http://shootout.alioth.debian.org/u64q/benchmark.php?test=threadring&lang=all

2.3Go 6g 8g #3 20.6120.613,272347  0% 100% 0% 0%

Interestingly, the CPU load is kind of comical for something spawning so many threads.

Anyone good at optimized D feel like commiting a version of this?

On Fri, Dec 11, 2009 at 12:53 PM, S < <S@s.com> wrote:
> http://shootout.alioth.debian.org/u64q/benchmark.php?test=threadring&lang=all
>
> 2.3Go 6g 8g #3 20.6120.613,272347  0% 100% 0% 0%
>
> Interestingly, the CPU load is kind of comical for something spawning so many threads.
>
> Anyone good at optimized D feel like commiting a version of this?

D used to be there, but the folks running the shootout de-listed it
for some reason.
Maybe it was lack of a 64-bit compiler?

--bb

Bill Baxter:
> D used to be there, but the folks running the shootout de-listed it
> for some reason.
> Maybe it was lack of a 64-bit compiler?

I think the notgentle person that manages the Shootout site has removed D because D programs were too much similar to the C programs. He is looking for very different code, not slight variations.

Bye,
bearophile

On Fri, Dec 11, 2009 at 1:17 PM, bearophile <bearophileHUGS@lycos.com> wrote:
> Bill Baxter:
>> D used to be there, but the folks running the shootout de-listed it
>> for some reason.
>> Maybe it was lack of a 64-bit compiler?
>
> I think the notgentle person that manages the Shootout site has removed D because D programs were too much similar to the C programs. He is looking for very different code, not slight variations.

Hmm.  And Go qualifies because it has some CSP implementation built in? Or is it more because it has a billion-dollar company built in?  Kinda makes you wonder.

--bb

== Quote from bearophile (bearophileHUGS@lycos.com)'s article
> Bill Baxter:
> > D used to be there, but the folks running the shootout de-listed it
> > for some reason.
> > Maybe it was lack of a 64-bit compiler?
> I think the notgentle person that manages the Shootout site has removed D
because D programs were too much similar to the C programs. He is looking for very different code, not slight variations.
> Bye,
> bearophile

Yeah, I think it would be much more interesting (for C++ as well as D) to see benchmarks in idiomatic D style.  This will likely be slower than code written in the C-like subset, as there is really not much reason why C-like D code should be any slower than C.  However, it will give a better indication of how much the abstractions present in D really cost if you choose to use them.

Bill Baxter Wrote:

> Hmm.  And Go qualifies because it has some CSP implementation built in? Or is it more because it has a billion-dollar company built in?  Kinda makes you wonder.
> 
> --bb

Yeah, and if you look at the "Alternatives" to C++ on this page, C and Go are mentioned, but not D because of some bias.

http://harmful.cat-v.org/software/c++/

== Quote from Bill Baxter (wbaxter@gmail.com)'s article
> Maybe it was lack of a 64-bit compiler?

Hint, hint, hint.  I will be forever grateful to whoever makes it so I can use
>4GB of RAM in D2.  (Ancient versions of D2 don't count.)  Then again, these
benchmarks are focused on D1, for which a 64-bit compiler has existed for a while.

bearophile wrote:
> Bill Baxter:
>> D used to be there, but the folks running the shootout de-listed it
>> for some reason.
>> Maybe it was lack of a 64-bit compiler?
> 
> I think the notgentle person that manages the Shootout site has removed D because D programs were too much similar to the C programs. He is looking for very different code, not slight variations.
> 
	Which doesn't prevent him from having 3 versions for the C program
plus 2 for C++, not counting the Java and C# versions...

		Jerome
-- 
mailto:jeberger@free.fr
http://jeberger.free.fr
Jabber: jeberger@jabber.fr

On Fri, 11 Dec 2009, "J?r?me M. Berger" wrote:

> bearophile wrote:
> > Bill Baxter:
> > > D used to be there, but the folks running the shootout de-listed it
> > > for some reason.
> > > Maybe it was lack of a 64-bit compiler?
> > 
> > I think the notgentle person that manages the Shootout site has removed D because D programs were too much similar to the C programs. He is looking for very different code, not slight variations.
> > 
> 	Which doesn't prevent him from having 3 versions for the C program
> plus 2 for C++, not counting the Java and C# versions...
> 
> 		Jerome

Sorry for replying deep into the thread rather than the earlier post, but this is the one I have handy...

This community has a lot of bad habits.  The particularly dangerous one demonstrated well in this thread is that of making assumptions about intent.  D isn't listed in the shootout, it once was.  We don't know the facts about why so we like to make the up.  Its human.  But it's really harmful.

Another really good example:  Silence on a top tends to be equated with rejection, especially when it comes to why whatever it was hasn't been implemented yet.  The majority of the time that's not the reason for the silence.. it's merely a matter of 'haven't gotten to it yet' or 'didn't see it amongst the hundreds of other things'.

Later,
Brad

dsimcha:

> Yeah, I think it would be much more interesting (for C++ as well as D) to see benchmarks in idiomatic D style.  This will likely be slower than code written in the C-like subset, as there is really not much reason why C-like D code should be any slower than C.  However, it will give a better indication of how much the abstractions present in D really cost if you choose to use them.

I agree. For example D programs compiled with LDC are about as fast as C programs compiled with GCC if those D programs use only C features.
But essentially most D1 features that are not present in C lead to slow or very slow programs, when compiled with ldc (like interfaces, dynamic array appends, etc, etc).

Anyway, the Shootout allows you to have more than one version of the same code, so you can put two versions of the same D program, one high level and one lower level. This was a C-style version of the D k-nucleotide benchmark:
http://shootout.alioth.debian.org/debian/benchmark.php?test=knucleotide&lang=dlang&id=3
And my higher level version of the same benchmark (in this case the higher level is faster, but usually it's the opposite):
http://shootout.alioth.debian.org/debian/benchmark.php?test=knucleotide&lang=gdc&id=1

Bye,
bearophile