3 ms·

I wonder if the circuit could be modularized into a quantum ALU. Where many of the otherwise identical components in the original layout, couldn't just be arran

I wonder if the circuit could be modularized into a quantum ALU. Where many of the otherwise identical components in the original layout, couldn't just be arranged as operations time, over a reduced number of distinct components. Most digital algorithms would explode in terms of hardware needed, for increasing N, if we didn't distribute that computation in time, as well as across elements.

cubefox 1y ago
Is this the same idea as using universal Turing machines (CPUs that can execute software) rather than conventional fixed-function Turing machines (ASICs/FPGAs)?
- Nevermark 1y ago
  Yes, exactly
  - cubefox 1y ago
    This sounds like what's described here: https://www.quora.com/What-is-the-universal-quantum-computer-and-how-does-it-work https://www.quora.com/What-is-the-universal-quantum-computer... Perhaps this is what's already done in practice.
adgjlsfhk1 1y ago
unfortunately this is kind of fundamentally impossible. the whole power of quantum computation is that big quantum computers can do computation on a massive state space "for free". that benefit only exists if you have enough qbits to hold the state space.
- Nevermark 1y ago
  Holding the state space and iteratively transforming it (in a reversible way, so the state space is preserved), seem like a solution to that. This would still require more hardware than digital circuits (which can be made reversible for energy efficiency, but that is rarely done). While still reducing the number of operation components, and reusing them.
  - adgjlsfhk1 1y ago
    This is already done. Qbits is the size of the state space and gates is the measure of operations (like flops but quantum operations). https://arxiv.org/pdf/2505.15917 https://arxiv.org/pdf/2505.15917 shows a factoring algorithm using 1 million qbits and ~10^10 gates.
Strilanc 1y ago
In most quantum computer designs, gates are signals generated on demand at runtime rather than material deposited at fabrication time. In this regime, the concept of an ALU makes no sense. Instead of just sending pulses doing the exact gates you know need, why would you instead expand that short sequence into long sequences that emulate potentially applying every arithmetic operation to every input and then mux out the result you already knew you needed. It's a lot of extra work for the same final result. A quantum ALU would also be substantially harder to design, because of the need to maintain reversibility. For example, every operation would have to run as slow as the slowest operation (or else the timing side channel would measure which operation occurred).

cubefox 1y ago

Is this the same idea as using universal Turing machines (CPUs that can execute software) rather than conventional fixed-function Turing machines (ASICs/FPGAs)?

Nevermark 1y ago
Yes, exactly
- cubefox 1y ago
  This sounds like what's described here: https://www.quora.com/What-is-the-universal-quantum-computer-and-how-does-it-work https://www.quora.com/What-is-the-universal-quantum-computer... Perhaps this is what's already done in practice.

adgjlsfhk1 1y ago

unfortunately this is kind of fundamentally impossible. the whole power of quantum computation is that big quantum computers can do computation on a massive state space "for free". that benefit only exists if you have enough qbits to hold the state space.

Nevermark 1y ago
Holding the state space and iteratively transforming it (in a reversible way, so the state space is preserved), seem like a solution to that. This would still require more hardware than digital circuits (which can be made reversible for energy efficiency, but that is rarely done). While still reducing the number of operation components, and reusing them.
- adgjlsfhk1 1y ago
  This is already done. Qbits is the size of the state space and gates is the measure of operations (like flops but quantum operations). https://arxiv.org/pdf/2505.15917 https://arxiv.org/pdf/2505.15917 shows a factoring algorithm using 1 million qbits and ~10^10 gates.

Strilanc 1y ago

In most quantum computer designs, gates are signals generated on demand at runtime rather than material deposited at fabrication time. In this regime, the concept of an ALU makes no sense. Instead of just sending pulses doing the exact gates you know need, why would you instead expand that short sequence into long sequences that emulate potentially applying every arithmetic operation to every input and then mux out the result you already knew you needed. It's a lot of extra work for the same final result. A quantum ALU would also be substantially harder to design, because of the need to maintain reversibility. For example, every operation would have to run as slow as the slowest operation (or else the timing side channel would measure which operation occurred).