06-parallelism.md

06 - Parallelism

Goal

Use multiple CPU cores for CPU-bound work.

Task

Implement prime counting.

public interface IPrimeCounter
{
    int CountPrimes(int fromInclusive, int toInclusive);
}

Create:

• SequentialPrimeCounter
• ParallelForPrimeCounter
• PlinqPrimeCounter

Step-by-step

1. Implement a simple IsPrime method.
2. Count primes sequentially.
3. Use Parallel.For.
4. Use PLINQ.
5. Benchmark each approach.
6. Try different degrees of parallelism.

Benchmark parameters

• 1 to 10,000
• 1 to 100,000
• 1 to 1,000,000

Questions to answer

• Is parallel always faster?
• What happens for small inputs?
• What happens when degree of parallelism is too high?
• What is the difference between CPU-bound and I/O-bound work?

Key takeaway

Parallelism helps when the work is CPU-heavy enough to justify scheduling and coordination overhead.