SYCL QPU: an LLVM-based QPU simulation framework built using DPC++

Current limitations in quantum hardware have made high-fidelity simulation an essential tool for prototyping and validating quantum algorithms, particularly those involving hybrid quantum-classical workflows. However, building flexible and scalable simulators that support heterogeneous hardware remains a technical challenge. This work introduces a QPU simulation framework based on LLVM and implemented using Intel’s open-source DPC++ (Data Parallel C++) compiler, which extends the SYCL programming model for heterogeneous computing. It builds on key abstractions in SYCL to target native CPU backends and CUDA devices while allowing fine-grained control over compilation and execution. This design empowers developers to focus on algorithm development without needing to manage low-level hardware details, while leveraging scalable simulation capabilities to bridge the current gap between classical simulation and future QPU execution. This framework enables hybrid execution on both CPUs and NVIDIA GPUs, supporting seamless integration with Qiskit AER.

keywords: Heterogeneous computing, Compilers, Quantum computing