Abstract
We report, in a sequence of notes, our work on the Alibaba Cloud Quantum Development Platform (AC-QDP). AC-QDP provides a set of tools for aiding the development of both quantum computing algorithms and quantum processors, and is powered by a large-scale classical simulator deployed on Alibaba Cloud. In this note, we report the computational experiments demonstrating the classical simulation capability of AC-QDP. We use as a benchmark the random quantum circuits designed for Google's Bristlecone QPU. We simulate Bristlecone-70 circuits with depth $1 + 32 + 1$ in $0.43$ second per amplitude, using $1449$ Alibaba Cloud Elastic Computing Service (ECS) instances, each with $88$ Intel Xeon (Skylake) Platinum 8163 vCPU cores @ 2.5 GHz and $160$ gigabytes of memory. By comparison, the previously best reported results for the same tasks are $104$ and $135$ seconds, using NASA's HPC Pleiades and Electra systems, respectively. Furthermore, we report simulations of Bristlecone-70 with depth $1+36+1$ and depth $1+40+1$ in $5.6$ and $580.7$ seconds per amplitude, respectively. To the best of our knowledge, these are the first successful simulations of instances at these depths.
Introduction
As the field of quantum computing approaches the milestone of "quantum supremacy," classical simulation serves as a critical benchmark. To validate the performance of quantum processors, it is necessary to understand the limits of classical algorithms and hardware. Simulating large-scale quantum circuits is computationally intensive, as memory and processing requirements grow exponentially with the number of qubits and circuit depth. This work addresses the challenge of simulating the Bristlecone-70 architecture, a 70-qubit grid designed by Google, which has previously pushed the world's most powerful supercomputers to their limits.
Methodology
The simulation engine, Tai-Zhang, is deployed on Alibaba Cloud and utilizes tensor network contraction. In this framework, a quantum circuit is represented as a network of tensors, and calculating an output amplitude involves a series of multi-dimensional matrix multiplications. The computational complexity is primarily governed by the treewidth of the circuit graph.
Key technical innovations include:
- Distributed Preprocessing: A centralized agent node decomposes the massive contraction task into thousands of independent subtasks.
- Cloud Orchestration: These subtasks are distributed across 1,449 Alibaba Cloud Elastic Computing Service (ECS) instances via an Object Storage Service (OSS) hub, enabling massive parallelization without shared memory constraints.
- Refactored Codebase: Transitioning to a "tensor-as-node" model improved abstraction and the handling of complex networks.
Experimental Setup
The study utilizes the Random Quantum Circuit (RQC) benchmark designed for Google’s Bristlecone QPU. Specifically, we focus on Bristlecone-70, a 70-qubit subset of the 72-qubit lattice. The circuits include layers of Hadamard gates, CZ gates in alternating configurations, and random single-qubit gates. The performance of AC-QDP was compared against records set on NASA’s Pleiades and Electra supercomputers, as well as the Tianhe-2 and Sunway TaihuLight systems.
Results and Validation
The AC-QDP demonstrated significant performance gains over traditional high-performance computing (HPC) clusters. For a Bristlecone-70 circuit at depth $1+32+1$, the system calculated an amplitude in just 0.43 seconds, representing a ~240x speedup over the 104–135 seconds required by NASA's systems. Furthermore, the team achieved the first successful simulations at depths of 36 and 40.
To ensure the accuracy of the simulation, the researchers calculated 200,000 amplitudes and verified that they followed the Porter-Thomas distribution, the expected statistical signature for truly random quantum circuits.
Conclusion
While the results demonstrate the power of cloud-based distributed simulation, a fundamental constraint remains: the exponential scaling of tensor contraction with circuit treewidth. Adding even a few layers of gates can significantly increase simulation difficulty. Nonetheless, AC-QDP provides a robust sandbox for quantum developers to verify algorithms and processor designs before deployment on physical, noisy quantum hardware.
Summary updated