Accelerated Quantum-Inspired Optimization for the 0-1 Knapsack Problem: A CUDA and SYCL Benchmark

At present, Noisy Intermediate-Scale Quantum (NISQ) era technology is prevalent but very restrictive. These are quantum computers with limited qubits and significant noise. While lacking the error correction of fully fault-tolerant quantum computers, NISQ devices offer potential for exploring near-term applications and provide a glimpse into the future of quantum computing. Waiting for higher qubit counts and fidelity is not an option in this quantum computing race; software and algorithm development must proceed independently of hardware advancements.

In our research, oneAPI is used to deploy our Quantum-Inspired algorithm to solve NP-Hard combinatorial optimization problems with discrete variables and constraints. Specifically, we tackle the 0-1 Knapsack problem (0/1 KP) for a large dataset and validate our algorithm, accelerated on NVIDIA GPUs using CUDA, with a classical approach based on dynamic programming. The SYCLomatic tool is used to convert the CUDA code to SYCL, which is benchmarked on various architectures to ensure universal portability, and the results are compared.

×


Register for the oneAPI DevSummit hosted by UXL:

Register Now