opit

OPIT COMP-5002 Parallel and Distributed Computing course index

Use Python multiprocessing (Pool) to speed up a CPU-bound task, compare sequential vs parallel map, implement cpu_intensive_task, run and record timings, and explain why processes bypass the GIL compared with threads.

Implement Python threading to demonstrate a race condition, fix it with locks, and compare CPU-bound vs I/O-bound performance to observe the GIL; includes unsafe/safe counter examples, a performance script, and an analysis file.

Implement a distributed Pi estimator with mpi4py collectives. Broadcast N, partition intervals per rank, compute local trapezoid sums, and reduce to a global result. Run with mpiexec for multiple process counts, record timings, and answer analysis on bcast vs reduce/allreduce and workload decomposition.

Implement a bounded buffer (producer/consumer) in Python using threading.Lock and threading.Condition. Complete put/get with while+wait/notify loops, run multithreaded producers and consumers, observe waiting messages, and write a brief analysis on condition variables, wait loops, and mutual exclusion.

Use Dask DataFrame to parallelise a groupby-mean aggregation on a large synthetic dataset, compare timings with Pandas, run locally with a Dask LocalCluster, and answer analysis on performance, API similarity, lazy evaluation, and what Dask abstracts vs multiprocessing/MPI.