GPU-Accelerated Host-Aware Dead-Measurement Detection in Hybrid Quantum–Classical Programs: Full Version

Jul 16, 2026

Upload by Eden

Hybrid quantum–classical programs couple a quantum circuit with a classical host that consumes measurement outcomes to compute final results. When a measured outcome is syntactically read by the host but semantically noncontributory, gate optimizers that rely on circuit-local structure fail to see the resulting deadness. The work proposes a semantics-aware static analysis using abstract interpretation, proves soundness, and evaluates 24 application-faithful hybrid workloads. Compared with a syntactic liveness baseline, it finds over 4× more non-contributory measurements and removes 37.98% of gates on average, further accelerating analysis via GPU-parallel SSA-style lowering with up to 6.53× speedups.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages44

Reading Time1 h 51 min

Abstract

Introduction

Semantic non-contributory measurements in hybrid programs

Motivation from circuit fidelity and readout traffic

Prior work and limitations

arXiv :2607 .09976v1 [ quant-ph] 10 Jul 2026 GPU-Accelerated Host-Aware Dead-Measurement Detection in Hybrid Quantum–Classical Programs: Full Version Yanbin Chen 1[0000−0002−1123−1432], Qunyou Liu2[0000−0002−7410−502X], Yu Wang 1[0009−0004−3972−4388], Christian B. Mendl 1[0000−0002−6386−0230], and Helmut Seidl 1[0000−0002−2135−1593] 1 TUM School of CIT, Technical University of Munich, Boltzmannstr. 3, 85748 Garching, Germany {yanbin.chen, [18yu.wang](18yu.wang), christian.mendl,[helmut.seidl}@tum.de](helmut.seidl}@tum.de) 2 School of Engineering, École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland [qunyou.liu@epfl.ch](qunyou.liu@epfl.ch) Abstract. Hybrid programs combine a quantum circuit with a classical host program that consumes measurement outcomes. In such programs, an outcome may be syntactically read by the host but semantically noncontributory: changing the outcome cannot change the returned value. Such outcomes obscure gates that are dead only relative to the host semantics, and are therefore invisible to circuit-local optimizers. We present a semantics-aware host-side static analysis that identifies non-contributory measurement outcomes by abstract interpretation, and prove its soundness. We implement the analysis and evaluate it on 24 application-faithful hybrid workloads across quantum chemistry, optimization, quantum machine learning, and quantum finance. Compared with a syntactic liveness baseline, our analysis identifies more than 4× as many non-contributory measurements, and it standalone enables the removal of 37.98% of total gates on average. Even after the state-of-theart optimizers like Qiskit, t|ket⟩, and PyZX have already optimized the circuits, our analysis still enables removal of more than 30% of the postoptimized gates, showing that the host-semantic opportunities exposed by our analysis are not subsumed by circuit-local optimization. To scale our analysis, we further lower host programs to an SSA-style levelized intermediate representation that exposes level-wise parallelism for GPU execution, and implement a CUDA backend. We prove that this lowering preserves the analysis result, and the evaluation shows speedups of up to 6.53 × over a sequential baseline as structural parallelism increases. 1 Introduction Hybrid quantum–classical programs execute a quantum circuit whose measurement outcomes are consumed by a classical host program to compute the final result [6, 7, 20, 29, 31, 42, 46] . Although every measured qubit is typically read 2 Y. Chen et al. and its value flows through the host code, not all of those values truly influence the final return of the program. Contributions of these values could be semantically neutralized. As exemplified by Example 1, a value may cancel algebraically with others, vanish under an int (·) truncation, or appear inside a sub-expression whose effect is later overwritten. Example 1 . In Fig. 1 the quantum circuit produces outcomes o0 , o 1 , o2 which the host binds to a,b,c. Although a is read multiple times and even guards a branch, the returned value is independent of the initial value feeding a. If the then-branch is taken, then u+v+w = (a+b)(c-a) + a(b-c) + a 2 = bc, where alla–terms cancel. If the else-branch is taken, with η∈(0 .1 , 0.5) and a,b,c ∈{0, 1}, we have int (η a+bc) = bc, and u+v+w = bc. Thus, despite being syntactically used, the initial value originating from o0 is non-contributory, i.e., changing o0 does not influence the semantics of the program. Remark 1 . In Example 1, if the else-branch is taken, then semantically a = 0 and we get int (η a+bc) = int (bc) = bc without even accessing the range of η . However, this would need a mechanism of propagating the branch-local constraint into branches, which we have not implemented within our analysis. The standard liveness analysis [3, 23] is not enough to detect the semantic deadness of a in Fig. 1, as the liveness analysis is syntactic-ba

FAQs

What problem does the paper address in hybrid quantum–classical programs?

It addresses situations where a measurement outcome is syntactically used by the host but semantically does not affect the returned value, making corresponding gates effectively dead.

Why do circuit-local optimizers miss these dead measurements?

Because the deadness is defined relative to host semantics, and syntactic liveness or circuit-local structure cannot capture cases where changing an outcome cannot change the final program semantics.

How does the proposed method detect non-contributory measurement outcomes?

It uses a semantics-aware host-side static analysis based on abstract interpretation, identifies non-contributory outcomes, and includes a correctness proof of soundness.

Jul 16, 2026

Upload by Eden