Leapfrogging Sycamore: Harnessing 1432 GPUs for 7$\times$ Faster Quantum Random Circuit Sampling (2406.18889v1)

Abstract: Random quantum circuit sampling serves as a benchmark to demonstrate quantum computational advantage. Recent progress in classical algorithms, especially those based on tensor network methods, has significantly reduced the classical simulation time and challenged the claim of the first-generation quantum advantage experiments. However, in terms of generating uncorrelated samples, time-to-solution, and energy consumption, previous classical simulation experiments still underperform the \textit{Sycamore} processor. Here we report an energy-efficient classical simulation algorithm, using 1432 GPUs to simulate quantum random circuit sampling which generates uncorrelated samples with higher linear cross entropy score and is 7 times faster than \textit{Sycamore} 53 qubits experiment. We propose a post-processing algorithm to reduce the overall complexity, and integrated state-of-the-art high-performance general-purpose GPU to achieve two orders of lower energy consumption compared to previous works. Our work provides the first unambiguous experimental evidence to refute \textit{Sycamore}'s claim of quantum advantage, and redefines the boundary of quantum computational advantage using random circuit sampling.

• The paper demonstrates a GPU-driven algorithm that generates quantum samples 7× faster than Sycamore with higher XEB scores and improved energy efficiency.
• It employs innovative techniques like partial network contraction and parallel tensor contraction to optimize simulation fidelity while reducing computational complexity.
• Empirical results show 3 million uncorrelated samples produced in 86.4 seconds using only 13.7 kWh, challenging claims of quantum computational advantage.

Harnessing GPUs for Efficient Quantum Random Circuit Sampling

The paper "Leapfrogging Sycamore: Harnessing 1432 GPUs for 7× Faster Quantum Random Circuit Sampling" addresses the challenges and breakthroughs in classical simulation of quantum random circuit sampling (RCS). As quantum computing advances, particularly through experiments like Google's Sycamore processor, there has been an increased emphasis on demonstrating quantum computational advantage (QCA). However, this paper reevaluates and challenges those assertions by employing a robust classical simulation framework utilizing 1432 GPUs, delivering competitive performance metrics.

Key Contributions

The authors present a comprehensive simulation approach that leverages an energy-efficient classical algorithm, enhanced by parallel processing capabilities afforded by GPUs. This method achieves uncorrelated sample generation with a linear cross-entropy score superior to that of previous classical simulations. Notably, it surpasses Google's Sycamore in terms of speed, being seven times faster in generating quantum samples from a 53 qubit system. The utilization of GPUs also brings about a significant reduction in energy consumption, achieving two orders of magnitude lower energy usage compared to prior work.

Methodological Advancements

The paper details several methodological improvements:

• Post-processing Algorithm: The authors introduce a post-processing algorithm which effectively reduces computational complexity while increasing the fidelity of the cross-entropy benchmark. This facilitates obtaining a higher linear cross-entropy score (XEB) using lower fidelity samples, thereby optimizing the time-to-solution.
• Partial Network Contraction: By employing partial network contraction, the simulation complexity is reduced, allowing faster execution without extensively compromising simulation fidelity.
• Parallel Tensor Contraction: The paper capitalizes on an advanced parallel tensor contraction algorithm across multiple GPUs, dramatically scaling accessible storage space and reducing computational complexity. This effectively uses 8-GPU parallelism for improved throughput.

Numerical Results

Empirical results validate the framework's efficacy, with the proposed method delivering 3 million uncorrelated samples in just 86.4 seconds. The XEB obtained is $2 \times 10^{-3}$, surpassing Sycamore's performance while consuming only 13.7 kWh. This positions the classical simulation as a formidable competitor in the ongoing discourse on quantum advantage claims.

Implications and Future Prospects

The implications of this work extend both theoretically and practically, asserting that advancements in classical algorithms can bridge the gap with quantum computational claims. The approach underscores the potential of utilizing classical resources efficiently, possibly setting new boundaries for what constitutes quantum advantage.

Future research may explore the integration of even larger GPU clusters and the application of similar algorithms to other quantum computing paradigms. The progress in simulation techniques also presents an opportunity to reexamine the energy efficiency debate within quantum computing, advocating for sustainable development alongside performance enhancements.

In conclusion, this paper contributes significantly to the discourse on quantum and classical computational capabilities, providing a nuanced perspective on the evolving landscape of quantum supremacy experiments. By leveraging high-performance computing resources, it serves as a reminder of the competitive nature of technological advancements, illustrating the dynamic interplay between theoretical potential and practical execution.