Roadrunner Supercomputer using 12,960 CELL Processors Hits 1 PetaFlop(1000 TeraFlops) of double-precision FP Performance

In comp.sys.ibm.pc.hardware.chips Cydrome Leader wrote in part:
I find it unlikely they still don't know how nuclear weapons
work, especially considering they're mature technology and
they've been around for decades.

A two-stage thermonuclear warhead is a surprisingly complex
device -- look up Teller-Ulam. To work properly, the design has to
transfer enough light energy to ignite and burn the fusion secondary
before the fission primary shock waves disassemble the device.
There's _lots_ to simulate here in at least 2D over many timeslices.

Yes, we know how to make them go bang. Just follow the recipe.
But we don't always know the critical parts of that recipe, and
what parts we could change.

In general, the whole field of Finite-Element computation is
still short of cycles and can swallow everything available.
Multi-CPU clusters are still being built. Imagine being able
to simulate vehicle collisions -- designers would be able
to determine where metal could be added or other changes to
improve occupant survival.

-- Robert

Robert Redelmeier wrote:
Multi-CPU clusters are still being built. Imagine being able
to simulate vehicle collisions -- designers would be able
to determine where metal could be added or other changes to
improve occupant survival.

No need to imagine this:

We have operated a Linux cluster for nearly 5 years now that Hydro (the
Aluminium company) uses to simulate the energy transfer/structure
deformation that happens when a (sports) car using one of their
aluminium frames with integrated "crash box" structures get into an
accident.

Yes, it is a _lot_ cheaper than running real crash tests, even though
you still have to verify the theoretical results before the car will be
certified for road use.

Terje

--
-
"almost all programming can be viewed as an exercise in caching"

On Jun 13, 11:10 am, Terje Mathisen
wrote:


Yes, it is a _lot_ cheaper than running real crash tests, even though
you still have to verify the theoretical results before the car will be
certified for road use.


Cost isn't the real advantage. You can't instrument an experiment the
way that you can a simulation.

The disadvantage, of course, is that you have mountains of information
and sometimes no clue as to what to do with it. The science, the
insight, the talent, and everything else worthwhile are in finding
those clues and not in the petaflops or exabytes.

Robert.

On Jun 13, 3:33 pm, (Eugene Miya) wrote:
In article ,
Robert Myers wrote:

The disadvantage, of course, is that you have mountains of information

data, not info

I hesitated at the word data, which is the word that computer
scientists would use. Outside the world of computers, the distinction
between data and predictions is not to be trifled with, especially if
you're in the same room with someone who actually measures things.
Theoreticians have a way of wanting to keep their work separate from
what comes out of computers, as if theory could do anything without
computation, automatic or otherwise.

Robert.

In article ,
Robert Myers wrote:
On Jun 13, 11:10 am, Terje Mathisen
wrote:
Yes, it is a _lot_ cheaper than running real crash tests, even though
you still have to verify the theoretical results before the car will be
certified for road use.
You want to catch potential problems in advance.
The normal cost of locating an error later in a life cycle holds
for cars as well as software.
Cost isn't the real advantage. You can't instrument an experiment the
way that you can a simulation.

Some one noted the cost of crash test dummies has gone way up (like 5
orders of magnitude) as we learn more about crashes. They are no longer
mere mannequins. They get filled with cameras, accelerators, etc.
The firm which makes them see a bright future. Car companies don't take
this cost lightly anymore. A sim has to compete and complement this
advance in instrumentation.

The disadvantage, of course, is that you have mountains of information
data, not info
and sometimes no clue as to what to do with it. The science, the
insight, the talent, and everything else worthwhile are in finding
those clues and not in the petaflops or exabytes.

That's Hamming's Insight.

--

In article ,
Robert Redelmeier wrote:
In comp.sys.ibm.pc.hardware.chips Cydrome Leader wrote in part:
I find it unlikely they still don't know how nuclear weapons work,

You'd be amazed.
I was just at a funeral in Livermore yesterday.

especially considering they're mature technology and
eee. Hardly.
they've been around for decades.
That's part of the problem.

A two-stage thermonuclear warhead is a surprisingly complex
device -- look up Teller-Ulam. To work properly, the design has to
transfer enough light energy to ignite and burn the fusion secondary
before the fission primary shock waves disassemble the device.
There's _lots_ to simulate here in at least 2D over many timeslices.

Complex yes, but that's not the problem. The unclassified problem is
stockpile stewartship. Look that up not Teller Ulam.

Yes, we know how to make them go bang. Just follow the recipe.
But we don't always know the critical parts of that recipe, and
what parts we could change.

In general, the whole field of Finite-Element computation is
still short of cycles and can swallow everything available.
Multi-CPU clusters are still being built. Imagine being able
to simulate vehicle collisions -- designers would be able
to determine where metal could be added or other changes to
improve occupant survival.

The problem is that regardless of how any one feels about weapons, WWII
introduced a substance on the face of the planet we don't understand.
Rather than suggest that one buy or borrow a copy of the Pu-metallurgy
book, J. Bernstein has a new book titled "94". There's also a slightly
older simpler IEEE Spectrum article. You can't use common sense to
deduce how it will behave. It's not stable like other substances in
common every day experience. We are only now learning how it ages
(poorly). Weaponized Pu is constantly degrading. Most people haven't a
clue that these weapons have known shelf-lives (amazingly short before
they requirement maintenance). And you can't merely think about what's
in your nation's stockpile. You have to think about the stability of
the other guys' (note plural possessive) stockpile.


I once asked an old boss of mine if he had ever designed a dud. *
He was a new boss at the time, and I didn't have a clearance (still
don't). And he answered: Technology walks a very fine line.
Years later he admitted Yes. He had. But that didn't stop others from
using his code which went into the bunker buster in current use.
He's proud of that.


* Why the design of this I'll likely never know. I can think of people
who would know, but the vast majority of the semi-interested public are
too biased by early WWII weapons history to reason rationally and
consequentally have no Need to Know. I don't expect any further answer
from him so he can retain his Q-clearance.

--

On Jun 11, 3:36 pm, Cydrome Leader wrote:
I wonder what they really do with these computers.

Well, once there was a company in Milpitas, CA named Cydrome.
While they didn't make a super fast machine, they made a comparatively
fast machine for a short period of time. They studied the market.

It's mostly long running sims, but there's the odd occsional Ask Marilyn
question which some one runs using a programmable calculator.


I find it unlikely they still don't know how nuclear weapons work,
especially considering they're mature technology and they've been around
for decades.

Oh, you were the guy to ask that question.
It's not mature.

In article ,
Robert Myers wrote:
Try these google searchs: ~simulation site:lanl.gov
~simulation site:llnl.gov

That's 1 set of apps.

Those are merely the publically known/released examples.

I get over 30000 page hits. If you surf around and even ponder how to
zero in on the nuclear weapons work, you will have your answer.

Excepting Cary's and Hanson's, and DOE funded sites, not much depth in
most. You have to get on the classified network to see current problems.

Question: We've discovered a warehouse full of Viet Nam-era artillery
shells. Should we just ship them to a war zone, or count on them
working in case of war? I mean, we *do* know how artillery shells
work, don't we?

You realize that about 10% will be duds.

--

In article ,
Robert Myers writes:
|
| Theoreticians have a way of wanting to keep their work separate from
| what comes out of computers, as if theory could do anything without
| computation, automatic or otherwise.

Hmm. How many pure mathematicians do you know?


Regards,
Nick Maclaren.

On Jun 13, 4:21 pm, (Nick Maclaren) wrote:
In article ,Robert Myers writes:

|
| Theoreticians have a way of wanting to keep their work separate from
| what comes out of computers, as if theory could do anything without
| computation, automatic or otherwise.

Hmm. How many pure mathematicians do you know?


Picky, picky. I know a few. The boundary is irrevocably blurred,
even in pure mathematics.

Robert.

In article ,
Robert Myers writes:
|
| | Theoreticians have a way of wanting to keep their work separate from
| | what comes out of computers, as if theory could do anything without
| | computation, automatic or otherwise.
|
| Hmm. How many pure mathematicians do you know?
|
| Picky, picky. I know a few. ...

Well, what do you expect? :-)

Yes, I agree that some pure mathematicians are known to use computers!


Regards,
Nick Maclaren.