Re: 1teraflops cell processor possible?

> > There is not a thing a compiler can do about that.  Between operations
> > (except in the case of an FMAC, which is an example of streaming),
> > operands have to sit in registers and they have to be ferried back and
> > forth.

Not true.  If the data is used soon,
it gets caught on the bypass path.
Typically more than half of all operands.
With a bit of work, using what I call
a bypass cache, more still.

Registers are not a performance limiter, because
they are bypassed.  They may be wasting power.
Dead value elimination misses part of this.