Instructor: Kevin Green (University of Saskatchewan)

Title: Scientific computing with PETSc

Introduction to scientific computing using PETSc.

Objectives:

• Give general intro to numerical linear algebra with PETSc.
• Introduce PETSc solvers for ODEs and PDEs: how to set up and solve linear, nonlinear, and initial value problems in PETSc.
• Introduce HPC with PETSc, look at PETSc data structures built on top of MPI.
• Provided with a series of examples and practical exercises, the attendees will run code on the cluster.

Contents:

• Morning:
  • what is PETSc, and why it is useful
  • finding documentation for PETSc
  • loading PETSc on Cedar
  • the main data structures we will be working with (Vec, Mat, DMDA) and how to manipulate them in parallel
  • solving linear systems by preconditioned iterative methods
  • solving linear systems by direct methods
• Afternoon:
  • setting up and solving nonlinear problems
  • something went wrong! checking your Jacobians
  • setting up and solving initial value problems
  • profiling your PETSc code. Scaling
  • estimating memory usage

Duration: 6 hours

Level: intermediate

Prerequisites: Introduction to HPC session and programming experience in C or Fortran