COMP-5002 – Lab 3 • Parallel Computation with Process Pools

Module Module 4. Parallelism via Multiprocessing
Objective Build a parallel program using multiprocessing.Pool to accelerate a CPU-bound task and compare it with sequential execution.

Prerequisites

• Python 3 installed.
• Git installed and basic familiarity with clone, add, commit, push.
• Concepts from Module 4:
  • Why multiprocessing can help CPU-bound work.
  • Process creation (multiprocessing.Process).
  • Process pools (multiprocessing.Pool) and map.
  • Pool lifecycle and the with context manager.
  • Pickling and inter-process communication.
  • The need for an if __name__ == "__main__": guard in multiprocessing programs.

Files Provided

• README.md this file
• lab3_parallel_map.py starter with function skeletons and a main block
• analysis.md where you record timings and answer the analysis questions

Tasks

General instructions

• Clone your GitHub Classroom repository.
• Modify lab3_parallel_map.py to complete the tasks.
• Keep the supplied benchmark structure so sequential and parallel runs process the same data.
• Record results in analysis.md.
• Commit frequently and push before the deadline.

---

Task 1 — Implement the CPU-intensive task

1. Open lab3_parallel_map.py.
2. Complete cpu_intensive_task(n) with a deterministic CPU-bound calculation.
3. A suitable implementation is math.factorial(n).
4. Keep the worker function at module scope. Functions passed to a process pool must be serializable by the multiprocessing machinery.

Do not add sleeps or I/O to this function. The purpose of this lab is to measure CPU-bound parallelism.

---

Task 2 — Implement sequential execution

In run_sequential(data):

1. Process every value in data with cpu_intensive_task.
2. Preserve the input order.
3. Return the complete list of results.

The sequential run and the pool run must perform the same work.

---

Task 3 — Implement parallel execution with Pool.map

In run_parallel_map(data, pool_size):

1. Create a multiprocessing.Pool with pool_size workers using a with block.
2. Apply pool.map(cpu_intensive_task, data).
3. Return the complete result list.

Do not create one process manually for every input value. This task is specifically about process pools.

---

Task 4 — Run and verify

1. Run python lab3_parallel_map.py.
2. The script reports:
   • Python version;
   • multiprocessing start method;
   • number of tasks;
   • pool size;
   • sequential time;
   • parallel time;
   • speedup.
3. The script also verifies that the sequential and parallel result lists are identical.
4. If your machine is unusually slow or fast, you may adjust VALUES_UPPER_BOUND modestly. Keep NUMBER_LIST_SIZE large enough to provide multiple tasks to the pool.

The supplied defaults are intended to make process-pool overhead small enough for the experiment to be meaningful while keeping runtime practical on typical student hardware. A speedup is not guaranteed on every machine.

---

Task 5 — Analysis (analysis.md)

Answer the following:

1. Speedup Compute Sequential / Parallel. Interpret the result, including the possibility of no speedup.
2. Why processes can help Explain how separate Python processes can execute CPU-bound work concurrently and how this differs from normal GIL-enabled threading.
3. Pool.map Explain why Pool.map is convenient for data-parallel workloads and what ordering guarantee it provides.
4. Overheads Identify process startup, task scheduling, serialization/pickling, inter-process data transfer, result collection, and load imbalance.
5. Pool size Explain why creating more worker processes than useful CPU resources or tasks can reduce performance.
6. Correctness Explain why timing alone is insufficient and why the sequential and parallel outputs must be checked for equality.

---

Submission

1. Ensure lab3_parallel_map.py runs successfully and the result verification passes.
2. Ensure analysis.md includes your recorded environment, timings, speedup, and answers.
3. Stage: git add lab3_parallel_map.py analysis.md (or git add .)
4. Commit: git commit -m "Complete Lab 3 Parallel Map"
5. Push: git push origin main (or your default branch)
6. Verify on GitHub that lab3_parallel_map.py and analysis.md are updated.