Processors in memory... NOT

The thread that started off talking about Teraflops processors,
(Re: 1teraflops cell processor possible?), which, amongst other stuff,
deviated to talking about stream processors (for which
I share Robert Myers interest[*]) reminds me to post
about a more theoretical topic:

Several times I have heard people say "of course, the limit,
the ultimate, is processor in memory - if the computation
were actually free, it would be done next to the memory
containing the data to be operated on".

I think this is NOT true.

Most computation involves more than one data item.
I think this implies that the optimum place to perform
a computation will be at something resembling the
centroid of the data items involved - weighted by
access frequency.  With some allowance for treating
instructions as data. During the life of a comnputation
this "computational centroid" would shift over time.

Parallel computations would be performed at something
resembling a Viterbi constallation diagram.

---

Re streams:  would you build a stream instrction set, or a
vector instruction set?

Obviously, vectors can be built on top of streams.
Vice versa is a bit harder.