Overview
What this challenge is about.
Profile the serial solver to identify the hot loops (likely: pressure-Poisson solve, advection step, BC update). Decompose the domain via 2D block-structured partitioning. Implement MPI halo-exchange with non-blocking sends + receives. Use a parallel Krylov solver (PETSc) for the pressure step. Run strong-scaling experiments at 16, 32, 64, 128, 256 ranks on the cluster. Produce a roofline analysis (peak FLOPS vs memory bandwidth). Deliver MPI source code, scaling-study CSV + plots, and a 10-page writeup including roofline + bottleneck analysis.
The Brief
What you'll do, and what you'll demonstrate.
Parallelize a serial CFD solver with MPI to 256 cores, achieving at least 60 percent parallel efficiency with documented strong-scaling + roofline evidence.
Earning criteria — what you'll demonstrate
- Apply domain decomposition to a real PDE-based solver
- Implement non-blocking MPI halo exchange correctly
- Use parallel Krylov solvers (PETSc) for elliptic problems
- Conduct strong-scaling + roofline analysis to identify bottlenecks
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.