Version française
Home     About     Download     Resources     Contact us    
Browse thread
about OcamIL
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: -- (:)
From: Jon Harrop <jonathandeanharrop@g...>
Subject: RE: [Caml-list] about OcamIL
Xavier Clerc wrote:
> Jon Harrop <jonathandeanharrop@googlemail.com> a écrit :
> > Xavier Clerc:
> >> Le 14 mai 2010 à 12:40, Jon Harrop a écrit :
> >> > Xavier Clerc wrote:
> >> >> Limiting myself to the JVM...
> >> >> Moreover, at least Scala and Bigloo deliver excellent
> performances.
> >> >
> >> > I have benchmarks where the JVM is well over 10x slower than .NET.
> So
> >> > I do not regard any JVM-based language as "high performance".
> >>
> >> Quite ironically, by scratching the surface, one would discover that
> >> both quoted projects can also target .NET (not tested that though).
> >
> > Does Bigloo.NET support value types? Does Scala.NET use .NET (2.0)
> > generics?
> > Not AFAICT. Name dropping them in the context of "high performance"
> > language
> > implementations is more than a little bit silly...
> 
> First off, public insult seems quite superfluous.

I was not trying to insult you. Your examples are silly because they are
incomplete and untested. Do you even have either of them working right now?
AFAICT, Scala.NET is known not to work and Bigloo.NET is still have dozens
of core bugs fixed.

> We should be able to handle a heated debate without resorting to that.

I don't think this is heated at all. We were talking about "high
performance" languages and you cited a bunch of languages that get whipped
by Python on this benchmark:

 
http://flyingfrogblog.blogspot.com/2009/04/f-vs-ocaml-vs-haskell-hash-table.
html

> And I still wait for a clear statement of your level for "high
> performance",

Within 2x of ANSI C compiled with gcc on all practically-relevant benchmarks
without dropping to low-level code, e.g. GHC's FFI in Haskell.

> and references to benchmarks that back up your claims in this thread.

  http://fsharpnews.blogspot.com/2010/05/java-vs-f.html

> As you seem to come from an academic background, I expect facts
> and references, and not ad hominem attacks and fuzzy unbacked claims.

An ad-hominem attack is an attack against a person. I attacked your
examples, not you.

> Unless you show that neither Bigloo nor Scala meet your (to be defined)
> criteria for "high performance", my counterexamples still stand.

Are you talking about Bigloo.NET and Scala.NET or have you gone back to the
original discussion about JVM-based languages?

Scala on the JVM is 7x slower than C++ on this benchmark:

 
http://shootout.alioth.debian.org/u64q/benchmark.php?test=all&lang=scala&lan
g2=gpp

The JVM's hash table is 17x slower than .NET's on this benchmark:

  http://fsharpnews.blogspot.com/2010/05/java-vs-f.html

I think that is not "high performance" by any reasonable definition and this
reflects fundamental deficiencies in the VM itself, so there is no hope of
working around it in general.

I have not been able to get Bigloo to run: it was deleted from Debian and
Ubuntu (and the shootout) and the source distribution barfs during
configuration with " ./install-gc-7.1: 39: patch: not found".

> It may just end up that we have different perceptions of "high
> performance", and of the trade-offs we are going to make in our
> language / platform choices.

Probably. What languages do not you not consider to be high performance?

Cheers,
Jon.