More specifically, assuming we manage to make an efficient quantum switch, some kind of "quansistor"(quantum transistor) and manage to resolve the problem of joining a bunch of them together without losing the quantum properties that we want, where in a modern computer would these quansistors come in to create a quantum computer? As in would the entire computer be made of quansistors? or would only some parts be quansistors and the rest be standard eletronics. would the logic of the processor(the ALU and such...) be made with quansistors? would the storage? down to what level? just the registers or the RAM too? I imagine that having the part of the storage that stores the instructions of the running program be affected by the randomness of quantum behaviour would cause some problems. Really my question comes down to where in the standard computer processing pipeline does the quantumness come in, in order to produce the intended theoretical behaviour of quantum computer?


closed as off-topic by Sasho Nikolov, Jan Johannsen, Daniel Apon, Emil Jeřábek, Robin Kothari Jun 29 '17 at 17:33

This question appears to be off-topic. The users who voted to close gave this specific reason:

  • "Questions must demonstrate a minimal understanding of the problem being solved. Tell us what you've tried to do, why it didn't work, and how it should work. See also our question checklist." – Sasho Nikolov, Daniel Apon, Robin Kothari
If this question can be reworded to fit the rules in the help center, please edit the question.

  • $\begingroup$ Integrating QM into CUDA (Compute Unified Device Architecture) seems to be a likely candidate. Note that this 'classical' parallel pipeline architecture only serves to initialize inputs and recover outputs at scale. CUDA treats devices as black-boxes, which QMs most singularly are. I tried to include this as an answer, but it got trolled by dilettantes. Note that IBM already provides a similar cloud architecture, have fun: quantumexperience.ng.bluemix.net/qx/editor $\endgroup$ – Dominic Cerisano Jun 30 '17 at 23:23

Really my question comes down to where in the standard computer processing pipeline does the quantumness come in, in order to produce the intended theoretical behaviour of quantum computer?

The critical part to get the intended theoretical (scaling) behaviour of quantum computer is probably to achieve and maintain the coherence between the different qubits in the computer.

Let me illustrate this with a hypothetical architecture. Assume that you have ready made modules with a fixed number of qubits (say 49=7x7). Further assume that each module for itself is able to maintain coherence between its qubits (by using quantum error correction), and reliably operate on its qubits controlled by a classical computer (my mental model is similar to this simple model of a finite quantum coprocessor I asked about before, but it doesn't really matter for this illustration1).

Now assume that you scale your quantum computer by putting many of those modules together, using some suitable topology. The modules cannot talk directly to each other, but we may assume that the classical computer controlling those modules is able to freeze any quantum operation of the modules and then permute qubits between those modules according to the topology.

Now assume that the quantum operation of each module randomly changes its global phase in a way uncontrollable and unmeasurable by the controlling classical computer. This is not even true decoherence (as would occur in practice - the density matrix formalism could be used for a better model), but I guess it will be sufficient to destroy the intended theoretical (scaling) behaviour of quantum computer. But if the qubits of the different modules stay perfectly coherent with each other, and the global phase differences between the modules are known to the classical computer, then I guess it will be possible to achieve the intended theoretical (scaling) behaviour of quantum computer.

  1. Edit Actually, the part of the model about initialization and measurement does matter a bit. Even if we assume that we know the global phase differences between the modules after initialization, we certainly have to assume that measurement is restricted to each module alone, and that coherence with other modules is lost after measurement. I still don't think that this will destroy the quantum speedup, since the global state can be (quantum/unitarily) transformed suitably before the measurement, hence one should still be able to perform any desired single measurement at the end of the quantum computation.
  • $\begingroup$ Using a parallel pipeline architecture like CUDA to black-box the QM (generalized quantum device) would overcome your concerns. A QM must have its input seeded, and output harvested very quickly. The resulting architecture would look like the QM is a bottleneck to its huge arrays of classical input and output devices. It is very non-intuitive in the classical sense to deliberately design a bottleneck. A QM is classically non-intuitive. $\endgroup$ – Dominic Cerisano Jun 30 '17 at 0:49
  • 1
    $\begingroup$ @DominicCerisano Gill Kalai has concerns (gilkalai.wordpress.com/2016/04/22/…)., I just tried to illustrate one difficulty to scale a quantum computer with a simple model that can both be described rigorously by QM and nevertheless degrades gracefully when describing it less rigorously for an audience less familiar with the formalisms of QM. $\endgroup$ – Thomas Klimpel Jun 30 '17 at 10:37
  • $\begingroup$ Separation of concerns: the advantage of black-boxing a QM in a classical parallel pipeline is that it is described by its inputs and outputs only. It becomes just another generalized device - a black box with no need to describe its internal workings. The goal of scaling in such an architecture is to saturate the input and output capacity of the black box, which could mean arrays of millions of classical input and output devices for a single QM. $\endgroup$ – Dominic Cerisano Jun 30 '17 at 22:52

Not the answer you're looking for? Browse other questions tagged or ask your own question.