Parallelism in Numerical Simulations

In this lecture we cover Demmel's lectures 11 and 12 and lecture 13.

We will examine the following four broad categories of simulations.

  • Discrete event systems
  • Particle systems
  • Lumped variables depending on continuous parameters
  • Continuous variables depending on continuous parameters
    • For each category, we give application examples and discuss parallelism. We will identify the following computation bottlenecks in simulations.

  • Matrix-vector multiplication
  • Solving linear systems (dense or sparse)
  • Graph partitioning
  • Finding eigenvalues