Exposing the hidden layers and interplay in the quantum software stack
Vlad Stirbu, Arianne Meijer-van de Griend, Jake Muff
TL;DR
The paper addresses the opacity of the quantum software stack on NISQ hardware and argues that hidden layer interactions hamper efficient resource use. By distinguishing transpilation from compilation and highlighting the role of the vendor-specific control hardware, it shows how backends, calibration, and hardware topology jointly shape performance. Through a concrete Qiskit-based example on the VTT Q5 Helmi device, the authors illustrate the cross-layer dependencies among the classical API, transpilation, backend selection, and hardware behavior. The authors emphasize that exposing calibration state, device topology, and supported operations via explicit interfaces is essential to enable principled algorithm design and cross-layer optimization. Although hardware execution occurs in the microsecond range ($\mu s$) per shot, the total runtime is dominated by shot-based statistics, making stack transparency crucial for practical quantum software engineering.
Abstract
Current and near-future quantum computers face resource limitations due to noise and low qubit counts. Despite this, effective quantum advantage can still be achieved due to the exponential nature of bit-to-qubit conversion. However, optimizing the software architecture of these systems is essential to utilize available resources efficiently. Unfortunately, the focus on user-friendly quantum computers has obscured critical steps in the software stack, leading to ripple effects into the stack's upper layer induced by limitations in current qubit implementations. This paper unveils the hidden interplay among layers of the quantum software stack.