Re: Statically AND Dynamically Typed Language ??

On Jan 6, 1:30 pm, Marco van de Voort <mar...@xxxxxxxx> wrote:
On 2009-01-06, Isaac Gouy <igo...@xxxxxxxxx> wrote:

Criticising theshootoutis not difficult, I mention a few points below.
Criticism is not difficult. Good criticism is difficult.

Well, I'm not going to debate about that.

I'll just mention some background: I liked theshootoutoriginally. I still
do, at least in principle. In the past I routinely used it to get a feel for
things. The first bunch of remarks were just about some pitfalls I found
when interpreting the results in the past. Some of the rankings simply
didn't scale beyond the simple benchmarks, or at least needed to be severely
nuances.

Of course one can react defensively by making some ridiculous analogies, and
do a bit selective reacting, but that doesn't help me one bit with theshootout, or change my opinion about it.

More recently, I haven't been able to do even that in my daily programming
practice. The last remarks testify of that frustration. There seems to be no
really useful ranking left.


That's a strange claim - the old default ranking for Pentium 4 seems
pretty much the same as the current default ranking for 32bit single
core Q6600 (except there are a lot fewer languages now).

http://shootout.alioth.debian.org/gp4/benchmark.php?test=all&lang=all

http://shootout.alioth.debian.org/u32/benchmark.php?test=all&lang=all


-snip-
If a computer program isn't "computational" what is it?

You really can't imagine programs that are not some math or
parsing problem in one or two loops?

Which do you think is a "parsing problem"?


-snip-
What do you mean by "overall application performance, and not
numeric or visual computational work"?

That means that I haven't found a language except maybe a lisp
interpreter that really has a rate determining speed depending
on tail recursion, simple vectorization etc.

4 years ago there were 3 different recursive problems: ackermann,
fibonacci and tak.

2 years ago they were combined into 1 problem: recursive.

There has only been 1 recursive problem for the last 2 years (and no
there are none) so I have no idea why you keep suggesting there's been
a heavy focus on recursion.

If the language implementation doesn't provide a way to use simple
vectorization...


-snip-
  o Some benchmarks obviously target certain very specific
    optimizations like tail recursion optimization.

Look and you'll see that "recursive" isn't measured anymore.

Good. Unfortunately, most seemed to have gone. And the old
ranking was more useful to get a feel for languages than the
current one (which seems to benchmark more the creativity of
cheap, but simplified in lanuage threading than the language
itself)

That's a strange claim - the old default ranking for Pentium 4 seems
pretty much the same as the current default ranking for 32bit single
core Q6600

http://shootout.alioth.debian.org/gp4/benchmark.php?test=all&lang=all

http://shootout.alioth.debian.org/u32/benchmark.php?test=all&lang=all


-snip-
If I'm bad at criticizing, what is this? Changing a debate into
a lousy analogy contest?

It's making obvious that those complaints amount to saying that the
benchmarks game does not even try to compare programming language
implementations in all the ways it might be useful to compare
programming language implementations.

If there was any confusion about that we could just read the top of
the FAQ

http://shootout.alioth.debian.org/u32/faq.php#scored


-snip-
  Typically they are only added to
  give functional languages a chance to excel at anything.

What exactly are you claiming was "added to give functional
languages a chance to excel"?

The old focus on recursion together with concise notation.

What "old focus on recursion"? There was one benchmark which focused
on recursion - no there are none.

What focus on concise notation?


-snip-
Some or all? If all, then why is there still a nearly 18 month
FPC 2.2.0 running, while there has been a new version (2.2.2)
since early august ?

No one has expressed any interest in FPC.


-snip-
I'll adapt some of the criticism on the page anyway, since this
is the first real feedback.

Why would you have expected feedback - did you post those comments in
the benchmarks game forum where they'd be seen?


-snip-
Compiler directives are allowed - on the command line ...

FPC/Delphi in general stuff them in source. Since you don't do
the maintaining, but throw it back at the contributors, I don't
understand this.

When there need to be multiple source code files for a program that is
identical apart from compiler directives - I have to maintain multiple
source code files for no benefit.
.

Relevant Pages
Loading