[Beowulf] Re: vectors vs. loops

[Toon Moene]

Vincent Diepeveen wrote:

> If for example a scientist or director writes that a machine must for
> example have a blue color, then you know they will sign with ibm, whatever
> they ask, and so does IBM know.

I am sorry for intervening in this Jerry Springer show of HPC, but this 
is not true.

I am the Dutch Representative on the Technical Advisory Committee of the 
European Centre for Medium Range Weather Forecasts (www.ecmwf.int).  In 
2002, there was a real threat that IBM would take over the complete 
array of computational gear at ECMWF - the supercomputer, the 
workstations of the researchers and the storage stuff (tape robots and 
such).

Contracts were worded such that it would remain easy to dis-entangle 
this mess and make ECMWF independent of this single provider.

The IBM staff recognised this danger itself and proposed remedies.

Disclaimer: The Dutch National Weather Service (KNMI) does not use IBM 
equipment for its own computations to a significant amount.  We don't 
need to - there's sufficient competition from others.

We *will* use the LOFAR BlueGene/L machine (www.lofar.org) for research, 
though - it's hard to resist free machine time; we've plenty of things 
we want to find out given unlimited resources :-)

[ As an aside, GCC vectorises our code nicely - 4.0.0 ]

-- 
Toon Moene - e-mail: toon at moene.indiv.nluug.nl - phone: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
A maintainer of GNU Fortran 95: http://gcc.gnu.org/fortran/
News on GNU Fortran 95: http://gfortran.org/