🚀 go-pugleaf

In article <6fljqyg@lynx.unm.edu> john@aquarius.unm.edu (John Prentice) writes:
>
>In article <1992Apr2.160434.1774@fcom.cc.utah.edu> alex@fff.chem.utah.edu (Alex Dommasch) writes:
>>
>>Basically, the purpose of this group is to discuss the use of computers in  
>>scientific research (including theoretical work).  This is admittedly broad,  
>>but the idea is to catch topics that aren't covered elsewhere.  If your work  
>>uses the computer as a tool, it probably belongs here.  If your work involves  
>>research *of* computers, then there's probably a better place for it
>>
>
>Okay, I guess I qualify :-) .  I use computers in research in alot of
>different capacities.  I am what you would call a full contact programmer,
>I consume supercomputing time like it was water.  I am a theoretical

Like water?  Well, I only use about 100 hours a year.  My colleagues 
doing Lattice Gauge with dynamical fermions have dedicated 1/2 of our 
CM-2 for continuous use (that is, they are given about 2*8000 hours of 
1/4 machine time) and they get 6 Gflops -- so I consider myself a piker. 

My supercomputing is on nuclear structure problems; see march issue of 
Phys. Rev. C for a nicely written exposition on that work.  (I think it 
might be more accesible to the non-specialist than the Letter we wrote, 
since space is available and it was targeted at the experimenters in 
the field.)  My other computing is on nuclear scattering problems, so 
some aspects of your problem interest me....

>         ...        My research is in low energy atom/surface scattering. 
>This is where you want to scatter something like a helium atom off a 
>crystal at thermal energies  ....
>     ...                 My research is aimed at developing a theory that
>is correct, makes few assumptions, yet is computationally tractable.  

 Discussion of this physics might belong is sci.physics however.

>In my case I am solving are large systems of Fredholm integral
>equations of the second kind.  I convert these to linear equations using
>a Nystrom method and are left with a dense, complex valued set of linear
>equations of order 5,000 and up.  We can currently solve problems of
>up to order 5,000 pretty easily on a Cray Y-MP, but the real interest
>is in problems of order 10,000 and up.  This is pushing the state of the
>art in a big way for dense systems.  The equations are devoid of the
>usual structure that would make them easily amenable to iterative schemes
>or other obvious ways of reducing the computational load associated with
>solving them.  This is an interesting problem because of this however.

I am curious what your Cray performance is.  What does hpm give you for 
your Mflop rate and what is the vector/scalar breakdown?  

I am not familiar with your matrix problem, but it might have some common 
features with the one encountered by the Lattice group.  They use a 
preconditioned conjugate gradient method to solve their system.  What 
method do you use?  And whose Cray?

--
J. A. Carr                                    |  "The New Frontier of which I  
jac@gw.scri.fsu.edu                           |  speak is not a set of promises
Florida State University  B-186               |  -- it is a set of challenges."
Supercomputer Computations Research Institute |   John F. Kennedy (15 July 60)