Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.programming.threads > #2414 > unrolled thread

About scimark2 ...

Back to article view | Back to comp.programming.threads

Contents

  About scimark2 ... aminer <aminer@toto.net> - 2014-06-06 04:22 -0700
    Re: About scimark2 ... aminer <aminer@toto.net> - 2014-06-06 04:50 -0700
    Re: About scimark2 ... aminer <aminer@toto.net> - 2014-06-06 04:57 -0700
    Re: About scimark2 ... Melzzzzz <mel@zzzzz.invalid> - 2014-06-06 23:04 +0200
      Re: About scimark2 ... aminer <aminer@toto.net> - 2014-06-06 05:22 -0700
      Re: About scimark2 ... aminer <aminer@toto.net> - 2014-06-06 05:42 -0700
        Re: About scimark2 ... Melzzzzz <mel@zzzzz.com> - 2014-06-07 00:56 +0200
        Re: About scimark2 ... Kaz Kylheku <kaz@kylheku.com> - 2014-06-06 23:11 +0000
          Re: About scimark2 ... Melzzzzz <mel@zzzzz.invalid> - 2014-06-07 01:58 +0200
            Re: About scimark2 ... aminer <aminer@toto.net> - 2014-06-06 09:08 -0700
              Re: About scimark2 ... Melzzzzz <mel@zzzzz.invalid> - 2014-06-07 04:44 +0200

#2414 — About scimark2 ...

Hello,

Look at the following scimark2 benchmark:

http://delphihaters.blogspot.ca/2011/03/scimark2-and-delphi.html


You will notice that Delphi is almost 2x times slower than
than C on theis scimark2 benchmark...

And you will notice that the poeple on this webpage don't seem
to understand that the big difference of speed between Delphi
and C comes from the fact that the C and C++ compilers uses
"auto-vectorization" , this "auto-vectorization" allows
the C and C++ scimark2 programs to use SIMD instructions such as SSE2,
and i think that this is what is making all the difference,
so to shorter the gap by much between C and Delphi on the
scimark2 , you have to use external libraries that uses
SIMD instructions such as SSE2 etc. and this why i have just
included LinMath inside my multiple libear program , this
LinMarth libraries support SIMD instructions such us SSE2,
and it is also multithreaded, so it is very fast, and my Multiple 
libnear program do uses LinMath so it uses SIMD SSE2 instructions, so 
it's fast.

So you can download LinMath from here(i have included it inside
the zipfile of my multiple linear regression and it now supports
Delphi7 to Delphi XE6 and it supports also FreePascal:


https://sites.google.com/site/aminer68/multiple-linear-regression




Thank you,
Amine Moulay Ramdane.

[toc] | [next] | [standalone]

#2415

Hello,

I have looked at the scimark2  benchmark here:

http://delphihaters.blogspot.ca/2011/03/scimark2-and-delphi.html


And look carefully at the Borland C++ scores , it is almost scoring
the same as Delphi, this information is important, cause we can infer 
from this that the Borland C++ do not support "auto-vectorization", but
Microsoft C++ and Intel C++ and GCC C++ do support "auto-vectorization"
and i thinkthat the actuel C++ and Delphi compilers do not support
"auto-vectorization", so i think that you have to ask Embarcadero and 
FreePascal to support auto-vectorization.



Thank you,
Amine Moulay Ramdane.

[toc] | [prev] | [next] | [standalone]

#2416

Hello again,

I think the actual Embarcadero C++ and Delphi compilers don't
supports this kind of auto-vectorization:

http://locklessinc.com/articles/vectorize/


FreePascal don't support it either.


So i think you have to ask Embarkadero and FreePascal to support
auto-vectorization, cause Microsoft C++ and Intel C++ and GCC do
already support "auto-vectorization".



Thank you,
Amine Moulay Ramdane.

[toc] | [prev] | [next] | [standalone]

#2417

On Fri, 06 Jun 2014 04:22:25 -0700
aminer <aminer@toto.net> wrote:

> Hello,
> 
> Look at the following scimark2 benchmark:
> 
> http://delphihaters.blogspot.ca/2011/03/scimark2-and-delphi.html
> 
> 
> You will notice that Delphi is almost 2x times slower than
> than C on theis scimark2 benchmark...
> 
> And you will notice that the poeple on this webpage don't seem
> to understand that the big difference of speed between Delphi
> and C comes from the fact that the C and C++ compilers uses
> "auto-vectorization" , this "auto-vectorization" allows
> the C and C++ scimark2 programs to use SIMD instructions such as SSE2,

Have you tried on some other architecture besides x86?



-- 
Click OK to continue...

[toc] | [prev] | [next] | [standalone]

#2418

On 6/6/2014 2:04 PM, Melzzzzz wrote:
> On Fri, 06 Jun 2014 04:22:25 -0700
> aminer <aminer@toto.net> wrote:
>
>> Hello,
>>
>> Look at the following scimark2 benchmark:
>>
>> http://delphihaters.blogspot.ca/2011/03/scimark2-and-delphi.html
>>
>>
>> You will notice that Delphi is almost 2x times slower than
>> than C on theis scimark2 benchmark...
>>
>> And you will notice that the poeple on this webpage don't seem
>> to understand that the big difference of speed between Delphi
>> and C comes from the fact that the C and C++ compilers uses
>> "auto-vectorization" , this "auto-vectorization" allows
>> the C and C++ scimark2 programs to use SIMD instructions such as SSE2,
>
> Have you tried on some other architecture besides x86?



I didn't tried, but i think that from the above webpage we can infer
that the actuel FreePascal  and Embarcado C++ and Delphi  don't support 
auto-vectorization.


You can try to do your research on that and i think you will come to the 
same conclusion...



Thank you,
Amine Moulay Ramdane.

[toc] | [prev] | [next] | [standalone]

#2419

On 6/6/2014 2:04 PM, Melzzzzz wrote:
> On Fri, 06 Jun 2014 04:22:25 -0700
> aminer <aminer@toto.net> wrote:
>
>> Hello,
>>
>> Look at the following scimark2 benchmark:
>>
>> http://delphihaters.blogspot.ca/2011/03/scimark2-and-delphi.html
>>
>>
>> You will notice that Delphi is almost 2x times slower than
>> than C on theis scimark2 benchmark...
>>
>> And you will notice that the poeple on this webpage don't seem
>> to understand that the big difference of speed between Delphi
>> and C comes from the fact that the C and C++ compilers uses
>> "auto-vectorization" , this "auto-vectorization" allows
>> the C and C++ scimark2 programs to use SIMD instructions such as SSE2,
>
> Have you tried on some other architecture besides x86?
>
>
>



I have benchmarked Delphi and FreePascal against GCC C++
and Microsoft C+ on scimark2, and i have found that without the GCC
-Ox optimization options the auto-vectorization will not be enabled so 
Delphi and FreePascal and GCC C++ scored the same speed on scimark2, but 
when you give -O1 or -O2,  GCC C++ uses auto-vectorization and it scores 
almost 2X times the speed of Freepascal or Delphi on scimark2 on x86.


Thank you,
Amine Moulay Ramdane.

[toc] | [prev] | [next] | [standalone]

#2420

On Fri, 06 Jun 2014 05:42:19 -0700
aminer <aminer@toto.net> wrote:

> On 6/6/2014 2:04 PM, Melzzzzz wrote:
> > On Fri, 06 Jun 2014 04:22:25 -0700
> > aminer <aminer@toto.net> wrote:
> >
> >> Hello,
> >>
> >> Look at the following scimark2 benchmark:
> >>
> >> http://delphihaters.blogspot.ca/2011/03/scimark2-and-delphi.html
> >>
> >>
> >> You will notice that Delphi is almost 2x times slower than
> >> than C on theis scimark2 benchmark...
> >>
> >> And you will notice that the poeple on this webpage don't seem
> >> to understand that the big difference of speed between Delphi
> >> and C comes from the fact that the C and C++ compilers uses
> >> "auto-vectorization" , this "auto-vectorization" allows
> >> the C and C++ scimark2 programs to use SIMD instructions such as
> >> SSE2,
> >
> > Have you tried on some other architecture besides x86?
> >
> >
> >
> 
> 
> 
> I have benchmarked Delphi and FreePascal against GCC C++
> and Microsoft C+ on scimark2, and i have found that without the GCC
> -Ox optimization options the auto-vectorization will not be enabled
> so Delphi and FreePascal and GCC C++ scored the same speed on
> scimark2, but when you give -O1 or -O2,  GCC C++ uses
> auto-vectorization and it scores almost 2X times the speed of
> Freepascal or Delphi on scimark2 on x86.
> 
> 
Haahahhahahh. No, you try -m32 (x64 requires sse2) and -O2 with -mno-sse
-mno-sse2.
Without -O is actually without optimisations at all ;)
I think that this particular benchmark does not use auto vectorization
much ;)


> Thank you,
> Amine Moulay Ramdane.
> 
> 
> 
> 
> 
> 



-- 
Click OK to continue...

[toc] | [prev] | [next] | [standalone]

#2421

On 2014-06-06, aminer <aminer@toto.net> wrote:
> I have benchmarked Delphi and FreePascal against GCC C++
> and Microsoft C+ on scimark2, and i have found that without the GCC
> -Ox optimization options the auto-vectorization will not be enabled so 
> Delphi and FreePascal and GCC C++ scored the same speed on scimark2, but 
> when you give -O1 or -O2,  GCC C++ uses auto-vectorization and it scores 
> almost 2X times the speed of Freepascal or Delphi on scimark2 on x86.

Without -O the GNU compilers do not optimize at all. The generated
code is quite bad.

If removing -O options brings the performance down to Delphi, and a mere -O1
can beat Delphi, that suggests that Delphi doesn't optimize, at least with
regard to the program being tested.

Or something else is wrong, which makes the programs not comparable.
(That's always a problem when comparing programs written in different
languages. They really have to have a very similar data and code organization
for that to be valid as a compiler comparison.)

Anyway, nobody in their right mind who is past the end of their puberty, or
past the year 1992, whichever came first, uses Delphi.

The last programmer I met who swore by Delphi was a 19 year old co-op student
back in 1999, doing QA work.  He said brilliant things like "Unicode is
gay with all these space-wasting wide characters; everyone should just learn
English and use ASCII".

Was that you?

[toc] | [prev] | [next] | [standalone]

#2422

On Fri, 6 Jun 2014 23:11:59 +0000 (UTC)
Kaz Kylheku <kaz@kylheku.com> wrote:

> On 2014-06-06, aminer <aminer@toto.net> wrote:
> > I have benchmarked Delphi and FreePascal against GCC C++
> > and Microsoft C+ on scimark2, and i have found that without the GCC
> > -Ox optimization options the auto-vectorization will not be enabled
> > so Delphi and FreePascal and GCC C++ scored the same speed on
> > scimark2, but when you give -O1 or -O2,  GCC C++ uses
> > auto-vectorization and it scores almost 2X times the speed of
> > Freepascal or Delphi on scimark2 on x86.
> 
> Without -O the GNU compilers do not optimize at all. The generated
> code is quite bad.

I think that gcc performs auto vectorization only at -O3 .

> 
> If removing -O options brings the performance down to Delphi, and a
> mere -O1 can beat Delphi, that suggests that Delphi doesn't optimize,
> at least with regard to the program being tested.
> 
> Or something else is wrong, which makes the programs not comparable.

Most probably pascal benchmark uses more complex code. Eg OO features...

-- 
Click OK to continue...

[toc] | [prev] | [next] | [standalone]

#2423

Hello,

Read this:

http://beeography.wordpress.com/2011/05/04/which-programming-language-is-best/



Thank you,
Amine Moulay Ramdane.

[toc] | [prev] | [next] | [standalone]

#2425

On Fri, 06 Jun 2014 09:08:05 -0700
aminer <aminer@toto.net> wrote:

> 
> Hello,
> 
> Read this:
> 
> http://beeography.wordpress.com/2011/05/04/which-programming-language-is-best/
> 
> 
> 
> Thank you,
> Amine Moulay Ramdane.
"
The most disappoinment come from the fact that C# and Java, both
interpreted languages, beat FPC as native programming language
"

C# and Java are *not* interpreted languages. They are run time
compiled and they have *very good* compilers. Though seems
that Java does not optimize integer division by constant...
Don;t know about C#.

-- 
Click OK to continue...

[toc] | [prev] | [standalone]

Back to top | Article view | comp.programming.threads

csiph-web