Eugene Leitl (
Tue, 17 Nov 1998 10:38:22 +0100

christophe delriviere writes:

> you used the term "computronium" for optimized computational matters, i
> find it a nice one... did you coined it or does it appear in some book
> or sci-fi litterature, culture ?

N. Margolus, "Crystalline Computation', to appear in The Feynman Lectures on Computation, volume 2 (Anthony Hey, ed.), Addison-Wesley (1998). Preprint (46 pages; .ps.gz) on the web.

Thus in the distant future I expect that our most powerful large-scale general purpose computers will be built out of macroscopic crystalline arrays of identical invertible computing elements. We would make such large arrays out of identical elements because they will then be easier to control, to design, to build and to test. These will be the distant descendants of todays SIMD and FPGA computing devices: when we need to perform inhomogenous computations, we will put the irregularities into the program, not the hardware. The problem of arranging the pieces of a computation in space will be part of the programming effort: architectural ideas that are used today in physical hardware may reappear as data structures within this new digital medium. With molecular scale computing elements, a small chunk

of this computronium[63] would have more memory and processing power

than all of the computers in the world today combined, and high Reynold's number CA MD calculations of fluid flow would be practical on such machines.

Note that I don't expect our highest performance general purpose computers to be quantum spin computers of the sort discussed in Section 1.1. In such a machine, the whole computer operates on a superposition of distinct computations simultaneously. This kind of quantum parallelism is very delicate, and the overhead associated with the difficult task of maintaining a superposition of computations over a large spatial scale will be such that it will only be worth doing in a very specialized situations -- if it is possible at all[74]. This won't be something that we will do in our general purpose computers. [...]

[63] Margolus, N., "Fundamental physical constraints on the computational process", Nanotechnology: Research and Perspectives (B.C. Crandall and J. Lewis eds.), MIT Press (1992).

[74] Preskill, J. "Fault-tolerant quantum computation", in [54] and quant-ph/9712048 ( )