High-Performance Statevector Simulation for TKET and Selene with NVIDIA cuStateVec
ORAL
Abstract
Exact statevector simulation is a vital tool for the development of quantum algorithms and the benchmarking of quantum hardware. As the scale of these systems increases, the demand for fast, scalable, and accessible simulation libraries becomes critical for both quantum software developers and hardware engineers.
Here, we introduce two GPU-accelerated statevector simulation backends for Quantinuum’s software stack based on NVIDIA’s cuStateVec library: pytket-custatevec and selene-custatevec. The former is a community-driven extension for pytket; the latter provides a state-of-the-art backend for Selene, an emulation framework supporting Quantinuum's next-generation stack.
We detail the architecture of both backends and present performance benchmarks illustrating the acceleration enabled by cuStateVec, with portability from single-GPU to multi-GPU settings. We also discuss how these capabilities provide a practical path from prototype to hardware within Quantinuum’s software stack and outline general directions for extending backend coverage and workflow features.
Here, we introduce two GPU-accelerated statevector simulation backends for Quantinuum’s software stack based on NVIDIA’s cuStateVec library: pytket-custatevec and selene-custatevec. The former is a community-driven extension for pytket; the latter provides a state-of-the-art backend for Selene, an emulation framework supporting Quantinuum's next-generation stack.
We detail the architecture of both backends and present performance benchmarks illustrating the acceleration enabled by cuStateVec, with portability from single-GPU to multi-GPU settings. We also discuss how these capabilities provide a practical path from prototype to hardware within Quantinuum’s software stack and outline general directions for extending backend coverage and workflow features.
–
Presenters
-
Fabian Finger
- Quantinuum