Overview
What this challenge is about.
Implement Jones-Plassmann graph coloring in CUDA (or HIP if AMD hardware available). Input: a 12-million-node graph in CSR format (compressed sparse row). Output: a valid coloring (no adjacent same-color nodes) with low color count (target: within 1.2x of greedy CPU baseline). Benchmark on an Nvidia A100 (or roughly equivalent) and report time-to-color, memory bandwidth used, and quality (color count) vs the CPU greedy baseline. Deliver source, a Jupyter benchmark notebook, and a 5-page memo on shipping decision + cases where GPU underperforms (small graphs, irregular structure).
The Brief
What you'll do, and what you'll demonstrate.
Reduce 12-million-node graph coloring from 90s CPU to under 10s GPU while keeping color count within 1.2x of the CPU baseline.
Earning criteria — what you'll demonstrate
- Reason about GPU memory hierarchy and coalesced access for graph workloads
- Implement a parallel graph algorithm that respects independence constraints
- Benchmark GPU kernels honestly (kernel time, transfer time, end-to-end)
- Recommend shipping decisions that respect where the new approach degrades
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.