[ocaml-biz] strategy bullet point list

[Tony Edgin]

>> On Mon, 30 Aug 2004, William D. Neumann wrote:
>> I really know jack about C#, so there may be a simple,
>> valid counterpoint, but if one looks at the recently
>> revived language shootout
>> <http://shootout.alioth.debian.org/index.php>, C# ranks
>> ahead of OCaml in raw speed.  So if we're going to claim
>> that C# does not offer performance, we're going to have to
>> come up with a valid explanation of why it ranks ahead of
>> OCaml on that scorecard...
>
> Brian Hurt wrote:
> Well, first of all, no one is optimizing the Ocaml code-
> while other languages are getting optimized.  From my
> following of the mailing list, C++ and C# are especially
> active.  Second, I'm wondering how csharp got a score of
> 39.9, when the next non-natively-compiled language
> (Java/Kaffe) came in 10 steps and 13.1 points.  Sun's Java
> is another 4 steps and 0.5 points down from Kaffe.  I think
> the question is not "why is Ocaml lower than C#", the
> question is "why is C# so much higher performance than Java,
> Ocaml byte code, and other byte code interpreted languages?"
>
> I suspect that some proscribed optimization is being
> performed- like the environment has noticed that results are
> simply be thrown away, and thus it isn't computing them.  I
> don't know this for sure, but I suspect it.
>
> My other answer to that is:
>http://shootout.alioth.debian.org/craps.php?xcpu=0&xmem=1&xloc=0&ackermann=1&ary=3&echo=5&except=1&fibo=2&hash=1&hash2=4&heapsort=4&hello=1&lists=3&matrix=3&methcall=5&moments=2&nestedloop=4&objinst=5&prodcons=1&random=3&regexmatch=4&reversefile=4&ringmsg=1&sieve=4&spellcheck=4&strcat=2&sumcol=3&wc=3&wordfreq=5
> and:
>http://shootout.alioth.debian.org/craps.php?xcpu=0&xmem=0&xloc=1&ackermann=1&ary=3&echo=5&except=1&fibo=2&hash=1&hash2=4&heapsort=4&hello=1&lists=3&matrix=3&methcall=5&moments=2&nestedloop=4&objinst=5&prodcons=1&random=3&regexmatch=4&reversefile=4&ringmsg=1&sieve=4&spellcheck=4&strcat=2&sumcol=3&wc=3&wordfreq=5

If you weight all three measure sets evenly, plus weight each
test evenly, Ocaml native and byte codes do surprisingly well.
 In summary, C comes in with a 48, Ocaml native with a 42,
C++ with a 34, Ocaml byte with a 32, C# with a 26, and Java
with a 19.  Although the shoot out shouldn't be taken too
seriously, maybe this could be used as part of the answer too.  

http://shootout.alioth.debian.org/craps.php?xcpu=1&xmem=1&xloc=1&ackermann=1&ary=1&echo=1&except=1&fibo=1&hash=1&hash2=1&heapsort=1&hello=1&lists=1&matrix=1&methcall=1&moments=1&nestedloop=1&objinst=1&prodcons=1&random=1&regexmatch=1&reversefile=1&ringmsg=1&sieve=1&spellcheck=1&strcat=1&sumcol=1&wc=1&wordfreq=1

cheers,
Tony Edgin

? 2004 Cbeyond Communications. All Rights Reserved.