Overview
What this challenge is about.
Use PETSc (or Eigen's IterativeLinearSolvers + a custom AMG) to benchmark: (1) GMRES + ILU(0), (2) GMRES + AMG (BoomerAMG), (3) BiCGStab + ILU(0), (4) BiCGStab + AMG, on three meshes (around 2M, 6M, 12M unknowns) from the supplier's anonymized crash-load matrix set. Report iteration count, wall time, memory peak, and parallel scaling from 1 to 64 cores. Provide a 7-page recommendation memo with a chosen default + per-mesh fallback policy, and a reproducible benchmark harness the simulation team can rerun on new mesh classes.
The Brief
What you'll do, and what you'll demonstrate.
Benchmark Krylov solver + preconditioner combinations on three crash-sim meshes and recommend a default that halves inner-solver wall time.
Earning criteria — what you'll demonstrate
- Compare direct vs Krylov-with-preconditioner solvers on real industrial sparse systems
- Configure and tune ILU(0) and algebraic multigrid (AMG) preconditioners
- Measure parallel scaling and identify the strong-scaling knee for each solver
- Recommend a solver policy that respects per-mesh characteristics
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Career mappings coming soon.