The project benchmarks quantum circuit simulation performance across different hardware architectures, including CPUs, GPUs, and FPGAs. Initial testing on CPUs and GPUs establishes a performance baseline and identifies opportunities for optimization, after which FPGA-based accelerators, specifically Alveo Cards such as the U55, are introduced to significantly accelerate quantum circuit simulations. Simulations are carried out in double precision to ensure high numerical accuracy, and the FPGA's parallel processing capabilities make it well suited to complex quantum workloads involving large datasets and intricate computations. The implementation leverages multiple HBM banks for high-bandwidth memory access and supports multi-card execution to scale simulations across several FPGAs.

Alongside benchmarking, the project develops a QSIM library tailored for hybrid quantum algorithms that combine classical and quantum computing techniques. The library incorporates seven configurable noise models for realistic simulation of quantum behaviour, and is being extended with kernel-level native gate implementations to maximize performance on the underlying hardware. Once optimized for FPGA-based acceleration, these hybrid quantum algorithms can be ported to and executed on FPGAs, improving simulation speed while making quantum simulations more practical for real-world applications. The ultimate goal is an FPGA-accelerated quantum simulation environment that efficiently supports hybrid quantum-classical algorithms and enables cross-architecture benchmarking, contributing to advancements in both quantum computing research and hardware utilization.

• Cross-Architecture Benchmarking (CPU vs GPU vs FPGA): The testbed provides a framework for comparing execution time, speedup and accuracy across CPU (AMD EPYC 7742), GPU (NVIDIA RTX A4000) and FPGA backends. Benchmarking shows the FPGA delivering acceleration of up to roughly 12 qubits over CPU and GPU and recording the most performance wins across the tested workloads (FPGA 12, GPU 7, CPU 2), helping teams select the right hardware for a given circuit size.
• Optimization in Finance and Logistics: Quantum simulators allow for the development and testing of quantum algorithms to solve optimization problems such as portfolio optimization, risk management, and logistics. Algorithms like Grover’s search and QAOA can be tested on simulators, enabling industries to explore quantum solutions to large-scale, complex optimization tasks without needing quantum hardware immediately.
• Noise modelling using density matrix backend: Simulating quantum field theories and particle interactions is a formidable challenge for classical computers. Quantum simulators allow physicists to explore fundamental aspects of particle interactions, high-energy physics, and quantum states evolution phenomena like energy transfer in quantum batteries providing insights into areas that are otherwise computationally intractable.
• Climate Modeling and Weather Forecasting: Quantum simulators can be used to model chaotic and highly complex systems involved in climate science, such as fluid dynamics and atmospheric processes. Simulating these systems may lead to more accurate predictions in climate modeling and weather forecasting, potentially offering improved long-term and short-term predictions through quantum-enhanced models.

• High-Performance Quantum-Classical Algorithm Simulation: Enables quantum researchers to simulate and benchmark complex quantum circuits efficiently using a hybrid CPU-GPU-FPGA setup, for algorithm development.
• Interoperability with different quantum languages: The integration and testing of quantum machine learning (QML) models, such as quantum neural networks (QNNs), across different hardware architectures to optimize performance.
• Benchmarking and Performance Analysis: Provides a framework for comparing execution times and accuracy of quantum simulations on CPU, GPU, and FPGA, supporting hardware selection and optimization for specific workloads.
• FPGA-Accelerated Quantum Computing: Leverages FPGA hardware to accelerate computationally intensive quantum operations, enabling faster processing and reduced energy consumption.
• Interactive Development Environment: Offers a Python-based interface compatible with Jupyter notebooks, VS Code, and Colab for accessible coding, debugging, and optimization of quantum circuits.

• Hardware: AMD EPYC CPU, NVIDIA RTX A4000 GPU with CUDA Driver 523.0, Xilinx Alveo U55C/U250 FPGA.
• Software: Python 3.8.10, Qiskit 0.46, PYNQ 3.0.1, CUDA/cuBLAS 12.6, Vitis, Vivado.
• Communication: PCIe protocol for CPU-GPU and CPU-FPGA data transfer.
• Memory: 16GB HBM2 for FPGA, 16GB VRAM for GPU.

• QSIM open source library Version V1.
• Linux (Ubuntu 20.04 recommended) with support for Python 3.8.10.
• Backed by C-DAC high-performance computing infrastructure with FPGA accelerators for quantum simulation through Qniverse.in .
• Simulation can also be run on a CPU Aer simulator for smaller circuits with the option for statvector and density matrix backend supports CPU and GPU other than FPGA.

Abhishek Tiwari, Scientist F

Embedded Systems Group,

C-DAC, Noida

Email: abhishek[at]cdac[dot]in