Accelerating scalable QPU adoption as a vital computing resource in HPC data center

David Teik-Hui Lee^1*, Chii-Dong Chen¹

¹Institute of Physics, Academia Sinica, Taipei, Taiwan

* Presenter:David Teik-Hui Lee, email:takehuge@gate.sinica.edu.tw

Leveraging a diverse computing resource has been critical to tackle problems with higher complexity. The integration of GPU with CPU has served to be a bedrock to the development of the most capable Artificial Intelligence (AI) models. Since quantum processor unit (QPU) harness superposition and entanglement to abstract away certain degree of complexity, it naturally provides an edge over classical compute in non-polynomial problem space. However, the nascent QPU is still too noisy to solve any practically hard problem directly. Hence in this study, we explore how to progressively integrate our superconducting QPU with the classical HPC. We first develop Cross Entropy Benchmarking (XEB) as a crucial metric by observing how well our tunable-coupler enabled superconducting 5q-QPU handles random quantum circuits and comparing the output against that churned out from the HPC. This way we can “learn” to better match the quantum control at the pulse level and quantum circuit at the algorithm level. Next, we will further explore how to use HPC to speed up calibration process and enable some level of error-correction while we continue to scale up our QPU in terms of both qubit count, architecture and topology. To allow for practical adoption of growing QPU into HPC, we are providing testbeds to seek to optimize future quantum computer footprints in the data center, and to nurture future talents in developing quantum applications.

Keywords: quantum algorithm, quantum control, XEB, QPU, HPC