs biases the results somewhat. There is an interesting
study at http://wwwipd.ira.uka.de/~prechelt/Biblio/jccpprtTR.pdf that
attempts to use a larger body of programmers and average over the results,
but Ruby isn't one of the tested languages.

[1] Before you ask, I haven't gotten around to submitting the new versions.
[2] I.e. each instead of a for loop over array indices, but not using
    default array values to initialize the array since array access speed
    is the question.

-- 
Jonathan Conway 	   The thing about Unix is that all the hoops are
rise@knavery.net           flaming, so at least you know where they are...