[Shootout-list] request to put MLton on "The Usual" page

[John Skaller]

On Tue, 2005-06-07 at 15:20 -0700, Brent Fulgham wrote:

> Do other people think we should use 'The Usual' to
> show
> the best performers for all languages, or would it
> be better to show the 'best known' implementations
> for various languages?

both?

> I'm sort of leaning towards the former (which would
> mean moving MLton to 'The Usual' instead of SML/NJ).

In that case will you move Felix into 'The Usual' since
it is not only 'the best known' Felix implementation,
it is also faster than gcc on some tests on some
platforms (it outperforms gcc *with optimisation
switches on* for Ackermann's function on an x86).

> Are there any contrary views?

Sort of. You/the community has to decide the 'purpose'
of the Shootout.

If it is a serious performance testing platform,
you would only show the results of fast translators.
This means throwing out Perl and Python, but you
would include both gcc and Intel C, for example.

If it is intended as a 'showcase' of different
interesting systems, you would not bother to show Intel C.
Who cares about Intel C?

In *both* cases you'd put Felix in 'the main
Shootout' now, since it is now extremely fast,
and I would contend it is certainly interesting.

In *both* cases MLton would be included as well,
since it is (a) the fastest SML implementation,
and (b) unlike most other compilers, MLton
is a whole program analyser. It is certainly
interesting to *me* to compare the performance
of a modern compiler design which throws out
the archaic concept of separate compilation
and linkage, with systems that continue
to support such outdated notions <hehe .. > 

Felix is also a whole program analyser. .. <grin>

I personally tend towards the 'interesting' languages
approach. I think inclusion of Intel C spoils the
main Shootout -- I don't give a hoot about how Intel C
compares with Ocaml or Haskell or Python. Yes, I may
be interested in how Intel C compares with gcc .. but
I'm even *more* interested in how gcc compares
with gcc -fomit-frame-pointer, and how gcc-3.4 compares
with gcc-4.0.



-- 
John Skaller, skaller at users.sf.net
PO Box 401 Glebe, NSW 2037, Australia Ph:61-2-96600850 
Download Felix here: http://felix.sf.net