Benchmarking Quantum Chemistry Simulations on Heterogeneous Systems using CUDA Quantum

The integration of GPUs in quantum chemistry simulations is crucial due to complex computational demands, particularly when studying molecular structures of intermediate and large sizes such as CO, H2O, NH2-, and CO2, etc. Utilizing the CUDA Quantum software development kit, we enable scalable simulations on heterogeneous systems. Our research, performed on NERSC's Perlmutter system, highlights large-scale multi-GPU simulations and compares the performance between commonly used approaches such as VQE, the Adaptive VQE and the Number-Preserving Ansatz algorithms. Furthermore, we leverage domain-specific knowledge to enhance the efficiency of quantum computations, thereby advancing the scope and precision of our simulations.