Fixstars and University of Osaka Achieve Large-Scale Quantum Chemistry Simulation on 1,024 GPUs
Summarize this article with:
Fixstars Corporation and the University of Osaka have jointly achieved an advance in quantum chemistry simulation, successfully modeling quantum circuits with up to 1,024 GPUs, exceeding the previously established 40-qubit limit. This expands the scale of molecular systems available for developing and validating quantum algorithms intended for future, more powerful computers, with potential applications in areas like drug discovery and materials science. Researchers utilized the “chemqulacs-gpu” quantum circuit simulator and a newly developed parallel computing technology on AIST’s ABCI-Q system to overcome computational hurdles and achieve this larger simulation scale. “Large-scale simulation of quantum circuits using 1,024 GPUs is technically demanding, and within the 48-hour computation window we repeatedly encountered unexpected issues,” said Professor Wataru Mizukami of the University of Osaka, highlighting the challenges overcome in this work. 1,024 GPUs Enable Record 42-Spin-Orbital Quantum Chemistry Simulation A simulation leveraging 1,024 graphics processing units (GPUs) has achieved a record 42-spin-orbital quantum chemistry calculation, significantly extending the capabilities of classical simulations for validating future quantum algorithms. This achievement addresses a critical need for robust testing of quantum algorithms before the advent of fully functional, fault-tolerant quantum computers.
The team focused on Iterative Quantum Phase Estimation (IQPE), a method requiring fewer qubits than standard Quantum Phase Estimation, and implemented it within the “chemqulacs-gpu” quantum circuit simulator. They also developed novel parallel computing technology to maximize performance across the massive GPU cluster. The simulations successfully calculated a 42-spin-orbital system for a water molecule, utilizing qubit reduction technology, and a 41-qubit circuit for an iron sulfide molecule as a benchmark of circuit scale. These calculations were performed on AIST’s ABCI-Q system, overcoming longstanding computational bottlenecks that previously restricted the size of simulatable quantum circuits. “I am delighted that the team, led by Yusuke Teranishi and Shoma Hiraoka, persevered throughout the effort, and that, with prompt support from the ABCI-Q operations staff, we were able to achieve one of the largest results in this area,” said Mizukami. This collaborative effort saw QIQB focusing on the simulation methods and interface development, while Fixstars Corporation contributed GPU performance profiling and optimization expertise, ultimately resolving complex communication issues between GPUs and enabling efficient circuit simulation. Simulator & Parallel Computing Optimize IQPE Circuits Researchers have expanded the scope of classical simulations for quantum chemistry, achieving a milestone in simulating iterative quantum phase estimation (IQPE) circuits with up to 1,024 GPUs; previously, state-vector-based simulations were largely limited to 40 qubits. Achieving these results necessitated the development of novel parallel computing techniques optimized for large-scale GPU clusters, allowing researchers to bypass longstanding computational bottlenecks. “I hope this accomplishment will help accelerate the development of quantum algorithms,” said Mizukami, highlighting the potential for this work to drive progress in the field. Large-scale simulation of quantum circuits using 1,024 GPUs in unison is technically demanding, and within the limited 48-hour computation window we repeatedly encountered unexpected issues. Fe₂S₂ Molecule Benchmarks Exceed 40-Qubit Simulation Limit This accomplishment surpasses the previously established 40-qubit limit for state-vector-based quantum circuit simulations, opening new avenues for validating quantum algorithms intended for future fault-tolerant quantum computers. Implementing IQPE within the “chemqulacs-gpu” quantum chemistry simulator, they tackled a 42-spin-orbital system for a water molecule, leveraging qubit reduction technology to manage computational demands. The breakthrough relied on harnessing the power of 1,024 NVIDIA H100 GPUs on AIST’s ABCI-Q system, a feat that required overcoming substantial inter-GPU communication bottlenecks. This parallel computing approach allowed the researchers to extend the scale of quantum circuit simulations beyond prior constraints, enabling the analysis of more complex molecular systems relevant to drug discovery and materials science.
The team not only benchmarked circuit scale with the iron sulfide molecule but also calculated a 42-spin-orbital system for water, demonstrating the versatility of their approach. I am delighted that the team, led by two young researchers, Yusuke Teranishi and Shoma Hiraoka, persevered throughout the effort, and that, with prompt support from the ABCI-Q operations staff, we were able to achieve one of the world’s largest results. Source: https://www.prnewswire.com/news-releases/worlds-largest-quantum-circuit-simulation-for-quantum-chemistry-achieved-on-1-024-gpus-302730945.html Tags: