Highlights of LeanCP

  • All the parallel 3D-FFTs involved in computing the density are invoked at once. As a parallel 3D-FFT enters a communication phase, another begins its compute phase.
    At the
  • same time the 3D-FFTs are computing the density, the non-local and kinetic energy of non-interacting electrons are, also, available for assignment to physical processors.
    The bottleneck occurs when the states are evolved using the solver and orthnormality is enforced.
    • A sophisticated orthnormalization scheme is required at large numbers of processors.

  • Published by Lotus® Freelance Graphics®