Ten miles from Long Island’s Atlantic coast, Shinjae Yoo, a computational scientist and machine learning group lead at the U.S. Department of Energy’s Brookhaven National Laboratory, is preparing to run quantum computing simulations on a supercomputer for the first time. Yoo is using the Perlmutter supercomputer at the National Energy Research Scientific Computing Center (NERSC) and the latest version of PennyLane, a quantum programming framework developed by Xanadu. By using PennyLane in conjunction with the NVIDIA cuQuantum software development kit, Yoo can run simulations on high-performance clusters of NVIDIA GPUs.
The performance of the software is crucial because Yoo and other researchers need to process large datasets. Yoo plans to run his programs on up to 256 NVIDIA A100 Tensor Core GPUs, simulating about three dozen qubits. This is twice the number of qubits that most researchers can currently model. The multi-node version of PennyLane, combined with the NVIDIA cuQuantum SDK, simplifies the task of accelerating massive simulations of quantum systems. This enables even interns to run large simulations, making it an exciting advancement for Yoo’s team.
The Perlmutter supercomputer is not just being used by Yoo. There are at least four other projects at NERSC this year utilizing the multi-node version of PennyLane. These projects include efforts from NASA Ames and the University of Alabama. Researchers in various fields, such as chemistry, physics, and machine learning, are using PennyLane to study complex problems that classical computers cannot handle. The software allows them to extend their capabilities and prepare for running algorithms on future large-scale quantum computers.
PennyLane is a unique quantum programming framework that incorporates popular deep learning techniques and tools to program quantum computers. It gained traction in the quantum community shortly after its introduction in 2018, thanks to its adaptability across different types of quantum computers. The demand for more qubits is increasing, and the PennyLane forum often receives requests for increased qubit availability. Xanadu is actively working on scaling PennyLane’s performance and plans to simulate more than 40 qubits by the end of the year.
The integration of cuQuantum’s multi-node capability with PennyLane has allowed for significant improvements in performance. Early experiments with splitting quantum programs across multiple GPUs have shown promising results, and the team is continuously collecting data to further optimize the software. This boost in performance will benefit not only researchers but also companies designing quantum computers, as it enables them to test ideas for building better systems. The collaboration between Xanadu, NVIDIA, and the quantum community is driving advancements in both software and hardware, creating a virtuous cycle of innovation.