> I find this particularly interesting because these are bast-case results for
> OCaml and C++. Specifically, the machine is a 2x Athlon 64 (OCaml does
> worse on Intels) and the F# is running in 32-bit Win XP but OCaml and g++
> are in 64-bit Linux.

> Having been using F# for many months now and typically seeing a slowdown
> when porting OCaml code to F#, I was pleasantly surprised to find that F#
> is already much faster than OCaml on some interesting benchmarks.
>
> Specifically, the numerically-intensive spectral-norm and n-body benchmarks
> from the shootout. I get:
>
> ocamlopt g++ -O3 F#
> Spectral-norm 14.94s 9.34s 9.37s
> N-body 9.23s 8.21s 6.87s
>
> I find this particularly interesting because these are bast-case results for
> OCaml and C++. Specifically, the machine is a 2x Athlon 64 (OCaml does
> worse on Intels) and the F# is running in 32-bit Win XP but OCaml and g++
> are in 64-bit Linux.

> Having been using F# for many months now and typically seeing a slowdown
> when porting OCaml code to F#, I was pleasantly surprised to find that F#
> is already much faster than OCaml on some interesting benchmarks.
>
> Specifically, the numerically-intensive spectral-norm and n-body benchmarks
> from the shootout. I get:
>
> ocamlopt g++ -O3 F#
> Spectral-norm 14.94s 9.34s 9.37s
> N-body 9.23s 8.21s 6.87s
>
> I find this particularly interesting because these are bast-case results for
> OCaml and C++. Specifically, the machine is a 2x Athlon 64 (OCaml does
> worse on Intels) and the F# is running in 32-bit Win XP but OCaml and g++
> are in 64-bit Linux.
>
> Any have data running F# on 64-bit Windows?
>
> -- ...

> I think you're blazing new ground!
>
> I don't remember seeing side-by-side programming language timings
> where the measurements were made on different OSs.
>
> Should we look forward to side-by-side comparions where OCaml was
> measured on Athlon and F# was measured on Xeon? ;-)

> I think that's awesome, but how useful is performance these days? I
> don't think people use python because it's fast. *I* certainly don't
> use ocaml because it *is* fast. I use it because it's damn faster to
> write code in than anything else. Testing and debugging python code
> totally sucks. I rarely test/debug ocaml code. When I do, it's
> fairly simple and straightforward.

> Isaac Gouy wrote:

>> I think you're blazing new ground!
>>
>> I don't remember seeing side-by-side programming language timings
>> where the measurements were made on different OSs.
>>
>> Should we look forward to side-by-side comparions where OCaml was
>> measured on Athlon and F# was measured on Xeon? ;-)

>
> I'll be sure to learn how to benchmark properly from you, Isaac:
>
> http://shootout.alioth.debian.org/sandbox/benchmark.php?test=binarytrees&lang=ocaml&id...
> - "De-optimized by Isaac Gouy"

> baguasquirrel@gmail.com wrote:

>> I think that's awesome, but how useful is performance these days? I
>> don't think people use python because it's fast. *I* certainly don't
>> use ocaml because it *is* fast. I use it because it's damn faster to
>> write code in than anything else. Testing and debugging python code
>> totally sucks. I rarely test/debug ocaml code. When I do, it's
>> fairly simple and straightforward.

>
> Well, performance is still vital for a wide variety of important
> applications in science and engineering. So the performance of languages in
> their respective environments is significant for anyone wanting to do such
> work.

> Oh, come on. This is becoming stale. Not that I could say that of if
> I.G. can benchmark or does benchmark right. We know that is a pet
> peeve of yours, but instead of focusing on what Isaac does wrong, do
> it better. It shouldn't be difficult to build a better benchmarking
> side _if_ you do it right, that means (a) don't compare apples with
> oranges, (b) keep the proper scientific spirit regarding the openness
> of results and (c) state clearly what the purpose of the benchmarks
> is.

> Specifically, the numerically-intensive spectral-norm and n-body benchmarks
> from the shootout. I get:
>
> ocamlopt g++ -O3 F#
> Spectral-norm 14.94s 9.34s 9.37s
> N-body 9.23s 8.21s 6.87s