HIGH PERFORMANCE COMPUTING AND SIMULATIONS

• Deterministic particle simulation algorithms
  • Survey of molecular dynamics (MD) simulation: spatiotemporal data locality in MD
  • Spatiotemporal multiscale algorithms
  • Fast computation of electrostatic interaction: O(N) fast multipole method
  • Multiple time stepping: fuzzy cluster dynamics
• Parallel computing frameworks
  • Parallel algorithm design: divide-conquer-"recombine" parallelization, spatial vs. particle vs. force vs. tuple decomposition, data-driven parallelization
  • Load balancing: wavelet-based computational space decomposition, recursive spectral bisection, spacefilling-curve decomposition, load diffusion
  • Scalability analysis
  • New architectures: multicore, graphics processing unit (GPU), artificial-intelligence accelerators, and quantum scientific computing
  • Hybrid message-passing + multithreading + data-parallel accelerator programming: MPI, OpenMP, CUDA, OpenMP target, SYCL, Qiskit
  • Performance profiling and optimization: roofline model
  • Optimization of parallel scientific applications
• Deterministic continuum simulation algorithms
  • Survey of quantum dynamics (QD) simulation
  • Fast solutions of partial differential equations (PDE): O(NlogN) fast Fourier transform, O(N) wavelet transform, O(N) multigrid method
  • O(N) Lanczos and Davidson algorithms for the eigenvalue problem
  • Newton Krylov-subspace solvers for nonlinear equations
  • Algebraic BLASification
• Multiscale particle-continuum simulation
  • Hybridization techniques: minimizing model-boundary artifacts, modular algorithm design, adaptive hybridization
  • O(N) multiscale optimization
  • Space-time multiscaling
• Stochastic simulation algorithms
  • Survey of Monte Carlo (MC) simulation: estimator, importance sampling, Markov chain, Metropolis algorithm
  • Simulated annealing
  • Kinetic MC: master equation, Poisson process
• Distributed scientific computing
  • Grid/cloud programming
  • Grid/cloud enabling parallel applications: virtualization-aware scientific algorithms based on data-locality principles
  • Distributed MC applications: parallel replica and replica exchange MC
• Scientific data visualization and learning
  • Interactive visualization of large datasets in immersive virtual environment: hierarchical/probabilistic culling algorithms
  • Topology analysis: shortest-path circuits, parallel graph algorithms
  • Scientific machine learning
  • Data compression
  • Singular value decomposition for low-rank approximations
  • Integration of simulation, visualization and machine-learning workflows on Grid/cloud
• Advanced scientific computing methods
  • Local and global optimization in molecular dynamics: physically-based preconditioning of iterative solvers, basin-hopping algorithms, disconnectivity-graph analysis of the energy landscape
  • Accelerated long-time dynamics: path-integral sampling, ensemble mean-field method, hyper dynamics, activation-relaxation metadynamics
  • Explorative search: pathfinders