Re: 1teraflops cell processor possible?

In article <[EMAIL PROTECTED]>,
Del  Cecchi <[EMAIL PROTECTED]> wrote:
>
>You need to get your timeline straight.  If Blue Gene/L is running today,
>the design was started long before 4/2002.  One doesn't need a crystal ball
>to figure out that all this bio stuff would be a huge market, nor to look at
>the calculations to be made and define systems that folks knew how to build
>and knew how to program that could begin to do those calculations.

Grrk.  We all know that the design was started well before then - it
has been said in public often enough.  But that is not the only point.

One DID and DOES need a crystal ball to know how to define appropriate
systems, because it was and is clear that most such calculations are
being done using ghastly methods.  That means that there are likely
to be massive gains (perhaps tenfold, perhaps a thousandfold, perhaps
much more) in sorting out the algorithms.

That doesn't mean that Blue Gene/L isn't a reasonable project, but it
does mean that its design was much more based on guesswork than careful
analysis of the requirements.  Which military strategist said something
like "Any action, even if it turns out to be incorrect, is better than
no action at all"?


Regards,
Nick Maclaren.