🚀 go-pugleaf

In article <1991May12.160205.9784@bronze.ucs.indiana.edu>
ury@mossad.huji.ac.il (ury segal) writes:
>How you define a supercomputer ? ( >100 MIPS ???)

You have to ask the United States Department of Commerce.

--eugene miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov
  Resident Cynic, Rock of Ages Home for Retired Hackers
  8^)
  {uunet,mailrus,other gateways}!ames!eugene

In article <1991May13.051215.8101@nas.nasa.gov> eugene@amelia.nas.nasa.gov (Eugene N. Miya) writes:
>In article <1991May12.160205.9784@bronze.ucs.indiana.edu>
>ury@mossad.huji.ac.il (ury segal) writes:
>>How you define a supercomputer ? ( >100 MIPS ???)
>
>You have to ask the United States Department of Commerce.
>

- I think this list should just about cover it %-)!

A supercomputer has the following characteristics:

- Lots of memory, but not quite enough to run your problem.

- It should be difficult to program. Assembly is the prefered
language of supercomputers. Fortran or C compilers are acceptable,
any machine with a lisp compiler can not be a supercomputer. 

- The ablity to run the LINPACK benchmark at near the maximum speed
after months of massaging assembly routines. The inablity to run 
portable code at greater than 1/10th the machine speed. 

- Non portable language extensions are a must.!!! Any code that is
optimized for one machine should require several months to get running
on any other supercomputer. 

- Frequent crashes just as your job is finishing up it's output. 

- Of course , we can't forget I/O bottlenecks. 

- Hundreds / thousands of users competing for cpu time. 

- Cryptic operating systems. Or if they use a commonly available one
it should be at least 2-3 revisions behind what you are used to. 

 
REMEMBER, this is a joke !!!!

- Booker C. Bense                    
prefered: benseb@grumpy.sdsc.edu	"I think it's GOOD that everyone 
NeXT Mail: benseb@next.sdsc.edu 	   becomes food " - Hobbes

In article <377@nic.cerf.net> benseb@grumpy.sdsc.edu (Booker Bense) writes
a very funny article:
>>In article <1991May12.160205.9784@bronze.ucs.indiana.edu>
>>ury@mossad.huji.ac.il (ury segal) writes:
>>>How you define a supercomputer ? ( >100 MIPS ???)

He probably asked this in all seriousness.
	"The fastest machine(s) at any given point in time." --S.F.

There is also the paper "'Supercomputer' is just a marketing term."

%A Sidney Fernbach, ed.
%T Supercomputers, Class VI Systems, Hardware and Software
%I North-Holland
%D 1986
%K book, text, cray, cdc cyber, data flow, NEC SX-2, Fujitsu VP-200,
Hitachi 810/20, vector processing,
%X A collection of papers surveying existing computer architectures
rather than newer proposed supercomputer architectures.
%X A book from one of the men who set up the "Class system" of the DOE.

I answered (only half serious):
>>You have to ask the United States Department of Commerce.

I posted this because of the author's "il" address.  It harkens back to
a chance meeting in which I gave DEC's Washington DC lawyer a ride
in Yosemite.  Seems she wanted to sell a VAX 9000 to IL.  DOC said
the 9000 was a supercomputer.  DOD apparently didn't think it was.
"Was I willing to say the VAX 9000 was not a super computer?" "So it
(the 9000) the fastest machine around?"

DOC won.  So I guess IL is getting a SUPRENUM.  And this story is what I
get for giving two hitchhikers a ride.  So I guess the 9000 is a supercomputer
in the eyes of the DOC, and Sid Fernbach's definition was wrong.
So the answer still stands: that you have to ask the US DOC.

Added featuritis:

>- I think this list should just about cover it %-)!
>
>A supercomputer has the following characteristics:
>
>- Lots of memory, but not quite enough to run your problem.

- Virtual memory?  What's that?
  [One could make jokes about parity....]

>- It should be difficult to program. Assembly is the prefered
>language of supercomputers. Fortran or C compilers are acceptable,
>any machine with a lisp compiler can not be a supercomputer. 

This gets rid of Wayne Anderson's Portable Standard Lisp efforts.
Does this disqualify the TMC CM line?

>- The ablity to run the LINPACK benchmark at near the maximum speed
>after months of massaging assembly routines. The inablity to run 
>portable code at greater than 1/10th the machine speed. 
>
>- Non portable language extensions are a must.!!! Any code that is
>optimized for one machine should require several months to get running
>on any other supercomputer. 

Exotic new languages are acceptable. (excepting LISP)

>- Frequent crashes just as your job is finishing up it's output. 

Exotic new hardware technology: cooling, packaging, etc.

>- Of course , we can't forget I/O bottlenecks. 

	The Batcher definition.  ("A Supercomputer is any machine
	which turns a CPU bound job into an I/O bound job.")

>- Hundreds / thousands of users competing for cpu time. 

>- Cryptic operating systems. Or if they use a commonly available one
>it should be at least 2-3 revisions behind what you are used to. 

>REMEMBER, this is a joke !!!!

It was funny!  I hope you keep the amended list and post it regularly.

--eugene miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov
  Resident Cynic, Rock of Ages Home for Retired Hackers
  {uunet,mailrus,other gateways}!ames!eugene

Actually, a REAL supercomputer has to cost about 20megaBuck$ to 
be State-of-the-Art.

In article <377@nic.cerf.net> benseb@grumpy.sdsc.edu (Booker Bense) writes:
In article <1991May13.051215.8101@nas.nasa.gov> eugene@amelia.nas.nasa.gov (Eugene N. Miya) writes:
>In article <1991May12.160205.9784@bronze.ucs.indiana.edu>
>ury@mossad.huji.ac.il (ury segal) writes:
>>How you define a supercomputer ? ( >100 MIPS ???)
>
>You have to ask the United States Department of Commerce.
>
My favorite definition (courtesy of John Levesque, Pacific-Sierra (I think)):

supercomputer: a device for converting a cpu-bound problem into an I/O-bound
               problem.

-Steve

In article <377@nic.cerf.net> benseb@grumpy.sdsc.edu (Booker Bense) writes:
>>>How you define a supercomputer ? ( >100 MIPS ???)
>>You have to ask the United States Department of Commerce.
>
>- I think this list should just about cover it %-)!

My favorite two definitions:

	The fastest machine $5 million can buy.
	Any machine faster than IBM's top mainframe.

Or how about:

	A machine that can complete an infinite loop in < 10 seconds.

:^) :^) :^) :^)


-- 
Jim Armstrong			  "The nonpayment and subsequent abuse of
armstron@cs.arizona.edu		  socially powerless athletes is simply a
uunet!arizona!armstron            form of modern-day slavery" --Rick Telander

A supercomputer is a machine for which the decision to buy is made for
political reasons, not technical.
--

Donald C. Wells             Associate Scientist        dwells@nrao.edu
National Radio Astronomy Observatory                   +1-804-296-0277
Edgemont Road                                     Fax= +1-804-296-0278
Charlottesville, Virginia 22903-2475 USA            78:31.1W, 38:02.2N 

What do you mean "This is a joke" !!????....looks fairly accurate to me!

In article <DWELLS.91May14220342@fits.cx.nrao.edu> dwells@fits.cx.nrao.edu (Don Wells) writes:
>A supercomputer is a machine for which the decision to buy is made for
>political reasons, not technical.

If that's true, then most of the IBM mainframes out there are "supercomputers".
To quote the industry cliche: "Nobody ever got fired for buying from IBM...".

- Steve (just another ex-Crayon...)

A supercomputer is any machine which is faster than IBM's current
top-of-line mainframe. 

So by definition, a 3090 is not a supercomputer. :-)

In article <4098@uc.msc.umn.edu> alan@msc.edu (Alan Klietz) writes:
>A supercomputer is any machine which is faster than IBM's current
>top-of-line mainframe. 
>
>So by definition, a 3090 is not a supercomputer. :-)

So Alan, the MSC document I just got which mentioned the
3090-600, this latter machine isn't a supercomputer?  Okay. ;^)

This reminds me when when we were cross-posting the Kahaner reports
in this news group.  First David had me post the list of Japanese Crays
and ETAs, then Patrick asked whether or not Convex-en were supers
(David said no).  Then I remembered a Connection Machine was delivered
(Yes).  Then came hypercubes (No) and 3090s.  "Well, IBM didn't
provide him with a list of 3090 centers in Japan, so No."

P.S. David enjoyed giving his presentations last month in Oregon
(Salishan and DMCC6), and he enjoyed meeting a host of readers.

--eugene miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov
  Resident Cynic, Rock of Ages Home for Retired Hackers
  {uunet,mailrus,other gateways}!ames!eugene

In article <1991May14.230507.8959@mnemosyne.cs.du.edu> sgombosi@isis.UUCP (Stephen O. Gombosi) writes:
>In article <377@nic.cerf.net> benseb@grumpy.sdsc.edu (Booker Bense) writes:
>In article <1991May13.051215.8101@nas.nasa.gov> eugene@amelia.nas.nasa.gov (Eugene N. Miya) writes:
>>In article <1991May12.160205.9784@bronze.ucs.indiana.edu>
>>ury@mossad.huji.ac.il (ury segal) writes:
>>>How you define a supercomputer ? ( >100 MIPS ???)
>>
>>You have to ask the United States Department of Commerce.
>>
>My favorite definition (courtesy of John Levesque, Pacific-Sierra (I think)):
>
>supercomputer: a device for converting a cpu-bound problem into an I/O-bound
>               problem.
>
>-Steve

I forgot where I got this definition (possibly from Neil Lincoln, ex-ETA,
now at SSESCO):

Supercomputer: Any computer which is just two orders of magnitude less
	       powerful than what users really want. ;-)

Dave

A definition I remember, and which seems to me to make sense: a supercomputer
is a system architected and designed for PERFORMANCE, to the *detriment* of
PRICE/performance.  The curve you can draw with price on X axis and
performance on y axis will generally have an inflection point: below that,
you are below optimal price/performance because you are not exploiting
economies of scale, and above that, you are below optimal price/performance
because of diminishing-returns effects.  
I always like this idea, incomplete as it is, because to me it says something
about the ('politically' motivated) "wisdom" of buying a supercomputer, and
then overloading it with timesharing jobs...
-- 
Alex Martelli - CAD.LAB s.p.a., v. Stalingrado 53, Bologna, Italia
Email: (work:) martelli@cadlab.sublink.org, (home:) alex@am.sublink.org
Phone: (work:) ++39 (51) 371099, (home:) ++39 (51) 250434; 
Fax: ++39 (51) 366964 (work only), Fidonet: 332/407.314 (home only).