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

Groups > sci.physics > #532127 > unrolled thread

Floating Point Test Suite

Back to article view | Back to sci.physics

Contents

  Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-11 17:58 +0000
    Re: Floating Point Test Suite Poutnik <poutnik4nntp@gmail.com> - 2015-11-11 23:44 +0100
      Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-11 22:54 +0000
        Re: Floating Point Test Suite Poutnik <poutnik4nntp@gmail.com> - 2015-11-12 00:23 +0100
    Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-11 23:16 +0000
      Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-11 23:48 +0000
        Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-12 00:32 +0000
        Re: Floating Point Test Suite Poutnik <poutnik4nntp@gmail.com> - 2015-11-12 07:52 +0100
          Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-12 21:13 +0000
            Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-12 21:46 +0000
              Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-12 22:27 +0000
                Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-12 23:06 +0000
                  Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-12 23:40 +0000
                    Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-13 00:12 +0000
                      Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-13 00:24 +0000
                        Re: Floating Point Test Suite dvus <dven1@adelphia.net> - 2015-11-13 06:10 -0500
                          Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-13 18:56 +0000
                          Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-13 20:03 +0000
                            Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-13 20:40 +0000
                              Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-13 20:51 +0000
                                Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-13 21:10 +0000
                                Re: Floating Point Test Suite benj <none@gmail.com> - 2015-11-13 21:29 -0500
                                  Re: Floating Point Test Suite Mahipal <mahipal7638@gmail.com> - 2015-11-13 19:35 -0800
                                    Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-14 03:56 +0000
                                      Re: Floating Point Test Suite Mahipal <mahipal7638@gmail.com> - 2015-11-13 20:34 -0800
                                        Re: Floating Point Test Suite "Bob Sutter" <invalid@exampl.com> - 2015-11-14 13:40 -0800
                                      Re: Floating Point Test Suite benj <none@gmail.com> - 2015-11-14 02:15 -0500
                              Re: Floating Point Test Suite noTthaTguY <abu.kuanysh05@gmail.com> - 2015-11-13 13:15 -0800
                              Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-14 03:50 +0000
                                Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-14 04:47 +0000
                                  Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-14 05:12 +0000
                                    Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-14 07:13 +0000
                                      Re: Floating Point Test Suite dvus <dven1@adelphia.net> - 2015-11-14 19:20 -0500
                                        Re: Floating Point Test Suite Fabian Russell <root@localhost.localdomain> - 2015-11-15 00:37 +0000
                                          Re: Floating Point Test Suite jimp@specsol.spam.sux.com - 2015-11-15 00:56 +0000
                                          Re: Floating Point Test Suite dvus <dven1@adelphia.net> - 2015-11-15 10:28 -0500
                                            Re: Floating Point Test Suite noTthaTguY <abu.kuanysh05@gmail.com> - 2015-11-15 10:47 -0800
                                              Re: Floating Point Test Suite dvus <dven1@adelphia.net> - 2015-11-15 16:24 -0500
                                                Stevin "made the switching device noTthaTguY <abu.kuanysh05@gmail.com> - 2015-11-16 13:39 -0800
                            Re: Floating Point Test Suite "hanson" <hanson@quick.net> - 2015-11-13 18:05 -0800

Page 1 of 2  [1] 2  Next page →

#532127 — Floating Point Test Suite

Greetings physics enthusiasts:

C'mon, you lazy laggards.  Let's get some discussion going.  Get off
your moribund asses and put forth some comments.

Modern science depends heavily, if not entirely, on the computer, and
the most important element of scientific computation is undeniably
the FLOATING POINT NUMBER.

The floating point standard for all computing machines, from supercomputers
to desktop workstations, is codified in the standards IEEE754/854.

How well does your machine perform in executing floating point calculations?

To help you answer this question I have uploaded a tarball of the
rigorous FP test suite known as UCBTEST, which was first developed
at the University of California at Berkeley.

Original source: http://www.netlib.org/fp/

The original source code is quite ancient, ca. 1995, and may not function
on modern machines.  Consequently I have provided a patched version to
bring the code up to date.  The patched source should run on any x64 Linux
machine (it certainly runs on mine).

Patched source: 
http://s000.tinyupload.com/index.php?file_id=05747896726892930617

Note that the original source is somewhat disorganized and not structured
according to the GNU/Linux style to which you may be accustomed.  But
the fault is not mine.  I did not write the code; I only patched it.

But execution should be foolproof if you follow these instructions.

1) Unpack the tarball anywhere.

2) You will see a directory called "ucbtest-src"

3) Change into that directory and execute the script:

	ucbREADME/linux.shlinux.sh > linux.log 2>&1 

4) The program requires several minutes to complete.  Results are
contained in the "ucbtest-results" directory as several *output
files as well as in the file "linux.log."

Documentation for the UCBTEST is included in the "ucbREADME" directory.

The tests require the Linux "time" utility which I have included
in the "ucbREADME" dir.  If this version of "time" fails for some
reason you will have to supply your own.  It is a common utility.

If you experience problems, check the compiler opt flags and the
other variables that are defined in ucbREADME/linux.sh.  Exactly
what to look for is explained in the README file.

Also feel free to post any questions, issues, results, or failures
to the list.

My machine performs excellently.  How does yours?

MS Windows users are out of luck, but MS Windows is junk anyway,
and the average MS Windows user could not distinguish a floating
point number from a hole in the ground.

[toc] | [next] | [standalone]

#532207

Dne 11/11/2015 v 18:58 Fabian Russell napsal(a):
> Greetings physics enthusiasts:
> 
> C'mon, you lazy laggards.  Let's get some discussion going.  Get off
> your moribund asses and put forth some comments.
> 
> Modern science depends heavily, if not entirely, on the computer, and
> the most important element of scientific computation is undeniably
> the FLOATING POINT NUMBER.
> 

Rather,
the most important element of scientific computation
is estimation of input data errors,
knowledge of accuracy of used physical constants
and knowledge of rules of error propagation.

As one of the classics said,
that nothing manifests ignorance in physics better,
then untamed precision of scientific computations.

-- 
Poutnik ( the Czech word for a wanderer )

Knowledge makes great men humble, but small men arrogant.

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

#532209

On Wed, 11 Nov 2015 23:44:17 +0100, Poutnik wrote:

> 
> Rather,
> the most important element of scientific computation
> is estimation of input data errors,
> knowledge of accuracy of used physical constants
> and knowledge of rules of error propagation.
> 

Yes, but each of those factors is underlain by the floating
point format.

Although error propagation is evaluated with pure mathematics,
the implementation of the theory rests on FP.

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

#532218

Dne 11/11/2015 v 23:54 Fabian Russell napsal(a):
> On Wed, 11 Nov 2015 23:44:17 +0100, Poutnik wrote:
> 
>>
>> Rather,
>> the most important element of scientific computation
>> is estimation of input data errors,
>> knowledge of accuracy of used physical constants
>> and knowledge of rules of error propagation.
>>
> 
> Yes, but each of those factors is underlain by the floating
> point format.

rather on real and complex number abstraction.
> 
> Although error propagation is evaluated with pure mathematics,
> the implementation of the theory rests on FP.
> 
But not on FP implementation.

-- 
Poutnik ( the Czech word for a wanderer )

Knowledge makes great men humble, but small men arrogant.

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

#532221

Fabian Russell <root@localhost.localdomain> wrote:
> Greetings physics enthusiasts:
> 
> C'mon, you lazy laggards.  Let's get some discussion going.  Get off
> your moribund asses and put forth some comments.
> 
> Modern science depends heavily, if not entirely, on the computer, and
> the most important element of scientific computation is undeniably
> the FLOATING POINT NUMBER.
> 
> The floating point standard for all computing machines, from supercomputers
> to desktop workstations, is codified in the standards IEEE754/854.
> 
> How well does your machine perform in executing floating point calculations?

That depends on the hardware and the compiler; the OS has very little to
do with it.

You don't seem to have a clue how such things work.


-- 
Jim Pennino

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

#532227

On Wed, 11 Nov 2015 23:16:38 +0000, jimp wrote:

> 
> That depends on the hardware and the compiler; the OS has very little to
> do with it.
> 
> You don't seem to have a clue how such things work.
>

I never mentioned OS.

And I know how it works far, far, FAR better than you will ever.

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

#532242

Fabian Russell <root@localhost.localdomain> wrote:
> On Wed, 11 Nov 2015 23:16:38 +0000, jimp wrote:
> 
>> 
>> That depends on the hardware and the compiler; the OS has very little to
>> do with it.
>> 
>> You don't seem to have a clue how such things work.
>>
> 
> I never mentioned OS.
> 
> And I know how it works far, far, FAR better than you will ever.

From your postings it appears you don't know how much of anything works.

In any case, the original statement stands, so what is your point you
think you are trying to make?


-- 
Jim Pennino

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

#532289

Dne 12/11/2015 v 00:48 Fabian Russell napsal(a):
> On Wed, 11 Nov 2015 23:16:38 +0000, jimp wrote:
> 
>>
>> That depends on the hardware and the compiler; the OS has very little to
>> do with it.
>>
>> You don't seem to have a clue how such things work.
>>
> 
> I never mentioned OS.
> 
> And I know how it works far, far, FAR better than you will ever.

What a humble statement...

Note that there are specialized compilers / interpreters,
that are focused on software implementation of deliberate precision,
that is not supported directly by CPU/FPU.

-- 
Poutnik ( the Czech word for a wanderer )

Knowledge makes great men humble, but small men arrogant.

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

#532462

On Thu, 12 Nov 2015 07:52:33 +0100, Poutnik wrote:

> 
> Note that there are specialized compilers / interpreters,
> that are focused on software implementation of deliberate precision,
> that is not supported directly by CPU/FPU.
>

So what?  Any arbitrary precision representation has to be subject
to the same behavior as currently specified in IEEE754.

William Kahan stated that IEEE754 was likely one of the greatest
human accomplishments of the 20th century.

You need to understand why.

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

#532484

Fabian Russell <root@localhost.localdomain> wrote:
> On Thu, 12 Nov 2015 07:52:33 +0100, Poutnik wrote:
> 
>> 
>> Note that there are specialized compilers / interpreters,
>> that are focused on software implementation of deliberate precision,
>> that is not supported directly by CPU/FPU.
>>
> 
> So what?  Any arbitrary precision representation has to be subject
> to the same behavior as currently specified in IEEE754.
> 
> William Kahan stated that IEEE754 was likely one of the greatest
> human accomplishments of the 20th century.

Nope, the invention of antibiotics, which has saved the lives of millions
upon millions of people.

Almost nobody knows what IEEE754 is or is effected by it.

 

-- 
Jim Pennino

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

#532505

On Thu, 12 Nov 2015 21:46:46 +0000, jimp wrote:

> 
> Almost nobody knows what IEEE754 is or is effected by it.
>

Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha!

You deserve the Supreme Idiot Award for that statement!

IEEE754 is so deeply infused in our society that nothing could
really exist without it.  William Kahan heralded it as one of
the greatest achievements of the 20th century.

Make sure to check out your window regularly.  They should be
coming to lock you up for such a preposterous claim.

Whooooooooh haaaa!  What a whopper!

Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha!

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

#532527

Fabian Russell <root@localhost.localdomain> wrote:
> On Thu, 12 Nov 2015 21:46:46 +0000, jimp wrote:
> 
>> 
>> Almost nobody knows what IEEE754 is or is effected by it.
>>
> 
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha!
> 
> You deserve the Supreme Idiot Award for that statement!
> 
> IEEE754 is so deeply infused in our society that nothing could
> really exist without it.  William Kahan heralded it as one of
> the greatest achievements of the 20th century.
> 
> Make sure to check out your window regularly.  They should be
> coming to lock you up for such a preposterous claim.
> 
> Whooooooooh haaaa!  What a whopper!
> 
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha,
> Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha, Ha, ha, ha, ha, ha!

Try asking someone on the street how IEEE754 has effected them
and see what happens.

 

-- 
Jim Pennino

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

#532547

On Thu, 12 Nov 2015 23:06:52 +0000, jimp wrote:

> 
> Try asking someone on the street how IEEE754 has effected them
> and see what happens.
>

The man on the street always needs a "mouthpiece" to handle
for him the complexity of things beyond his grasp, e.g. a
lawyer.

Thus, if we asked the average man on the street, through his
mouthpiece, how he has been effected, his mouthpiece would
exclaim:

"In all ways possible, baby!"

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

#532557

Fabian Russell <root@localhost.localdomain> wrote:
> On Thu, 12 Nov 2015 23:06:52 +0000, jimp wrote:
> 
>> 
>> Try asking someone on the street how IEEE754 has effected them
>> and see what happens.
>>
> 
> The man on the street always needs a "mouthpiece" to handle
> for him the complexity of things beyond his grasp, e.g. a
> lawyer.

Non sequitur, red herring.


-- 
Jim Pennino

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

#532559

On Fri, 13 Nov 2015 00:12:20 +0000, jimp wrote:

> 
> Non sequitur,
>

Non sequitor?  Even a six-year-old could see the connection
in that one -- and while half asleep -- and with both arms
tied.

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

#532692

On 11/12/2015 7:24 PM, Fabian Russell wrote:
> On Fri, 13 Nov 2015 00:12:20 +0000, jimp wrote:
>
>>
>> Non sequitur,
>>
>
> Non sequitor?  Even a six-year-old could see the connection
> in that one -- and while half asleep -- and with both arms
> tied.

I'd hazard to say that if asked most people would place the toilet way 
ahead of IEEE754 in a list of useful human accomplishments. But then 
again, what do people know?

-- 
dvus

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

#532788

dvus <dven1@adelphia.net> wrote:
> On 11/12/2015 7:24 PM, Fabian Russell wrote:
>> On Fri, 13 Nov 2015 00:12:20 +0000, jimp wrote:
>>
>>>
>>> Non sequitur,
>>>
>>
>> Non sequitor?  Even a six-year-old could see the connection
>> in that one -- and while half asleep -- and with both arms
>> tied.
> 
> I'd hazard to say that if asked most people would place the toilet way 
> ahead of IEEE754 in a list of useful human accomplishments. But then 
> again, what do people know?

Most people don't have the foggiest clue what floating point means let
alone IEEE754.

 

-- 
Jim Pennino

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

#532805

On Fri, 13 Nov 2015 06:10:03 -0500, dvus wrote:

> 
> I'd hazard to say that if asked most people would place the toilet way 
> ahead of IEEE754 in a list of useful human accomplishments.
> 

Yeah, but more and more toilets are now incorporating microprocessors
(with IEEE754) to control such things as flush intensity, bidet sprays,
and other features that augment the traditional rather barren toilet
"experience."

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

#532812

Fabian Russell <root@localhost.localdomain> wrote:
> On Fri, 13 Nov 2015 06:10:03 -0500, dvus wrote:
> 
>> 
>> I'd hazard to say that if asked most people would place the toilet way 
>> ahead of IEEE754 in a list of useful human accomplishments.
>> 
> 
> Yeah, but more and more toilets are now incorporating microprocessors
> (with IEEE754) to control such things as flush intensity, bidet sprays,
> and other features that augment the traditional rather barren toilet
> "experience."

Smart toilets are few and far between and a toilet has no need to use
floating point arithmetic; 8 bit integer arithmetic is more than
sufficient for such an application.
 
That give a one part in 256 resolution for flush intensity, bidet sprays,
etc.


-- 
Jim Pennino

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

#532815

On Fri, 13 Nov 2015 20:40:40 +0000, jimp wrote:

> 
> Smart toilets are few and far between and a toilet has no need to use
> floating point arithmetic; 8 bit integer arithmetic is more than
> sufficient for such an application.
>  

The person that writes the code for the toilet controller will
likely use a very high level language that is not strongly typed
and therefore all numerical quantities that are specified will
be translated only into floating point.

You lose again.

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

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | sci.physics

csiph-web