ada is getting spanked :(

ada is getting spanked :(

[cl1]

I just got wind of this site. It would seem that ada, while not at
the very bottom, is one of the slowest languages in this benchmark
competition:

http://shootout.alioth.debian.org

Re: ada is getting spanked :(

[Martin Krischik]

cl1 wrote:

> I just got wind of this site. It would seem that ada, while not at
> the very bottom, is one of the slowest languages in this benchmark
> competition:
> 
> http://shootout.alioth.debian.org

Looks worse then last time I looked. Seems that some language advocates
really optimised there implementations.

And there is still one test missing for Ada.

Martin
-- 
mailto://krischik@users.sourceforge.net
Ada programming at: http://ada.krischik.com

Re: ada is getting spanked :(

[Georg Bauhaus]

On Sat, 2006-10-28 at 19:45 +0200, Martin Krischik wrote:
> cl1 wrote:
> 
> > I just got wind of this site. It would seem that ada, while not at
> > the very bottom, is one of the slowest languages in this benchmark
> > competition:
> > 
> > http://shootout.alioth.debian.org
> 
> Looks worse then last time I looked. Seems that some language advocates
> really optimised there implementations.

Don't think so, a more recent GNAT's output is speeding
up, almost keeping pace. For example, comparing the C and Ada versions
of the terribly recursive program using GCC 4.2 on -march=pentium,
the C advantage shrinks: 6.7s (C) : 7.3s (Ada).

The Debian / AMD listing shows 2.68 : 4.00,
the Gentoo / P4 listing shows 2.63 : 10.61

This says something about the volatility of speed comparisons
(probably documented on site). :-)

Re: ada is getting spanked :(

[Georg Bauhaus]

On Sat, 2006-10-28 at 19:45 +0200, Martin Krischik wrote:
 
> > http://shootout.alioth.debian.org
> 
> Looks worse then last time I looked. Seems that some language advocates
> really optimised there implementations.

In order to optimize GNAT's standing, I have made a small
but quite effective change to one of Jim Rogers' programs
(regex-dna #2), the factor is 14 (fourteen). Nothing submitted
yet, please use this as a starting point if you like.

See here, under the heading Shootout:
http://home.arcor.de/bauhaus/Ada/

I have some more ideas but cannot try them right now.

regards,
Georg Bauhaus

Re: ada is getting spanked :(

[Jeffrey R. Carter]

Georg Bauhaus wrote:
> 
> In order to optimize GNAT's standing, I have made a small
> but quite effective change to one of Jim Rogers' programs
> (regex-dna #2), the factor is 14 (fourteen). Nothing submitted
> yet, please use this as a starting point if you like.

I've found a small change that improves the nsieve program (also Jim 
Rogers). Change the line

S (I) := False;

to

if S (I) then
    S (I) := False;
end if;

on my computer (3.2 GHz Pentium 4 HT), compiled with MinGW GNAT 3.4.2 with

-O3 -gnatnp -march=pentium4 -fomit-frame-pointer

the time was reduced from about 1.8 s for the original to about 1.1 s 
with this modification, a speedup of 38.9%. Applied to the 2.71 s 
reported time, that would give GNAT 1.66 s, which would move it to 3rd 
place.

I hope no one thinks we're picking on Jim.

-- 
Jeff Carter
"Your mother was a hamster and your father smelt of elderberries."
Monty Python & the Holy Grail
06

Re: ada is getting spanked :(

[Isaac Gouy]

Jeffrey R. Carter wrote:
> Georg Bauhaus wrote:
> >
> > In order to optimize GNAT's standing, I have made a small
> > but quite effective change to one of Jim Rogers' programs
> > (regex-dna #2), the factor is 14 (fourteen). Nothing submitted
> > yet, please use this as a starting point if you like.
>
> I've found a small change that improves the nsieve program (also Jim
> Rogers). Change the line
>
> S (I) := False;
>
> to
>
> if S (I) then
>     S (I) := False;
> end if;
>
> on my computer (3.2 GHz Pentium 4 HT), compiled with MinGW GNAT 3.4.2 with
>
> -O3 -gnatnp -march=pentium4 -fomit-frame-pointer
>
> the time was reduced from about 1.8 s for the original to about 1.1 s
> with this modification, a speedup of 38.9%. Applied to the 2.71 s
> reported time, that would give GNAT 1.66 s, which would move it to 3rd
> place.
>
> I hope no one thinks we're picking on Jim.
>
> --
> Jeff Carter
> "Your mother was a hamster and your father smelt of elderberries."
> Monty Python & the Holy Grail
> 06

Thanks for the reminder that I should throwout/fix the programs that
use a test to avoid doing what the benchmark description asks them to
do :-)

Re: ada is getting spanked :(

[Isaac Gouy]

Isaac Gouy wrote:
> Jeffrey R. Carter wrote:
> > Georg Bauhaus wrote:
> > >
> > > In order to optimize GNAT's standing, I have made a small
> > > but quite effective change to one of Jim Rogers' programs
> > > (regex-dna #2), the factor is 14 (fourteen). Nothing submitted
> > > yet, please use this as a starting point if you like.
> >
> > I've found a small change that improves the nsieve program (also Jim
> > Rogers). Change the line
> >
> > S (I) := False;
> >
> > to
> >
> > if S (I) then
> >     S (I) := False;
> > end if;
> >
> > on my computer (3.2 GHz Pentium 4 HT), compiled with MinGW GNAT 3.4.2 with
> >
> > -O3 -gnatnp -march=pentium4 -fomit-frame-pointer
> >
> > the time was reduced from about 1.8 s for the original to about 1.1 s
> > with this modification, a speedup of 38.9%. Applied to the 2.71 s
> > reported time, that would give GNAT 1.66 s, which would move it to 3rd
> > place.
> >
> > I hope no one thinks we're picking on Jim.
> >
> > --
> > Jeff Carter
> > "Your mother was a hamster and your father smelt of elderberries."
> > Monty Python & the Holy Grail
> > 06
>
> Thanks for the reminder that I should throwout/fix the programs that
> use a test to avoid doing what the benchmark description asks them to
> do :-)

And that has now been done (although it will be a little while before
the ripple reaches the Debian/AMD measurements).

Re: ada is getting spanked :(

[Jeffrey R. Carter]

For the Nsieve problem, is there any reason not to do it like this:

-- The Great Computer Language Shootout
-- http://shootout.alioth.debian.org/
--
-- Contributed by Jim Rogers
-- Modified by Jeffrey R. Carter

with Ada.Command_Line;    use Ada.Command_Line;
with Ada.Text_IO;         use Ada.Text_IO;
with Ada.Integer_Text_IO; use Ada.Integer_Text_IO;

procedure Nsieve is
    procedure Count (N : in Natural) is
       M : constant Natural := 2 ** N * 10_000;

       type Boolean_Array is array (2 .. M) of Boolean;

       M2 : constant Natural := M / 2;
       M3 : constant Natural := M2 / 2;

       C1 : Natural := 0;
       C2 : Natural := 0;
       C3 : Natural := 0;
       S : Boolean_Array := (others => True);
       I : Positive;
    begin -- Count
       for K in S'Range loop
          if S (K) then
             C1 := C1 + 1;
             I := K + K;

             if I <= M2 then
                C2 := C2 + 1;

                if I <= M3 then
                   C3 := C3 + 1;
                end if;
             end if;

             while I <= M loop
                if S (I) then
                   S (I) := False;
                end if;

                I := I + K;
             end loop;
          end if;
       end loop;

       Put ("Primes up to ");
       Put (Item => M, Width => 8);
       Put (" ");
       Put (Item => C1, Width => 8);
       New_Line;
       Put ("Primes up to ");
       Put (Item => M2, Width => 8);
       Put (" ");
       Put (Item => C2, Width => 8);
       New_Line;
       Put ("Primes up to ");
       Put (Item => M3, Width => 8);
       Put (" ");
       Put (Item => C3, Width => 8);
       New_Line;
    end Count;

    N : constant Natural := Natural'Value (Argument (1) );
begin -- Nsieve
    Count (N);
end Nsieve;

?

This is much faster (0.6 s on my machine, which gave about 1.8 s for the 
existing version), and should put us in 1st place.

-- 
Jeff Carter
"Your mother was a hamster and your father smelt of elderberries."
Monty Python & the Holy Grail
06

Re: ada is getting spanked :(

[Jeffrey R. Carter]

Jeffrey R. Carter wrote:

Minor correction. Change

>             I := K + K;
> 
>             if I <= M2 then
>                C2 := C2 + 1;
> 
>                if I <= M3 then
>                   C3 := C3 + 1;
>                end if;
>             end if;
> 
>             while I <= M loop

to


 >             if I <= M2 then
 >                C2 := C2 + 1;
 >
 >                if I <= M3 then
 >                   C3 := C3 + 1;
 >                end if;
 >             end if;
 >
 >             I := K + K;
 >
 >             while I <= M loop

-- 
Jeff Carter
"Your mother was a hamster and your father smelt of elderberries."
Monty Python & the Holy Grail
06

Re: ada is getting spanked :(

[Jeffrey R. Carter]

Jeffrey R. Carter wrote:

I'm not thinking very well tonight. Make that

>  >             if K <= M2 then
>  >                C2 := C2 + 1;
>  >
>  >                if K <= M3 then
>  >                   C3 := C3 + 1;
>  >                end if;
>  >             end if;
>  >
>  >             I := K + K;
>  >
>  >             while I <= M loop

I apologize for any confusion. The timings remain the same.

-- 
Jeff Carter
"Your mother was a hamster and your father smelt of elderberries."
Monty Python & the Holy Grail
06

Re: ada is getting spanked :(

[Jeffrey R. Carter]

Another source of speedup for the nsieve program is to pack the array 
type. We're still dealing directly with Booleans, so I don't see that 
this is not allowed by the benchmark description.

-- 
Jeff Carter
"I'm a kike, a yid, a heebie, a hook nose! I'm Kosher,
Mum! I'm a Red Sea pedestrian, and proud of it!"
Monty Python's Life of Brian
77

Re: ada is getting spanked :(

[Jeffrey R. Carter]

Also for nsieve, M is defined to be a multiple of 2, and so we know it 
is not prime. We can omit it from consideration by changing the loop 
condition to

while I < M loop

This is done by the C++ program in 2nd place, for example:

for (size_t i=2; i<m; ++i)
   if (isPrime[i])
     {
       for(size_t k=i*2; k<m; k+=i)
         isPrime[k] = false;
       ++count;
     }

which only looks at 2 .. M - 1.

I don't expect much of an improvement from this, though.

-- 
Jeff Carter
"I'm a kike, a yid, a heebie, a hook nose! I'm Kosher,
Mum! I'm a Red Sea pedestrian, and proud of it!"
Monty Python's Life of Brian
77

Re: ada is getting spanked :(

[Martin Krischik]

Georg Bauhaus wrote:

> On Sat, 2006-10-28 at 19:45 +0200, Martin Krischik wrote:
>  
>> > http://shootout.alioth.debian.org
>> 
>> Looks worse then last time I looked. Seems that some language advocates
>> really optimised there implementations.
> 
> In order to optimize GNAT's standing, I have made a small
> but quite effective change to one of Jim Rogers' programs
> (regex-dna #2), the factor is 14 (fourteen). Nothing submitted
> yet, please use this as a starting point if you like.
> 
> See here, under the heading Shootout:
> http://home.arcor.de/bauhaus/Ada/
> 
> I have some more ideas but cannot try them right now.

Wikibooks Ada - the sourceforce project [1] has a place for the shootout [2]
code. Having the code inside a Subversion archive means we can colaborate.
And one can use the resulting code as example repository for the book ;-).

Martin

[1] http://wikibook-ada.sourceforge.net/
[2]
http://wikibook-ada.svn.sourceforge.net/viewvc/wikibook-ada/trunk/demos/Source/Language_Shootout
-- 
mailto://krischik@users.sourceforge.net
Ada programming at: http://ada.krischik.com

Re: ada is getting spanked :(

[Georg Bauhaus]

Martin Krischik wrote:
> Georg Bauhaus wrote:

>> See here, under the heading Shootout:
>> http://home.arcor.de/bauhaus/Ada/
>>
>> I have some more ideas but cannot try them right now.
> 
> Wikibooks Ada - the sourceforce project [1] has a place for the shootout [2]
> code. Having the code inside a Subversion archive means we can colaborate.
> And one can use the resulting code as example repository for the book ;-).

> http://wikibook-ada.svn.sourceforge.net/viewvc/wikibook-ada/trunk/demos/Source/Language_Shootout

OK, let's build the cathedral, a good one. However, this particular
program isn't mine, I'd not want to say where it goes, license permitting
or not. The source mentioned above is put there for the sake of this
discussion. It is supposed to vanish later. Surely I hope to see it as part.

BTW, is there a way to retrieve old contributions to the Shootout?
I remember some very elegant programs for example in ML that
I can no longer find there.

Re: ada is getting spanked :(

[Isaac Gouy]

Georg Bauhaus wrote:
-snip-
> BTW, is there a way to retrieve old contributions to the Shootout?
> I remember some very elegant programs for example in ML that
> I can no longer find there.

No. However if the programs were contributed in the normal way they
will still be in the "Contribute Programs" tracker system - provided by
Alioth GForge. Unhappily those webpages have been "improved" in the
past few days and now you have to - 1) Login 2) Build and save a query
- before you can seach on a specific "language implementation" category
like SML or MLton.

Re: ada is getting spanked :(

[Björn Persson]

Georg Bauhaus wrote:
> In order to optimize GNAT's standing, I have made a small
> but quite effective change to one of Jim Rogers' programs
> (regex-dna #2), the factor is 14 (fourteen).

Ah, the one that uses Spitbol. Doesn't it really belong under 
"interesting alternative programs"? The requirements clearly state that 
regular expressions should be used.

I fixed up the regex version some time ago and achieved a dramatic 
improvement, but I didn't touch the Spitbol version. They were very 
similar before that, and as I recall they had similar performance too.
My changes are now in "regex-dna Ada 95 GNAT #3".

-- 
Bj�rn Persson                              PGP key A88682FD
                    omb jor ers @sv ge.
                    r o.b n.p son eri nu

Re: ada is getting spanked :(

[Georg Bauhaus]

On Sun, 2006-10-29 at 16:27 +0000, Björn Persson wrote:
> Georg Bauhaus wrote:
> > In order to optimize GNAT's standing, I have made a small
> > but quite effective change to one of Jim Rogers' programs
> > (regex-dna #2), the factor is 14 (fourteen).

The speed factor is now close to 18 :-). After some simplifications,
and after (I think) more closely reflecting the benchmark description.

http://home.arcor.de/bauhaus/Ada/

The program now also works more like the others (in my view).

> Ah, the one that uses Spitbol. Doesn't it really belong under 
> "interesting alternative programs"? The requirements clearly state that 
> regular expressions should be used.

The pattern strings in all programs all look regular to me,
including the GNAT.SPITBOL ones. But that doesn't mean
that the various programs' calls to RE routines such as
.findall, -all, m//g global subst(...) etc. imply "normal"
naive regular expression processing. So GNAT.SPITBOL is no
exception here. (In particular, I have left out obvious
standard SPITBOL improvements in order to reflect the benchmark
description.)
 I guess that the benchmark is also about how well an
implementation deals with just simple REs.

Anyway, I think that all of the RE, PCRE, and SPITBOL patterns
I've seen in many of the contributions reflect the spirit of the
benchmark, as is required (literally).
How could a notion of REs be both precise and precisely
applicable to the ways in which various PLs implement them?
 E.g. how can you turn off Boyer-Moore string searching when
one implementation has it, just so that only some specified
internal way of pattern matching is compared?

So my rule was: Simple patterns, no tricks.

> I fixed up the regex version some time ago and achieved a dramatic 
> improvement, but I didn't touch the Spitbol version. They were very 
> similar before that, and as I recall they had similar performance too.
> My changes are now in "regex-dna Ada 95 GNAT #3".

I did have a look at this program, however with program #3 as is
I keep getting segmentation faults due to the

  Sequence_Lines (1 .. 1_000_000) of type Unbounded_String.

ulimit -s 10000 didn't help, reducing the number of lines did.


regards,
Georg 

Re: ada is getting spanked :(

[Björn Persson]

Georg Bauhaus wrote:
> The speed factor is now close to 18 :-). After some simplifications,
> and after (I think) more closely reflecting the benchmark description.

You're probably going to beat #3, because Gnat.Spitbol works with 
unbounded strings but Gnat.Regpat requires fixed strings.

> How could a notion of REs be both precise and precisely
> applicable to the ways in which various PLs implement them?
>  E.g. how can you turn off Boyer-Moore string searching when
> one implementation has it, just so that only some specified
> internal way of pattern matching is compared?

I wasn't talking of internal implementations. As I understand it, 
"regex" is a name of a family of similar notations for expressing 
patterns. If you don't write "[ABC]*" for "any number of A's, B's and 
C's in any order", then it isn't regex.

The requirement to use one particular pattern notation may of course be 
unintentional. Maybe the person who wrote the requirements didn't know 
that other notations existed.

> I did have a look at this program, however with program #3 as is
> I keep getting segmentation faults due to the
> 
>   Sequence_Lines (1 .. 1_000_000) of type Unbounded_String.
> 
> ulimit -s 10000 didn't help, reducing the number of lines did.

Now that you mention it, I remember having that problem too. I didn't 
change that part, and as it had obviously worked on the contest servers 
I figured it should keep working there. I suppose #2 falls below that 
limit on your box because it calculates how big the array needs to be.

-- 
Björn Persson                              PGP key A88682FD
                    omb jor ers @sv ge.
                    r o.b n.p son eri nu

Re: ada is getting spanked :(

[Dmitry A. Kazakov]

On Sun, 29 Oct 2006 23:50:24 GMT, Bj�rn Persson wrote:

> Georg Bauhaus wrote:

>> How could a notion of REs be both precise and precisely
>> applicable to the ways in which various PLs implement them?
>>  E.g. how can you turn off Boyer-Moore string searching when
>> one implementation has it, just so that only some specified
>> internal way of pattern matching is compared?
> 
> I wasn't talking of internal implementations. As I understand it, 
> "regex" is a name of a family of similar notations for expressing 
> patterns. If you don't write "[ABC]*" for "any number of A's, B's and 
> C's in any order", then it isn't regex.

Actually RE is not a notation, it is a formal language. SNOBOL patterns
represent a different wider class of languages. For example SNOBOL can
matching balanced strings like (a(b))(c). Technically, a highly optimized
implementation of RE should beat SNOBOL.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Re: ada is getting spanked :(

[Georg Bauhaus]

On Tue, 2006-10-31 at 12:32 +0100, Dmitry A. Kazakov wrote:

> Actually RE is not a notation, it is a formal language. SNOBOL patterns
> represent a different wider class of languages. For example SNOBOL can
> matching balanced strings like (a(b))(c). Technically, a highly optimized
> implementation of RE should beat SNOBOL.

Why? Boyer-Moore and similar algorithms may well be part
of any string scanning machine, no matter what languages
it can recognize. By the AS-IF rule, as long as RE semantics
are followed, I don't see a reason why optimized RE
scanning should be faster than optimized CF scanning.

Re: ada is getting spanked :(

[Dmitry A. Kazakov]

On Tue, 31 Oct 2006 14:12:41 +0100, Georg Bauhaus wrote:

> On Tue, 2006-10-31 at 12:32 +0100, Dmitry A. Kazakov wrote:
> 
>> Actually RE is not a notation, it is a formal language. SNOBOL patterns
>> represent a different wider class of languages. For example SNOBOL can
>> matching balanced strings like (a(b))(c). Technically, a highly optimized
>> implementation of RE should beat SNOBOL.
> 
> Why? Boyer-Moore and similar algorithms may well be part
> of any string scanning machine, no matter what languages
> it can recognize. By the AS-IF rule, as long as RE semantics
> are followed, I don't see a reason why optimized RE
> scanning should be faster than optimized CF scanning.

Yes, if you (the compiler) *can* understand semantics.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Re: ada is getting spanked :(

[Martin Krischik]

Bjï¿œrn Persson wrote:

> The requirements clearly state that
> regular expressions should be used.

But they did not say "perl-regex", "sed-regex", etc. pp.

Martin
-- 
mailto://krischik@users.sourceforge.net
Ada programming at: http://ada.krischik.com

Re: ada is getting spanked :(

[Isaac Gouy]

Martin Krischik wrote:
> cl1 wrote:
>
> > I just got wind of this site. It would seem that ada, while not at
> > the very bottom, is one of the slowest languages in this benchmark
> > competition:
> >
> > http://shootout.alioth.debian.org
>
> Looks worse then last time I looked. Seems that some language advocates
> really optimised there implementations.
>
> And there is still one test missing for Ada.
>
> Martin
> --
> mailto://krischik@users.sourceforge.net
> Ada programming at: http://ada.krischik.com

The positions for the compiled languages shuffled around a bit after we
made an effort to use the same gcc compiler options on all the language
implementations that would take them.

And at least on Gentoo, after upgrading to gcc 4.1.1 I'm not that sure
which version of GNAT we should be using?

Re: ada is getting spanked :(

[Martin Krischik]

Isaac Gouy wrote:

> 
> Martin Krischik wrote:
>> cl1 wrote:
>>
>> > I just got wind of this site. It would seem that ada, while not at
>> > the very bottom, is one of the slowest languages in this benchmark
>> > competition:
>> >
>> > http://shootout.alioth.debian.org
>>
>> Looks worse then last time I looked. Seems that some language advocates
>> really optimised there implementations.
>>
>> And there is still one test missing for Ada.
>>
>> Martin
>> --
>> mailto://krischik@users.sourceforge.net
>> Ada programming at: http://ada.krischik.com
> 
> The positions for the compiled languages shuffled around a bit after we
> made an effort to use the same gcc compiler options on all the language
> implementations that would take them.

I see. perhaps we should have a look at the options used for Ada - there are
a few performace options for GNAT only.

> And at least on Gentoo, after upgrading to gcc 4.1.1 I'm not that sure
> which version of GNAT we should be using?

4.1.* is usually good for Ada - and as we recently found out improves
performance as well.

Martin
-- 
mailto://krischik@users.sourceforge.net
Ada programming at: http://ada.krischik.com

Re: ada is getting spanked :(

[Isaac Gouy]

Martin Krischik wrote:
> Isaac Gouy wrote:
>
> >
> > Martin Krischik wrote:
> >> cl1 wrote:
> >>
> >> > I just got wind of this site. It would seem that ada, while not at
> >> > the very bottom, is one of the slowest languages in this benchmark
> >> > competition:
> >> >
> >> > http://shootout.alioth.debian.org
> >>
> >> Looks worse then last time I looked. Seems that some language advocates
> >> really optimised there implementations.
> >>
> >> And there is still one test missing for Ada.
> >>
> >> Martin
> >> --
> >> mailto://krischik@users.sourceforge.net
> >> Ada programming at: http://ada.krischik.com
> >
> > The positions for the compiled languages shuffled around a bit after we
> > made an effort to use the same gcc compiler options on all the language
> > implementations that would take them.
>
> I see. perhaps we should have a look at the options used for Ada - there are
> a few performace options for GNAT only.
>
> > And at least on Gentoo, after upgrading to gcc 4.1.1 I'm not that sure
> > which version of GNAT we should be using?
>
> 4.1.* is usually good for Ada - and as we recently found out improves
> performance as well.
>
> Martin
> --
> mailto://krischik@users.sourceforge.net
> Ada programming at: http://ada.krischik.com

The Gentoo/Intel website has now been updated, and the Ada measurements
are for GNAT 4.1.1

There were several small performance improvements but the largest
improvement was for recursive.

Re: ada is getting spanked :(

[jtg]

Martin Krischik wrote:

> 
> And there is still one test missing for Ada.
> 

Some time ago I started to implement it. However, the problem requires 
usage of big integers, which are not supported by Ada.
However, gcc has very fast library (gmp), which can be used in c and c++ 
programs just by inserting "# include <gmp.h>" into a program file and 
"-lgmp" into a compiler command-line.
Is there any standard Ada binding to the GMP library? By "standard 
binding" I mean a binding included in gnat/gcc.

Re: ada is getting spanked :(

[george]

jtg wrote:
> Is there any standard Ada binding to the GMP library? By "standard
> binding" I mean a binding included in gnat/gcc.

You can take a look here:
https://bugs.gentoo.org/show_bug.cgi?id=132119
As you can see, this is on my list, but first I need to finish with the
transition to split gnat compilers (should be soon finally - only a few
libs left) - in Gentoo you have the ability to have multiple gnat
implementation installed side-by-side and switch them on the fly..

>From the description it seems that the link in the last posting may be
in decent shape, however I did not look at it in detail yet.. 

George

Re: ada is getting spanked :(

[Jeffrey R. Carter]

jtg wrote:
> 
> Some time ago I started to implement it. However, the problem requires 
> usage of big integers, which are not supported by Ada.
> However, gcc has very fast library (gmp), which can be used in c and c++ 
> programs just by inserting "# include <gmp.h>" into a program file and 
> "-lgmp" into a compiler command-line.
> Is there any standard Ada binding to the GMP library? By "standard 
> binding" I mean a binding included in gnat/gcc.

How big do they need to be? Signed or unsigned?

-- 
Jeff Carter
"Mr. President, we must not allow a mine-shaft gap!"
Dr. Strangelove
33

Re: ada is getting spanked :(

[Gautier]

jtg:

> Some time ago I started to implement it. However, the problem requires
> usage of big integers, which are not supported by Ada.

Sure they are! Only on this page you have 5 flavours of them:
  http://www.chez.com/bignumber/
A quick look on Google seems to give some more.

> However, gcc has very fast library (gmp), which can be used in c and c++
> programs just by inserting "# include <gmp.h>" into a program file and
> "-lgmp" into a compiler command-line.
> Is there any standard Ada binding to the GMP library? By "standard
> binding" I mean a binding included in gnat/gcc.

The AdaGMP bindings seem to be no more available, however, you can give
a try with my package which doesn't even require dynamic allocation:

  http://homepage.sunrise.ch/mysunrise/gdm/gsoft.htm#mathpaqs

see the mathpaqs.zip archive, mupreint.ad* files.
There are also a test and a RSA cypher using it.

HTH
Gautier
_______________________________________________________________ 
Ada programming -- http://www.mysunrise.ch/users/gdm/gsoft.htm 

NB: For a direct answer, e-mail address on the Web site!

Re: ada is getting spanked :(

[Martin Krischik]

jtg schrieb:
> Martin Krischik wrote:
> 
>>
>> And there is still one test missing for Ada.
>>
> 
> Some time ago I started to implement it. However, the problem requires 
> usage of big integers, which are not supported by Ada.
> However, gcc has very fast library (gmp), which can be used in c and c++ 
> programs just by inserting "# include <gmp.h>" into a program file and 
> "-lgmp" into a compiler command-line.
> Is there any standard Ada binding to the GMP library? By "standard 
> binding" I mean a binding included in gnat/gcc.

I too have looked into implementing the test. I found out that there are 
several Big_Integer implementations - some a binding - some native.

Really one would need to implement the test with all of them to find out 
which is the most effective Big_Integer implementation.

To much for a lone developer but a collaboration could do it.

Martin

Re: ada is getting spanked :(

[Jeffrey R. Carter]

Martin Krischik wrote:
> 
> I too have looked into implementing the test. I found out that there are 
> several Big_Integer implementations - some a binding - some native.
> 
> Really one would need to implement the test with all of them to find out 
> which is the most effective Big_Integer implementation.

I have implemented the Pi-digits program using Jerome Delcourt's 
Big_Number package

http://www.chez.com/bignumber/info.htm

2500 digits takes about 7 s on my 3.2 GHz Pentium 4 HT/WinXP machine, so 
that would be about 9 s on the comparison machine, putting Ada in the 
vicinity of Python.

I looked at Gautier de Montmollin's package, but there doesn't seem to 
be a way to convert a big-num to an Integer.

Based on the Free Pascal entry, binding to gmp would be much faster.

-- 
Jeff Carter
"I soiled my armor, I was so scared."
Monty Python & the Holy Grail
71

Re: ada is getting spanked :(

[Gautier]

Jeffrey R. Carter:

> I looked at Gautier de Montmollin's package, but there doesn't seem to 
> be a way to convert a big-num to an Integer.

Mmmh, you looked at multi002.zip on Jerome's page, which appears to be 
outdated, and not at mathpaqs.zip, where you have:

   -- Convert basic_int to multi_int
   function Multi(small: Basic_int) return Multi_int;

   -- Convert multi_int to basic_int (when possible, else: Cannot_fit raised)
   function Basic(large: Multi_int) return Basic_int;

I'd be keen to give a try. Do you have a piece of code I could
continue with that pi test ?
Cheers, Gautier
______________________________________________________________
Ada programming -- http://www.mysunrise.ch/users/gdm/gsoft.htm

NB: For a direct answer, e-mail address on the Web site!

Re: ada is getting spanked :(

[Jeffrey R. Carter]

Gautier wrote:
> 
> Mmmh, you looked at multi002.zip on Jerome's page, which appears to be 
> outdated, and not at mathpaqs.zip, where you have:
> 
>   -- Convert basic_int to multi_int
>   function Multi(small: Basic_int) return Multi_int;
> 
>   -- Convert multi_int to basic_int (when possible, else: Cannot_fit 
> raised)
>   function Basic(large: Multi_int) return Basic_int;

Yes, that's correct.

> I'd be keen to give a try. Do you have a piece of code I could
> continue with that pi test ?

I based mine on the Pascal version.

-- 
Jeff Carter
"I soiled my armor, I was so scared."
Monty Python & the Holy Grail
71

Re: ada is getting spanked :(

[Martin Krischik]

Jeffrey R. Carter schrieb:
> Martin Krischik wrote:
>>
>> I too have looked into implementing the test. I found out that there 
>> are several Big_Integer implementations - some a binding - some native.
>>
>> Really one would need to implement the test with all of them to find 
>> out which is the most effective Big_Integer implementation.
> 
> I have implemented the Pi-digits program using Jerome Delcourt's 
> Big_Number package
> 
> http://www.chez.com/bignumber/info.htm
> 
> 2500 digits takes about 7 s on my 3.2 GHz Pentium 4 HT/WinXP machine, so 
> that would be about 9 s on the comparison machine, putting Ada in the 
> vicinity of Python.

Would you mind adding it to the wikibook ada subversion archive

http://sourceforge.net/projects/wikibook-ada

So we will never loose it.

> I looked at Gautier de Montmollin's package, but there doesn't seem to 
> be a way to convert a big-num to an Integer.
> 
> Based on the Free Pascal entry, binding to gmp would be much faster.

But still faster then no entry as no entry is treated as slower then the 
slowest in the overall statistic.

Martin

Re: ada is getting spanked :(

[Jeffrey R. Carter]

Martin Krischik wrote:
> 
> Would you mind adding it to the wikibook ada subversion archive
> 
> http://sourceforge.net/projects/wikibook-ada
> 
> So we will never loose it.

Loose, an adjective, is the opposite of Tight. Lose is a verb.

I don't have subversion or the time right now to get it and figure out 
how to do this. I'll stick the code here so you can copy it and add it.

I compiled with

-O3 -gnatnpws -march=pentium4 -fomit-frame-pointer

-gnatws (suppress all warnings) because there's unreachable code in the 
big-num packages that generates a bunch of warnings.

Here it is:

-- Shootout header should go here.

with Ada.Command_Line;
with Ada.Text_IO;
with Ada.Integer_Text_IO;
with Big_Number;
procedure Pi_Digits is
    type Big_Index is mod 2 ** 16;

    package Bignum is new Big_Number
       (Index_Type => Big_Index, Nb_Item => 3300);

    subtype Big is Bignum.Big_Signed;

    use Bignum.Signed_Number;

    function To_Big is new
    Bignum.Generic_Conversion.Int_Number2Big_Signed
       (Int_Generic => Integer);
    function To_Float (Num : Big) return Long_Long_Float renames
    Bignum.Conversion.Big_Signed2Long_Long_Float;

    procedure Output_Pi (Num_Digits : in Integer) is
       -- State : transformation matrix components
       Q : Big := Bignum.Big_Signed_One;
       R : Big := Bignum.Big_Signed_Zero;
       S : Big := Bignum.Big_Signed_Zero;
       T : Big := Bignum.Big_Signed_One;

       -- Work variables
       U     : Big;
       V     : Big;
       W     : Big;
       X_Big : Big;
       K_Big : Big;
       K2    : Big;
       K4    : Big;

       K     : Natural := 0;
       Digit : Integer;

       Ten : constant Big := To_Big (10);

       function Extract (X : Natural) return Integer is
          -- null;
       begin -- Extract
          X_Big := To_Big (X);
          U := Q * X_Big + R;
          V := S * X_Big + T;
          W := U / V;

          return Integer (To_Float (W) );
       end Extract;

       function Safe (Digit : Integer) return Boolean is
          -- null;
       begin -- Safe
          return Digit = Extract (4);
       end Safe;
       pragma Inline (Safe);

       procedure Produce is
          -- null;
       begin -- Produce
          V := T * To_Big (-10 * Digit);
          R := Ten * R + V;
          Q := Ten * Q;
       end Produce;

       procedure Consume is
          -- null;
       begin -- Consume
          K := K + 1;
          K_Big := To_Big (K);
          K2 := To_Big (2 * K + 1);
          K4 := To_Big (4 * K + 2);
          U := Q * K4;
          R := R * K2 + U;
          V := S * K4;
          T := T * K2 + V;
          S := S * K_Big;
          Q := Q * K_Big;
       end Consume;

       subtype Line_Str is String (1 .. 10);

       I    : Natural := 0;
       C    : Natural := 0;
       Line : Line_Str;
    begin -- Output_Pi
       loop
          exit when I >= Num_Digits;

          loop
             Digit := Extract (3);

             exit when Safe (Digit);

             Consume;
          end loop;

          Produce;
          C := C + 1;
          Line (C) := Character'Val (Character'Pos ('0') + Digit);
          I := I + 1;

          if C >= 10 then
             C := 0;
             Ada.Text_IO.Put (Item => Line & ASCII.HT & ':');
             Ada.Integer_Text_IO.Put (Item => I, Width => 1);
             Ada.Text_IO.New_Line;
          end if;
       end loop;

       if C /= 0 then
          Ada.Text_IO.Put (Item => Line (1 .. C) &
                                   String'(C + 1 .. 10 => ' ') &
                                   ASCII.HT & ':');
          Ada.Integer_Text_IO.Put (Item => I, Width => 1);
          Ada.Text_IO.New_Line;
       end if;
    end Output_Pi;

    N: Positive := 2500;
begin -- Pi_Digits
    if Ada.Command_Line.Argument_Count > 0 then
       N := Integer'Value (Ada.Command_Line.Argument (1) );
    end if;

    Output_Pi (Num_Digits => N);
end Pi_Digits;

-- 
Jeff Carter
"We call your door-opening request a silly thing."
Monty Python & the Holy Grail
17

Shootout / pidigits [Re: ada is getting spanked :(]

[Gautier]

There is now a pi_digits_gmp.adb submitted for the shootout, using GMP.
Took around 1 hour from the Pascal version: P2Ada, rewriting the Import 
pragmata, some clean-up...

   https://alioth.debian.org/tracker/?atid=411646&group_id=30402&func=browse

Gautier

Martin Krischik wrote:
> cl1 wrote:
> 
>> I just got wind of this site. It would seem that ada, while not at
>> the very bottom, is one of the slowest languages in this benchmark
>> competition:
>>
>> http://shootout.alioth.debian.org
> 
> Looks worse then last time I looked. Seems that some language advocates
> really optimised there implementations.
> 
> And there is still one test missing for Ada.
> 
> Martin
______________________________________________________________
Ada programming -- http://www.mysunrise.ch/users/gdm/gsoft.htm

NB: For a direct answer, e-mail address on the Web site!

Re: ada is getting spanked :(

[Larry Kilgallen]

In article <1162052997.664967.135910@e3g2000cwe.googlegroups.com>, "cl1" <charles.w.lambert@gmail.com> writes:
> I just got wind of this site. It would seem that ada, while not at
> the very bottom, is one of the slowest languages in this benchmark
> competition:
> 
> http://shootout.alioth.debian.org

It is not possible to measure the speed of a language, only the speed
of an implementation.

The position you ascribe to Ada actually belongs to GNAT.

Re: ada is getting spanked :(

[Leif Holmgren]

cl1 wrote:
> I just got wind of this site. It would seem that ada, while not at
> the very bottom, is one of the slowest languages in this benchmark
> competition:

So what?

Has anyone ever tried to measure the important programming language 
speed aspects? What I mean is
- Time to implement a complex algo by an average programmer, including 
bug-fixing.
- Time for an average programmer to understand the code written for a 
complex algorithm?
- Same stuff for complex control system.

These are the issues that are important, not that the execution speed is 
of no importance, but for most of my software I don't care if it takes 
1us or 1ms to open a hydraulic valve, but it sure has to happen.

Re: ada is getting spanked :(

[Simon Wright]

Leif Holmgren <nobody@nowhere.com> writes:

> cl1 wrote:
>> I just got wind of this site. It would seem that ada, while not at
>> the very bottom, is one of the slowest languages in this benchmark
>> competition:
>
> So what?
>
> Has anyone ever tried to measure the important programming language
> speed aspects? What I mean is
> - Time to implement a complex algo by an average programmer, including
> bug-fixing.
> - Time for an average programmer to understand the code written for a
> complex algorithm?
> - Same stuff for complex control system.
>
> These are the issues that are important, not that the execution
> speed is of no importance, but for most of my software I don't care
> if it takes 1us or 1ms to open a hydraulic valve, but it sure has to
> happen.

I agree, though my current project is the sort in which it _must not_
take more than 1 ms!

Re: ada is getting spanked :(

[jtg]

Leif Holmgren wrote:

> Has anyone ever tried to measure the important programming language 
> speed aspects? What I mean is
> - Time to implement a complex algo by an average programmer, including 
> bug-fixing.

There is one natural environment for such tests: programming 
competition. For example at topcoder (www.topcoder.com) you can see some 
statistics (average coding time, best coding time, correct programs 
percentage etc.) for several languages used in these competitions. 
Unfortunately, the languages are only C++, C#, java and Visual Basic.

> - Time for an average programmer to understand the code written for a 
> complex algorithm?

At TopCoder there is a "challenge phase", when you can look into other 
contestants' programs, look for bugs, and prepare a test case to prove 
the program is incorrect.
There is also a "system test phase" to catch the bugs not found by 
contestants.

Re: ada is getting spanked :(

[Leif Holmgren]

jtg wrote:
> Leif Holmgren wrote:
> 
>> Has anyone ever tried to measure the important programming language 
>> speed aspects? What I mean is
>> - Time to implement a complex algo by an average programmer, including 
>> bug-fixing.
> 
> 
> There is one natural environment for such tests: programming 
> competition. For example at topcoder (www.topcoder.com) you can see some 
> statistics (average coding time, best coding time, correct programs 
> percentage etc.) for several languages used in these competitions. 
> Unfortunately, the languages are only C++, C#, java and Visual Basic.
> 

Without having looked I believe the problem is that such sites will only 
attract the really talented programmers. And a talented programmer can 
write fortran in any programming language.

What's really interesting is the result from those coders that go home 
after their 8 hours in front of the terminal and dont spend five minutes 
thinking about software until they are back at work the next day.

Re: ada is getting spanked :(

[Harald Korneliussen]

Leif Holmgren wrote:
> jtg wrote:
> > Leif Holmgren wrote:
> >
> >> Has anyone ever tried to measure the important programming language
> >> speed aspects? What I mean is
> >> - Time to implement a complex algo by an average programmer, including
> >> bug-fixing.
> >
> >
> > There is one natural environment for such tests: programming
> > competition. For example at topcoder (www.topcoder.com) you can see some
> > statistics (average coding time, best coding time, correct programs
> > percentage etc.) for several languages used in these competitions.
> > Unfortunately, the languages are only C++, C#, java and Visual Basic.
> >
>
> Without having looked I believe the problem is that such sites will only
> attract the really talented programmers. And a talented programmer can
> write fortran in any programming language.
>
> What's really interesting is the result from those coders that go home
> after their 8 hours in front of the terminal and dont spend five minutes
> thinking about software until they are back at work the next day.

Topcoder is a bit of a swindle, really. They have competitions to
design and implement various component, but even if you should win, the
hourly wage would be lousy. Topcoder programmers are either obsessed
about their reputation, or dirt-poor out of work eastern
europeans/asians. Sometimes both. The site is made for the sponsors,
not the participants, and it shows everywhere, not only in language
alternatives.

Re: ada is getting spanked :(

[jtg]

Harald Korneliussen wrote:

> Topcoder is a bit of a swindle, really.

It is not swindle, because it has clear and stable rules.
It is not for making money by the participants. It is for improving 
skills, assessing skills and providing some kind of skills certificate. 
Usually you have to pay for these services.
Participants are mostly students, but there are also programmers looking 
for a better job.

Re: ada is getting spanked :(

[Harald Korneliussen]

Ok, then, I moderate myself somewhat. Topcoder isn't a swindle, it's a
rip-off. They get people to develop software components for them for
almost nothing, by calling it a competition. I don't think the alleged
gains to reputation and experience are valuable enough to compensate
for this - don't you get that at regular employers too, if you do a
reasonably good job?

No, there is one software competition I have way more respect for, and
that is the ICFP. No prizes, but extremely impressing displays of skill
(as well as truly original challenges). It's may be a little biased for
tasks that functional languages are suited for (despite what they say),
but notetheless.