CS 5220
Applied High-Performance and Parallel Computing
CS 5220 introduces performance optimization and parallel programming for scientific and engineering applications. Topics include serial performance tuning, parallel architectures, and programming with MPI, OpenMP, and CUDA, as well as emerging systems such as the Cerebras Wafer-Scale Engine. Students learn key parallel patterns, including data partitioning, synchronization, and load balancing, through examples from numerical and scientific computing. Basic C/C++ programming skills and familiarity with computer architecture (e.g., CS 3410) are expected; prior parallel programming experience is not required.
Topics
- Analyze the performance of a serial or parallel program by measuring runtime, identifying bottlenecks, and relating observed behavior to the underlying architecture and memory hierarchy.
- Apply performance models, including the Roofline model and strong and weak scaling analysis, to predict achievable performance and to determine whether a kernel is compute-bound or memory-bound.
- Implement parallel solutions to scientific and data-intensive problems using distributed-memory and GPU programming models, and explain how these extend to multi-GPU execution.
- Evaluate design trade-offs in data partitioning, synchronization, and load balancing, and justify an implementation choice using empirical scaling results.
- Students in CS 5220 additionally complete HW5, Programming emerging accelerators: Cerebras CS-3 (simulator) with CSL.
Official roster learning outcomes
Prerequisites & course information
background equivalent to: CS 3410.
Sources
- Course catalogue ↗2026-2027
- Catalogue entry ↗
Catalogue variants
- Applied High-Performance and Parallel Computing · 4.0 · Graduate