First, I would like to express my sincere gratitude for your excellent work, which has been tremendously helpful in our research.
I am currently performing STEM image simulations with the following parameters using an NVIDIA A40 GPU. While the GPU utilization reaches approximately 90%, the computation time for a single simulated image remains between 1-2 hours.(5×5×5 Supercell)
Could you please advise on potential optimization methods to improve computational efficiency?
potential = Potential(
atoms,
gpts=2048,
projection='finite',
slice_thickness=0.5,
parametrization='lobato'
)

    probe = Probe(energy=300e3, semiangle_cutoff=21.3, defocus=0)
    probe.grid.match(potential)

    haadf = AnnularDetector(inner=50, outer=200)
    scan_end = (potential.extent[0], potential.extent[1])
    gridscan = GridScan(start=(0, 0), end=scan_end, sampling=0.3)