NVIDIA cuQuantum SDK: Accelerating Quantum Circuit simulation II – cuStateVec
发布时间：2024年6月9日 08:41
87
(37:44) Dr. Shinya Morino, Nvidia -- Presentation from 2023 Workshop on Quantum Computing: Devices, Cryogenic Electronics and Packaging (QC-DCEP) ...
Summary: The cuStateVec library is a part of NVIDIA’s cuQuantum SDK and accelerates state vector simulations on NVIDIA GPUs. It provides APIs to accelerate gate applications, sampling, and other common operations in state vector simulations on single GPUs. The library also provides APIs to scale out state vector simulations on multiple GPUs and multiple nodes. Two cuStateVec-based simulators are released in the NVIDIA cuQuantum Appliance for IBM’s Qiskit/Qiskit-Aer with multi-node support and Google’s Cirq/qsim with multi-GPU support. This presentation will describe the use cases of the cuStateVec API in these simulators and the optimizations to realize large-scale high-performance state vector simulations.
Shinya Morino is a principal software engineer in the CUDA Math Libraries group at NVIDIA. He joined NVIDIA in 2013 and engaged CUDA trainings and developer support in various industrial fields as a solution architect. As a developer of a quantum computing simulator, he has prototyped a GPU-accelerated state vector simulator in the NVIDIA AI Technology center. With the launch of cuQuantum, he joined cuQuantum development and moved to the CUDA Math Libraries group to drive cuStateVec development. Shinya holds a Ph.D in engineering from the University of Tokyo, Japan.
Summary: The cuStateVec library is a part of NVIDIA’s cuQuantum SDK and accelerates state vector simulations on NVIDIA GPUs. It provides APIs to accelerate gate applications, sampling, and other common operations in state vector simulations on single GPUs. The library also provides APIs to scale out state vector simulations on multiple GPUs and multiple nodes. Two cuStateVec-based simulators are released in the NVIDIA cuQuantum Appliance for IBM’s Qiskit/Qiskit-Aer with multi-node support and Google’s Cirq/qsim with multi-GPU support. This presentation will describe the use cases of the cuStateVec API in these simulators and the optimizations to realize large-scale high-performance state vector simulations.
Shinya Morino is a principal software engineer in the CUDA Math Libraries group at NVIDIA. He joined NVIDIA in 2013 and engaged CUDA trainings and developer support in various industrial fields as a solution architect. As a developer of a quantum computing simulator, he has prototyped a GPU-accelerated state vector simulator in the NVIDIA AI Technology center. With the launch of cuQuantum, he joined cuQuantum development and moved to the CUDA Math Libraries group to drive cuStateVec development. Shinya holds a Ph.D in engineering from the University of Tokyo, Japan.
Additional videos from the QC-DCEP Workhop can be accessed at https://attend.ieee.org/qc-dcep.
