Session 09 — Load Balancing & Case Studies

Session 09 applies Unit III to real algorithms. These interactive demos make the load-balancing idea physical — open them full-screen and show them live in class.

▶ Open the Interactive Slides

1. Static vs Dynamic Load Balancing

Eight uneven tasks, four workers. Watch static leave workers idle while one is swamped — and dynamic keep everyone busy. The core lesson, animated.

2. Parallel Matrix Multiply

Watch an output matrix fill in parallel across workers, each owning a band of rows. The embarrassingly-parallel classic (Lab 1 & 3 preview).

3. Parallel Prime Generation — Lab 4 (OpenMP / C)

Real code, not an animation. Counts primes to 10,000,000 three ways — serial, schedule(static), schedule(dynamic) — and prints a per-thread breakdown. Static leaves 27% of the machine idle; dynamic leaves 0%. Same answer every time.

Running Lab 4: gcc -O2 -fopenmp 03-parallel-primes.c -o primes then OMP_NUM_THREADS=4 ./primes. No compiler? Paste it into OnlineGDB and add -fopenmp to the compiler flags. All three runs must report 664,579 primes — only the clock changes.

The first two demos need no installs and run in any modern browser. Lab 4 is your first taste of the real OpenMP code that Unit IV is built on.