From: Eugen Leitl (eugen@leitl.org)
Date: Thu Nov 28 2002 - 04:43:43 MST
On Wed, 27 Nov 2002, Peter C. McCluskey wrote:
> I suspect that the closest to what Gene is claiming is something like
> the methods mentioned in:
> http://www.hpc.csiro.au/External/NEC/1998Aug/quantum.htm
No, it was work similiar to http://bifrost.lanl.gov/MD/MD.html
They still keep churning out papers, lastly a big one on shockwave
propagation in copper.
You'll see plenty of multi-billion atom MD stuff under
Today's cluster have up to 10^4 CPUs. Blue Gene class of machines will
achive of >10^6 CPUs, which should give you at least 1 s: 1 ps time time
domain for 10^9 atoms, or slower performance for 10^12 atoms (assuming
each CPU has a few megawords on-die memory, as is to be expected).
This can involve pretty accurate forcefields, similiar to Brenner's.
Mechanosynthesis is restricted to sufficiently small regions so it can
profit from a deeper leverl of theory (ab initio, in some cases). It is
also parametrizable, once the parameter space has been sampled
sufficiently.
This archive was generated by hypermail 2.1.5 : Wed Jan 15 2003 - 17:58:28 MST