Architectural Evaluation of Quantum Computers
Architectural Foundation
To understand quantum computer architecture, it is crucial to understand the qubits,and superposition.In classical computers, information is represented as the binary digits 0 or 1. Computer hardware understands the 1-bit as an electrical current flowing through a wire (in a transistor) while the 0-bit is the absence of an electrical current in a wire. These electrical signals can be thought of as “on” (the 1-bit) or “off” (the 0-bit). Computers then decode the classical 1 or 0 bits into the human-readable format such as words, graphics etc.
Quantum bits or qubits are similar to bits in that there are two measurable states called the 0 and 1 states. However, unlike classical bits, qubits can also be in a superposition state of these 0 and 1 states. We can think of it like of a state of the coin while it is in the air. It is both heads and the tails until it lands. When it lands it has a definite state, meaning any particular way that the system can possibly described. For example, the coin can be either heads, or tails, or a combination of heads or tails while flipped in the air. All of these cases are called states of the coin system. While the coin is being flipped it is in a state of superposition. When we observe the coin, we are making a measurement which destroys the superposition.
Process is same for qubits, actually for all particles. Qubits are in superposition until we make an observation. That enables the Quantum Parallelism, A system of n qubits can represent all possible 2^n combinations of classical bits simultaneously while a classical computer must evaluate each input one at a time. But superpositions are not alone that grants the power of quantum computers, the real source of the quantum computers comes from the entanglement.
Entanglement occurs when qubits become correlated in a way that their states cannot be described independently. Measuring one qubit instantly determines the other, no matter the distance. So entanglement allows n qubits to represent exponentially more information than n classical bits. While a classical computer needs 2^n bits to store all possible states of n bits.
As the classical computers manipulate bits using logic gates, quantum computers manipulate qubits using quantum gates. However their logic differs. Classical gates operate on definite states (0 or 1) while Quantum gates operate on probability amplitudes. Quantum gates manipulate superpositions and generate entanglement, enabling algorithms like Shor’s and Grover’s.
Comparison Between Classical and Quantum Computer Architectures
As we mentioned previous section, traditional computers rely on the Boolean Algebra(deterministic), while quantum computers exploit quantum superposition and entanglement for parallel computation (probabilistic). As the bit manipulation differs, architecture between the Traditional and Quantum Computers are different.
In traditional computers , Harvard and Von Neumann architectures are the fundamental theoretical approaches for CPUs and are still influential today. In practice, modern processors are a hybrid of the two. For example, an Intel Core i7 processor behaves like Von Neumann at the high level, while taking advantage of the Harvard architecture at the cache levels. This is called "Modified Harvard Architecture" . And the ~%85-90 of the modern computers work with the x86 instruction set architecture. In addition, ARM-based chips (especially the Apple M series) have a rapidly growing share.
It is crucial to have common ISA in computing ecosystems. Programs compiled for a specific ISA (e.g., x86-64) can run on any compatible CPU (Intel, AMD, or even emulated on ARM). That way, chip designers can improve performance without breaking software, and old software works on new CPUs. But in the quantum computers, gates are not deterministic, due to nature of quantum computers, they do not fit into the classical ISA concept. Quantum hardware (ion trap, superconductor, photonics...) have significant differences in terms of gate times, error rates, and topologies. Therefore, abstract, high-level but hardware-translatable middleware languages are preferred. They are still called ISAs because they define the instructions to be given to the machine. But they do not directly correspond to the actual hardware. So unlike classical ISAs, quantum ISAs are more of a protocol for high-level control, not a direct correspondence to physical operations.
Figure 1: Table of Quantum Middlewares
The number of qubits is limited (e.g. 5 to several hundred qubits). Qubits are in a constant state of superposition and entanglement, so there is no possibility of stable datastorage like classical RAM. Memory management focuses on keeping track of which qubits quantum circuits are temporarily using. Techniques like Qubit Allocation, Routing, Ancilla Management are used to perform memory management. Quantum computers do not have a cache structure in the classical sense, because: Qubits do not hold classical data; they represent wave functions.
Every measurement process disrupts (collapses) the state of the qubit. Data is transient and cannot be stored until measured. Addressing in quantum systems is not done like classical addressing, but rather with control signals. One common technique is physical qubit addressing, where each qubit resides on a fixed location within a processor's 2D lattice or similar physical geometry. The arrangement and physical layout of these qubits directly influence how quantum circuits are mapped and executed. To perform operations on specific qubits, gate targeting is used, which involves sending control pulses to apply quantum gates precisely to selected qubits without disturbing others. In superconducting qubit systems, frequency addressing plays a critical role; each qubit operates at a unique resonance frequency, allowing selective control through microwave pulses tuned to those frequencies. This method helps minimize crosstalk and improve gate fidelity. Additionally, due to the limited connectivity between qubits in many quantum hardware architectures, topology-aware scheduling is employed to plan the order and location of gate operations based on the physical qubit layout. This optimization reduces the need for additional operations like qubit swapping, ultimately improving the efficiency and reliability of quantum computations.
Development of actual quantum computers is still in its infancy. Hardware diversity is high (photonic, superconductor, ion trap, neutral atom, etc.). It is not certain which physical system will be scalable, cheap and error-free.
Figure 2: Quantum Hardwares by different Companies
So the "x86 moment" hasn't happened in quantum hardware yet. But it will, most likely with a low-error, scalable architecture. In software, there will be standardization around middleware like OpenQASM and QIR. Just like the CUDA-Nvidia relationship, we could see closed but efficient systems where software and hardware work together in quantum.
Advantages and Limitations of Quantum Computers
As we mentioned in the previous chapters, the quantum technology is still in its nascentstages, facing significant hardware and physical limitations. This composition explores the types of problems where quantum computers are effective, the current hardware and physical constraints, and the bottlenecks compared to today's classical systems, supported by recent developments and expert analyses.
Quantum computers are particularly adept at solving complex optimization problems that are computationally intensive for classical systems. For instance, D-Wave's quantum annealing technology has been applied to real-world optimization tasks, such as logistics and scheduling, demonstrating potential advantages over classical approaches. Simulating quantum systems is inherently challenging for classical computers due to the exponential growth of the quantum state space. Quantum computers can naturally model these systems, aiding in the development of new materials, drugs, and understanding fundamental physics. Quantum algorithms, like Shor's algorithm, can factor large integers efficiently, posing a threat to current cryptographic schemes. While practical implementation requires more advanced quantum hardware, this application underscores the disruptive potential of quantum computing in cybersecurity.
While quantum computers excel in specific domains, many problems do not benefit from quantum speedups. Classical computers remain more efficient for a broad range of tasks, and developing quantum algorithms that outperform classical counterparts is an ongoing research challenge.
Also, transferring data between classical and quantum systems is a bottleneck. Encoding classical data into quantum states and retrieving results is non-trivial, often requiring complex procedures that can negate the speed advantages of quantum computation. And Implementing error correction in quantum systems demands significant overhead, with thousands of physical qubits needed for a single logical qubit. This contrasts with classical systems, where error correction is more straightforward and less resource-intensive.
Scaling quantum systems to a large number of qubits is hindered by increased noise, crosstalk, and the complexity of control systems. For example, IBM's 433-qubit Osprey processor still struggles with error rates that make most real-world applications impractical. Quantum hardware often requires extreme conditions, such as temperatures near absolute zero, to function correctly. These requirements necessitate complex and costly infrastructure, limiting the practicality and accessibility of quantum computers.
Quantum computing holds promise for revolutionizing specific computational tasks, particularly in optimization, simulation, and cryptography. However, significant hardware and physical limitations currently constrain its broader application.
Future Oriented Design and Modeling Ideas
Quantum computing lacks a standardized low-level instruction set architecture (ISA). Current efforts like IBM’s OpenQASM and Google’s Cirq intermediate representations serve this role to some extent, but they are still higher-level than traditional machine code. Development of quantum ISA that tightly maps quantum gates to hardware, support for hybrid instructions to mix classical control flow with quantum gates may change the world, enabling seamless integration of classical control flow with quantum operations.
Quantum systems require cryogenic environments and complex control electronics.These are orders of magnitude more expensive and energy-intensive than classical chips. We need to be able to run quantum computers more stably in more normal environments.
Microsoft's work on Majorana-based qubits represents a potentially groundbreaking shift in this area. By using Majorana quasiparticles, which are theorized to enable topological qubits, Microsoft aims to dramatically improve qubit stability and scalability. They claim this approach could unlock access to one million qubits, a scale far beyond what is currently feasible. If successful, this could represent the discovery of a new phase of matter, with properties uniquely suited for fault-tolerant quantum computing.
However, there is a healthy degree of skepticism in the scientific community. While Microsoft has demonstrated some experimental evidence of detecting and manipulating Majorana particles, they have not yet succeeded in constructing a functional topological qubit. In fact, a widely publicized 2018 claim regarding the discovery of Majorana particles was later retracted, leading researchers to be more cautious about similar announcements.
In summary, the future of quantum computing depends on breakthroughs across multiple layers: from low-level ISAs that can fully exploit hardware capabilities, to physical systems that can operate reliably and economically at scale. The road ahead is filled with scientific and engineering challenges but also with transformative potential.
Bibliography
[1] C. Hughes, J. Isaacson, A. Perry, R. F. Sun, and J. Turner, Quantum Computing for the Quantum Curious.Springer, 2021.
[2] K. B. Rao, “Computer Systems Architecture Vs Quantum Computer,” in 2017 International Conference on
Intelligent Computing and Control Systems (ICICCS), Madurai, India, Jun. 2017, pp. 763–767.
[3] P. B. Upama, M. J. H. Faruk, M. Nazim, M. Masum, H. Shahriar, G. Uddin, S. Barzanjeh, S. I. Ahamed, and A. Rahman, “Evolution of Quantum Computing: A Systematic Survey on the Use of Quantum Computing Tools,” in 2022 IEEE 46th Annual Computers, Software, and Applications Conference (COMPSAC), 2022, pp. 1299–1304.
[4] M. Kjaergaard et al., “Programming a quantum computer with quantum instructions,” Nature, vol. 593, no. 7858, pp. 205–210, 2021.
[5] Y. Ding, P. Gokhale, S. F. Lin, R. Rines, T. Propson, and F. T. Chong, “Systematic Crosstalk Mitigation forSuperconducting Qubits via Frequency-Aware Compilation,” in Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), 2023.
[6] M. Weiden, J. Kalloor, J. Kubiatowicz, E. Younis, and C. Iancu, “Wide Quantum Circuit Optimization with Topology Aware Synthesis,” arXiv preprint arXiv:2209.12948, 2022. [7] A. Butko, I. Turimbetov, G. Michelogiannakis, D. Donofrio, D. Unat, and J. Shalf, “TIGER: Topology-aware Assignment using Ising Machines Application to Classical Algorithm Tasks and Quantum Circuit Gates,” in 2022 IEEE International Conference on Quantum Computing and Engineering (QCE), 2022.
[8] Microsoft Azure Quantum, “Interferometric single-shot parity measurement in InAs–Al hybrid devices,” 2022. [Online]. Available: https://www.microsoft.com/en-us/quantum
[9] C. Seitz, “D-Wave Claims ‘Quantum Supremacy,’ Beating Traditional Computers,” The Wall Street Journal, May 2023. [Online]. Available: https://www.wsj.com/articles/d-wave-claims-quantum-supremacy-beating-traditional-computers-155ca634
[10] B. Kluger, “The Quantum Computing Revolution Is Coming,” TIME Magazine, Feb. 2023. [Online]. Available: https://time.com/6249784/quantum-computing-revolution/
[11] “Shor’s algorithm,” Wikipedia. [Online]. Available: https://en.wikipedia.org/wiki/Shor%27s_algorithm
[12] McKinsey & Company, “Potential and Challenges of Quantum Computing Hardware Technologies,” 2023. [Online]. Available: https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/tech- forward/potential-and-challenges-of-quantum-computing-hardware-technologies
[13] Milvus Blog, “What are the Limitations of Current Quantum Computing Hardware?”, 2023. [Online]. Available: https://blog.milvus.io/ai-quick-reference/what-are-the-limitations-of-current-quantum-computing-hardware
[14] “x86 instruction listings,” Wikipedia. [Online]. Available: https://en.wikipedia.org/wiki/X86_instruction_listings
[15] “ARM architecture family,” Wikipedia. [Online]. Available: https://en.wikipedia.org/wiki/ARM_architecture_family
[16] D. Evans, “x86 Assembly Guide,” University of Virginia. [Online]. Available: https://www.cs.virginia.edu/~evans/cs216/guides/x86.html
[17] “Quil (instruction set architecture),” Wikipedia. [Online]. Available: https://en.wikipedia.org/wiki/Quil_(instruction_set_architecture)