Speed-oriented quantum circuit backend
1 st Sören Wilkening · 2026
Read Speed-oriented quantum circuit backend by 1 st Sören Wilkening (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-21656.txt.
Introduction
Sören Wilkening presents a new software backend built for fast quantum circuit generation, aimed at keeping classical preprocessing from eating away any quantum speedup. Benchmarked on the quantum Fourier transform, it builds circuits for systems up to 2000 qubits ahead of popular frameworks like Qiskit and Q#. The package also supplies high-level primitives for bit and integer manipulation, easing integration with higher-level quantum programming languages.
Catalog record
| CREATOR | 1 st Sören Wilkening |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23821 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.21656
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Speed-oriented quantum circuit backend 1 st Sören Wilkening Abstract We present a new software package for efficient quantum circuit generation, designed to achieve optimal runtime performance. Despite being in an early stage of development, our implementation demonstrates significant advantages over existing tools. Using the quantum Fourier transform (QFT) as a benchmark, we show that our backend can generate circuits for systems with up to 2000 qubits faster than widely used frameworks such as Qiskit and Q#. This improvement is particularly relevant for applications where classical preprocessing time, including circuit generation, must be minimized to not diminish any potential quantum advantage - for example, in combinatorial optimization tasks. Additionally, our software provides high-level primitives for bit- and integer-level manipulations, offering a simplified interface for integration with high-level quantum programming languages. I Introduction Over the past decade, significant progress has been made in the development of quantum algorithmic frameworks [ 26 ] , including quantum algorithms [ 25 ] , hardware architectures, and full-stack quantum computing pipelines [ 11 ] , advancing the field towards achieving practical quantum advantage [ 39 , 13 ] . A key component of this effort is the design of high-level quantum programming languages [ 18 , 6 , 7 , 16 , 40 ] , which abstract away low-level implementation details and accelerate the discovery and deployment of new quantum algorithms. The usefulness of such high-level languages critically depends on the availability of efficient backends capable of generating, storing, and optimizing quantum circuits [ 31 , 27 , 14 , 33 ] . In the near-term era of noisy intermediate-scale quantum (NISQ) devices, optimization often focuses on minimizing qubit counts and gate depth [ 20 , 12 ] . However, in the longer term — when fault-tolerant devices with abundant qubits become available — other considerations arise. In particular, for many applications, the runtime cost of circuit generation itself becomes a limiting factor, and a backend that emphasizes high throughput while producing circuits with “sufficient” levels of optimization becomes essential. This requirement is especially evident in quantum algorithms for combinatorial optimization [ 42 , 9 , 4 , 24 , 41 ] , where classical preprocessing, including circuit generation, must not dominate the total runtime. To address this challenge, we present a software package for quantum circuit generation (QCG), introduced in [ 45 ] , which is designed for near-optimal runtime performance. Using the quantum Fourier transform (QFT) as a benchmark, we demonstrate that our backend can generate circuits for systems with up to 2000 qubits significantly faster than existing tools. The QFT is chosen as a benchmark because it is a well-known routine that appears in various applications, such as quantum integer arithmetic [ 32 ] , quantum amplitude estimation [ 8 ] , and Shor’s algorithm [ 34 ] , among others. Beyond basic gate-level construction, our package supports assembly-like instructions for integer and Boolean operations, enabling more expressive circuit generation and providing a foundation for high-level quantum programming languages [ 18 , 6 , 7 ] . Although the current implementation is an early-stage prototype, it already supports key arithmetic operations such as addition, multiplication [ 32 ] , and division [ 38 ] , as well as bitwise operations on quantum and hybrid classical–quantum data. These capabilities facilitate rapid implementation of Grover-type search algorithms [ 3 , 44 , 43 ] and other applications that require structured data manipulation. Our approach contrasts with existing software packages such as Qiskit [ 27 ] , Cirq [ 14 ] , and others [ 18 , 23 , 2 , 30 , 5 , 35 , 10 , 37 , 22 ] , which typically combine circuit generation with simulation or execution features. Our software focuses exclusively on efficient circuit construction, making it easy to integrate with high-level languages, optimizers, or simulators as needed. In summary, we propose a fast and lightweight backend for quantum circuit generation, designed for future large-scale quantum algorithms. By decoupling circuit generation from execution and emphasizing runtime efficiency, our work lays the groundwork for scalable quantum software stacks, supporting the development of more complex, high-level quantum programming paradigms. II Methods Due to the desired performance and minimal requirements for data storage, we chose the programming language C [ 21 ] for our implementation, which has no external dependencies. For this reason, provided code and data structures are written in C. \ket{q_{0}} H H R_{2} R_{2} R_{3} R_{3} R_{4} R_{4} R_{5} R_{5} \ket{q_{1}} H H R_{2} R_{2} R_{3} R_{3} R_{4} R_{4} \ket{q_{2}} H H R_{2} R_{2} R_{3} R_{3} \ket{q_{3}} H H R_{2} R_{2} \ket{q_{4}} H H Figure 1: Circuit of a 5-qubit QFT without swaps. To store a quantum algorithm, we require certain data structures. In principle, a quantum gate can be characterized by a few integer and floating-point values, as shown in Data Structure . The minimum information required to describe a gate is its type (e.g., X , H , P , …), the qubit indices it acts on (targets), and the qubit indices it is controlled by (controls). Parametric gates additionally store the respective parameter value. Some gates may require multiple parameters, but in our description we consider only single-parameter gates. The implementation can easily be extended to support gates with multiple parameters. Structure to store data of a quantum gate. ⬇ 1 typedef struct gate_t { 2 int * controls ; // index of control qubits 3 int num_controls ; 4 int * targets ; // index of target qubits 5 int num_targets ; 6 int type ; // type of gate (X, H, ..) 7 double value ; // value of parametric gate 8 } gate_t ; A straightforward approach to representing a quantum circuit is to store all applied gates sequentially in a one-dimensional list. While this implementation is simple, it becomes inefficient when performing circuit optimization. For instance, consider the case where a newly inserted gate is the inverse of a previously applied gate. Identifying the redundant pair requires searching through the entire list, which incurs a time complexity of \mathcal{O}(N) , where N is the current circuit size. Moreover, such cancellation is only possible if the two gates occur in consecutive layers, further complicating the optimization process. To address these limitations, we adopt a layered circuit representation. In this model, a layer is defined as a set of mutually independent operations acting on disjoint qubits that can be executed in parallel. Organizing gates into layers enables optimization algorithms to identify and eliminate redundancies locally within each layer, thereby significantly reducing computational overhead. While a layer-sorted representation can still be implemented using a linear array, this approach would require frequent element rearrangements whenever new gates are inserted, resulting in additional runtime costs. Structure to store data of a quantum circuit. ⬇ 1 typedef struct circuit_t { 2 gate_t ** gates ; // [layer][gates] 3 int ** gate_index ; // [layer][qubit] 4 int * gates_per_layer ; 5 int ** last_layer_of_qubit ; // [qubit][index] 6 int * head_last_layer_of_qubit ; 7 int num_layer ; 8 } circuit_t ; Instead, we employ a two-dimensional data structure (Data Structure ), where each row corresponds to a circuit layer and each element within a row stores the gates assigned to