QUESTION: Why do we need a parallel computer to do good Computational Physics?
ANSWER: MANY interesting problems need -
LONG run times to reach equilibrium or steady state,
MANY different random realizations,
MANY different temperatures or pressures !
There is a limit to processor speed and so the only way to get the above is
to connect many processors with a fast communication switch between them.
Either the system itself,(demanding fast communication between the parts) or the
realizations are shared out. There is a limit to a students ability to do bookeeping for many runs and a single machine with many identical nodes makes this realistic.
Example at left: to study a LONG nanotube for LONG runtimes - divide it between processors.
After years of delay we got first NANCO and now TAMNUN for the TECHNION
and as the icing on the cake Israel has recently become a partner in PRACE,
the European Supercomputer sharing community.
See the real TAMNUN
homepage on the Computer Center Site.
There will be 3 courses for TAMNUN users. A preliminary course on
computing in a remote LINUX environment and two courses to be given by SGI.
The preliminary course notes are at: http://phycomp.technion.ac.il/~tamnun/linenv
and the course will be given on 22nd/23rd April.
Acounts can be opened on TAMNUN by pointing your browser to the registration page
Some more details on the mailing list archive which is linked below.